UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Determination of small molecules in food matrices by molecularly imprinted polymers and surface enhanced… Gao, Fang 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2016_february_gao_fang.pdf [ 17.18MB ]
Metadata
JSON: 24-1.0223172.json
JSON-LD: 24-1.0223172-ld.json
RDF/XML (Pretty): 24-1.0223172-rdf.xml
RDF/JSON: 24-1.0223172-rdf.json
Turtle: 24-1.0223172-turtle.txt
N-Triples: 24-1.0223172-rdf-ntriples.txt
Original Record: 24-1.0223172-source.json
Full Text
24-1.0223172-fulltext.txt
Citation
24-1.0223172.ris

Full Text

Determination of Small Molecules in Food Matrices by Molecularly Imprinted Polymers and Surface Enhanced Raman Spectroscopy by  Fang Gao  B.Sc., Nankai University, 2011  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  Doctor of Philosophy in The Faculty of Graduate and Postdoctoral Studies (Chemistry)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  January 2016  © Fang Gao, 2016   ii Abstract  The research work focuses on developing novel methods for determining small molecules in food matrices using molecularly imprinted polymers (MIPs) and surface enhanced Raman spectroscopy (SERS). MIPs are synthesized as artificial antibodies towards target molecules utilizing interactions between templates and functional monomers to impress complementary binding sites on polymers. MIPs selectively isolate templates from food extracts. SERS technique provides rapid and sensitive detection of MIPs-separated molecules. Statistical analysis including unsupervised principal component analysis (PCA), supervised simple linear regression and partial least square regression (PLSR) is employed to analyze SERS spectra.     Chloramphenicol in milk and honey was determined using MIPs-packed solid phase extraction cartridge to isolate chloramphenicol from food matrices and dendritic silver to acquire SERS spectra of the eluted chloramphenicol. These spectra obtained from different spiked contents (0, 0.1, 0.5, 1, 5 ppm) of chloramphenicol in milk and honey were analyzed by PCA and PLSR (R > 0.9).      MIPs particles were spread onto a thin layer chromatography (TLC) plate to determine Sudan I in paprika powder.  Separation of Sudan I from paprika extract by an MIPs-TLC plate takes 30-40 s. SERS spectra obtained from Sudan I spot on the plate can be acquired within 1 s with gold colloid serving as SERS active substrate. A PLSR (R2 = 0.978) model was constructed based on spiking levels (5, 10, 40, 70 and 100 ppm) of Sudan I in paprika powder.      Histamine level in canned tuna was investigated using MIPs-polyvinyl chloride (PVC)-SERS method. MIPs-PVC films (recognition element) selectively extracted histamine from tuna extract. A gold colloid solution served as an eluting solvent to extract histamine from MIPs-PVC   iii film and conducted a SERS detection of histamine. A PLSR model (R2 = 0.947, RMSECV = 3.526) was built on SERS spectra of histamine with different spiking levels (3, 30 and 90 ppm) in canned tuna.     The spectral results suggest the powerful separation of MIPs and sensitive detection of SERS. With statistical analysis, we have confirmed that SERS signals obtained by this MIPs-SERS approach rapidly and accurately quantify chloramphenicol in milk and honey, Sudan I in paprika powder and histamine in canned tuna.   iv Preface  The following sections are partially based on my published publications and submitted manuscript. Besides the contents arising from my research work, I have published a book chapter as: Gao, F., Lu, X. (2015), Detection of pesticides in foods by enzymatic biosensors in: Yada, R.Y. (Ed.) Improving and tailoring enzymes for food quality and functionality. Woodhead Publishing.     Chapter 3 is mainly describing the work on my first publication: Gao, F., Feng, S., Chen, Z., Li-Chan, E., Grant, E., Lu, X. (2014), Detection and quantification of chloramphenicol in milk and honey using molecularly imprinted polymers: Canadian penny-based SERS nano-biosensor. Journal of Food Science, 79: 2542-2549. Dr. Lu designed the study. Dr. Grant and Dr. Li-Chan provided valuable suggestions and helped to edit the manuscript. I was responsible for partially designing, performing experimental work, analyzing the data and writing the manuscript.     A version of Chapter 4 has been published as: Gao F., Hu Y., Chen D., Li-Chan E.C.Y., Grant E., Lu X. (2015), Determination of Sudan I in paprika powder by molecularly imprinted polymers–thin layer chromatography–surface enhanced Raman spectroscopic biosensor. Talanta, 143: 344-352. I designed the study under the guidance of Dr. Lu, performed experimental work, analyzed the data and prepared the manuscript. Dr. Chen provided the portable Raman spectrometer. Dr. Lu and Dr. Grant helped to edit the manuscript.     Chapter 5 is based on the publication as: Gao F., Grant E., Lu X. (2015), Determination of histamine in canned tuna by molecularly imprinted polymers-surface enhanced Raman spectroscopy. Analytica Chimica Acta, 901: 68-75. I developed the research work and wrote the manuscript. Dr. Lu and Dr. Grant were responsible for editing the manuscript.   v     The published contents including text, figures and tables are used with permission.    vi Table of Contents  Abstract .......................................................................................................................................... ii	Preface ........................................................................................................................................... iv	Table of Contents ......................................................................................................................... vi	List of Tables ................................................................................................................................ ix	List of Figures ................................................................................................................................. x	List of Abbreviations ............................................................................................................... xviii	Acknowledgements .................................................................................................................... xxi	Dedication .................................................................................................................................. xxii	 Introduction ................................................................................................................1	Chapter 1:1.1	 Food analysis: common methods ....................................................................................... 1	1.2	 Raman spectroscopy .......................................................................................................... 6	1.2.1	 Raman scattering ......................................................................................................... 7	1.2.2	 Confocal Raman spectroscopy .................................................................................... 9	1.2.3	 Ways to improve Raman signals ............................................................................... 12	1.3	 Data analysis .................................................................................................................... 20	1.3.1	 Preprocessing ............................................................................................................ 21	1.3.2	 Multivariate statistical analysis ................................................................................. 24	1.4	 Molecularly imprinted polymers ...................................................................................... 29	1.4.1	 Mechanism and categories ........................................................................................ 29	1.4.2	 Applications .............................................................................................................. 35	1.5	 Objectives and outline ...................................................................................................... 37	  vii  Experimental details .................................................................................................40	Chapter 2:2.1	 Synthesis of MIPs and ways to optimize their properties ................................................ 40	2.2	 Preparation and discussion of SERS substrates ............................................................... 47	2.3	 Examples of data analysis ................................................................................................ 52	 Detection and quantification of chloramphenical in milk and honey using Chapter 3:molecularly imprinted polymers: Canadian penny-based SERS nano-biosensor .................59	3.1	 Introduction ...................................................................................................................... 59	3.2	 Experimental section ........................................................................................................ 62	3.2.1	 Synthesis and characterization of MIPs towards chloramphenicol .......................... 63	3.2.2	 Preparation of penny-based silver nano-substrate .................................................... 64	3.2.3	 Performance of molecularly imprinted solid phase extraction on food samples ...... 64	3.2.4	 Spectral collection and data analysis Section ........................................................... 66	3.3	 Results and discussion ..................................................................................................... 67	3.3.1	 Assessment of MIPs separation of CAP in foods ..................................................... 67	3.3.2	 Quantitative analysis of MIPs-SERS nano-biosensor ............................................... 78	3.4	 Conclusion ....................................................................................................................... 84	 Determination of Sudan I in paprika powder by molecularly imprinted Chapter 4:polymers-thin layer chromatography-surface enhanced Raman spectroscopic biosensor ...85	4.1	 Introduction ...................................................................................................................... 85	4.2	 Experimental section ........................................................................................................ 90	4.2.1	 Synthesis and characterization of MIPs towards Sudan I ......................................... 91	4.2.2	 Fabrication of MIPs–TLC plate and developing ...................................................... 92	4.2.3	 Synthesis of gold colloid ........................................................................................... 93	  viii 4.2.4	 SERS detection of Sudan I in paprika extract and data analysis .............................. 94	4.3	 Results and discussion ..................................................................................................... 97	4.3.1	 Assessment of MIPs–TLC plate and developing ...................................................... 97	4.3.2	 SERS spectral acquisition and data analysis ........................................................... 103	4.3.3	 Portable Raman analysis ......................................................................................... 112	4.4	 Conclusion ..................................................................................................................... 113	 Determination of histamine in canned tuna by molecularly imprinted polymers Chapter 5:and surface enhanced Raman spectroscopy ............................................................................114	5.1	 Introduction .................................................................................................................... 114	5.2	 Experimental section ...................................................................................................... 118	5.2.1	 Synthesis of MIPs, fabrication and characterization of MIPs-PVC film ................ 119	5.2.2	 MIPs-PVC film: adsorption & desorption of histamine in tuna ............................. 120	5.2.3	 Preparation of gold colloid ...................................................................................... 121	5.2.4	 Determination of histamine by SERS and data analysis ......................................... 121	5.3	 Results and discussion ................................................................................................... 123	5.3.1	 MIPs-PVC film: adsorption & desorption of histamine in tuna ............................. 123	5.3.2	 SERS spectral collection and data analysis ............................................................ 132	5.4	 Conclusion ..................................................................................................................... 138	 Conclusion and outlook .........................................................................................140	Chapter 6:Bibliography ...............................................................................................................................143	   ix List of Tables  Table 1.1: Advantages and disadvantages of MIPs. ..................................................................... 29	Table 2.1: Examples of functional monomers and template molecules in non-covalent MIPs. .. 41	Table 2.2: Pairs of functional monomers and cross-linking monomers. ...................................... 43	Table 3.1: Spiked recoveries: molecularly imprinted solid phase extraction (MISPE) and non-imprinted solid phase extraction (NISPE). ................................................................................... 74	Table 3.2: Partial least squares regression (PLSR) models for the prediction of chloramphenicol concentration in foods. .................................................................................................................. 82	Table 4.1: Sudan I recoveries of extraction from paprika powder determined by HPLC (n = 3, value shown: mean ± standard error of the mean). ..................................................................... 100	Table 4.2: Peak assignments of Sudan I in normal Raman spectrum. ....................................... 105	Table 4.3: Comparison of silica gel TLC-SERS and MIPs-TLC-SERS with Sudan I standard solution (100 ppm, 2 µL) and Sudan I (100 ppm) spiked paprika extract (2 µL). ..................... 112	Table 5.1: Peak assignment of histamine in normal Raman spectrum. ...................................... 134	Table 5.2: Comparison of univariate linear regression models constructed based on different peaks and their intensities. .......................................................................................................... 138	   x List of Figures  Figure 1.1: Schematic illustration of a biosensor (recognition element, transducer and detector). 6	Figure 1.2: Diagram of Raman scattering (Stokes Raman with energy loss and anti-Stokes Raman with energy gain) and Rayleigh scattering. ........................................................................ 8	Figure 1.3: Comparison of SERS spectra (x axis: Raman shift in cm-1, y axis: Raman intensity in arbitrary units) of same sample under different incident lasers, SERS spectrum on the left shows high fluorescence background that swamps Raman peaks in contrast to the SERS spectrum on the right displays distinct Raman peaks with relatively low fluorescence background. (Left: SERS spectrum of Sudan I at 10 ppm obtained with 532 nm laser at 100 mW; Right: SERS spectrum of Sudan I at 10 ppm acquired from 785 nm laser at 25 mW.) ....................................... 9	Figure 1.4: (A) Confocal Raman spectroscopy; (B) Stray light induce large sample illumination with the oversized first pinhole; (C) Out of focus photons contribute to detector without the second aperture. ............................................................................................................................ 11	Figure 1.5: Optical excitation of localized surface plasmon resonance by oscillating electric vector of the incident electromagnetic wave. ............................................................................... 17	Figure 1.6: (A) Spontaneous Raman scattering under monochromatic incident light; (B) Coherent Raman scattering (coherently interacted pump and Stokes lasers excite coherently vibrating molecules, next, probe laser strikes Raman scattering of these vibrating molecules). . 20	Figure 1.7: Covalent imprinting employing forming (A) boronate ester bond, (B) imine groud and (C) ketal bond between template molecules and functional monomers. ............................... 31	Figure 2.1: Comparison of Raman spectra acquired from MIPs and NIPs. (each spectrum is averaged from 5 raw spectra) (A) histamine: 0.15 mmol (0 mmol for NIPs), methacrylic acid:   xi 0.65 mmol, ethylene glycol dimethylacrylate: 3.90 mmol; (B) histamine: 0.15 mmol (0 mmol for NIPs), methacrylic acid: 0.65 mmol, ethylene glycol dimethylacrylate: 3.28 mmol. .................. 46	Figure 2.2: SEM images of dendritic silver. ................................................................................ 47	Figure 2.3: TEM images of AuNPs. ............................................................................................ 49	Figure 2.4: Comparison of UV spectra. [Black: AuNPs, red: AuNPs diluted by ethanol (v/v, 1:1), green: AuNPs and 0.8 ppm histamine in ethanol (v/v, 1:1), blue: AuNPs and 4.0 ppm histamine in ethanol (v/v, 1:1), brown: AuNPs and tuna extract in ethanol (v/v, 1:1)] ................ 51	Figure 2.5: SERS spectra of undiluted tuna extract (concentration of histamine in this extract: 30 ppm) and AuNPs (v/v, 1:1).  (Four samples of undiluted tuna extract were mixed with AuNPs to acquire SERS spectra, respectively.) ............................................................................................ 52	Figure 2.6: Comparison of polynomial fitted SERS spectra (pink and green) at different polynomial orders to raw SERS spectra of Sudan I (black and red) acquired from the samples with the same concentration. [High orders (fourth to eighth) show good background removal compared to lower orders (first to fourth) displaying biased resultant spectra.] .......................... 55	Figure 2.7: Representatives of first and second derivatives of SERS spectra. Derivatives of raw SERS spectra of Sudan I (black and red) acquired from the samples with the same concentration show high noise and difficulty to differentiate them. Derivatives of SERS spectra after 7-point Savitzky-Golay smoothing mainly emphasize useful Raman peaks of Sudan I. .......................... 56	Figure 2.8: Comparison of PCA plots. [(A): PCA result (PC1: 99.56%, PC2: 0.25%) of raw SERS spectra of histamine at different concentrations. (B): PCA plot (PC1: 59.72%, PC2: 14.72%) of smoothed and background removed SERS spectra of these samples.] PCA plot of preprocessed spectra shows better separated clusters based on concentrations. .......................... 58	  xii Figure 3.1: Adsorption capacity of MIPs and NIPs to CAP at 50 ppm after 3 h incubation as a function of methanol percentage in aqueous solution. .................................................................. 68	Figure 3.2: (A) Isotherm of binding properties of molecularly imprinted polymers (MIPs) and non-imprinted polymers (NIPs); (B) Scatchard plot of MIPs and NIPs. Values are shown as mean ± standard deviation (n = 3) (initial concentration of CAP in 10% methanol: 25, 50, 100, 200, 300 ppm; time: 3 h). .............................................................................................................. 70	Figure 3.3: Kinetic study of molecularly imprinted polymers (MIPs) and non-imprinted polymers (NIPs) as a function of incubation time. Values are shown as mean ± standard deviation. (n = 3) (Initial concentration of CAP in 10% methanol: 50 ppm; time points: 10, 20, 30, 45, 60, 90, 120, 180 min). ....................................................................................................... 71	Figure 3.4: Illustration of molecularly imprinted solid phase extraction (MISPE) for chloramphenicol. ........................................................................................................................... 74	Figure 3.5: Scanning electron microscope image of silver dendrite formed on Canadian penny........................................................................................................................................................ 76	Figure 3.6: Representative Raman spectra of chloramphenicol [from top to bottom: MIPs-SERS spectrum of extract of CAP-spiked (1 ppm) skim milk, SERS spectrum of standard CAP solution (1 ppm), and normal Raman spectrum of solid crystal]. ............................................................... 77	Figure 3.7: Schematic model of orientation of chloramphenicol on silver dendrite. ................... 78	Figure 3.8: Schematic illustration of the development of MIPs-SERS nano-sensor to detect chloramphenicol in selective food matrices. ................................................................................. 79	Figure 3.9: Representative 2-dimensional principal component analysis (PCA) model for differentiation of whole milk containing various concentrations of chloramphenicol (PC1:   xiii 95.3%; PC2: 2.1%); (B) representative partial least squares regression (PLSR) model to predict actual concentrations of chloramphenicol in whole milk. ............................................................ 80	Figure 3.10: (A) Representative two-dimensional principal component analysis (PCA) model for separation of skim milk containing different concentrations of chloramphenicol (PC1: 93.8%; PC2: 2.8%); (B) representative partial least squares regression (PLSR) model to predict actual concentrations of chloramphenicol in skim milk. ......................................................................... 81	Figure 3.11: (A) Representative two-dimensional principal component analysis (PCA) model for separation of honey containing different concentrations of chloramphenicol (PC1: 73.3%; PC2: 9.5%); (B) representative partial least squares regression (PLSR) model to predict actual concentrations of chloramphenicol in honey. ............................................................................... 82	Figure 4.1: Schematic mechanism of developing molecularly imprinted polymers towards Sudan I. .................................................................................................................................................... 91	Figure 4.2: Representative HPLC chromatogram of Sudan I in 5 ppm spiked paprika extract. [Additional info: retention time: 5.665 min; column conditions: C18 column (µBondapak-C18, 10 µm, 3.9 mm × 300 mm, Waters, Milford, MA, USA); temperature: 25 °C; injection volume: 20 µL; gradient method: 0-6 min, acetonitrile : DI water =85 : 15, flow rate: 1 mL/min, 6-16 min, acetonitrile : DI water =95 : 5, flow rate: 1 mL/min, 16-20 min, acetonitrile, flow rate: 1.2 mL/min, 20-25 min, acetonitrile : DI water =95 : 5, flow rate: 1 mL/min.] ................................. 93	Figure 4.3: Schematic illustration of SERS spectral collection by portable Raman spectrometer (top) and corresponding SERS spectral features of Sudan I in range of 350-1650 cm-1 (bottom). [From bottom to top: averaged SERS spectra (n = 3) of Sudan I spiked paprika extract: 0 ppm (1), 1 ppm (2), 5 ppm (3), 10 ppm (4), 40 ppm (5), 70 ppm (6), 100 ppm (7); SERS spectra were collected by portable Raman spectrometer from the sample spots after MIPs-TLC developing.   xiv SERS spectrum of Sudan I spiked paprika extract from the sample spots after MIPs-TLC developing: 100 ppm (8) and normal Raman spectrum acquired from dried standard solution of 2 µL of 2000 ppm (9); spectra were collected by bench-top Raman spectrometer.] ....................... 95	Figure 4.4: (A) Isotherm of rebinding assays of molecularly imprinted polymers (MIPs) and non-imprinted polymers (NIPs). (B) Scatchard plot of MIPs. (Additional info: initial concentration: 5, 10, 20, 30, 45 ppm; time 3 h; values are shown as mean ± standard error of the mean, n = 3.) ................................................................................................................................. 99	Figure 4.5: Kinetic plots of MIPs and NIPs. (Additional info: initial concentration: 10 ppm; time intervals: 1, 5, 10, 20, 30, 60, 90, 120 min; values are shown as mean ± standard error of the mean, n = 3.) ............................................................................................................................... 100	Figure 4.6: Absorbance spectrum of gold colloid. ..................................................................... 101	Figure 4.7: Illustration of working mechanism of MIPs-TLC-SERS sensor. (A: plate before developing, B: plate after developing, C: plated after addition of Au colloid; spot 1: 0 ppm Sudan I spike paprika extract, spot 2: 100 ppm Sudan I standard solution, spot 3: 100 ppm Sudan I spiked paprika extract.) ............................................................................................................... 102	Figure 4.8: Average SERS spectra of Sudan I (collected in 49 s) in the wavenumber range of 350-1650 cm-1. [From bottom to top: normal Raman spectrum acquired from dried standard solution of 2 µL of 2000 ppm on clean glass slide (1), normal Raman spectrum from dried 2 µL of 1000 ppm of Sudan I standard solution on MIPs-TLC plate (2), SERS spectrum of 0 ppm (3) spiked paprika extract, 1 ppm (4) standard solution, 1 ppm (5) spiked paprika extract, 10 ppm (6) standard solution, 10 ppm (7) spiked paprika extract, 100 ppm (8) standard solution, 100 ppm (9) spiked paprika extract; SERS spectra were collected from the sample spots after MIPs-TLC developing; n = 8 for each concentration.] ................................................................................. 104	  xv Figure 4.9: Time profile of Sudan I SERS spectra of 100 ppm spiked paprika extract. (Additional info: from top to bottom: spot 1: acquisition 1: 0-49 s, acquisition 2: 50-98 s, acquisition 3: 99-147 s; spot 2: acquisition 1: 0-49 s, acquisition 2: 50-98 s, acquisition 3: 99-147 s. SERS spectra were collected from the sample spots after MIPs-TLC developing.) ............... 106	Figure 4.10: Averaged SERS spectra of Sudan I (collected in 1 s) in the wavenumber range of 703-1262 cm-1. [Sudan I spiked paprika extract: 0 ppm (7), 1 ppm (6), 5 ppm (5), 10 ppm (4), 40 ppm (3), 70 ppm (2), 100 ppm (1); SERS spectra were collected from the sample spots after MIPs-TLC developing; n = 8 for each concentration.] ............................................................... 107	Figure 4.11: Illustration of principal component analysis (PCA) models: (A) spectra in the wavenumber range of 350-1650 cm-1 (PC1: 57.4%, PC2: 34%; n =4 at each concentration); (B) spectra in the wavenumber range of 703-1262 cm-1 (PC1: 50%, PC2: 23.9%; n = 8 at each concentration). [Sudan I spiked paprika extract: 0 ppm (black), 1 ppm (red), 5 ppm (blue), 10 ppm (olive), 40 ppm (pink), 70 ppm (green), 100 ppm (purple); SERS spectra were collected from the sample spots after MIPs-TLC developing.] (n=5) ....................................................... 109	Figure 4.12: Illustration of regression models: (A) established PLSR regression based on spectra in range of 350-1650 cm-1; (B) established PLSR regression based on spectra in range of 703-1262 cm-1; (C) established linear regression based on band height at 721 cm-1 of spectra in range of 703-1262 cm-1 (values shown in: mean ± standard error of the mean, n = 8 at each concentration.). [Sudan I spiked paprika extract: 0 ppm (black), 1 ppm (red), 5 ppm (blue), 10 ppm (olive), 40 ppm (pink), 70 ppm (green), 100 ppm (purple); SERS spectra were collected from the sample spots after MIPs-TLC developing.] ................................................................. 110	Figure 4.13: Silica gel TLC plate developing. [Additional info: left: after spotting, right: after developing; on each plate, left: 2 µL of 100 ppm Sudan I standard solution, right: 2 µL of Sudan   xvi I (100 ppm) spiked paprika extract; distances of Sudan I spots to bottom edge were shown in cm.] ............................................................................................................................................. 112	Figure 5.1: Illustration of working mechanism of MIPs-PVC-SERS. ....................................... 118	Figure 5.2: Schematic illustration of polymerization of MIPs towards histamine. ................... 119	Figure 5.3: (A) Kinetic plots of MIPs-PVC and NIPs-PVC towards histamine at 100 ppm. (B) Isotherm of 1 h static study on MIPs-PVC and NIPs-PVC to histamine. (Values are shown as mean ± standard error of the mean, n = 3.) ................................................................................. 124	Figure 5.4: (A) The ionization of histamine at different pH. (B) The structure of L-histidine. 125	Figure 5.5: The comparison of adsorption capacities of MIPs-PVC and NIPs-PVC towards histamine and L-histidine at 100 ppm after 1 h incubation. (Values are shown as mean ± standard error of the mean, n = 3.) ............................................................................................................ 126	Figure 5.6: The adsorption capacities of MIPs-PVC and NIPs-PVC toward histamine at 100 ppm as a function of different concentrations of ammonium acetate in the solution after 1 h incubation. (Values are shown as mean ± standard error of the mean, n = 3.) ........................... 128	Figure 5.7: Averaged SERS spectra (n = 4) of histamine desorbed by AuNPs from MIPs-PVC film that was incubated with tuna extract (final concentration of histamine in this tuna extract: 30 ppm) for 5 min at different time points. (Desorption time: 1: 30 s, 2: 60 s, 3: 90 s, 4: 120 s, 5: 150 s, 6: 180 s; Raman shift range: 1082-1603 cm-1; exposure time: 1 s.) ........................................ 130	Figure 5.8: SERS spectra (4 samples) of histamine and interferents desorbed by AuNPs from NIPs-PVC film that was incubated with 30 ppm tuna extract for 5 min. (Desorption time: 2.5 min; Raman shift range: 1082-1603 cm-1; exposure time: 10 s.) ........................................................ 131	Figure 5.9: SERS spectra (n = 4 at each concentration) of histamine desorbed by AuNPs from MIPs-PVC film that was incubated with tuna extract at different spiking levels for 5 min. (Final   xvii concentration of histamine in tuna extract: A: 30 ppm, B: 10 ppm, C: 1 ppm, D: 0 ppm, E: blank MIPs-PVC film treated by pure ethanol; desorption time: 2.5 min; Raman shift range: 1082-1603 cm-1; exposure time: 10 s.) .......................................................................................................... 132	Figure 5.10: Raman spectra of histamine. [A: SERS spectra (n = 4) of histamine at 10 ppm, B: normal Raman spectra (n = 4) of histamine at 10 × 103 ppm.] ................................................... 133	Figure 5.11: SERS spectra (4 samples) of L-histidine at 1 × 103 ppm. (Raman shift range: 1082-1603 cm-1; exposure time: 10 s.) ................................................................................................. 134	Figure 5.12: PCA plot of SERS spectra of histamine acquired from different tuna extract. (Raman shift range: 1082-1603 cm-1, exposure time: 10 s.) ....................................................... 136	Figure 5.13: PLSR model of SERS spectra of histamine acquired from different tuna extract. (Raman shift range: 1082-1603 cm-1, exposure time: 10 s.) ....................................................... 137	Figure 5.14: Univariate linear regression models based on different peaks and their intensities. (A: 1267 cm-1, B: 1304 cm-1, C: 1317 cm-1, D: 1576 cm-1.) ....................................................... 138	   xviii List of Abbreviations  AFM Atomic force microscopy AIBN 2,2’-azobis(isobutyronitrile)  AM Acrylamide CAP Chloramphenicol CARS Coherent anti-Stokes Raman spectroscopy  CCD Charge-coupled devices  DA Discriminant analysis  DI Deionized DPPP Di-phenyl-1-pyrenylphosphine EGDMA Ethylene glycol dimethacrylate  ELISA Enzyme-linked immunosorbent assay  EM Electromagnetic  FT Fourier transform  GC Gas chromatography  HA Histamine  HAuCl4 Chloroauric acid  HPLC High performance liquid chromatography  HPTLC High-performance thin layer chromatography  IF Imprinting factor   xix IR Infrared  LC Liquid chromatography  LDA Linear discriminant analysis  LOD Limit of detection LSPR Localized surface plasmon resonance MAA Methacrylic acid  MIPs Molecularly imprinted polymers MISPE Molecularly imprinted solid phase extraction  MLR Multiple linear regression  MS Mass spectrometry  MWNTs Multi-walled carbon nanotubes  NIPs Non-imprinted polymers NMR Nuclear magnetic resonance  β-L,L-ZAPM N-(benzyloxycarbonyl)-β-L-aspartyl-L-phenylalanine methyl ester  PC Principal component  PCA Principal component analysis  PCR Principal component regression  PDA Photo diode arrays  PDMS Polydimethylsiloxane  PLSR Partial least square regression  PVC Polyvinyl chloride    xx Q2 Predictive squared correlation coefficient  QCM Quartz crystal microbalance  RMSECV Root mean square error of cross-validation  RMSEP Root-mean-square error of prediction  RR Ridge regression SERRS Surface enhanced resonance Raman spectroscopy SERS Surface enhanced Raman spectroscopy SLE Solid–liquid extraction  S/N Signal to noise  SPE Solid phase extraction SPME Solid-phase microextraction  SPR Surface plasmon resonance SRS Stimulated Raman spectroscopy TERS Tip enhanced Raman spectroscopy  THF Tetrahydrofuran  TLC Thin layer chromatography  UV-vis Ultraviolet-visible     xxi Acknowledgements  I would like to give many thanks to my co-supervisors: Dr. Xiaonan Lu and Dr. Ed Grant for their valuable help and support in my study here. In addition, I owe thanks to my committee members: Dr. Eunice Li-Chan, Dr. Russ Algar and Dr. Takamasa Momose for their great advice through my study.     Dr. Pedro Aloise, Dr. Zhiwen Chen, Dr. Xiumin Chen, Dr. Guangtao Meng, Benny Chan, Huey Kuan, Hilda Kisielewski and Peter Hoffman sincerely helped to solve the technical problems and provide valuable discussion. Besides research help, the generous and sincere encouragement from Drs. Pedro Aloise, Zhiwen Chen, Xiumin Chen, Isabelle Lacroix and Guangtao Meng can’t be more important to me. I really appreciate these great people that I met here.     At last, deepest thanks to my parents for their endless patience and inspiration, to my loyal friends for being there for me.   xxii Dedication  To my parents. “Schließlich, morgen ist ein anderer Tag!”     ----Margaret Mitchell    1  Introduction Chapter 1:  Determination of food components, additives and contaminants has received extensive attention due to the growing global need for high quality agri-food products [1]. It is well known that public health can be traced to chemical food contaminants (in large doses or continuous ingestion of low doses) [2], food additives [3] and foodborne pathogens [4]. Public concerns have driven the development of various analytical methods to detect these contaminants in foods. The following section in this chapter introduces and describes the existing methods to analyze food contaminants and additives.  1.1 Food analysis: common methods Food analysis is essential for assessing food quality and safety, especially in relation to additives and chemical contaminants. During food processing such as cooking, curing and canning and storage, heating or microbial spoilage can cause protein degradation, leading to the release of amines and other byproducts [5]. The formed amines can be toxic and cause potential risk to human health (e.g. carcinogenic N-nitrosamines [6]). Food additives such as artificial food colors and sweeteners can be concerns as well because it has been suggested that they may induce hyperactivity in children, including overactive, inattentive and impulsive behaviors [7]. Additionally, agricultural and environmental contaminants such as pesticides and veterinary drugs can have a significant impact on food safety [8]. Therefore, sensitive and accurate determination of these compounds is vital to prevent the distribution of contaminated foods or wide spread infections [9].     2     Quantification of molecules of interest in food systems usually follows procedures of sampling, sample preparation, and analysis that includes purification, detection, and data analysis [10]. Among these, sampling contributes most to the variation of quantification. The ideal food sample must be representative of bulk food materials. Sampling uncertainty is sometimes caused by the heterogeneity of bulk food materials. Therefore, a large number of samples are required to reduce the variation. It has been reported that ~80% of time for sample analysis is used for sampling and sample preparation [9]. To significantly reduce the time for food sample analysis including sampling, sample preparation and analysis, high-throughput sample preparation and rapid screening methods are desired. Indeed, there are a variety of solutions to deal with sampling uncertainty in sampling. Random sampling is simple, but may induce significant bias, whereas systematic sampling deals well with the changes that are subject to time, temperature or other treatments [10]. Here, the following section only focuses on the sample preparation and analysis.      Proper sample preparation is required in order to assure satisfactory precision and accuracy of the analysis. Homogenization and reduction of sample size are the most common pretreatment methods. With respect to the detection of small molecules in food systems, the matrix effect is a major challenge that affects precision and accuracy [11, 12]. Chromatographic, spectroscopic, and enzymatic methods are three representative quantification methods [13]. Chromatography, including gas chromatography (GC), liquid chromatography (LC) (commonly combined with mass spectrometry (MS) or other detectors such as ultraviolet-visible (UV-vis) lamps or fluorescence detectors) draws much attention because of its separation and analytical ability.     GC is mainly applied for the characterization and quantification of volatile and semi-volatile compounds in foods [14]. These compounds can be identified in raw food materials (e.g.   3 metabolites during ripening, harvest) or during food processing (e.g. storage) and it is also possible that they are from additives or contaminants [15]. They are usually used as an indicator to evaluate food quality. LC is mostly conducted to analyze weak volatile or thermally unstable compounds [15]. El-Saeid and Selim have analyzed 86 pesticides in a variety of fresh vegetables using GC-MS and reported that the retention time of these molecules ranged from 7.211 min for dichlorvos to 58.440 min for deltamethrin, with 85% of these pesticides presenting a retention time over 20 min [16]. The limit of quantification has been shown to be between 0.01 mg/kg and 0.10 mg/kg, with 92% of the compounds displaying a limit of quantification <0.05 mg/kg. In another review paper, parameters including LC phase, limit of detection, and sample matrix have been discussed with the pesticides determined by the LC-MS/MS [17]. The limit of detection has been reported to be as low as several parts-per-billion [17].     However, these chromatographic instruments don’t allow the direct injection of food sample matrices. As a result, food samples must first be prepared by distillation, extraction with solvents or solid-phase extraction [9]. Food sample preparation involving multiple steps is not preferred, especially for analysis of trace amount of analytes, because large amounts of sample and organic solvents are required, and loss of analytes is not avoidable. Moreover, the process is usually time consuming, which is not suitable for unstable compounds [9] and is impractical for the analysis of a large number of food samples. Especially nowadays, the demand for food safety assessment grows rapidly. For example, more than $30 billion was spent on pesticides in 2000, while almost $40 billion was spent in 2007 due to the global need for high quality foods [18, 19]. Because of the wide use of pesticides, more chemical contaminants are introduced into groundwater and soil, and ultimately to agri-food products [18]. The research cost of pesticide testing is not   4 acceptable by dealing with such a large number of samples. Furthermore, in-field detection of pesticide contamination using chromatographic-based technique is not feasible.     In addition to traditional chromatographic methods, microextraction approaches have been developed to facilitate sample preparation using less solvent, smaller amount of a sample and shorter time. Solid-phase microextraction (SPME) can improve the extraction and concentration of analytes in the gaseous or aqueous phase of food samples using fibers coated by fused silica, polydimethylsiloxane or other materials, and it is recognized as a solvent-free method [20]. Sometimes coated fibers can be fragile, therefore the in-tube SPME has been developed and employed to shorten analysis time and allow automatic analysis [21]. Moreover, dispersive microextraction such as liquid-liquid microextraction is another simple and rapid technique to extract analytes from food samples [22].     Spectroscopy also provides rapid and relatively simple detection of food chemical hazards and contaminations. Fluorometry, infrared (IR) spectrometry, and Raman spectrometry are widely used spectroscopic techniques [13]. For example, hydroperoxides in foods can oxidize di-phenyl-1-pyrenylphosphine (DPPP) to DPPP oxide. The DPPP oxide emits strong fluorescence that can then be utilized to determine the amount of hydroperoxide in foods [23]. However, sample preparation and derivatization associated with the fluorescence spectroscopic-based method is time consuming. Near-IR (750-2500 nm) (studying overtones and combinations of fundamental molecular vibrations) and mid-IR (2500-23500 nm) (studying fundamental molecular vibrations) can readily determine the major compositions (e.g. fat, protein, water and carbohydrates) of food samples and require little sample preparation. Detection can usually be finished within 90 s per sample and sometimes can be conducted on-line [24]. Peak position and peak shift provide rich structural information of chemical compounds in food samples. Multivariate statistical analysis is   5 normally associated with IR detection [25] to reduce the data dimensions of IR spectra and extract the most useful information from IR spectra. Besides IR spectroscopy, Raman spectroscopy is another vibrational spectroscopic technique to determine the vibrational modes of functional groups of chemicals. Detailed introduction and discussion of Raman spectroscopy are shown in the following section. Nuclear magnetic resonance (NMR) spectroscopy (e.g. 1H and 13C NMR) is the technique to characterize structural information for molecules in food samples, such as protein 3D structure. 1H, 13C, and 31P NMR spectroscopies have been applied to assess the quality of olive oils by analyzing diglyceride content [26]. However, the derivatization reaction of the quality index (i.e. diglyceride content) is necessary because of the labile hydrogen from hydroxyl groups in diglycerides and usually takes around 30 min. Atomic absorption spectroscopy is a very fast technique that only deals with elemental analysis without providing structural information as IR, Raman, and NMR spectroscopies do [13].     In recent years, various enzymatic biosensors have been developed for the detection of food chemical hazards and contamination, and different recognition elements have been employed in biosensors, as shown in Figure 1.1. Based on the working mechanism of an enzyme, inhibition-based biosensors and direct biosensors are developed. Because the most essential aspect of a biosensor is the signal generation or reaction catalyzed by an enzyme, the principle is to follow changes in the concentrations of reactants and products and evaluate the corresponding activity of the enzyme (inhibited or activated, depending on the type of enzyme employed).  For example, the substrate acetylthiocholine has been enzymatically hydrolyzed to produce thiocholine, which generates a proton on the electrode (oxidization reaction) due to its electroactive property with a limit of detection as low as 0.4 pM of paraoxon [27]. In another study, organophosphates (paraoxon, diisopropylfluorophosphate, chlorpyriphos, and   6 chlorfenvinphos) could be identified down to 0.02 mg/L by quartz crystal microbalance based on an inhibitory effect, and the whole incubation and regeneration time took about 25 min [28]. Furthermore, the pesticide carbofuran (a carbamate) could be detected at 1.30 × 10-9 M with cholinesterase (pesticides substrate) activity inhibited with 5 min incubation by quartz crystal microbalance, and the sample can be reused five times [29]. However, antibodies are recognized as unstable compounds and some of them do not naturally occur.    Figure 1.1: Schematic illustration of a biosensor (recognition element, transducer and detector).  1.2 Raman spectroscopy Raman spectroscopy, firstly discovered by C.V. Raman in 1928 [30], is a well-known non-destructive analysis technique that requires little sample preparation such as dissolution, compression and grinding, allows contact-free analysis and realizes on-line detection. All these characteristics allow the rapid screening of food products in both solid and liquid phases, as Raman spectroscopy provides well-resolved peaks and shows limited sensitivity to water in contrast to IR spectroscopy. Moreover, unlike fluorometry, Raman can be employed without the concern for the need of fluorophore in fluorometry [31].    7  1.2.1 Raman scattering Raman spectroscopy is a type of vibrational spectroscopy based on inelastically scattered radiation [32]. It can be applied to characterize molecular species as this spectroscopic technique provides structural information. Scattering radiation is caused by the interactions between incident photons and particles (e.g. molecule and dust) because the interactions change the direction of incident photons. The redirected photons can be categorized into some species based on particle size. Mie scattering is obtained when the particle size in diameter (d) and the wavelength (λ) of the incident photon are comparable [33]. In contrast, if d is smaller than 10% of λ, Rayleigh scattering dominates. Both of these are elastic scattering due to the unchanged energy of radiation. When Rayleigh scattering takes place, it always comes along with a small portion of inelastic scattering known as Raman scattering. Inelastic scattering exhibits a shift of wavelength of the photon scattering off a molecular bond. The process of Raman scattering is described in Figure 1.2. Basically, a molecule at 0th ground state is excited to a virtual state. Then, the unstable excited molecule tends to return to the ground state and in the meantime emits a photon with less energy expressed in the Stokes line. The anti-Stokes line displays more energy than the incident photon because the molecule possesses the vibrational energy before excitation and finally relaxes back to the most stable state (i.e. 0th ground state). At room temperature, it is less probable to have many excited molecules. Therefore, the anti-Stokes Raman scattering is much less than Stokes Raman scattering. Among all the incident photons, about one in a thousand of those incident photons can result in Rayleigh scattering, whereas, only one out of 10 million of those photons undergo Raman scattering. Thus, weak signal intensity is the inherent problem of Raman scattering.   8       Figure 1.2: Diagram of Raman scattering (Stokes Raman with energy loss and anti-Stokes Raman with energy gain) and Rayleigh scattering.      However, vibrational and even rotational information from molecular bonds can be provided by Raman shift (i.e. νr). It has been recognized that Raman spectroscopy is a valuable complementary tool for IR spectroscopy as Raman spectroscopy can describe stretching and ring motions of the symmetric covalent bonds (e.g. O=C=O, C=C, S-S and N=C) well, but polar bonds or molecule like water only show very weak Raman signals. In addition, Raman spectroscopy can be applied to detect a portion of a molecule with proper excitation wavelength if necessary (e.g. large biomolecules [34]). Furthermore, some characteristic Raman bands of specific molecules are usually sharp, which could contribute to assay the presence of those molecules and assist quantitative analysis. Besides small molecules, Raman spectroscopy is also applicable for the recognition of cellular components in cells and tissues.    9 1.2.2 Confocal Raman spectroscopy As shown in Figure 1.2, if the incident photon possesses sufficient energy, the electronic transition of the molecule will take place and fluorescence (often intense) will be emitted. As a result, weak but informative Raman signals are prone to be swamped (Figure 1.3). Another source of fluorescence may be generated by impurities from analyzed samples. This drives continuous research on reducing the fluorescence background and enhancing Raman scattering signals. One of these studies is focused on confocal Raman microscopy [35] with regard to instrumental improvement by using two apertures to reject stray light from light sources and out-of-focus light from samples. Other optimization methods towards Raman spectra will be discussed later.     Figure 1.3: Comparison of SERS spectra (x axis: Raman shift in cm-1, y axis: Raman intensity in arbitrary units) of same sample under different incident lasers, SERS spectrum on the left shows high fluorescence background that swamps Raman peaks in contrast to the SERS spectrum on the right displays distinct Raman peaks with relatively low fluorescence background. (Left: SERS spectrum of Sudan I at 10 ppm obtained with 532 nm laser at 100 mW; Right: SERS spectrum of Sudan I at 10 ppm acquired from 785 nm laser at 25 mW.)      The concept of confocal microscope was first come up by Dr. Marvin Minsky in 1950s [35] with the goal of imaging unstained tissues and thus understanding biological events better. In   10 1980s [35], the first scanning confocal microscope was developed, contributing significantly to modern research. Generally, a piece of confocal aperture is placed in front of the light source, after which the observer can focus the diffraction-limited laser to illuminate a single point of the sample instead of conventional wide-field illumination, avoiding stray light sources thanks to the first pinhole. To increase the spatial resolution (2D or 3D), a second aperture is introduced in front of the detector, aiding in rejecting the out-of-focus light from the sample (Figure 1.4).       11  Figure 1.4: (A) Confocal Raman spectroscopy; (B) Stray light induce large sample illumination with the oversized first pinhole; (C) Out of focus photons contribute to detector without the second aperture.    12     A multi-channel detector, such as a charge-coupled device (CCD) or photodiode arrays (PDA), collects a spectrum that contains the chemical information relating to the composition of a sample. The spatial resolution of a Raman image is as high as 1 µm or even better, as reported in some studies [36, 37]. As the first aperture reduces analysis volume of the sample, and the fluorescence background generated by analyte itself (but expressing a large difference from Raman signals in terms of energy) is attenuated by the second pinhole, Raman signals with better signal to noise ratio are achieved. By scanning the sample in x-y-z dimension and collecting spectral information at the same time, the Raman image of the sample including internal profile can be acquired; Raman peak position distinguishes molecular structures and its intensity accounts for the concentration, additionally the peak width ascribes to crystallinity and phase [38]. Raman mapping has broad applications, such as differentiating tumor cells from normal cells and opening possibility of diagnosing cancer in the early stages [39], controlling the synthesis of carbon nanotubes by acquiring the spectra of their different electronic structures [40] and finding the components (including distribution and homogeneity) of pills [41], etc. It is noteworthy that the performance of a confocal Raman microscope is related to laser wavelength, objective properties, pinhole size, sample and instrumental design.   1.2.3 Ways to improve Raman signals Unlike Rayleigh scattering, Raman effect contains rich structural information of molecules but the signals are inherently weak. Moreover, intense fluorescence background is another major challenge. Researchers are therefore continuously seeking methods to remove fluorescence background and improve Raman signal intensity. Some of these methods have been proved useful. Raman signal intensity is inversely proportional to the 4th power of the incident   13 wavelength [42]; normally, the Raman cross section (σ) is employed to describe Raman scattering efficiency.    𝐼𝑠𝑐= 𝑁σ𝐼0 (1.1) in which, I0  and  Isc are the numbers of the incident photons and scattered photons per unit time,  respectively, and N represents the number of molecules under unit area [42].      Consequently, the relationship between cross section and incident wavelength functions in the same way as that between Raman signal intensity and incident wavelength does. Therefore, despite the consideration given to fluorescence, an excitation laser with a short wavelength and high intensity can produce improved Raman signal intensity. For instance, the Raman scattering intensity derived from a 532 nm laser can be greater than that of a 785 nm laser by a factor of 4 to 5 on the basis of σ ∝ 1/λ4.  As mentioned above, even though the analyte is free of interferents, fluorescence can still hinder the distinguishment of Raman signals when the analyte molecule itself contains a fluorophore (Figure 1.3). A confocal Raman microscope shows the advantage of decreasing the fluorescence effect by the second pinhole in front of the detector for the situation that their energy levels (Raman and fluorescence) differ to some extent. It has been found that lasers lying in the visible range are most likely to induce high fluorescence, while UV or near-IR lasers somewhat avoid it, owing to the fact that increased laser frequency falls within the bandwidth of an electronic transition except for deep UV (below 260 nm) that leads to the Raman scattering far above the emission frequency of fluorescence due to the vibrational relaxation prior to fluorescence emission [43]. Nevertheless, lasers with short wavelengths that carry high energy per photon damage samples easily and the cost of lasers and lenses is usually high. Lasers at 532, 633 and 785 nm are the standard configuration in Raman spectrometers to characterize general chemical and biological samples. Blue (488 nm) and green (514 and 532   14 nm) lasers are more applicable for inorganic materials. UV lasers are preferred by the determination of biomolecules such as nucleic acids, proteins and interaction between proteins and drugs [44].   Fourier transform Raman spectroscopy Fourier transform (FT) Raman is regarded as a good solution to deal with the fluorescence problem. By using a near-IR laser (e.g. Nd:YAG at 1064 nm) that is far away of electronic transitions, the absorption in the fluorescence process can be avoided. Additionally, thermal decomposition of samples caused by high-frequency lasers is circumvented. But the Raman intensity is decreased because the Raman scattering is inversely proportional to the 4th power of the incident wavelength (Equation 1.1). Therefore, high laser power is usually applied to compensate for the loss in Raman cross section due to the long wavelength incident laser, subsequently reducing illumination time. The dielectric long pass filter efficiently removes Rayleigh scattering, therefore decreasing the noise through the whole spectrum. Furthermore, accurate spectral subtraction is easy to obtain given the advantage of interferometric instruments with respect to high frequency precision due to the internal calibration and Fourier transformation of interferograms employing mathematical method to re-construct Raman spectra requires no dispersing element and provides high through-put signal intensities [45].  Resonance Raman spectroscopy  Resonance Raman spectroscopy takes place when the excitation laser overlaps with the molecular electronic absorption, resulting in as high as 106 augmentation of the Raman signal intensity comparing to that of normal Raman spectroscopy [46]. Such wavelengths usually lie in   15 the UV-vis range to provide sufficient energy but in the meantime provoke fluorescence emission as the molecule or part of the molecule (i.e. chromophores) accounting for resonance Raman activity is fluorescently active as well. Those chromophores are typically responsible for coloring the molecules. Visible resonance Raman spectroscopy is often the option for determining the authenticity of artwork by analyzing pigments and dyes. Carotenoids, the light receptor components in biological samples, can also be analyzed by resonance Raman spectroscopy [47]. Another molecule that can be characterized by resonance Raman spectroscopy is the metalloprotein containing active electron transfer sites that involve intense electronic transition [48]. The most promising laser for resonance Raman spectroscopy should be tunable to gain the maximal intensity of resonance Raman signals without changing lasers, due to the fact that the electronic absorption energies of different molecules vary a lot. It has been found, however, that Raman signal intensity is impressively enhanced with a laser line not identical but very close to those energies; in addition, the tunable laser complicates the spectrometer and increases cost [49].      It is well known that fluorescence is always the enemy of both non-resonance and resonance Raman spectroscopy. The difference between resonance Raman spectroscopy and fluorescence spectroscopy is related to the involved time scale and intermediate state. It usually takes nano-seconds for the excited molecule to undergo emission of fluorescence; in contrast, scattering a Raman photon from the virtual state takes less than 10-12 s, which is considered to be simultaneous to the excitation [50, 51]. As a result, some techniques have been developed to reduce or suppress the fluorescence. Sometimes, the low fluorescence background can be caused by the intermediate states of analytes. For example, the excited state of crystal violet undergoes non-radiative relaxation due to the twisting of the arms in the molecule [52]. Time-gating   16 techniques have been used by taking advantage of the length of lifetimes of Raman scattering and fluorescence. The Kerr-gating method transmits the Raman scattering but blocks the slower fluorescence by controlling the property of the Kerr medium and the associated pulsed laser [50, 53]. However, the pulsed laser has a high power that it may degrade the sample.   Surface enhanced Raman spectroscopy and tip enhanced Raman spectroscopy Surface enhanced Raman spectroscopy (SERS) employs nanoparticles or nano-roughened surfaces of noble metals (e.g. Au, Ag and Cu) to obtain the enhanced Raman signals. Surface plasmon resonance (SPR) of those metallic nanoparticles is produced upon exciting light that matches the inherent oscillation frequency of metal electrons [54]. The resultant SPR decreases exponentially as a function of distance away from the surface (a few nanometers), and is known as “non-propagating” or “localized.” Furthermore, the SPR intensity also relates to the incident angle of the light and the dielectric constant (interface material). As described in Figure 1.5, the electronic vector of the incident electromagnetic wave polarizes the nanoparticle and redistributes the surface electrons of the metallic nanoparticle, finally resulting in a new electromagnetic wave. Noble metal nanoparticles including gold and silver are mostly employed because they possess intense localized SPR fields and their excitation wavelengths lie within the visible range [55]. Analyte staying in the “non-propagating” electromagnetic field will be excited to a virtual state and instantly release a scattered photon that resonates with the new electromagnetic field, therefore, an enhanced Raman spectrum can be collected. In other words, the incident light and the scattered light are both enhanced by the “non-propagating” electromagnetic field of noble metallic nanoparticles. This process is ascribed to the “electromagnetic enhancement” mechanism.   17   Figure 1.5: Optical excitation of localized surface plasmon resonance by oscillating electric vector of the incident electromagnetic wave.       “Charge transfer” as the chemical enhancing method is proposed, too. The formation of a metal-analyte complex deforms the analyte and causes changes in polarizability of the analyte, and consequently contributes to the Raman signal enhancement.     As described in some studies [46, 56], the intensity of a Raman signal is proportional to the square of the electric dipole moment P. Additionally,   𝑃 = 𝛼𝐸 (1.2) in which, α is the polarizability of the analyte and E represents the electric field that the analyte experiences. Usually, electromagnetic enhancement (E enhancing) coincides with chemical enhancement (α enhancing); in addition, when resonance Raman scattering (E enhancing) acts as the additional enhancment technique, about 14 orders of magnitude of the total enhancement can be acquired (known as surface enhanced resonance Raman spectroscopy) [57]. This advances the field of sensitive detection and realizes single molecule detection. SERS can even aid in overcoming fluorescence when the intermediate of metal-molecule undergoes non-radiative relaxation [58].   18     SERS has been widely used in sensitive analysis of trace amounts of compounds. It has been reported that the number of publications regarding chemical analysis in food samples determined by SERS keeps increasing [59, 60]. Various approaches have been developed for the separation and SERS detection of analytes in food samples. Lin and coworkers have made use of common extractions (i.e. solid-liquid and liquid-liquid extractions) to separate the melamine in various food samples (gluten, chicken feed and processed foods) and Klarite™ SERS-active substrates to carry out SERS measurement [61], whereas Chueng and coworkers have performed these multiple extractions as well to purify Sudan I in chili powder and SERS detection with gold colloid as SERS active substrate [62]. Solid phase extraction (SPE) cartridges with different packing materials (e.g. reverse phase, normal phase and ion exchange packings) are one of the most popular tools to extract analytes from complex samples. Farquharson and coworkers have firstly extracted chlorpyrifos-methyl from orange juice by reverse phase SPE cartridge and then developed a silver-doped sol-gel capillary to further facilitate sample preparation and measurement of a series of pesticides [63]. Antibody [64] or nucleic acid aptamer-based [65] separation approaches provide good separation as well, furthermore, silver dendrite have been employed as SERS active substrate. Molecularly imprinted polymer is another option used to isolate analytes of interest from food matrices (e.g. α-tocopherol in vegetable oils [66], melamine in milk [67] and theophylline in green tea drinks [68]), afterward, silver dendrite or silver nanoparticles functions as SERS substrates to enhance Raman signals of these compounds.      As peak width, shift and appearance (or disappearance) are unique features of SERS due to the affinity and physisorption between analytes and SERS active substrates, quantitative analysis by SERS can be challenging. Chemically produced nanoparticles are usually good SERS substrates for sensitive detection, nevertheless, physically deposited nano-roughened surfaces or chemically   19 etched surfaces sacrifice sensitivity for better accuracy and precision [69].      Tip enhanced Raman spectroscopy (TERS) is the combined result of SERS and atomic force microscopy (AFM), substituting SERS surface substrates (i.e. nanoparticles or nanoroughened surfaces of noble metals) with an AFM tip [70]. TERS can assist quantitative analysis to some extent compared with SERS and provide a high spatial resolution (~100 nm) given the fine size of the tip [46]. However, maintenance and mass-production of reproducible tips challenge TERS detection.  Coherent anti-Stokes Raman spectroscopy  From Figure 1.2, we can observe that in spontaneous Raman scattering, there is a small fraction of anti-Stokes scattering photons with higher frequency. Instrumental configuration of coherent anti-Stokes Raman spectroscopy (CARS) is different than that in a normal Raman spectrometer. CARS requires two lasers with fixed wavelengths, one acting as pump laser (νpm) while the other serving as probe laser (νpb), and a tunable laser as Stokes laser (νsk) which matches the Stokes frequency of a Raman mode of the molecule (Figure 1.6). In fact, CARS can be equipped with one fixed laser due to the same frequency of pump laser and probe laser. When the Stokes laser line is tuned to interact with the pump laser coherently, a beat frequency (νpm – νsk) is produced to strike the lowest ground state molecule to a vibrational state (e.g. symmetric stretching of CH2: ~2800 cm-1); the probe laser is then introduced to excite the vibrational molecule to obtain anti-Stokes scattering [71]. CARS targets only one Raman peak of interest and has the signal intensity to obtain as high as 105 of the enhancement factor [72]. Since only one peak is enhanced, the monochromator is not necessarily required. CARS has found great application in cell and tissue imaging, as it can provide chemical information without a fluorescence tag and   20 achieve high resolution in sub-micron scale; furthermore, it is a fast technique and can minimize fluorescence background [73].  Figure 1.6: (A) Spontaneous Raman scattering under monochromatic incident light; (B) Coherent Raman scattering (coherently interacted pump and Stokes lasers excite coherently vibrating molecules, next, probe laser strikes Raman scattering of these vibrating molecules).      Stimulated Raman spectroscopy (SRS) is another example of non-linear Raman spectroscopy that can enhance a particular Stokes Raman peak. Unlike spontaneous Raman spectroscopy, the transition rate of SRS between molecular states is significantly higher owing to the fact that the Stokes laser initially occupies the Raman emitting channel, and consequently the SRS Raman signal is 4 to 5 orders of magnitude higher than that of the spontaneous Raman spectroscopy. SRS microscopy is widely used to image biomedical samples due to the advantage of free background [42]. Surface enhanced resonance Raman spectroscopy (SERRS) combines the advantages of SERS and resonance Raman spectroscopy so the resultant spectrum is not only highly enhanced but also resembles the resonance Raman spectra [74].  1.3 Data analysis As aforementioned in Section 1.1, great demand for food and the assessment of food quality leads to a large number of samples to be studied by analytical techniques. Besides instrumental   21 analysis, experimental data analysis is another elementary task in food analysis. Dealing with massive data, especially with many variables or features, in a relatively short time requires employing multivariate statistical analysis to extract the most useful information from data. The collected raw data could be signals derived from various measurements: pH, concentration, chromatograms, spectra etc. Multivariate statistical analysis usually consists of data preprocessing, model construction using training dataset, validation and prediction to classify existing samples and predict unknown samples based on the built model [75]. The following sections will provide detailed descriptions for better presenting, classifying and quantifying data acquired by Raman spectroscopy.  1.3.1 Preprocessing In general, noise, background, baseline drift and overlapping peaks are common issues in Raman spectra of samples. Data preprocessing can remove or reduce these effects by employing mathematical techniques such as polynomial fitting, Savitzky-Golay smoothing and derivatization [76]. These developed mathematical techniques make use of the relationship between variables (e.g. collinearity) to correct those undesired effects [76]. Therefore, a better input to multivariate statistical analysis could be provided by data preprocessing. Afterward, better linear or non-linear models describing relationships between variables and corresponding responses could be constructed.   Smoothing Smoothing mainly reduces random noise, therefore signal-to-noise (S/N) ratio can be improved, and subsequently the appearance of the spectra is enhanced. However, smoothing always gives   22 rise to reduced resolution of spectra. One of the simplest ways of smoothing is ensemble average that is mostly used in NMR and FT-IR spectroscopies. Ensemble average smoothes random noise by taking the average of hundreds or thousands of spectra from repetitive scans as the final spectrum. The improved S/N ratio of a spectrum may be obtained by signal accumulation owing to the fact that S/N ratio is proportional to the root square of the number of scans. Another way to smooth noise is to employ moving window algorithm on a single spectrum. A moving average simply takes the average of a fixed number of data points (i.e. 2n+1) as the smoothed point and then moves to next data point. The process is repeated to produce more smoothed points. Next, all smoothed points are plotted to obtain the smoothed spectrum. However, this method degrades too much underlying information especially when the window width (2n+1) is greater than the peak width of interest. Alternatively, the Savitzky-Golay algorithm performs a least-square polynomial fitting in the moving window; the greater the window width and the lower the polynomial order, the better the smoothing. It is unavoidable that Savitzky-Golay algorithm loses information as well, but less than that of the moving average algorithm.  Background removal As aforementioned, fluorescence is an enemy of Raman signals as the fluorescence intensity can be a few orders of magnitude higher than that of Raman scattering. Besides instrumental approaches to removing background of Raman spectra such as changing excitation light (e.g. deep UV and NIR lasers) and employing time-gated Raman spectroscopy (transferring Raman light and blocking late-arriving fluorescence light), computational implementation is another option. Polynomial fitting is the method that mathematically estimates the fluorescence background of a Raman spectrum, allowing the subsequent subtraction of the estimation of   23 fluorescence background from the raw spectrum to take place. Attention should be paid to the order of polynomial fitting and the spectral range when carrying out polynomial fitting because it may cause errors in further analysis without considering those two parameters. After the optimal values for those parameters having been entered, the fluorescence background is rejected as much as possible, at the same time, the possibility of subtracting Raman signals is minimized [77]. Modified multi-polynomial fitting is developed to deal with the problem that one has to manually select the polynomial order and spectral range [78]. In modified multi-polynomial fitting, a comparison between original value and fitted result is carried out on each data point of a Raman spectrum. If the original value is higher than the fitted result, the latter will be stored, nevertheless, the former will be kept for the next round of polynomial fitting if the fitted result is higher. The fitting will stop when the desired setting is fulfilled. Some problems can still be found in the modified approach. For example, high noise could be falsely treated as small Raman peaks and intense peaks may cause bias during fitting. Another disadvantage of modified polynomial fitting is time consuming (usually 20 to 500 times of iterations to return fitted result). In addition to simple and modified polynomial fitting, a new idea has been brought up by Zeng and coworkers [78] that the standard deviation of a Raman spectrum after simple polynomial fitting is treated as noise and only the data greater than the standard deviation will be considered as Raman signals; in addition, the identified huge peaks are removed after the first round of polynomial fitting to avoid bias. Therefore, the approach to removing fluorescence background is improved. However, this approach is still somewhat subject to spectral range and polynomial order.     Moreover, the Fourier transform method breaks a spectrum into a sum of sine and cosine waves as a function of frequency, the low-frequency background and high-frequency noise can   24 therefore be distinguished from the middle-frequency signal. Furthermore, only the desired signal can be retained after filtration [79]. The first and second derivatives are utilized when the variables of a Raman spectrum are highly related to each other [76]. The first derivative is the result of subtracting two adjacent variables so that the signals displaying high slope are left but the neighboring data points showing similar values that most likely represent background are rejected. The second derivative involves the same operation to the first derivative on the result of first derivative. The resultant derivative spectrum emphasizes high slope signals in the original spectrum and deemphasizes the baseline with low slope. However, the high-frequency noise is accentuated due to the special feature of derivatives. For this reason, smoothing is always taking place to reject the effect of noise amplifying prior to derivatizations [76].   1.3.2 Multivariate statistical analysis The purposes of the aforementioned data preprocessing are manifold including better presentation and more accurate quantification as well as greater preparation for further data processing (e.g. classification or regression) of Raman spectra [25]. Multivariate statistical analysis deals well with large datasets or a data matrix with multiple rows, and expresses computational efficiency.     In general, a data matrix with n rows and m columns is composed of n samples that are analyzed and m variables that are recorded. In typical FTIR or Raman spectra, m represents the signal intensities at different wavenumbers, while n is the number of samples that are measured. These samples and the corresponding observations are stored into a data matrix. It is well known that dealing with this large quantity of data can be difficult, because not all observations are informative. For example, some variables are not independent as they co-vary with each other.   25 To simplify and speed up the analysis process and in the meantime maintain the major features of those samples, the matrix has to be compressed and only the most representative information is extracted out. Therefore, multivariate statistical analysis such as principal component analysis (PCA), discriminant analysis (DA) and partial least square (PLS) regression (PLSR) is employed to fulfill those requirements.     PCA is one of the most popular tools that provide a way to visualize the complicated data [76]. Matrix X with m × n dimensions can be mathematically written as below:  X = 𝐭! 𝐩𝟏𝐓 + 𝐭𝟐 𝐩𝟐𝐓 + … 𝐭𝐤 𝐩𝐤𝐓 + E (1.3) in which E is the matrix for residual, ti (m × 1) vectors are the scores describing the relationship between samples and pi (1 × n) vectors are the well-known eigenvectors that can be defined as directions. Furthermore, ti is the result of linear combination of the original variables of sample i (1, 2…m) described on the pi direction. The original data points are projected onto the pi direction. Among all the projected results, the one showing largest variance of the sample dataset indicates that the new data points are spreading out most on the pi direction. As a result, this pi vector is exploited to represent the x axis in a new coordinate. Afterward, those projected results carrying smaller variance values of the sample dataset are sorted from highest to lowest. The covariance of ti and tj displaying the second largest variance on the pj direction is zero, therefore, the pj vector assembles the y axis in the new coordinate. The third largest variance of tl on the pl direction is used to describe the data variation on the z axis, etc. All pi vectors are uncorrelated to each other. As a result, these scores that are the weighted sums of original variables are used to adequately represent the complicated original dataset in the new coordinate with much fewer dimensions because the dimensions with lower score values can be omitted. Those related   26 parameters including scores and eigenvectors can be calculated from the variance-covariance matrix that is derived from matrix X.      Cluster analysis is another classification approach that measures distances between samples and subsequently compares those distances, aiming at grouping [76]. A dendrogram is the best way to summarize the compared results of distances by plotting distances against samples. Euclidean distance (Equation 1.4) is preferred over Mahalanobis distance (Equation 1.5) for datasets with high dimensions. Noises contribute too much to variations in high-dimension datasets, thereby an overestimated classification result may show up as Mahalanobis distance is adjusted to unit variance. Cluster analysis can be inspected by PCA to validate classification results.  dij  =  (𝒙𝒊 − 𝒙𝒋)(𝒙𝒊 − 𝒙𝒋)𝐓 (1.4)  dij =  (𝒙𝒊 − 𝒙𝒋)𝐂!!(𝒙𝒊 − 𝒙𝒋)𝐓 (1.5) where, dij represents the distance between sample xi and xj, and C is the covariance matrix of the dataset.     Discriminant analysis (DA) is a supervised classification approach compared to PCA and cluster analysis. In DA, the classification rule is to find the greatest probability that we obtain the observation x in a given group i based on the assumption that the means of those groups are different and the data in each group are multivariate normally distributed. When all groups are homogenous so that they have the same variance-covariance matrix, linear discriminant analysis (LDA) is preferred, whereas quadratic discriminant analysis is applied if the groups are heterogeneous [76]. Unlike the unsupervised PCA that tries to find the largest total variance, LDA is used to find a projection direction on which the means of each group fall apart as much   27 as possible and the ratio of between-group variance and within-group variance is maximized [80]. PLS-DA belonging to LDA in a wider sense takes advantage of PLS to obtain noise reduction and variable extraction, therefore, PLS-DA deals well with high-dimension datasets [76].     Despite classification, regression aids in predicting unknown samples on the basis of comparing their pattern of measurements to that of known samples. Multiple linear regression (MLR) is the most straightforward regression approach that deals with the situation that one response variable correspondes to multiple variables by fitting a linear plot to those variables and response variables [76]. The linear model (Equation 1.6) describing the relationship between the response y of sample i and predictors xij can be concluded under the assumption that the residual 𝜺𝒊 are normally distributed and mean centered with a constant variance.  𝒚𝒊  = 𝛽! + 𝛽!𝒙𝒊𝟏 + 𝛽!𝒙𝒊𝟐 + … 𝛽!𝒙𝒊𝒏 + 𝜺𝒊 (1.6) However, it has been found that MLR is not satisfactory when variables co-vary with each other so that the constructed model is unstable and possible to be overfitted, therefore it fails to predict new samples (e.g. the correlation of peak intensities between adjacent wavelengths). Moreover, if the number of samples is less than the number of variables, underestimation of the regression may occur [81].     Attempts made to accommodate the collinearity of variables and insufficient numbers of samples involve extracting the most representative and uncorrelated new variables or putting a constant to the diagonal of the variance-covariance matrix of the original dataset that is known as ridge regression (RR) [82]. In RR, it is necessary to determine the optimal value for the “constant” to obtain a multivariate model with good performance. Despite RR, principal component regression (PCR) and PLSR are the approaches that account for the uncorrelated   28 information extraction. As a result, PCR or PLSR can stabilize the constructed model when the number of principal components (PCs) or latent variables (LVs) is less than that of samples [82].     PCR performs PCA first to retain the most variance in datasets by reducing the collinearity in high-dimension datasets, and in the second step, MLR is carried out on PCs and their response variables. Thus, PCR is the result of linear combination of PCs. As the selection of PCs is not related to response variables, there may be some factors that are not positively contributing to the prediction capability of the PCR model [76].     In contrast, PLSR identifies latent variables by decomposing the response variable dataset Y and independent variable dataset X simultaneously and maximizing the relationship between these two datasets (Equation 1.7 & 1.8) [81, 83]. Latent variables in the independent variable dataset may not necessarily explain the most variations in the dataset such as PCR.  X = 𝐓𝐏𝐓 + E (1.7)  Y = 𝐔𝐐𝐓 + F = 𝐓𝐑𝐐𝐓 + F (1.8) in which, T and U are the scores and P and Q represent loadings, E and F are the residuals. To minimize F, one has to find the weights R to build the relationship. Meanwhile, the root mean square error of cross-validation (RMSECV) is plotted as a function of the number of PCs or LVs to avoid overfitting in PCR and PLSR. This is done by leaving one or more samples out and creating a model with the rest of those samples, followed by fitting the pre-excluded samples into the existing model and calculating the residuals. This process is iterative via changing the pre-excluded samples until the least RMSECV (Equation 1.9) value is found. Larger numbers of PCs can explain more variation in a model but may cause an increase of RMSECV, consequently, the constructed model may fail to predict new samples.   29  RMSECV = !!!!! !!! !  (1.9) in which, yi shows known value, whereas 𝑦! contains the estimated value by the PLSR model.  1.4 Molecularly imprinted polymers Molecular recognition and separation are important in many areas such as detection of drugs, analysis in environment and foods and assessment of reaction progress. Enzymes, antibodies and nucleic acids are mostly involved as receptors of molecules of interest [72]. However, these compounds suffer from their instabilities under harsh or even mild chemical and physical conditions. Moreover, some of these receptors may not occur naturally, thereby often resulting in complex synthesis [72]. These drawbacks drive the development of new candidates for selective binding of target molecules. MIPs are one of the good artificial receptors (Table 1.1) and express great stabilities to acid, base and high temperature. MIPs exhibit rigidity as well. They are widely known as “plastic antibodies.”  Table 1.1: Advantages and disadvantages of MIPs. Advantages Disadvantages Cost effective Template leaking Easy preparation Grinding for monolith No animal needed Tailor-made products for different analytes Stability (thermally, chemically and physically) Low selectivity compared to immunoassay  1.4.1 Mechanism and categories Basically, functional monomers and target molecules will interact and generate complexes by self-assembly. The interactions can be various according to the nature of those “bonds” between functional monomers and target molecules. Cross-linking monomers mostly contribute to the rigidity of resultant MIPs, and solvents account for dissolving all components. Once the polymerization is done, MIPs can be obtained. Recognition binding sites on MIPs particles are   30 due to the memory left by target molecules. Interactions between functional monomers and target molecules can be of most importance in MIPs as they are directly responsible for the desired rebinding. Different types of imprinting will be discussed in this section.     Covalent imprinting has been studied well by the Wulff Research Group. Covalent binding usually makes use of functional monomers and target molecules bearing special functional groups [84]. For example, boronophtalide based monomers can form the readily reversible boronate ester bonds with imprinting molecules with well-separated hydroxyl group(s). They find broad applications in imprinting carbohydrate derivatives (Figure 1.7A). Another study regarding covalent imprinting is based on producing imine groups, which is suitable for monomers containing aldehyde group and amine targets, or vice versa (Figure 1.7B). In addition, Shea and coworkers have utilized the ketal covalent bond generated by diol and carbonyl groups to covalently imprint target molecules onto MIPs (Figure 1.7C) [85]. Covalent imprinting exhibits excellent rebinding selectivity after target molecules removal because: 1) the involved chemical reactions regarding the formation of reversible covalent bonds between functional monomers and template molecules prior to polymerization drive the production of homogenous binding sites during polymerization due to stoichiometry between functional monomers and template molecules and the reduction of non-specific binding sites; 2) afterward, template molecules are removed by chemical cleavage of those covalent bonds and the complementary binding sites are exposed for further rebinding of template molecules on the basis of forming covalent linkage [86]. Furthermore, it is beneficial for studying the structure of binding sites. However, the strong covalent binding causes unavoidable problems such as the difficulty in removing imprinting molecules and long rebinding time. In some cases, the removal of targets may destroy the binding sites, and moreover, the steric effect of polymers makes it difficult for   31 target molecules to approach the binding sites in rebinding [87]. To deal with these problems, non-covalent rebinding is employed and sometimes a sacrificial spacer group (e.g. carbonyl) is involved in polymerization but removed with imprinting molecules to increase the space of binding sites [87]. In general, some limitations of the covalent binding method cannot be ignored. For example, the narrow choice of monomers and imprinting molecules limits applications of MIPs to function as separation element or microreactor, and the water based hydrolysis mechanism required by target removal confines polymerization to some extent.    Figure 1.7: Covalent imprinting employing forming (A) boronate ester bond, (B) imine groud and (C) ketal bond between template molecules and functional monomers.       Non-covalent binding has gained much attention owing to its flexibility in terms of choosing functional monomers, target molecules and applications of MIPs. This approach, well studied by   32 Mosbach and his coworkers, has been recognized as more similar to those non-covalent natural interactions (e.g. antigen-antibody and substrates and enzyme) than covalent binding [88, 89]. Non-covalent imprinting mainly relies on attractive forces between functional monomers and target molecules, such as H-bonding, electrostatic force, dipole-dipole and even van der Waals interactions. These forces are not as strong as covalent bonds, therefore, the formed precursor (i.e. functional monomer-target complex) displaying different bond strength (i.e. H-bonding, electrostatic force, dipole-dipole and van der Waals interactions) is subject to time and temperature [88]. In general, longer time and lower temperature during the process of forming pre-polymerization solution are preferred owing to the consideration of the achievement of equilibrium and maximizing the yield of the precursor. The introduction of cross-linking monomers aims at anchoring the precursor structure to the polymer network.      Unlike covalent imprinting, a single binding site in non-covalent MIPs normally involves multiple interactions. For instance, the primary interaction H-bonding is sometimes followed by hydrophobic interaction [88]. Despite the multiple interactions in a single binding site, the ratio of functional monomer to target molecule is always more than 1 to strengthen the interaction. However, too many functional monomers may give rise to the increase of non-specific binding sites due to the possibility of producing dimers formed by functional monomers, especially for those with carboxylic acid groups [87].      In order to find the potential best candidates of functional monomers and stoichiometry for target molecules, UV-Vis spectroscopy and nuclear magnetic resonance spectroscopy are commonly employed by monitoring the chemical shifts when mixing functional monomers and target molecules and varying the ratio of them [90]. The most common ratio is 4:1 and further optimization has been discussed in many studies [91-93]. In spite of the vast choice of functional   33 monomers, there are some drawbacks that people have to take into account. Typically, the complementary binding sites display low yield and heterogeneity. In fact, those drawbacks are mainly caused by the instability of the precursor. During the polymerization, cross-linker monomer is present in the largest amount to form the rigid frame of MIPs. However, this may form a competing source for target molecules with functional monomers and consequently increase the non-specific binding. Secondly, the rigid frame may enclose complementary binding sites inside and make it difficult for rebinding, especially for the bulk polymerization as it produces one large piece polymer [87].      Porogen (i.e. polymerization solvent) is also very important for forming porousness during synthesis of polymers. The best approach is to find the proper solvent that not only assists in forming pores in polymers so that target molecules can find access to it in rebinding but also maintains the stability of the precursor. For example, polar protic solvent may interrupt the H-bonding, in contrast, polar aprotic solvent may be favored to overcome solubility problems in pre-polymerization system such as acetonitrile [94, 95].      In radical polymerization to produce MIPs, high temperatures in thermal initiation of polymerization may disrupt the precursor (i.e. complex of functional monomer and template molecule) and also result in fast polymerization that is difficult to control. Therefore, photo initiation of polymerization at room temperature has gained popularity [96]. Among all the reported methods, bulk polymerization resulting in one piece of polymer is easy to work with and the yield of polymer is promising. In some applications, however, MIPs in particulate form are required (e.g. sorbent materials or making target molecules to approach the binding sites in rebinding), but grinding leads to damage of some complementary binding sites and irregular shape of particles. In precipitation polymerization, the fine and uniform particles produced   34 provide more complementary binding sites due to the larger surface area and no grinding. Additionally, the higher amount of solvent can dissipate heat better, which may avoid disrupting non-covalent interactions between functional monomers and target molecules [97]. However, the yield of MIPs is lower and it is more difficult to obtain imprinted polymers as the large amount of solvent can break those non-covalent interactions. Emulsion [98] and suspension polymerization [99] methods can release heat well, and furthermore, these methods can drive the complementary binding sites to the surface of polymer particles under some conditions to ease the rebinding. But large volumes of solvent and complex systems are the disadvantages. Besides these two methods, surface imprinting can also be achieved by fixing target molecules at an inert surface followed by pouring solvated monomers to the surface to cover those molecules; after polymerization, the peeled off polymers have been imprinted. For this “physical” method of surface imprinting, solvent can be avoided in some cases. This method is good for imprinting large molecules including proteins and even bacteria, because the methods described above do not work well for large molecule imprinting [88]. Nowadays, controlled polymerization has been introduced to carry out the production of MIPs under control, and consequently better performance of MIPs can be observed [65, 100].     Transition metal-complex can also be used as functional monomers to coordinate to template molecules as they can provide some ligand sites to either charged or neutral templates [84, 87]. Cobalt (III) and Zinc (II) complexes are usually used for coordinating to template molecules containing nitrogen. Copper (II) complexes can bind to nitrogen as well as oxygen based template ligands. Besides traditional polymerization for MIPs, inorganic monomers have also been employed to fabricate MIPs. Inorganic monomers are usually the metal or semi-metal oxides (Si, Ti and Ge). These materials can precede gelation at low temperatures and work well   35 with aqueous systems. This polymerization is considered fast compared to organic polymerization.      In polymerization, attention should be paid to select suitable cross-linking monomers and solvents because in some cases, they can affect properties of MIPs [95]. Cross-linking monomers can form the steric effect at binding sites. Solvents possessing similar structures to target molecules have been found to have the possibility of false imprinting. In practice, there are many experimental problems that have to be addressed, and researchers have tried in various ways to solve them. For example, sometimes more than one kind of functional monomers or solvents are introduced into the system to provide better molecular imprinting effect; furthermore, it has been found that the addition of another “target” molecule can facilitate the dissolving of the molecule of interest, and hence the performance of MIPs has been improved [101]. In contrast, analogue of the molecule of interest imprinting is preferred in some cases, known as “dummy imprinting” owing to the fact that the target molecule may be expensive, toxic or unstable during polymerization. In addition, analogue imprinting can prevent the leakage of target molecules especially during rebinding of target molecules in trace amount because the imprinted target molecule is difficult to remove completely [102].  1.4.2 Applications Separation is a well-known application of MIPs. For this purpose, MIPs are usually employed as sorbent materials in chromatographic methods or recognition elements closely attaching to transducers in sensors [72]. MIPs particles functioning as sorbent materials can be packed into high performance liquid chromatography columns and solid phase extraction (SPE) cartridges, or spread onto thin layer chromatography (TLC) plates, thereby selectively binding target   36 molecules in samples. Besides post-treatment (e.g. grinding, sieving and packing) of those MIPs particles, in situ polymerization in a column, in a capillary tube or on a membrane can simplify the preparation of those MIPs based separation elements as aforementioned and avoid the potential risk of damaging complementary binding sites by grinding involved in bulk polymerization.      In sensors, once the binding between a target molecule and a recognition element (i.e. MIPs) takes places, it can cause subsequent signal changes. These changes can be transferred to a detector and finally translated to a readable format. For example, conductivity can be monitored during the binding by connecting a MIPs based recognition element to electrodes. Quartz crystal microbalance has also been used as a detector to determine the recognition of a target molecule due to the mass changes that occur upon MIPs binding [92]. An optical method has also been introduced by employing a detector using surface plasmon resonance that focuses on changes in SPR angles during MIPs binding [103]. Aforementioned sensors can be categorized as label-free approaches. In contrast, another optical method requiring fluorescence labeling either on analyte molecules or polymers detects signal changes in fluorescence [72]. These applications of MIPs regarding separation and recognition of analytes of interest are widely used in analyses involving complex sample matrices such as environmental analysis, food analysis and drug screening, because MIPs can simplify the pre-treatment of samples (e.g. multiple extractions) due to the selective interactions between MIPs and analytes.     In addition to separation, MIPs can also be utilized in organic chemistry in some aspects. Ye and coworkers have used N-(benzyloxycarbonyl)-β-L-aspartyl-L-phenylalanine methyl ester (β-L,L-ZAPM) imprinted MIPs to remove the produced enantiomer byproduct after the chemical synthesis of aspartame [104]. In fact, it is similar to the separation application of MIPs. In   37 addition to the byproduct separation, MIPs can work as a microreactor to optimize some reactions. In a microreactor, reactants approach the product imprinted binding sites, and subsequently orientate themselves in a favorable way to reduce the possibility of producing byproducts (e.g. enantiomers) and eventually improve the reaction efficiency [105].       Furthermore, a microreactor has been employed as a protector to shield some functional groups of sterols (multifunctional compounds) from reacting during the regioselective modification of some hydroxyl groups in sterols [106]. Moreover, MIPs can function as a catalyst to increase the rate of a reaction by imprinting the transition state analogue of the reaction to MIPs, thus the Gibbs free energy of activation is decreased [107].    1.5 Objectives and outline As discussed in Section 1.1, chromatographic instruments can effectively analyze food components and chemical contaminants in foods. Given the development of microextraction and capillary analytical columns, general concerns such as solvent wasting, tedious sample preparation and time-consuming detection are well solved; nevertheless, these instruments are still regarded as of high cost and inconvenient for in-field detection. Enzymatic and antibody-based analyses provide relatively simplified sample preparation due to the strength of specific separation; furthermore, enzymatic and antibody-based biosensors advance in-field detection. However, enzymes and antibodies are usually expensive and unstable. Spectroscopic approaches can be helpful to food analysis but most of them require extensive sample purification. Among them, even though IR and Raman spectroscopies allow little sample preparation (e.g. dissolution, compression and grinding) and realize in-field detection via portable spectrometers, the complex   38 fingerprint region in spectra complicates the identification of some compounds, especially in trace amounts.     MIPs possessing thermal and chemical stability offer the possibility to approach relatively simple separation similar to antibodies. The Raman spectroscopy, displaying sharp and well-resolved peaks in spectra, is a good candidate for fast detection. To improve weak Raman signals, the SERS technique has been used to greatly enhance sensitivity. Thus, the objective of this Ph.D. thesis project is to design an MIPs-SERS based approach to realize rapid, sensitive and high-throughput determination of small molecules in food samples including liquid, solid and semi-liquid (i.e. paste) food samples.     Chapter 2 describes the experimental details regarding the study of MIPs-SERS.     In Chapter 3, chloramphenicol imprinted MIPs particles have been synthesized via precipitation polymerization using similar amount of solvent as bulk polymerization does. The resultant particles exhibit uniformity so that manual grinding and the subsequent potential risk of destroying imprinting binding sites can be avoided. However, the variation of cartridges caused by manual packing could not be ignored. The Canadian penny, containing 98% copper, was employed as the reducing agent to generate dendritic silver functioning as a SERS substrate. It is considered to be cost-effective. In this method, chloramphenicol elution and detection are constructed separately, therefore it is not an integrated sensor.      To integrate those two elements (i.e. separation and detection) and facilitate the detection, Chapter 4 introduces a method that spreads Sudan I imprinted MIPs particles onto a TLC plate. This MIPs-TLC plate links separation and detection together and facilitate the determination. Due to the selective binding interaction, Sudan I has been retained at the original spot, whereas interferents from paprika extracts move forward after developing of the MIPs-TLC plate,   39 therefore, plate developing can be finished less than a minute. Rather than silver dendrite, gold colloidal solution has been deposited onto the Sudan I spot, then the plate is ready for SERS spectral acquisition. The drying of the colloid is fast, therefore diminishing the SERS effect; moreover, the commonly used “glues” to immobilize MIPs particles onto the TLC plate can generate intense SERS signals so that they haven’t been used in the fabrication of the MIPs-TLC plate, which results in unstable MIPs-TLC plate.     Chapter 5 introduces another convenient MIPs film-SERS approach to quantify histamine in canned tuna. Precipitation polymerization has been carried out to generate fine MIPs particles for better distribution when immobilizing particles with PVC to produce a stable film compared to those MIPs-TLC plate. Gold colloid serves as the eluting solvent as well as SERS substrate to simplify the total operation and provide sensitive detection of histamine in canned tuna.     To summarize, this Ph.D. thesis project focuses on determining small molecules in food samples by a MIPs-SERS approach aiming at developing an alternative way for rapid and accurate detection. Different food samples including liquid food samples (i.e. milk and honey), solid food samples (i.e. paprika powder) and paste food samples (i.e. canned tuna) have been studied. Different methods related to separation and detection have been studied and compared, and the obtained results have validated the feasibility of this novel approach.    40  Experimental details Chapter 2: Chapter 2 introduces some experimental procedures that I used in this thesis involving ways to improve the performance of MIPs and acquire high-quality SERS spectra aiming at providing some help to people who may carry out some similar work. This chapter presents and discusses some representative results including assessing properties of MIPs and SERS substrates (i.e. dendritic silver and colloidal gold). Further spectral analysis such as data preprocessing and statistical analysis are also introduced to describe some examples that may help to receive better analytical results.    2.1 Synthesis of MIPs and ways to optimize their properties  MIPs design MIPs design includes binding sites design, scaffold design and the step of designing the morphology of polymers. In my study, I mainly develop non-covalent binding MIPs. Therefore, in terms of non-covalent binding sites design, I am going to introduce some strategies of choosing functional monomers for different template molecules. Non-covalent binding mainly relies on electrostatic interaction and H-bonding. I usually categorize functional monomers to be acidic, basic and neutral on the basis of properties of functional groups that they bear (Table 2.1). Therefore, template molecules bearing functional groups such as amino group (basic), carboxylic group (acidic) or carbonyl (proton acceptor) indicate strong interaction with functional monomers containing proton-donating (acidic) or proton-accepting (basic) groups. Another example of non-covalent binding is hydrophobic interaction that is employed in MIPs. Therefore,   41 neutral and hydrophobic functional monomers (e.g. styrene) work well for template molecules containing less aforementioned functional groups.   Table 2.1: Examples of functional monomers and template molecules in non-covalent MIPs. Functional monomer Property Template example Binding interaction Reference  Bearing acidic group  Electrostatic & H-bonding  [108]  Bearing acidic group  Electrostatic & H-bonding [109]  Bearing neutral group  H-bonding [110]  Bearing neutral group  H-bonding [111]  Bearing neutral group  Hydrophobic [112] HNSOHOOON NNHNNHOOHONH2NHOOOOHOOH2NOOONHNClOClClPSH3CO OCH3  42  Bearing basic group  Electrostatic & H-bonding [89]  Bearing basic group  Electrostatic & H-bonding [113]      Next, it is the scaffold design of MIPs. Scaffold design involves using cross-linking monomers to form the rigid frame of MIPs. The strategy behind is to use cross-linking monomers bearing similar structure of functional monomers, thus, the polymerization property of these two types of monomers can be sort of maintained and disruption of interaction between functional monomers and template molecules by cross-linking monomers can be reduced. Table 2.2 concludes the some of those mostly used cross-linking monomers with respect to functional monomers.      NOOHONH2NCOOHOOOC5H11OHOOC  43 Table 2.2: Pairs of functional monomers and cross-linking monomers. Cross-linking monomer Functional monomer                    At last, it is the morphology design of MIPs. The morphology of MIPs is required by different applications of MIPs. In general, monoliths, beads, films, membranes, fibers and composites of MIPs are mostly employed. All these MIPs are achieved by polymerization methods (e.g. bulk polymerization-monolith, film and membrane, precipitation polymerization-beads), synthesis vessels (e.g. film, membrane and fibers) and post-processing such as grafting of MIPs to a substrate and gelation of MIPs and other polymers (e.g. film, composites).  NOOOOOHOHOOOOOOOOOOOONHONHONH2O  44 Synthesis of MIPs Synthesis of MIPs involves various polymerization methods aiming at yielding high-quality polymer products [84, 93, 95, 96]. Bulk and precipitation polymerizations draw much attention because they are easy to carry out [84]. Pre-polymerization before the addition of initiators at low temperature is recommended to sufficiently obtain relatively stable precursors consisting of template molecules, functional monomers and even cross-linking monomers. Usually it takes a few hours up to a day to prepare precursors and the temperature is maintained at 4 °C or room temperature. Polymerization is carried out around 60 °C because a temperature higher than that may cause instant polymerization, whereas a lower temperature cannot cleave initiators. I usually use strong solvents such as alcohol and acetonitrile with 10-20 % (v/v) acetic acid to wash the resultant polymers to thoroughly remove imprinted molecules and oligomers. Afterward, a second wash by alcohol or acetonitrile without addition of acid is carried out to remove acetic acid residue from first wash because this acid residue can affect further rebinding.   In the meantime, non-imprinted polymers (NIPs) without addition of template molecules during polymerization are also prepared as controls to MIPs.   Characterization of MIPs With respect to testing the property of MIPs and comparing MIPs with NIPs, related static and kinetic studies involve dispersing MIPs particles or other forms of MIPs into analyte molecule solutions. A good solvent with moderate solubility to analytes is preferred as stronger solvent may interrupt the interaction between analytes and MIPs and result in low or even no adsorption of MIPs towards analytes. In contrast, the dissolving solvent displaying less solubility to analytes may result in an increase of non-specific interactions such as hydrophobic interaction. Therefore   45 the adsorption of MIPs towards analytes is similar to that of NIPs, which makes it difficult to differentiate the imprinting effect of MIPs.      Raman spectroscopy can also be used to distinguish MIPs and NIPs. For example, histamine imprinted MIPs and the control polymer (i.e. NIPs) have been prepared as well as the polymers (i.e. MIPs and NIPs) with different amount of monomers. After template removal, Raman spectral acquisition takes place under a 785 nm laser line. As shown in Figure 2.1, most Raman peaks in NIPs are higher than that of MIPs indicating the different composition. The difference in peak intensity may be caused by the formation of precursors because the distribution of functional monomers and cross-linking monomers could be affected by the addition of template molecules and the presence of template molecules may have impact on the monomer activity regarding polymerization.   Moreover, the variation between MIPs and the corresponding NIPs has been highlighted. A peak showing up (1361 cm-1) in the Raman spectrum of NIPs cannot be found in that of MIPs. This may also be induced by the addition of template molecules.     Preparation of MIPs with satisfactory performance (e.g. selectivity to template molecules) is very important because MIPs serve as the separation element in the MIPs-SERS method to determine analytes of interest in food matrices. In this subsection, some experimental techniques have been introduced. Further optimization of MIPs regarding polymerization conditions (e.g. temperature, ratio of template and monomers and reaction time) and testing properties of MIPs (e.g. imprinting effect) can be carried out based on the aforementioned methods.    46  Figure 2.1: Comparison of Raman spectra acquired from MIPs and NIPs. (each spectrum is averaged from 5 raw spectra) (A) histamine: 0.15 mmol (0 mmol for NIPs), methacrylic acid: 0.65 mmol, ethylene glycol dimethylacrylate: 3.90 mmol; (B) histamine: 0.15 mmol (0 mmol for NIPs), methacrylic acid: 0.65 mmol, ethylene glycol dimethylacrylate: 3.28 mmol.   47  2.2 Preparation and discussion of SERS substrates  Preparation and characterization of silver dendrite nanoparticles Besides the preparation of MIPs, a high-performance SERS substrate is also important. In general, we have mainly used dendritic silver and colloidal gold as SERS active substrates. Producing silver dendrite on the Canadian penny is convenient. Scanning electron microscopy (SEM) is employed to obtain the image of silver nanoparticles in the dendritic structure (Figure 2.2). The reaction time between penny and silver nitrate has been studied. Silver nanoparticles start to appear immediately after deposition of AgNO3 solution but it is not completely covering the reaction spot yet until the reaction time reaches 20 to 30 s. Washing is necessary owing to the fact that the nitrate ion is SERS active (1057 cm-1) and may compete with analytes to interact with silver nanoparticles.   Figure 2.2: SEM images of dendritic silver.       As aforementioned, the preparation of silver dendrite involving galvanic reaction between silver nitrate and the Canadian penny (98% copper) is fast. Therefore, I have prepared a few   48 silver dendrite substrates with different reaction time (10 s, 20 s, 30 s, 40 s and 50 s) of the galvanic reaction. After testing, I have found that 10 s is not sufficient to produce silver dendrite that fully covers the reaction spot and the subsequent SERS detection shows relatively weak and non-repeatable Raman signals of chloramphenicol compared to silver dendrite produced in longer time. After 30 s, the reaction spot can be fully covered by silver dendrite with great SERS activity (e.g. good enhancement effect and reproducible spectra). To avoid oxidation of SERS substrate, longer reaction time (> 60 s) is omitted.  Preparation and characterization of gold colloid nanoparticles In terms of gold colloid, homogeneity of AuNPs is significantly important regarding spectral reproducibility. Therefore, transmission electron microscopy (TEM) images are acquired to assess the morphology of AuNPs (Figure 2.3). These two TEM images of AuNPs clearly display spherical, well-dispersed and homogeneous AuNPs with a mean diameter around 18 nm. I believe that the presented aggregation of AuNPs is caused by AuNPs deposition on TEM substrate.     49  Figure 2.3: TEM images of AuNPs.      Given the cost of acquiring TEM images is pretty high, UV spectrometry is employed to assess the absorbance of colloidal gold. A narrow absorption peak of AuNPs is preferred. Figure 2.4 presents the UV spectra from different samples. The UV spectrum of AuNPs presents a slim and isolated absorption peak, indicating the homogeneity of AuNPs. When we introduce 0.8 ppm histamine solution in ethanol into AuNPs (1:1, v/v), a shoulder peak clearly shows up compared to the AuNPs diluted by the same amount of ethanol, while a solution of histamine at higher concentration (4.0 ppm) mixing with AuNPs (1:1, v/v) provokes not only more red shift but also greater absorbance at the corresponding wavelength. For spectra derived from AuNPs with 0.8 ppm histamine and 4.0 ppm histamine, a decrease of AuNPs absorbance at 525 nm is observed. I believe that histamine successfully induces the colloid aggregation (e.g. decrease in peak intensity at original absorption peak and increase in band width) and it may indicate the 17.9	±	3.4	nm   50 subsequent SERS effect as well. However, the appearance of the UV spectrum acquired from the mixture of AuNPs and diluted tuna extract (1:1, v/v) differs from the rest (i.e. UV spectra of AuNPs, AuNPs and ethanol (1:1, v/v), AuNPs and 0.8 ppm histamine in ethanol and AuNPs and 4.0 ppm histamine in ethanol). Diluted tuna extract is used because the undiluted tuna extract can swamp the UV signal of AuNPs when mixing with AuNPs.      In addition, SERS activity of AuNPs can also be evaluated after introducing analytes of interest into AuNPs. Mixture of AuNPs and histamine solutions shows greatly enhanced and reproducible Raman signals of histamine (data not shown).  However, the SERS activity of AuNPs in undiluted tuna extract (concentration of histamine in this extract: 30 ppm) tends to be quenched (Figure 2.5). This may be because the viscous tuna extract may prevent AuNPs from approaching histamine to some extent. Therefore, initial purification of sample extract is highly required.        51   Figure 2.4: Comparison of UV spectra. [Black: AuNPs, red: AuNPs diluted by ethanol (v/v, 1:1), green: AuNPs and 0.8 ppm histamine in ethanol (v/v, 1:1), blue: AuNPs and 4.0 ppm histamine in ethanol (v/v, 1:1), brown: AuNPs and tuna extract in ethanol (v/v, 1:1)]   52  Figure 2.5: SERS spectra of undiluted tuna extract (concentration of histamine in this extract: 30 ppm) and AuNPs (v/v, 1:1).  (Four samples of undiluted tuna extract were mixed with AuNPs to acquire SERS spectra, respectively.)  2.3 Examples of data analysis  Background removal by polynomial fitting In many SERS spectra, lasers of different wavelength produce different envelopes of fluorescence emission, which gives rise to varying baseline. In addition, same incident laser can also produce different trends of baselines (Figure 2.6: black and red SERS spectra of Sudan I derived from the laser line at 785 nm). It is clearly shown that those two spectra (i.e. black and red SERS spectra of Sudan I) are remarkably different than each other despite the fact that they are acquired from the samples with the same concentration. Therefore, I have used initial spectral processing to remove the fluorescence background in order to better compare, present and   53 analyze spectra. Figure 2.6 demonstrates the strategy that I use to optimize the order of polynomial fitting. First order polynomial fitting is not sufficient to remove the fluorescence background here because the resultant spectra mismatch and the tails drift. To better match and flatten those spectra, I select second order polynomial fitting. The processed spectra provide a better similarity, but drifting tails are still obvious. By increasing the polynomial order, the similarity of spectra is optimized and drifting tails are reduced to some extent. In Figure 2.6, the fifth order fitting takes place and the obtained spectra are almost the same to each other, whereas, higher orders (i.e. sixth, seventh and eighth order) present subtle mismatch because those spectra are gently overfitted. It has been reported that the optimal polynomial order are usually 4 to 6 for SERS spectra processing but the spectral range and trend matter a lot [78]. Therefore, in some cases, the first order polynomial fitting may work best. For example, the SERS spectra of Sudan I that I have collected with spectral range 703-1262 cm-1 present relatively flat baseline, therefore, I have chosen first order polynomial fitting to remove fluorescence background. A comparison of processed spectra by polynomial fitting with different orders has to be carried out so that the best order of polynomial fitting could be identified.   54    55 Figure 2.6: Comparison of polynomial fitted SERS spectra (pink and green) at different polynomial orders to raw SERS spectra of Sudan I (black and red) acquired from the samples with the same concentration. [High orders (fourth to eighth) show good background removal compared to lower orders (first to fourth) displaying biased resultant spectra.]  Background removal by smoothing and derivatization Besides the polynomial fitting, derivatives are very helpful as well to remove spectral background. The aforementioned raw spectra (i.e. black and red SERS spectra of Sudan I in Figure 2.6) are processed and compared. Derivatives are generally carried out after spectral smoothing because derivatization can accentuate signals and noise simultaneously, which can make it difficult to differentiate spectra especially when the derivatized signal is not significant. As shown in Figure 2.7A and B, without Savitzky-Golay filter, neither first nor second derivatives present distinguishable signals. In contrast, after 7-point Savitzky-Golay smoothing, S/N ratio is greatly improved. Even though the noise is still strong, a broader Savitzky-Golay filter (more than 9 points) is not recommended as the filter window could possibly smooth narrow but true signals in SERS spectra. It is noteworthy that the background has been efficiently removed by the first or second derivatives.     56  Figure 2.7: Representatives of first and second derivatives of SERS spectra. Derivatives of raw SERS spectra of Sudan I (black and red) acquired from the samples with the same concentration show high noise and difficulty to differentiate them. Derivatives of SERS spectra after 7-point Savitzky-Golay smoothing mainly emphasize useful Raman peaks of Sudan I.  Statistical analysis: PCA As discussed above, the preprocessing is a cosmetic tool that can optimize spectra for a better presentation. In addition, spectral preprocessing can also be advantageous for further statistical treatment and may provide more accuracy to analytical results. To compare results of statistical analysis to raw spectra and preprocessed spectra derived from SERS, we have employed unsupervised PCA to visualize samples. As aforementioned, fluorescence background is not   57 preferred in Raman spectra because it can affect the S/N ratio as well as the statistical analysis of those spectra. Figure 2.8 shows how the spectral preprocessing can affect the unsupervised PCA results of histamine SERS spectra at different concentrations. PCA plot in Figure 2.8A is obtained from raw SERS spectra of histamine. As clearly shown below, SERS spectra of histamine at different concentrations present subtle overlap, in addition, the first PC accounts for 99.56% variation of samples. In contrast, these spectra (Figure 2.8B) can be better separated by unsupervised PCA after the fifth order polynomial fitting and PC1 (59.72%) and PC2 (14.72%) together explain 74.44% variation of samples. I have found the similar results (PC1 explains more than 90% variation of samples without fluorescence removal) in Chapter 3 when analyzing SERS spectra of chloramphenicol with PCA. I believe that the fluorescence background is the reason that causes high PC1 score.      Thus, we can conclude that the fluorescence background does contribute to data analysis results. Furthermore, attention should be paid to investigate the efficacy of preprocessing because spectral preprocessing can totally affect the analytical results, which can be validated by the statistical analysis.     58 Figure 2.8: Comparison of PCA plots. [(A): PCA result (PC1: 99.56%, PC2: 0.25%) of raw SERS spectra of histamine at different concentrations. (B): PCA plot (PC1: 59.72%, PC2: 14.72%) of smoothed and background removed SERS spectra of these samples.] PCA plot of preprocessed spectra shows better separated clusters based on concentrations.  59  Detection and quantification of chloramphenical in milk and Chapter 3:honey using molecularly imprinted polymers: Canadian penny-based SERS nano-biosensor  Chapter 3 is mainly describing the work on my first publication: Gao, F., Feng, S., Chen, Z., Li-Chan, E., Grant, E., Lu, X. (2014), Detection and quantification of chloramphenicol in milk and honey using molecularly imprinted polymers: Canadian penny-based SERS nano-biosensor. Journal of Food Science, 79: 2542-2549. In Chapter 3, MIPs were packed into SPE cartridge to separate chloramphenicol in food samples. This SPE cartridge provided easy operation, involving loading, washing and eluting. Once the separation was done, eluted chloramphenicol was then deposited onto silver dendrite to acquire SERS spectra.  3.1 Introduction Chloramphenicol (CAP) is a broad-spectrum antibiotic originally produced by the bacterium Streptomyces venezuelae and used to treat a variety of bacterial infections in humans. However, use of CAP may induce aplastic anemia and bone marrow suppression due to its toxic effect [114]. Drug abuse is the main factor that causes CAP residues in foods. Many countries, such as China, Canada, United States, and European Union, have strictly banned the use of CAP in food-producing animals [115]. Current methods applied to detect CAP in foods are microbiological test [116], sensors [117], chromatographic methods (for example, gas chromatography-mass spectrometry [118], Liquid chromatography–mass spectrometry [119]), and immunoassays [49]. Unfortunately, a few technical challenges exist for these methods. Specifically, laborious sample   60 pretreatment is required due to the interference from food matrices and high-throughput detection of a large amount of samples cannot be realized. Solid phase extraction (SPE) is commonly used as a clean-up method to isolate CAP from food matrices but with unsatisfied selectivity.     Molecularly imprinted polymers (MIPs) have been reported as a unique material to selectively recognize, separate, and enrich different chemical compounds including CAP [115, 120]. MIPs are recognized as “artificial antibodies” due to their ability to bind specific analytes [121]. In addition, MIPs have the advantages of low cost, ease of preparation, high chemical and thermal stability, and long shelf-life [89]. During the polymerization of MIPs, functional monomers interact with template molecules (for example, CAP) through covalent or noncovalent binding initially, followed by addition of cross-linking agents and initiators; MIPs containing template molecules are thus synthesized. Subsequent removal of the template molecules reveals complementary binding sites molded in the MIPs that are completely complementary to the structure of the template molecules and thus able to selectively recognize template molecules with relatively high separation and enrichment efficiency even in a complex matrix such as foods. MIPs have been incorporated as an important element of sensors for the determination of chemical and microbiological hazards in foods [122, 123]. In summary, MIPs can serve as a adsorbent for SPE to concentrate target molecules, resulting in lower limit of detection.     Surface-enhanced Raman spectroscopy (SERS) is a rapid, sensitive and nondestructive detection tool to analyze and characterize molecules. This application has overcome the technical barrier of normal Raman spectroscopy, which is the weak intensity of Raman scattering [44]. In SERS, the enhancement of Raman signal may be derived from either electromagnetic (EM) field or chemical effect. Localized surface plasmon resonance (LSPR) generated by the incident light   61 is believed to produce an EM field. Molecules adsorbed on or in the proximity to the surface of noble metal (for example, silver and gold) nanostructures would sense the EM field and reradiate the enhanced Raman signal generated by the vibrations of molecules [60]. Sharp surfaces and surface protrusions are most likely to produce enhanced signals with high order due to the probability of generating LSPR, explaining the wide application of roughened metallic surface for SERS. The EM field is the dominant contribution to the enhancement of Raman signal to the orders of 106 to 108, whereas chemical enhancement is associated with a smaller enhancement effect (approximately 102) [124]. The total enhancement factor can sometimes reach to 1014 to 1015 [125], leading to detection at the single molecule level. This effect has been validated to be feasible by both theoretical calculation based on classic EM theory [125] and research work for the detection of a single molecule of crystal violet in aqueous colloidal silver solution [57]. Recently, researchers have successfully manipulated the detection of single molecule by tip-enhanced Raman spectroscopy [126] and plasmon-enhanced Raman spectroscopy [127] by producing an even stronger EM field for the analyte molecules.     SERS is widely used in the detection of foodborne pathogens [128], chemical contaminants [129, 130], and food adulteration [131, 132], indicating feasible application in food safety inspection and bioterrorism prevention. A recent work was designed to apply SERS to detect enrofloxacin, furazolidone, and malachite green in fish products after a complicated sample pretreatment [133]. In another study, SERS was conducted to detect melamine in selective food matrices; only a simple sample pretreatment was required in this case, due to the unique ring breathing mode of triazine in the molecule of melamine, resulting in a unique Raman band at 680 cm−1 which could be easily distinguished from Raman bands derived from the remaining food components after sample pretreatment [61, 132]. Because SERS is a detection method and   62 acquires all vibrational modes from existing molecules, in most applications extensive sample pretreatment is required to remove interfering molecules from matrices before detection by SERS and this involves the usage of large amounts of solvents and multiple extraction procedures (for example, liquid–liquid extraction, SPE). To avoid laborious sample pretreatment, other promising separation methods have been proposed to eliminate interferences from matrices before SERS detection. Our group recently developed a microfluidics-based “lab-on-a-chip” system as the separation approach coupled with SERS to detect and differentiate methicillin-sensitive Staphylococcus aureus and methicillin-resistant S. aureus clinical isolates from China and the United States [128]. This innovative nano-sensor significantly shortened analysis time and realized high-throughput detection and analysis.     In this study, our objective was to integrate MIPs with SERS to develop a “2-step” nano-sensor for separation and detection of CAP in different food products, including skim milk, whole milk, and honey. The results demonstrate the potential of MIPs-SERS to serve as an innovative nano-sensor for rapid, high-throughput, and reliable detection of trace levels of chemical hazards in agricultural and food systems.  3.2 Experimental section  Chemicals and reagents Acrylamide (AM; 99% purity), ethylene glycol dimethacrylate (EGDMA; 98% purity), 2,2’-azobis(isobutyronitrile) (AIBN; 99% purity), CAP, silver nitrate, CAP base were purchased from Sigma-Aldrich (St. Louis, MO, USA). Methanol (high-performance liquid chromatography; HPLC grade), acetic acid, sodium chloride, ethyl acetate, and anhydrous sodium sulfate were   63 obtained from Thermo Fisher Scientific (Waltham, MA, USA). Florphenicol was purchased from AK Scientific (Union City, CA, USA). Deionized (DI) water (Millipore (Darmstadt, Germany), 18.2 MΩ/cm) was prepared.  3.2.1 Synthesis and characterization of MIPs towards chloramphenicol  Synthesis of MIPs CAP (template molecule, 0.5 mmol) was dissolved in 5 mL methanol and then 10 mmol EGDMA (cross-linking agent) and 142 mg AM (monomer) were added. The mixture was sonicated in an ice bath for 5 min, followed by the addition of 20 mg AIBN (initiator). After the solution was purged by nitrogen for another 5 min, polymerization was conducted in an oil bath at 60 °C for 20 h. The formed polymer was sieved through 200-mesh sieve. CAP was removed from the synthesized MIPs by Soxhlet extraction first with 300 mL methanol/acetic acid (9:1, v/v) for 24 h, followed by 300 mL methanol until no CAP could be detected based on UV absorbance at 277.5 nm. MIPs were vacuum dried at 22 °C for 4 h. Nonimprinted polymers (NIPs) were also synthesized following the same procedure but with no addition of CAP.  Static and kinetic adsorption capacities of MIPs To assess the static adsorption capacities, MIPs (20 mg) were added into 2 mL of each of the standard solutions (25, 50, 100, 200, 300 ppm CAP in 10% methanol). The maximum adsorption (λmax = 277.5 nm) was determined by scanning the standard solution from 200 to 800 nm. The calibration curve of absorbance against concentration was also plotted. The mixtures were shaken at 200 rpm for 3 h at 22 °C, and then centrifuged at 13000 ×g for 20 min. The   64 concentration of CAP in the supernatant was determined by UV absorbance at 277.5 nm. The kinetic adsorption study was carried out by mixing 160 mg MIPs and 16 mL 50 ppm CAP standard solution, followed by shaking at 200 rpm. At different time points (10, 20, 30, 45, 60, 90, 120, 180 min), 1 mL aliquots were taken, centrifuged (13000 ×g, 20 min), and the concentration of CAP determined by UV absorbance at 277.5 nm. The static and kinetic adsorption capacities of NIPs were determined following the same procedure as described above.     Interference study was performed by adding 2 mL of florphenicol or CAP base (50 or 300 ppm in 10% methanol aqueous solution) to 20 mg MIPs or NIPs, followed by shaking at 200 rpm for 3 h at 22 °C. The mixture was centrifuged at 13000 ×g for 20 min and the supernatant was collected and determined by UV spectrometry (224 and 272 nm for florphenicol and CAP base, respectively).  3.2.2 Preparation of penny-based silver nano-substrate The Canadian pre-1996 bronze pennies (98% copper) were selected and sonicated within methanol with 5% acetic acid for 30 min and then washed by methanol [134]. Once the penny was dried, 2 µL of 0.1 M silver nitrate solution was deposited onto the surface of the penny for 30 s displacement to form a silver dendrite spot. Excessive ions were gently washed off with DI water and methanol.  3.2.3 Performance of molecularly imprinted solid phase extraction on food samples Pretreatment of food samples Whole milk, skim milk, and honey products were obtained from local grocery stores. Sodium chloride (NaCl, 1 g) was added in 2 mL milk or honey (1 g dissolved in 1 mL DI water), and   65 then spiked with CAP (100 ppm in methanol) for a final concentration of 0, 100 ppb, 500 ppb, 1 ppm, and 5 ppm. Ethyl acetate (5 mL) was then added and the mixture was vortexed for 30 s and centrifuged at 10000 ×g for 2 min. Top ethyl acetate layer of liquid phase was run through an anhydrous sodium sulfate column to remove water. For milk samples, a sencond extraction with ethyl acetate was performed, and the two extracts were pooled. Dehydrated ethyl acetate was subsequently dried down by rota-vap and finally redissolved in 10% methanol aqueous solution.  High-performance liquid chromatography The analyses were conducted on a HPLC system consisting of Agilent 1100 series quaternary pump and a diode array detector. A sample amount of 20 µL (in 50% methanol aqueous solution) was injected on a C18 column (µBondapak-C18, 10 µm, 3.9 mm × 300 mm, Waters, Milford, Mass., U.S.A.) and separated by isocratic elution with a mixture of methanol and water (50:50, v/v) at 1.0 mL/min flow rate and 25 °C. The CAP was detected at 278 nm.  Molecularly imprinted SPE (MISPE) MIPs (100 mg) were packed into a pipette tip (1 mL), which was packed with glass fiber plug at the bottom. A thin layer of sea sand was applied on the top of MIPs. Methanol (3 mL) was first added, followed by 3 mL DI water to condition the MISPE. CAP spiked (0, 0.1, 0.5, 1, 5 ppm, respectively) solution (0.2 mL × 5 times) was loaded at 0.33 mL/min, followed by the addition of 1 mL of 15% methanol aqueous solution and 1 mL DI water to wash off interference at 1 mL/min. Finally, CAP was eluted out by 3 mL of 10% acetic acid in methanol solution at 1 mL/min. Flow rate was controlled by the application of vacuum to SPE manifold. The eluents (4 µL) were dried down and redissolved in 4 µL 50% methanol aqueous solution. The resultant   66 solution was directly deposited onto the silver dendrite spot formed on Canadian penny and blow-dried by nitrogen.  3.2.4 Spectral collection and data analysis Section  Raman spectroscopy SERS spectral collection (700 to 1700 cm−1) was conducted using a 785 nm laser (0.25 mW incident laser power) coupled with a 50× objective. The spectrometer (Renishaw, Gloucestershire, UK) has an entrance slit of 50 µm and confocal length of 300 mm and is equipped with a 1200-line/mm grating. Raman scattered light was detected by using a 578 by 385 pixel charge–coupled device array detector. Each spectrum was collected in 10 s. The size of each pixel is 22 µm × 22 µm. Spectra (N = 8) from different locations of a sample were collected to study spectral reproducibility.   Chemometric models and statistical analysis Raman spectra were binned (2 cm−1) and smoothed (9-point Savitzky-Golay algorithm) by using OMNIC version 7.1 (Thermo Electron Corp., Madison, Wis., U.S.A.). Unsupervised principal component analysis (PCA) was conducted to classify samples according to different concentrations of CAP (0, 0.1, 0.5, 1, 5 ppm). Partial least squares regression (PLSR) model was established and cross-validated via “leave-one-out” validation for the quantification and prediction of the actual contents of CAP in milk and honey products. The experiment was performed in at least 3 replicate trials. PCA and PLSR models were established by using Delight version 3.2.1 (D-Squared Development Inc., LaGrande, Oreg., U.S.A.).   67  3.3 Results and discussion   3.3.1 Assessment of MIPs separation of CAP in foods  Evaluation of MIPs performance Precipitation polymerization was carried out using methanol as the porogen. The resultant polymer showed uniform and fine powdery form instead of one-piece hard solid synthesized by bulk polymerization, saving tremendous time on grinding and possibly avoiding destroying complementary binding sites due to the irregular shape of particles caused by grinding [135]. Parallel polymerization experiment conducted without the addition of template molecule and functional monomer validated that methanol tends to break down the cross-linking degree and leads to precipitation polymerization without considering large amount of solvent to be used in precipitation polymerization. Size of the fine particle of MIPs may be manipulated by further studies of HPLC packing materials (for example, ratio of methanol to monomers).     Adsorption capacities (Q) of MIPs and NIPs are the major factors that can be used to evaluate the performance of MIPs. Further, imprinting factor (IF) is another important factor to assess the selectivity of MIPs over NIPs while Scatchard plot can be used to determine the affinity of MIPs and NIPs [136]. In brief, adsorption capacity is calculated as:  Q = (Ci – Cf) × V/W (3.1) where Ci is the initial concentration of CAP, Cf represents the final concentration of CAP after polymer adsorption, V is the volume of CAP solution, and W is the amount of the polymers used. IF is determined by the equation:   68  IF = QMIPs/QNIPs (3.2)     The Scatchard plot is developed according to the equation:  Q/Ci = −Q/Kd + Qmax/Kd (3.3) where Qmax is the saturated adsorption capacity and Kd is the dissociation constant.     Both static and kinetic studies were carried out using CAP in 10% methanol aqueous solution. Due to improved solubility of CAP in methanol, Q decreased as a function of increasing portion of methanol aqueous solution, in the cases of both MIPs and NIPs (Figure 3.1). The 10% methanol aqueous solution was selected because this solution provided better IF and an appropriate amount of methanol can further reduce the non-specific binding caused by low solubility of CAP in aqueous solution.  Figure 3.1: Adsorption capacity of MIPs and NIPs to CAP at 50 ppm after 3 h incubation as a function of methanol percentage in aqueous solution.   0 0.4 0.8 1.2 1.6 0% 20% 40% 60% 80% 100% Adsorption capacity (mg/g) Methanol portion in aqueous solution MIPs NIPs   69     The binding isotherms of MIPs and NIPs for CAP in the static adsorption studies are shown in Figure 3.2A. Both QMIPs and QNIPs increased as the initial concentration of CAP increased, but the QMIPs reached 5.10 mg/g while QNIPs was only 1.05 mg/g at 300 mg/L of the initial CAP solution, leading to an IF value of 4.9 at this concentration. The low adsorption capacity isotherm curve exhibited by NIPs indicate only low level of non-specific adsorption of NIPs toward CAP; these results also suggest limited non-specific binding sites in MIPs, indicating that most of the binding sites in MIPs are complementary to CAP. Both the higher QMIPs isotherm curve and the IF validated that MIPs have much better selectivity toward CAP than NIPs and are suitable for further application of MISPE.     Scatchard analysis showed a linear plot between Q/Ci and Q for MIPs (R2 = 0.97), whereas no linear relationship was observed for NIPs (Figure 3.2B). The Qmax of MIPs calculated according to the Scatchard plot was 13.14 mg/g. The linear Scatchard plot of MIPs indicates that the binding sites in MIPs are homogenous, which are hypothesized to be driven by hydrogen bonding between MIPs and the template molecule (CAP) [137]. A previous study validated that the strong retention of CAP by MIPs is mainly due to the interaction of 2 hydroxyl groups of CAP and the tertiary amine of 2-(diethylamino) ethyl methacrylate (monomer) [138]. For NIPs, we assumed that non-specific interactions between NIPs and CAP may take place and this non-specific interaction is mainly associated with hydrophobic interaction [65].    70   Figure 3.2: (A) Isotherm of binding properties of molecularly imprinted polymers (MIPs) and non-imprinted polymers (NIPs); (B) Scatchard plot of MIPs and NIPs. Values are shown as mean ± standard deviation (n = 3) (initial concentration of CAP in 10% methanol: 25, 50, 100, 200, 300 ppm; time: 3 h).      The results of kinetic adsorption studies are shown in Figure 3.3. The adsorption capacity of MIPs gradually increased along with time while QNIPs was almost constant with time. This validated that non-specific binding is fast but limited, whereas complementary binding sites were successfully imprinted in MIPs, resulting in longer times for CAP molecules to orientate themselves to fit into those sites in MIPs.  y = -0.0021x + 0.0276 R² = 0.9682 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0 1 2 3 4 5 6 Q/C (mL/mg) Q (mg/g)	MIPs NIPs A B 0 1 2 3 4 5 6 0 50 100 150 200 250 300 350 Adsorption capacity (mg/g) Initial concentration of chloramphenicol (ppm) MIPs NIPs   71  Figure 3.3: Kinetic study of molecularly imprinted polymers (MIPs) and non-imprinted polymers (NIPs) as a function of incubation time. Values are shown as mean ± standard deviation. (n = 3) (Initial concentration of CAP in 10% methanol: 50 ppm; time points: 10, 20, 30, 45, 60, 90, 120, 180 min).      An interference study was carried out by comparing the adsorption capacity of synthesized MIPs for CAP and 2 CAP structural analogs (that is, florphenicol and CAP base). The adsorption capacity of florphenicol at 50 ppm was 0.66 mg/g for MIPs and 0.10 mg/g for NIPs, respectively. For 300 ppm of florphenicol, the adsorption capacity was 2.70 mg/g for MIPs and 0.81 mg/g for NIPs, respectively. The adsorption capacities of CAP base were 0.03 mg/g for MIPs and 0.02 mg/g for NIPs at 50 ppm, while 1.32 mg/g for MIPs and 0.30 mg/g for NIPs at 300 ppm. Low adsorption capacity of CAP structural analogs indicates the selectivity of MIPs toward CAP. The molecular structure of florphenicol is quite similar to CAP, except -SO2CH3 replaces -NO2 and –F substituents for -OH. CAP base shows no amide tail. Thus, we conclude here both binding and spatial effect contribute to the special affinity between MIPs and CAP. 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 0 50 100 150 200 Adsorption capacity (mg/g) Time (min) MIPs NIPs   72     The current MIPs were designed and synthesized based upon noncovalent binding, which is easy to process and has a fast rebinding profile. The disadvantage of this type of MIPs is the inevitable non-specific binding compared to covalent binding-based design. Our developed MIPs significantly decrease the non-specific binding sites and provide better selectivity of MIPs to template molecules compared to other previous studies [139, 140]. Previous studies developed MIPs using an analog to CAP as template molecule, which avoids the drawback associated with MIPs of residual template bleeding [120]. However, the interaction between created MIPs cavities and analyte may be different from template (analyte analog), resulting in less selectivity (IF = 2.43 under HPLC condition [120]).  Assessment of MIPs separation of CAP in foods Honey is a complicated food matrix and therefore extensive sample pretreatment is required to effectively separate and enrich trace level of target molecules in honey and the sample pretreatment generally takes several hours [141]. Dissolving, hydrolysis, and complicated extraction steps are required to be conducted because honey contains mainly sugars (77% of monosaccharides and oligosaccharides in total) and some other minor components, such as organic acids, proteins, amino acids, and elements (K, P, Mg, Ca, Fe, Si, S, Cu, F, Zn, Mn, Co, Mo, Cr, and I) [141]. The complex nature of honey poses a technical challenge to separate specific analyte molecules and maintain a decently high level of recovery during sample pretreatment. Milk contains lactose, fats, proteins, vitamins, and salts. For separation of target molecules in milk, liquid–liquid extraction is usually performed, followed by basic hydrolysis to break down fats. After neutralization, the solution is concentrated down to a small volume, which is subsequently run through a silica SPE cartridge to further purify the analyte molecules   73 before determination by chromatographic method. The whole process takes about 1 h [142]. Taken together, both sample pretreatments involve large amounts of organic solvent and are time consuming. Our current sample pretreatment was performed by minimum operation to initially remove ions, fats, and proteins, saving a tremendous amount of time.      MIPs were packed into 1 mL pipette tip used as SPE sorbents (Figure 3.4). The thin cartridge, slow flow rate (0.33 mL/min) and multiple loading (200 µL × 5) offered CAP a longtime reservation for sufficient interaction with MIPs. Washing with 1 mL of 15% methanol aqueous solution could clean up the polar interferences and may wash the superficially absorbed CAP to the complementary binding sites to tightly bind with MIPs, followed by 1 mL water to further remove sugar. The final eluting solvent broke down noncovalent interaction between CAP and MIPs and eluted CAP out. Recoveries of both MISPE and nonimprinted SPE (NISPE) quantified by HPLC are summarized in Table 3.1. NIPs could barely hold CAP even at a low spiked level in foods and showed relatively low recovery (only approximately 20%). In contrast, MIPs show about 2 times the recovery of NIPs, validating the existence of selective binding interaction between MIPs and the target molecule (CAP). The relatively low recoveries of MIPs are believed to be due to the slow kinetics of MIPs.    74  Figure 3.4: Illustration of molecularly imprinted solid phase extraction (MISPE) for chloramphenicol.  Table 3.1: Spiked recoveries: molecularly imprinted solid phase extraction (MISPE) and non-imprinted solid phase extraction (NISPE).  skim milk recoveries whole milk recoveries honey recoveries CAP spiking (ppm) MISPE NISPE MISPE NISPE MISPE NISPE 0 N/A N/A N/A N/A N/A N/A 0.1 61.5% 0.0% 60.0% 0.0% 53.1% 0.0% 0.5 50.0% 22.6% 50.9% 9.1% 49.1% 24.7% 1 50.2% 16.3% 51.7% 15.8% 51.3% 29.1% 5 54.8% 19.2% 47.2% 16.2% 42.9% 22.0%      75 Optimization of process for generating SERS substrate on Canadian penny Silver dendrite structure has been applied to obtain SERS signals with high enhancement factor and good reproducibility [143, 144]. This nano-structure is generated by galvanic reaction between reducing metal (for example, zinc) and silver nitrate or silver fluoride solution. By varying the categories of metal as substrate, concentration of silver salts solutions, or reaction time, silver dendrite can be achieved with consistency [143, 145]. Gutés and coauthors fabricated a monolayer of silver dendrite produced by galvanic displacement from commercial aluminum foil by depositing it onto double-sided Scotch tape. Excellent SERS results from 1,2-benzenedithiol, 1-phenylethyl mercaptan and 2,2’-dithiodipyridine validated it to be a good candidate as SERS nano-substrate [146]. In another study, Mabbott and collaborators created a cost-effective method to generate silver dendrite on British 2p coins and demonstrated that it could enhance Raman signals of Rhodamine 6G and 3 drugs, namely 4-methylmethcathinone, 5,6-methylenedioxy-2-aminoindane and 3,4-methylenedioxy-N-methylamphetamine [134].     In this study, silver dendrite was synthesized by depositing silver nitrate solution onto a surface-cleaned Canadian penny. Pre-1996 Canadian penny was selected due to its high content of copper (98% purity). A spot of silver dendrite with 3-mm diameter was generated by using 2 µL of 1 mol AgNO3. It was found that 30 s of reaction time could produce the dendrite structure that has the best performance for the SERS detection of CAP molecules. The dendritic structure of silver can be clearly observed by a scanning electron microscope (Figure 3.5). Peeling off operation is not required and this leaves the structure intact and uniform. The roughened surface of the coin not only provided the substrate for dendritic silver, but also formed a uniform coat tightly adhering to the coin surface to avoid fragmentation, which was reported to occur when depositing onto a bare slide [147]. We blow-dried the solution of CAP spotted on silver dendrite   76 to shorten the drying time and this procedure could also prevent the generation of a “coffee ring” [148] of solution containing the target molecule (for example, CAP), which may produce normal Raman signal.    Figure 3.5: Scanning electron microscope image of silver dendrite formed on Canadian penny.      A representative normal Raman spectrum of CAP (solid crystal) in Figure 3.6 shows 3 major bands at 1107 cm−1 (N-H in-plane bending), 1345 cm−1 (NO2 symmetric stretching), and 1598 cm−1 (ring stretching) [149]. In comparison, a SERS spectrum of standard CAP solution (1 ppm) and MIPs-SERS spectrum of CAP spiked skim milk (1 ppm) are also shown on the top of normal   77 Raman spectrum. These 3 prominent peaks assigned to CAP were obviously showing up among samples. Only 2 distinct bands attributed to components from the milk sample after MIPs treatment were found in the MIPs-SERS spectrum; these were at 930 and 980 cm−1 (assigned to lactose in skim milk) [150]. The band at 1057 cm−1 is assigned to NO3− [143], which originated from the synthesis of silver dendrite. The enhanced intensity of Raman bands indicates that CAP anchors itself in its favorite orientation onto the silver dendrite. We believe that silver tends to interact with the electron deficient nitrobenzyl group (Figure 3.7).    Figure 3.6: Representative Raman spectra of chloramphenicol [from top to bottom: MIPs-SERS spectrum of extract of CAP-spiked (1 ppm) skim milk, SERS spectrum of standard CAP solution (1 ppm), and normal Raman spectrum of solid crystal].      78  Figure 3.7: Schematic model of orientation of chloramphenicol on silver dendrite.  3.3.2 Quantitative analysis of MIPs-SERS nano-sensor Figure 3.8 demonstrates the steps of the synthesis of MIPs and the application of this nano-sensor integrating MIPs with SERS. In Figure 3.6, the MIPs-SERS spectrum indicates the capability of cleaning interferences from food matrices by MIPs and provides the enhanced Raman signals of CAP. After spectral collection (700 to 1700 cm−1), chemometric models were constructed to further evaluate the feasibility of MIPs-SERS nano-sensor. Preprocessed SERS spectra (after bin and smooth) were used to establish unsupervised 2-dimension PCA models. A representative PCA model (Concentrations: 5; Spectra at each concentration: 8) of whole milk is shown in Figure 3.9A. Different clusters stand for different spiked levels of CAP (0, 0.1, 0.5, 1, 5 ppm) in whole milk. Clusters are well segregated, indicating PCA-based Raman spectroscopic analysis could successfully differentiate various CAP contents in whole milk. PCA models of skim milk and honey are shown in Figure 3.10 and 3.11, both of which were well established with tight clusters for each spiked concentration. PLSR was applied to the processed SERS spectra and 3 individual PLSR models were established and applied to 3 different food matrices. Figure 3.9B is the representative PLSR model (Concentrations: 5; Spectra at each concentration: 8) for the quantification of CAP in whole milk. The linearity of the plot (R = 0.972) indicates the   79 accurate prediction ability of this PLSR model. In addition, the PLSR model of skim milk (R = 0.975) (Figure 3.10 and Table 3.2) shows a good prediction of the actual concentration of CAP as well. The PLSR model of honey provided a less satisfactory quantification and prediction of the actual concentration of CAP (R = 0.908) (Figure 3.11 and Table 3.2) compared to the ones of whole milk and skim milk. This is probably due to a higher content of sugar residues after MIPs clean-up.    Figure 3.8: Schematic illustration of the development of MIPs-SERS nano-sensor to detect chloramphenicol in selective food matrices.    80   Figure 3.9: Representative 2-dimensional principal component analysis (PCA) model for differentiation of whole milk containing various concentrations of chloramphenicol (PC1: 95.3%; PC2: 2.1%); (B) representative partial least squares regression (PLSR) model to predict actual concentrations of chloramphenicol in whole milk.    81  Figure 3.10: (A) Representative two-dimensional principal component analysis (PCA) model for separation of skim milk containing different concentrations of chloramphenicol (PC1: 93.8%; PC2: 2.8%); (B) representative partial least squares regression (PLSR) model to predict actual concentrations of chloramphenicol in skim milk.          A B   82  Figure 3.11: (A) Representative two-dimensional principal component analysis (PCA) model for separation of honey containing different concentrations of chloramphenicol (PC1: 73.3%; PC2: 9.5%); (B) representative partial least squares regression (PLSR) model to predict actual concentrations of chloramphenicol in honey.  Table 3.2: Partial least squares regression (PLSR) models for the prediction of chloramphenicol concentration in foods. Food matrix Spiked concentration (ppm) Number of spectra Correlation coefficient RMSa error of estimation RMS error of validation    skim milk 0, 0.1, 0.5, 1, 5 40 0.975 0.300 0.413    whole milk 0, 0.1, 0.5, 1, 5 40 0.972 0.221 0.516    honey 0, 0.1, 0.5, 1, 5 40 0.908 0.538 0.799    aRMS: root-mean-square.      The currently developed MIPs-SERS nano-sensor could significantly shorten the separation and detection time of CAP in food samples without the usage of large amount of solvents.         A B   83 Minimum pre-extraction of CAP in food samples and further purification of CAP by running MISPE at most take 15 min. Even with MIPs as pretreatment method, common chromatographic measurement takes unavoidable much longer time (at least 10 min for isocratic method) instead of 10 s for individual SERS spectral collection. Therefore, this sensitive and high-throughput MIPs-SERS nano-sensor is more practicable in the circumstances (for example, in-field and on-line) of monitoring large amount of potentially contaminated agricultural and food samples in a short time period.     Different from chromatographic approach, which provides a second separation by column before detection [151], SERS records every probable vibration of molecules including interferences. Therefore, a separation technique is critical for SERS detection. Our group recently integrated MIPs with SERS for the determination of alpha-tocopherol in different vegetable oils [66]. In other studies, aptamer [65] and antibody [64] were introduced as separation methods before SERS detection of specific chemical compounds in liquid foods. However, developing a specific aptamer is laborious and the commercially available aptamers are still limited. Antibody is sensitive but difficult to be employed for in-field detection because proteins tend to denature under environmental conditions. In contrast, MIPs are easy to be fabricated, highly selective, and thermally stable, demonstrating to be a good candidate as separation method before SERS detection. Recently, a “one-step” MIPs-SERS nano-sensor has been developed by incorporating silver nanoparticles into MIPs matrices [68, 152]. Generally, 1 h was needed to reach equilibrium for the interaction between MIPs-Ag nanoparticles and food samples containing trace levels of target analytes.   84 3.4 Conclusion In conclusion, we developed an innovative MIPs-SERS nano-sensor, which can rapidly and accurately separate and detect CAP in foods with high throughput capability. The whole process including sample preparation, MIPs separation and SERS detection are finished in 15 min. MIPs are stable and reusable materials to selectively separate CAP from complicated food matrices. A cost-effective Canadian penny can be used to generate silver dendrite as a reducing agent. MIPs-SERS-sensor has the potential to determine trace level of chemical hazards in agricultural and food products.        85  Determination of Sudan I in paprika powder by molecularly Chapter 4:imprinted polymers-thin layer chromatography-surface enhanced Raman spectroscopic biosensor  A version of Chapter 4 has been published as: Gao F., Hu Y., Chen D., Li-Chan E.C.Y., Grant E., Lu X. (2015), Determination of Sudan I in paprika powder by molecularly imprinted polymers–thin layer chromatography–surface enhanced Raman spectroscopic biosensor. Talanta, 143: 344-352. This chapter introduces TLC based MIPs separation. During TLC developing, Sudan I spot was retained at original sample spot, whereas, interferents were washed forward. The developing was done less than a minute. Addition of gold colloid (i.e. SERS active substrate) to Sudan I spot facilitated detection. Chapter 4 introduces a way to integrate MIPs separation and SERS detection and therefore significantly improve the efficacy of determination of Sudan I in paprika powder compared to the method described in Chapter 3 with separated two elements: MIPs separation and SERS detection.  4.1 Introduction As an industrial dye, Sudan I is commonly used to color oils, waxes, and polishes. It is also added illegally to foodstuffs and cosmetics for color enhancement [153]. The compound is classified as a carcinogenic and mutagenic compound by International Agency for Research on Cancer [153]. In July 2003, Sudan I was detected in chili products imported from India by a French lab, and the European community immediately released the legislation to track and remove the contaminated food products [154]. In February 2005, the United Kingdom recalled   86 more than 470 food products contaminated with Sudan I [62]. A large adulteration incident with Sudan I spread globally later in the same year, causing tremendous panic and concern over food safety [62]. Monitoring Sudan I in foods is required due to the wide adulteration of this azo compound in a variety of natural agricultural products and processed foods, and rapid, low-cost, reliable and high-throughput methods are in high demand.     Chromatographic methods coupled with various detectors for the detection of Sudan I have been studied [154] and [155]. Liquid chromatography [156] has validated to be essential as a separation method because Sudan I is thermally unstable, making gas chromatography (GC) inappropriate without derivatization [157]. The complexity of typical food samples demands a long column running time (15 min or more) for separation; thus, LC is not suitable for rapid detection. In terms of detectors, UV–vis spectrometry [158] lacks specificity and leads to misidentification of analytes [154]. Mass spectrometry (MS) is substantially more reliable and shows a good limit of detection (LOD). Atmospheric pressure chemical ionization tandem MS combining isotope dilution was used to assay Sudan I in food samples with sensitivity (LOD of Sudan I in food extract: 0.025 ppm) 20 times better than that of a high-performance liquid chromatography (HPLC)-UV approach [154]. He et al. [157] performed GC–MS approach under electron impact mode or negative chemical ionization mode to determine Sudan I in eggs with an LOD of 2.0–2.1 µg/kg. However, MS is well known as an expensive detector. In sum, column-based chromatographic approaches were time-consuming and sometimes expensive (i.e. MS), resulting in less preference.     Alternatively, TLC or high-performance TLC (HPTLC) for the detection of Sudan I has been studied as well. Quantitative study of the analyte zone on a TLC plate often involves scraping the stationary phase off the plate and extracting the analyte from the sorbent in multiple steps.   87 Unfortunately, this may cause loss of analyte at each step, and requires a further off-line detection by either MS [156] or nuclear magnetic resonance [158]. Having an effective interface between plate and detector is needed [159]. To avoid the loss and achieve on-line detection, Kandler et al. [160] developed a fast and low-cost HPTLC-densitometry approach to determine Sudan I in food products. The LOD of this method was 30–150 ng/zone, but they realized the limited reliability of this approach due to sample matrix effect. In terms of stationary phase of the chromatographic approaches, reverse phase C18 were most common [154], [155] and [157]. Further, a novel stationary phase (i.e. MIPs) to extract Sudan I from food samples has recently been developed [161], [162] and [163]. MIPs are often regarded as an “artificial antibody” because MIPs can form “lock and key” bonds with specific template molecules. The precursor (i.e. the newborn monomer derived from the interaction between a functional monomer and a template molecule) copolymerizes with a cross-linking monomer (i.e. monomer which serves to provide MIPs rigidity) to produce durable template-imprinted polymers. After the removal of the template, the complementary binding sites on MIPs are exposed and available for selective rebinding of the template from a complex sample matrix, such as foods. Functional monomers, cross-linking monomers, initiators, solvents and other parameters vary according to different template molecules. MIPs are widely used for separation and determination of the analytes in complicated samples [164] and [165]. Previous studies demonstrated that MIPs towards Sudan I were generated and packed either into solid phase extraction (SPE) cartridges [166] or HPLC analytical columns [167] or pre-columns [161] for separation, with a pre-treatment (e.g. solid–liquid extraction (SLE)) required for real samples. However, column conditioning is time consuming and sometimes expensive detectors are required.   88     Sudan I has been analyzed without chromatographic separation. For example, Lopez, [168] Di Anibal [169] and Cheung [62] individually conducted SLE and SERS detection of Sudan I in food samples, applying multivariate analysis as a screening tool. The LOD of these approaches were 10-7 M (Sudan I in extract) and 48 µg/kg (Sudan I in chili powder), respectively. In addition, electrochemical [e.g. carbon electrode modified by silver nanoparticles-graphene [170] or multi-walled carbon nanotubes (MWNTs) [171] and ionic liquid coupled with MWNTs [172]] and chemiluminescence (e.g. quenched fluorescence of quantum dots [173]) measurements can also be applied to determine Sudan I in food samples, but intensive sample pre-treatment was required.     As a derivative of vibrational spectroscopy, SERS offers extremely low LOD and is capable of qualitative and quantitative analysis. It is known as a fast, cost-effective and non-destructive detection tool [54]. SERS is used to substitute normal Raman mode due to the weak intensity of Raman scattering signals of molecules, with only 1 out of 10 million of the scattered photons experiencing Raman scattering when incident light interacts with the target molecules. Nanoparticles or nanoscale roughened surface of metals (e.g. gold, silver and copper) can be employed as a SERS substrate to enhance the normal Raman signals of molecules. Enhancement effect is contributed by two factors: chemical enhancement caused by charge transfer between the molecule and the substrate, and electromagnetic enhancement when the molecule is placed in the zone of localized surface plasmon resonance of SERS substrate. The distance between the molecule and surface of substrate should be less than 10 nm [174]. Total enhancement factor can exceed 1014 and molecules can be easily distinguished due to their unique SERS spectral pattern [125]. Extensive research with analytes in foods by SERS have been conducted over the past several years [59]. Tip enhanced Raman scattering [126] and plasmon enhanced Raman   89 scattering [127] are also rapidly developing because of the favor in single molecule detection. However, intensive sample pre-treatment to extract analytes was required to reduce the sample matrix effect before the detection of trace levels of chemical contaminants in foods by SERS. Otherwise, the detector of the Raman instrument [e.g. charge-coupled device (CCD)] will record Raman signals of all the molecules in a sample.     We have developed novel MIPs–SERS sensors to determine alpha-tocopherol [175], melamine [176] and chloramphenicol [177] in foods. MIPs were packed into SPE cartridges to separate analytes from food samples with minimum sample pre-treatment. Silver dendrite was used as the SERS substrate to enhance Raman signal intensity improving sensitivity. Xue et al. [152] have integrated MIPs and SERS substrate to generate surface-imprinted core–shell Au nanoparticles which can selectively detect bisphenol A in water and drinks. Low-cost, rapid and reliable features of these sensors were developed.     Compared to the aforementioned column-based chromatographic methods, TLC requires less time for separation. TLC coupled with SERS has been explored for on-line, sensitive and cost-effective detection of chemicals in complex sample matrix [174], [178], [179] and [180]. Ultra-thin layer chromatography was developed by laser ablating silver onto a plate substrate to generate an ultra-thin film of silver that is capable of separating a mixture of a few dyes and enhancing the corresponding Raman signals [174]. Freye et al. [178] cured polydimethylsiloxane (PDMS) polymers in a mold as the solid phase extractor and physically deposited silver nanoparticles onto the top of PDMS to receive separated compounds transferred from a C18 silica gel TLC plate. SERS spectra of the separated compounds were recorded. Malachite green isothiocyanate, Rhodamine 6G, methylene violet, known as SERS active molecules [174] and [178], were used to test the performance of TLC–SERS, and the mixture of a few   90 standard amino acids [180] was also tested; in addition, detection of aromatic pollutants in water by TLC–SERS has been reported as well [181]. Real samples were investigated, but sample pre-treatment was necessarily complex because these samples contained significant interferences [179]. Although the TLC–SERS technique is sometimes a feasible approach to determine analytes in real samples rapidly, the sensitivity is low, and separation still takes a couple of minutes [178].     Therefore, the objective of our current study was to develop an MIPs–TLC–SERS sensor by using MIPs as the stationary phase of a TLC plate. To the best of our knowledge, this was the first study of an integrated MIPs–TLC–SERS sensor to determine trace levels of food chemical hazards. We confirmed that the pre-treatment of food samples was simplified and TLC development was extremely fast (30–40 s) due to the “lock and key” principle between MIPs and Sudan I. Gold colloid was used to enhance Raman signals of Sudan I and the enhancement factor was ~4×104. The lowest content of Sudan I in paprika powder that can be detected was 1 ppm. By using a portable Raman spectrometer, in-field and on-line detection of Sudan I contamination in paprika powder could be realized.  4.2 Experimental section  Reagents Sudan I, methacrylic acid (MAA), ethylene glycol dimethylacrylate (EGDMA), azobisisobutyronitrile (AIBN), chloroauric acid (HAuCl4), trisodium citrate dihydrate, and acetonitrile (all in HPLC grade) were purchased from Sigma Aldrich (St. Louis, MO). Methanol, hexane, chloroform (all in HPLC grade) and acetic acid were purchased from Fisher Scientific   91 (Waltham, MA). Paprika powders (2 brands: paprika 1 and paprika 2) were obtained from different local grocery stores in Vancouver, British Columbia, Canada. Deionized (DI) water at 18.2 MΩ/cm was prepared by the Millipore system (Billerica, MA). The stock solution of Sudan I at 1000 ppm in acetonitrile was prepared.  4.2.1 Synthesis and characterization of MIPs towards Sudan I  The template Sudan I (0.3 mmol), functional monomer MAA (1.2 mmol), cross-linking monomer EGDMA (8 mmol) were mixed with methanol (3 mL) and chloroform (0.5 mL) in a glass vial. The mixture was sonicated for 10 min. The initiator AIBN (20 mg) was added, and the whole mixture was purged by N2 for 5 min. The polymerization was thermally initiated at 55 °C for 20 h (Figure 4.1). The resultant polymer was ground and sieved by a 200 mesh. MIPs in powder form (< 75 µm) were washed with methanol and acetic acid (9:1, v/v) to remove template by monitoring UV-Vis signal of Sudan I at 483 nm. Another wash by methanol was carried out to remove acetic acid. MIPs were dried under vacuum for 4 h. Non-imprinted polymers (NIPs) were prepared in the same manner without the addition of Sudan I.   Figure 4.1: Schematic mechanism of developing molecularly imprinted polymers towards Sudan I.  EGDMA Sudan I MAA 55 °C AIBN, solvent       NNNOOCH3OHNHOOCH3H OOHH3COOOOOOOO      Sudan I removal   92 A rebinding assay was conducted to evaluate the selective binding ability of MIPs to Sudan I. Static studies were performed by re-exposing 20 mg MIPs powder (or NIPs powder) to 2 mL Sudan I solution in acetonitrile and water (1:1, v/v) at 22 °C for 3 h. The concentrations of Sudan I solutions used were 5, 10, 20, 30 and 45 ppm, respectively. Supernatants were collected and centrifuged (2348 ×g, 5 min). UV-Vis spectra (483 nm) were acquired to determine the concentrations of analyte in the supernatants. Kinetic studies were performed by suspending 100 mg MIPs powder (or NIPs powder) to 10 mL Sudan I solution at 10 ppm. At each time point (1, 5, 10, 20, 30, 60, 90 and 120 min), 1 mL mixture was collected and centrifuged (2348 ×g, 5 min), and the supernatant was measured for the unbound template molecules by UV spectrometry at 483 nm.  4.2.2 Fabrication of MIPs–TLC plate and developing   Pre-treatment of paprika samples  Sudan I stock solution (1000 ppm) was spiked into paprika powder (0.1 g). Different amounts of the stock solution (0, 2, 10, 20, 80,140 and 200 µL) were mixed with untreated powder, and acetonitrile was added accordingly to make the total volume of Sudan I solution in each sample to 200 µL. These artificially contaminated paprika samples were then mixed with 1.82 mL of acetonitrile to the spiking levels of 0, 1, 5, 10, 40, 70 and 100 ppm, respectively, followed by sonication for 10 min to extract Sudan I. The extracts were centrifuged (2348 ×g, 5 min) and filtered. Concentrations of Sudan I in the extracts were determined by HPLC (Agilent 1100 series, Germany) equipped with a quaternary pump and a diode array detector. Column conditions were described in Figure 4.2.   93  Figure 4.2: Representative HPLC chromatogram of Sudan I in 5 ppm spiked paprika extract. [Additional info: retention time: 5.665 min; column conditions: C18 column (µBondapak-C18, 10 µm, 3.9 mm × 300 mm, Waters, Milford, MA, USA); temperature: 25 °C; injection volume: 20 µL; gradient method: 0-6 min, acetonitrile : DI water =85 : 15, flow rate: 1 mL/min, 6-16 min, acetonitrile : DI water =95 : 5, flow rate: 1 mL/min, 16-20 min, acetonitrile, flow rate: 1.2 mL/min, 20-25 min, acetonitrile : DI water =95 : 5, flow rate: 1 mL/min.]  Fabrication of MIPs-TLC plate and developing  MIPs powder (50 mg) was suspended in 0.5 mL DI water and 0.1 mL ethanol thoroughly. The slurry was spread on a clean glass slide with dimension of 25 mm × 75 mm and dried at 22 °C for 4 h. The thickness of the sorbent was controlled between 0.35-0.40 mm. TLC plate developing was followed by the conventional way: spotting (2 µL paprika extract), developing and drying. Hexane and chloroform (1:1, v/v) were used as the optimal developing solvent.  4.2.3 Synthesis of gold colloid  Au colloid was synthesized following the protocol by Cheung and coauthors [62]. HAuCl4 (1 mM, 50 mL) was brought to boil first, and sequentially reduced by trisodium citrate (38.8 mM, 5   94 mL). The mixture was kept boiling for 30 min. Both pH and UV-Vis spectrum were recorded after the mixture was chilled down. Gold colloid with a wine red color was ready to be used.  4.2.4 SERS detection of Sudan I in paprika extract and data analysis  Au colloid (12 µL for bench-top Raman spectrometer or 20 µL for portable Raman spectrometer) was applied onto the original spot. The laser was introduced immediately onto the original spot and Raman spectra were acquired. Both bench-top Raman spectrometer (Renishaw, Gloucestershire, UK) and portable Raman spectrometer (custom made) were set up for spectral collection. A near infrared laser (785 nm, 25 mW) was equipped with the bench-top Raman spectrometer, coupled with 1200-groove/mm grating to disperse scattered photons and a 578 by 385 pixel CCD array detector to convert those photons to electronic form as read out. The Raman spectral scan was conducted by a 50× Nikon objective (NA = 0.75, WD = 0.37 mm) of an optical microscope with a movable sample stage in x-y-z dimension (Leica Biosystems, Germany). A spectrum range of either 350-1650 cm-1 (10 s exposure time) or 703-1262 cm-1 (1 s exposure time) was selected.  Raman spectra were also recorded using the portable Raman system with a backscattering probe, with dimensions of 127×165×51 mm and weight of 1.2 kg (Figure 4.3). The instrument uses a 785 nm single-mode diode laser source with a power of 100 mW for illumination. The backscattered light was collected with free-space design, and a Volume-Phase Holographic grating disperses this light over an image plane measured in CCD pixels as 64 height × 1024 width, providing 1024 pixel number spanning a Raman shift interval from 200-2200 cm-1. Each Raman spectrum was collected in 0.1 s and data points were acquired every 3 cm-1.   95                            Figure 4.3: Schematic illustration of SERS spectral collection by portable Raman spectrometer (top) and corresponding SERS spectral features of Sudan I in range of 350-1650 cm-1 (bottom). [From bottom to top: averaged SERS spectra (n = 3) of Sudan I spiked paprika extract: 0 ppm (1), 1 ppm (2), 5 ppm (3), 10 ppm (4), 40 ppm (5), 70 350 550 750 950 1150 1350 1550 Normalized Raman intensity Raman shift (cm-1) 9 7 6 5 4 3 2 1 8   96 ppm (6), 100 ppm (7); SERS spectra were collected by portable Raman spectrometer from the sample spots after MIPs-TLC developing. SERS spectrum of Sudan I spiked paprika extract from the sample spots after MIPs-TLC developing: 100 ppm (8) and normal Raman spectrum acquired from dried standard solution of 2 µL of 2000 ppm (9); spectra were collected by bench-top Raman spectrometer.]  WiRE v3.4 was programmed for spectral collection and initial processing (e.g. baseline correction). Further data analysis and construction of chemometric models were performed in Delight 3.2.1 (D-Squared Development Inc., LaGrande, OR, USA). The wavenumbers of 740-1190 cm-1 in the 703-1262 cm-1 dataset (1 s exposure time) were negated for model development, whereas, 710-725 cm-1 and 1210-1235 cm-1 were selected for chemometric analysis of the dataset of 350-1650 cm-1 (10 s exposure time). Unsupervised principal component analysis (PCA) was conducted to reveal the variations among samples without a priori knowledge. A partial least squares regression (PLSR) model was constructed to predict sample concentration for sets of paprika powder samples spiked with different concentrations of Sudan I [128]. PLSR models were tested using leave-one-out cross-validation and 5 latent variables were selected. Root-mean-square error of cross-validation (RMSECV), root-mean-square error of prediction (RMSEP) and predictive squared correlation coefficient (Q2) were calculated based on the training dataset from paprika 1 and the testing dataset from paprika 2, to assess the predictive capability of the developed PLSR models [182]. Equations are shown as below:  RMSECV = !!!!! !!! !  (4.1)  RMSEP = !!!!! !!! !  (4.2)   97  Q2 = 1 − !!!!! !!! !!!!! !!!  (4.3) in which, n is the number of samples (i.e. spectra), 𝑐! is the real concentration of sample i, and 𝑐! is the predicted concentration by models, while 𝑐! is the arithmetic mean of all the testing dataset.  4.3 Results and discussion  4.3.1 Assessment of MIPs–TLC plate and developing   Properties of MIPs  Rebinding assays were carried out to confirm the selectivity of MIPs towards Sudan I by comparing the performance of MIPs with control polymers (i.e. NIPs). A few parameters, such as imprinting factor (IF), adsorption capacity (Q) and the Scatchard plot model, were calculated and derived from the created isotherm based on the static studies [183]. Calculations of Q (Equation 4.4), IF (Equation 4.5) and formula of the Scatchard plot (Equation 4.6) are shown as below:  Q = (Ci – Cf) × V/W (4.4)    IF = QMIPs/QNIPs (4.5)  Q/Cf = -Q/Kd + Qmax/Kd (4.6) in which, Q represents the adsorption capacity of MIPs or NIPs, Ci is the initial concentration of Sudan I before addition of polymer, Cf is the final concentration of Sudan I (in the supernatant) after rebinding, V is the volume of Sudan I solution (2 mL), and W represents the quantity of   98 polymer (20 mg). In Equation 4.6, Kd is the dissociation constant and Qmax represents the saturated adsorption capacity. Along with the increase in initial concentration of Sudan I, QMIPs increased almost linearly compared to QNIPs, which expressed limitation at high concentrations (Figure 4.4A). Non-specific binding is the speculated factor to explain the similar values of QNIPs to QMIPs when the concentration of Sudan I was lower than 20 ppm. In comparison, differences between QNIPs and QMIPs in higher concentrations of Sudan I were significant (P = 0.01) and this may be due to the imprinted binding sites that are complementary to Sudan I on MIPs rather than NIPs. IF is usually used to evaluate the selectivity of MIPs towards the template. It varies depending on the concentrations of analyte. At 45 ppm of Ci, IF was 1.46 (Figure 4.4A). We constructed the Scatchard plot of MIPs with a convex shape (Figure 4.4B), in contrast, no similar plot pattern could be constructed using the NIPs dataset. According to the Scatchard plot theory [184], non-linear regression indicates the non-identical and dependent binding sites; in other words, binding at one site will affect the binding at another site, thus the plots indicate the different properties of MIPs and NIPs. Selective binding sites on MIPs have been discussed with the conclusion that they can be complementary to template in size, shape, and interactions [164] (e.g. covalent, H-bonding, hydrophobic). In the current study, H-bonding and size exclusion are the two major factors (Figure 4.1). Thus, both physical and chemical elements contribute to selective binding. Taken together, binding sites on MIPs were selective towards Sudan I compared to NIPs.   99   Figure 4.4: (A) Isotherm of rebinding assays of molecularly imprinted polymers (MIPs) and non-imprinted polymers (NIPs). (B) Scatchard plot of MIPs. (Additional info: initial concentration: 5, 10, 20, 30, 45 ppm; time 3 h; values are shown as mean ± standard error of the mean, n = 3.)  QMIPs and QNIPs at different time points were plotted against the initial Sudan I concentration of 10 ppm for the kinetic studies. Instant binding and saturation were achieved within 5 min for control polymer (i.e. NIPs), while QMIPs gradually increased for the following 115 min although most binding was done for the first 5 min (Figure 4.5). Different Q values of MIPs and NIPs after saturation and the time to reach saturation further validated the existence of complementary binding sites on MIPs. Fast binding (86% in 5 min) was a good indicator for further TLC sample loading, in which potential loss could be limited. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 5 10 15 20 25 30 35 40 45 50 Adsorption capacity (mg/g) Initial concontration of Sudan I (ppm) MIPs NIPs 0 0.005 0.01 0.015 0.02 0.025 0 0.2 0.4 0.6 0.8 1 Q/Ci (mL/mg) Q (mg/g) MIPs NIPs A B   100  Figure 4.5: Kinetic plots of MIPs and NIPs. (Additional info: initial concentration: 10 ppm; time intervals: 1, 5, 10, 20, 30, 60, 90, 120 min; values are shown as mean ± standard error of the mean, n = 3.)  Pre-treatment of food samples  Minimum sample treatment (i.e. SLE, 10 min by sonicating) was performed to extract Sudan I from paprika powder samples. Unavoidable interferents were also extracted because of their solubility in acetonitrile. Recoveries were quantified by HPLC and are summarized in Table 4.1.  Table 4.1: Sudan I recoveries of extraction from paprika powder determined by HPLC (n = 3, value shown: mean ± standard error of the mean). 	 Spiking level (ppm) Recoveries (%) 100 70 40 10 5 1 0 Paprika 1 94.7 ± 0.5 94.7 ± 0.5 95.7 ± 1.4 91.8 ± 1.8 90.5 ± 0.5 84.8 ± 2.5 N/A Paprika 2 95.5 ± 1.4 95.1 ± 1.6 94.9 ±1.3 87.4 ± 3.4 92.6 ± 0.9 85.7 ± 0.6 N/A    0 0.04 0.08 0.12 0.16 0.2 0 30 60 90 120 150 Adsorption capacity (mg/g) Time (min) MIPs NIPs   101 Synthesis of Au colloid  The pH of Au colloid was 6.5 and a UV-Vis absorbance spectrum of Au colloid was shown in Figure 4.6, in which the wavelength of maximum absorption was 525 nm. These parameters could be referred to for further development of Au colloid in the same manner.  Figure 4.6: Absorbance spectrum of gold colloid.  Fabrication of MIPs-TLC plate and developing  The MIPs thickness and stationary phase area were optimized, given the fact that the gravity of sorbent may overcome friction between MIPs and glass slide, causing MIPs to fall off the plate during developing. Sample spotted within a small area was also desirable for further SERS detection. The amount of MIPs was optimized by using 50 mg and the thickness was controlled between 0.35-0.40 mm. A small portion of ethanol was added into the slurry for a better disperse of MIPs into liquid and avoiding generation of air bubbles. MIPs-TLC plate was ready to be used after it was completely dried under ambient temperature. Paprika extract (2 µL) was deposited to form a spot of ~3 mm in diameter and allowed to dry. For the developing, a mixture of hexane   102 and chloroform was selected without destroying the interaction between Sudan I and MIPs, but washing interferents away. The same MIPs-TLC plate at different developing stages is shown in Figure 4.7. Spots on the plate (from left to right) indicated 0 ppm spiked extract, 100 ppm Sudan I standard solution in acetonitrile and 100 ppm spiked extract, respectively. Spiked extract showed more intensive color than the standard solution of Sudan I, indicating the massive components (as interferents) in the paprika samples. After developing, pale but visible red spots were left at the original spots but not 0 ppm spiked extract, indicating that Sudan I was successfully captured by MIPs.    Figure 4.7: Illustration of working mechanism of MIPs-TLC-SERS sensor. (A: plate before developing, B: plate after developing, C: plated after addition of Au colloid; spot 1: 0 ppm Sudan I spike paprika extract, spot 2: 100 ppm Sudan I standard solution, spot 3: 100 ppm Sudan I spiked paprika extract.)     103 Due to the selectivity of MIPs towards Sudan I, the retardation factor is not a concern because Sudan I in the extract was fixed at the original sample spot, whereas interferents moved upward with the developing solvent. MIPs-TLC developing performance further validated the selective “lock and key” principle of MIPs towards Sudan I. In this case, the developing time was shortened to be 30-40 s depending on the distance between sample spot and slide edge and less solvent was used.   4.3.2 SERS spectral acquisition and data analysis  After TLC developing, Au colloid was applied onto the original sample spot on the MIPs-TLC plate. A preliminary experiment was conducted using different amounts of Au colloid for deposition and it was found that 12 µL of Au colloid could achieve a good enhancement of Sudan I Raman signal. Figure 4.8 showed the SERS spectra of standard Sudan I and paprika extract spiked at different concentrations of Sudan I (after TLC developing). Bands in the 350-1650 cm-1 range could be seen in SERS spectra at 100 ppm (9) compared to the normal Raman spectrum of Sudan I (1) and the SERS spectra of Sudan I obtained by Cheung and coauthors [62], indicating the SERS spectra we acquired are indeed from Sudan I. Bands at 461, 581 and 721 cm-1 were significantly enhanced. Some band shifts existed, which may be attributed to the affinity between covalent bonds and Au colloid particles and physisorption strain [69]. In addition to the bands assigned to Sudan I (Table 4.2), only limited interference bands appeared. Even at 1 ppm, bands at 461, 536, 581, 721, 980, 990,1224, 1377, 1482 and 1577 cm-1 (featured bands of Sudan I) were still distinct. In terms of SERS spectra from 0 ppm spiked extract, no featured bands can be observed. Thus, good separation efficiency of MIPs-TLC plate was validated.   104    Figure 4.8: Average SERS spectra of Sudan I (collected in 49 s) in the wavenumber range of 350-1650 cm-1. [From bottom to top: normal Raman spectrum acquired from dried standard solution of 2 µL of 2000 ppm on clean glass slide (1), normal Raman spectrum from dried 2 µL of 1000 ppm of Sudan I standard solution on MIPs-TLC plate (2), SERS spectrum of 0 ppm (3) spiked paprika extract, 1 ppm (4) standard solution, 1 ppm (5) spiked paprika extract, 10 ppm (6) standard solution, 10 ppm (7) spiked paprika extract, 100 ppm (8) standard solution, 100 ppm (9) spiked paprika extract; SERS spectra were collected from the sample spots after MIPs-TLC developing; n = 8 for each concentration.]   350 550 750 950 1150 1350 1550 Normalized Raman Intensity Raman Shift (cm-1) 9 7 6 5 4 3 2 1 8   105  Table 4.2: Peak assignments of Sudan I in normal Raman spectrum. Raman shift (cm-1) Assignments 1595 C-C of benzene ring scissoring; N=N stretching 1547 C-H of naphthalene ring in plane bending; N-H (equilibrium form) in plane bending; C-C of naphthalene ring stretching 1494 C-H of benzene ring in plane bending; N=N stretching; C-C of benzene ring stretching 1449 C-H of benzene ring in plane bending; N=N stretching 1386 C-H of benzene ring and naphthalene ring in plane bending; C=N (equilibrium form) out-of-plane bending 1337 C-H of naphthalene ring in plane bending; C-C of benzene ring in plane bending 1254 benzene ring ring breathing; C-C-H of naphthalene ring deformation vibration; C-N stretching 1226 C-C-H of naphthalene ring scissoring 1206 C-H of ring scissoring; C-N=N scissoring; N-N-H (equilibrium form) scissoring 1162 C-H of benzene ring scissoring; C-N stretching 1092 C-H of benzene ring scissoring 1072 C-H and C-C-C of naphthalene ring scissoring 998 C-H and C-C-C of benzene ring scissoring 983 C-H of ring twisting 752 ring breathing; C-N out of plane bending 722 naphthalene ring deformation vibration; benzene ring ring breathing  614 benzene ring deformation vibration 583 ring deformation vibration; C-N=N-C torsion 542 ring breathing; C-N=N-C deformation vibration 459 C=O-C=N (equilibrium form) torsion   However, the developed MIPs-TLC plate had to be used for spectral collection immediately after the addition of Au colloid. SERS band intensity could not be maintained once the Au colloid was dried. It took 49 s for the collection of each SERS spectrum in the range of 350-1650 cm-1. Collection of SERS spectra in the next cycle (another 49 s) or more resulted in a reduction of signal intensity. Time profiling was recorded with a laser illuminated at a single spot on the plate (Figure 4.9). Relative height (or peak area) of certain bands (i.e. 461, 721 and 1224 cm-1) was reproducible, but not all bands within the first 49 s. Moreover, the reduction of band   106 intensity was random, indicating the stringent time factor in terms of reliability of this MIPs-TLC-SERS sensor. Aggregation of Au colloid particles during drying may account for the weaker SERS signal as the morphology of nanoparticles changed, resulting in the change of localized surface plasmon resonance; thus, the enhancing ability of Au nanoparticles compromised accordingly [180].    Figure 4.9: Time profile of Sudan I SERS spectra of 100 ppm spiked paprika extract. (Additional info: from top to bottom: spot 1: acquisition 1: 0-49 s, acquisition 2: 50-98 s, acquisition 3: 99-147 s; spot 2: acquisition 1: 0-49 s, acquisition 2: 50-98 s, acquisition 3: 99-147 s. SERS spectra were collected from the sample spots after MIPs-TLC developing.)  Therefore, the ideal circumstance is to keep the Sudan I spot on MIPs in a dampened state. However, laser illumination generated heat and facilitated the drying of the Au colloid. Faster spectral collection is thus the strategy to eliminate random reduction of SERS signal intensity.   107 Compared to previous spectral collection in 49 s, the exposure time was shortened to 1 s with a narrow scanning wavenumbers (703-1262 cm-1) due to the limitation of the bench-top Raman instrument. A few featured bands (i.e. 721, 980, 990, and 1224 cm-1) could still be recorded in 50 s after the addition of Au colloid (Figure 4.10).   Figure 4.10: Averaged SERS spectra of Sudan I (collected in 1 s) in the wavenumber range of 703-1262 cm-1. [Sudan I spiked paprika extract: 0 ppm (7), 1 ppm (6), 5 ppm (5), 10 ppm (4), 40 ppm (3), 70 ppm (2), 100 ppm (1); SERS spectra were collected from the sample spots after MIPs-TLC developing; n = 8 for each concentration.]  1 5 4 3 2 7 6   108 Both univariate and multivariate analysis were conducted towards datasets of either shorter- or longer-wavenumber range (703-1262 cm-1 or 350-1650 cm-1) of Sudan I SERS spectra, aiming at testing the reproducibility and reliability of this MIPs-TLC-SERS biosensor. PCA plots were constructed to visualize spectral variations and leave-one-out cross-validation was employed to assess the performance of PLSR models. We also performed the linear regression towards the spectra with the shorter wavenumber range (703-1262 cm-1) as a comparison. In Figure 4.11A, the PCA plot of spectra with the longer wavenumber range (350-1650 cm-1) was capable of differentiating high concentrations of spiked Sudan I, and the lowest concentration that could be clearly separated was 5 ppm. However, the coefficient of determination (R2 = 0.788) of the PLSR model (Figure 4.12A) was not satisfactory. As we have discussed, the signal intensity of SERS spectra with 49 s collection time was strongly time dependent, indicating its inappropriate usage to quantify Sudan I. The PCA plot of spectra with 1 s collection time (Figure 4.11B) showed clear segregation clusters among most concentrations, with minor overlapping between the clusters of 1 ppm and 5 ppm. Corresponding PLSR model (Figure 4.12B) also showed good linear regression (R2 = 0.978).         109  Figure 4.11: Illustration of principal component analysis (PCA) models: (A) spectra in the wavenumber range of 350-1650 cm-1 (PC1: 57.4%, PC2: 34%; n =4 at each concentration); (B) spectra in the wavenumber range of 703-1262 cm-1 (PC1: 50%, PC2: 23.9%; n = 8 at each concentration). [Sudan I spiked paprika extract: 0 ppm (black), 1 ppm (red), 5 ppm (blue), 10 ppm (olive), 40 ppm (pink), 70 ppm (green), 100 ppm (purple); SERS spectra were collected from the sample spots after MIPs-TLC developing.] (n=5)  	 	 	 	 	 	 	 (B) 	 	 	 	 	 	 	 (A)   110  Figure 4.12: Illustration of regression models: (A) established PLSR regression based on spectra in range of 350-1650 cm-1; (B) established PLSR regression based on spectra in range of 703-1262 cm-1; (C) established linear regression based on band height at 721 cm-1 of spectra in range of 703-1262 cm-1 (values shown in: mean ± standard error of the mean, n = 8 at each concentration.). [Sudan I spiked paprika extract: 0 ppm (black), 1 ppm (red), 5 ppm (blue), 10 ppm (olive), 40 ppm (pink), 70 ppm (green), 100 ppm (purple); SERS spectra were collected from the sample spots after MIPs-TLC developing.] R2 = 0.788 (A) (B) R2 = 0.978 RMSECV = 7.695 RMSEP = 11.437 Q2 = 0.902   111 Due to the challenge in separating 1 ppm and 5 ppm spiked Sudan I samples using PCA plot, we next created a linear regression model because we observed the trend of increasing peak intensity along with the increase in concentrations (spiking levels). The featured SERS band at 721 cm-1 was selected and a good linear correlation (R2 = 0.983) was plotted between band intensity and corresponding concentration (spiking level) (Figure 4.12C). The prediction was conducted by incorporating one SERS spectrum of Sudan I at each spiked level in paprika 2 extract (i.e. the testing dataset) into the established models (Figure 4.12B & C). All the models established on the basis of 1 s collection time could validate the sensitivity and reliability of this sensor. In terms of LOD, 1 ppm (or 2 ng/spot) was defined. The diameter of laser spot of bench-top Raman spectrometer on the MIPs-TLC plate is ~10 µm, whereas sample spot on plate showed 3 mm in diameter with 0.35-0.40 mm thickness. An even lower LOD could be achieved with a better manipulation of sample spot size and plate thickness.  We also performed silica gel TLC-SERS to determine Sudan I in paprika extract and compared it with the performance of MIPs-TLC-SERS. Standard solution of Sudan I (100 ppm) and 100 ppm Sudan I spiked paprika extract were used and sample spotting amount was 2 µL. However, it took 3.5 min for the separation (developing solvent: hexane/chloroform/acetic acid, 1:1:0.02 v/v) and no SERS signal could be recorded on the TLC plate whether it was wet or dry. This may be due to the inappropriate SERS substrate such as the size and morphology of nanoparticles or different types of colloids that are employed [185]. Parameters such as separation time and LOD of these two approaches can be found in Table 4.3 and Figure 4.13.      112  Figure 4.13: Silica gel TLC plate developing. [Additional info: left: after spotting, right: after developing; on each plate, left: 2 µL of 100 ppm Sudan I standard solution, right: 2 µL of Sudan I (100 ppm) spiked paprika extract; distances of Sudan I spots to bottom edge were shown in cm.]  Table 4.3: Comparison of silica gel TLC-SERS and MIPs-TLC-SERS with Sudan I standard solution (100 ppm, 2 µL) and Sudan I (100 ppm) spiked paprika extract (2 µL).  Separation Developing solvent  Separation time LOD Silica gel TLC-SERS Good Hexane/chloroform 1:1, v/v  3.5 min  N/A MIPs-TLC-SERS Good Hexane/chloroform/acetic acid 1:1:0.02 v/v 30-40 s  1 ppm   4.3.3 Portable Raman analysis  More Au colloid (20 µL) was applied to sample spot on the MIPs-TLC plate given the fact that laser power with 100 mW from portable Raman spectrometer can generate much higher thermal energy, resulting in challenges in spectral collection. Spectra collected from portable Raman spectrometer were averaged (n = 3 at each concentration) and stacked in Figure 4.3. Less resolution was founded due to the inherent properties of a portable instrument that tend to reduce     4  1.9  0.5    113 sensitivity, but SERS spectra were still sufficient to assay the presence of Sudan I. Therefore, in-field or on-line screening of Sudan I can be conducted with a portable Raman spectrometer at least for screening purposes with further quantitative analysis performed by a bench-top Raman spectrometer.  4.4 Conclusion We developed an innovative approach of MIPs-TLC-SERS to determine Sudan I in paprika powder. The MIPs-TLC plate could separate Sudan I from paprika extract within 30-40 s with minimum sample pre-treatment required. Spectral collection by bench-top Raman spectrometer was finished in 1 s by adding Au colloid (as a SERS substrate) to a separated Sudan I spot on a TLC plate, aiming to obtain enhanced Raman spectra. The enhancement factor was ~4 × 104, based on the calculation of peak height at 721 cm-1 after baseline correction of Sudan I spectra collected from the MIPs-TLC plate. Chemometric and regression models showed that the detection of Sudan I was accurate and sensitive [LOD of 1 ppm (or 2 ng/spot)] using either bench-top or portable Raman spectrometers. This novel MIPs-TLC-SERS sensor had fast, reliable and low cost features for in-field and on-line screening of food chemical contamination and adulteration.      114  Determination of histamine in canned tuna by molecularly Chapter 5:imprinted polymers and surface enhanced Raman spectroscopy  Chapter 5 is based on the publication as: Gao F., Grant E., Lu X. (2015), Determination of histamine in canned tuna by molecularly imprinted polymers-surface enhanced Raman spectroscopy. Analytica Chimica Acta, 901: 68-75. Chapter 4 introduced MIPs-TLC method to significantly facilitate separation. However, MIPs particles were not tightly sticking to the plate, which lead to concerns with moving this plate back and forth. In Chapter 5, MIPs were immobilized onto a PVC film to provide stable and flexible MIPs-PVC film and therefore adsorb histamine in tuna extract. In addition, gold colloid served as eluting solvent and SERS active substrate to desorb histamine off MIPs-PVC film during incubation of gold colloid and MIPs-PVC film and later enhance Raman signals of histamine.   5.1 Introduction Histamine is a biogenic amine that acts to transmit signals from cell to cell in the skin, gut, and other organs of the immune system. Structural differences in receptors on cell membranes account for different histamine responses among individuals [186]. The interaction between histamine and H1 receptors can cause a drop in blood pressure and muscle contraction. H2 receptors to histamine are associated with the secretion of acid by the stomach [187]. The ingestion of histamine-rich foods can trigger histamine toxicity manifested as nausea, headache, diarrhea, and asthma. Reactions such as these are known as histamine poisoning or scombroid poisoning [188]. Among foods containing histamine, tuna is one of the most extensively   115 consumed. To prevent the potential risk provoked by histamine in foods, the European Community has directed that the average concentration of histamine in fish must be lower than 10 mg/100 g. For canned tuna, the limit of histamine level is normally 3 mg/100 g [189]. Studies have established the principal mechanism of histamine formation in tuna. Tuna contains a large amount of free histidine, and the enzyme, histidine decarboxylase, efficiently converts histidine to histamine. The common histamine-producing bacteria, Morganella morganii, Klebsiella pneumoniae, Hafnia alvei, Citrobacter freundii, and Escherichia coli all possess histidine decarboxylase [190]. High temperature supports the rapid growth of these histamine-forming bacteria [189, 191]. When the temperature is above 20 °C, histamine rapidly forms in several tuna species (e.g. bigeye, skipjack and yellowfin) and reaches an extremely high level within one or two days, accompanied by appearance changes in tuna meat (e.g. muscle color, texture and odor) [189, 191]. Guizani et al. reported that yellowfin could be stored for 17 days at 0 °C without exceeding the safety level of histamine [191]. In another study, the amount of histamine produced in bigeye reached the toxic level after six days of storage at 4 °C [189]. Accordingly, tuna fish must be chilled immediately after harvest to prevent histamine formation. However, low temperature alone is not sufficient to completely suppress bacterial growth and the continuing accumulation of histamine. Once formed, histamine in tuna meat is not destroyed by freezing, canning or cooking [191, 192]. Thus, assured food safety demands continuous monitoring of histamine levels in tuna. Conventional methods to determine the histamine level in tuna include high-performance liquid chromatography (HPLC), fluorometry, and enzymatic detection [e.g. enzyme-linked immunosorbent assay (ELISA)] [193-196]. HPLC and fluorometry entail time consuming protocols to derivatize histamine by o-phthalaldehyde or dansyl chloride, [194-196]. The   116 fluorometric assay requires methanol extraction, anion exchange column purification, and derivatization as pre-treatment. Because of the structural similarity of histidine to histamine, fluorescence measurement has a poor selectivity for histamine [197], even when aided by HPLC separation [198]. In contrast, enzyme based methods offer a rapid means of detection, but necessitate the use of unstable enzymes and expensive test kits, and tend to overestimate histamine [197]. In addition to the aforementioned conventional methods, recent studies have identified several other effective analytical approaches. Cohen et al. reported a good correlation between HPLC detection and rapid, simple and inexpensive ion mobility spectrometry for the quantification of histamine in tuna [199]. Previous work has explored molecularly imprinted polymers (MIPs), testing their affinity and selectivity for histamine [200-203]. Horemans et al. have employed MIPs as the separation element of sensors relying on electrochemical impedance spectroscopy using a quartz crystal microbalance (QCM) to analyze histamine in aqueous media [204]. Another MIPs-QCM sensor was used to determine histamine in canned fish [203], yielding a limit of detection in the nano-molar range. In these studies and others, MIPs function as a “plastic antibody” featuring thermal and chemical stability and low cost. During the synthesis of MIPs, functional monomers imprint for an interaction with analyte molecules (e.g. covalent interaction, H-bonding, electrostatic and hydrophobic interaction), while cross-linking monomers contribute to the rigid structure of MIPs [205]. Surface enhanced Raman spectroscopy (SERS) has attracted a great deal of attention owing to its high sensitivity [206]. A SERS substrate (i.e. noble metal nanoparticles or nanoscale roughened surface) amplifies the intensity of normal Raman signal to a remarkable degree. Enhancement arises from both electromagnetic and chemical (charge transfer) effects. Photo-  117 excitation of a noble metal in a nanostructured format generates a localized surface plasmon resonance, increasing the field that inelastically scatters from the sample [125]. Raman enhancement factors can surpass 1014 times [57, 125, 207]. The SERS technique has been applied in the detection of trace components in foods [59], but the food matrix contributes interfering features to the SERS spectrum, which presents a major challenge to its analytical utility. For this reason, several studies have developed an integrated MIPs-SERS approach to simultaneously separate and detect chemical hazards in environmental and food samples. For example, Xue et al. have developed MIPs-Au nanoparticles in a core-shell structure to detect bisphenol A in water and soft drinks [152]. In addition, MIPs have been tested as components of solid phase extraction (SPE) cartridges [66, 67, 208] or thin layer chromatography (TLC) plates [209] to separate chemical hazards in foods, followed by SERS detection of the eluent. The current study introduces a novel approach, combining MIPs with SERS to detect histamine in canned tuna. Dissolved polyvinyl chloride (PVC) serves to distribute MIPs fine particles and produce a MIPs-PVC film. Gold colloid suspended in the eluting solvent is used to desorb histamine off the MIPs and provides SERS enhancement. Methacrylic acid acts as the functional monomer for MIPs synthesis.  Cations, such as sodium ion in gold colloid aid elution by disrupting the interaction between the MIPs carboxylic group and histamine (Figure 5.1). We have confirmed that the SERS signal obtained by this approach rapidly and accurately quantifies histamine in canned tuna.     118  Figure 5.1: Illustration of working mechanism of MIPs-PVC-SERS.  5.2 Experimental section  Reagents and materials Histamine (HA), methacrylic acid (MAA), ethylene glycol dimethylacrylate (EGDMA), azobisisobutyronitrile (AIBN), PVC, chloroauric acid (HAuCl4), trisodium citrate dihydrate, and acetonitrile in HPLC grade were purchased from Sigma Aldrich (St. Louis, MO). L-histidine, hexane, acetic acid, and ammonium acetate (all in HPLC grade) were obtained from Fisher Scientific (Waltham, MA). Tetrahydrofuran (THF) for HPLC was purchased from VWR (Wayne, PA) and ethanol was purchased from Commercial Alcohols (Toronto, ON). Canned tuna fish products were obtained from a local grocery store in Vancouver BC. Deionized (DI) water (18.2 MΩ/cm) was prepared by the Millipore system (Billerica, MA).  Film & tuna extract MIPs & PVC Incubation MIPs-PVC film Drying Peeling off ethanol Washed by Incubation Film & AuNPs 50	×   119 5.2.1 Synthesis of MIPs, fabrication and characterization of MIPs-PVC film  Histamine (0.13 mmol), MAA (0.65 mmol), and EGDMA (3.28 mmol) were dissolved in 20 mL acetonitrile, and AIBN (20 mg) was then introduced. The mixture was incubated at 55 °C for 24 h to achieve polymerization. Afterward, the resulting polymer particles were washed with ethanol and acetic acid (9:1, v/v), followed by pure ethanol to remove any residual histamine captured by polymer particles (Figure 5.2). The polymer particles were vacuum dried to yield MIPs. The control polymers (i.e. non-imprinted polymers, NIPs) were prepared in the same manner without the addition of template molecule (i.e. histamine).  Figure 5.2: Schematic illustration of polymerization of MIPs towards histamine.  PVC (100 mg) was dissolved in 20 mL THF. A 0.5 mL aliquot of PVC solution was added to 5 mg MIPs powder (or NIPs powder) in a 25 mL beaker. After drying the mixture at room temperature to form a film, we peeled off the developed film and stored it in ethanol. A 1:2 (w/w) ratio of PVC to MIPs was judged to yield a film of optimum flexibility and integrity. Histamine was dissolved in 30 mM ammonium acetate, to form standard solutions with different concentrations (10, 30, 50 and 100 ppm). The dried film was introduced into 1 mL histamine solution at 100 ppm. A set of 5 samples was incubated at 25 °C for different time AIBN, ACN NNHNH55 ºC NNHNHOOHOOHOOOOOOOOHAMAAEGDMA2 Histamine removal   120 points (5, 15, 30, 60 and 120 min). The remaining solution was collected and quantified by HPLC-UV at 210 nm and the kinetics of the adsorption was investigated. For the study of static adsorption of MIPs-PVC (or NIPs-PVC) film towards histamine, the films and histamine solutions (10, 30, 50 and 100 ppm) were incubated at 25 °C for 1 h and then the remaining solutions were quantified by HPLC-UV. The selectivity test was performed by comparing the adsorption of MIPs-PVC (or NIPs-PVC) film towards histamine and L- histidine at 100 ppm after 1 h incubation. Each test for adsorption was repeated 3 times. An HPLC system (Agilent 1100 series, Germany) coupled with quaternary pump and a diode array detector was used to determine the concentration of histamine, with the use of a HILIC column (Luna@HILIC, 3 µm, 100 × 2 mm, Phenomenex, Torrance, CA, USA). The flow rate was 0.4 mL/min and mobile phase was a mixture of ammonium acetate (5 mM) and acetonitrile (1:9, v/v).  5.2.2 MIPs-PVC film: adsorption & desorption of histamine in tuna  Preparation of tuna samples We processed canned tuna meat for analysis immediately after opening the can. The meat was transferred to a blender and then completely homogenized at a high speed. The homogenized meat (2 g) was transferred to a falcon tube. All samples were subsequently spiked with histamine to different levels and vortexed for 30 s to mix completely. Ethanol (4 mL) was then added to the mixture, followed by another 30 s vortex for extraction. The sample was centrifuged at 10,000 ×g for 2 min and filtered to collect the extract. The clear extract was defatted by 8 mL of hexane via vortex (30 s) and centrifugation (10,000 ×g, 2 min). The hexane layer was discarded and the   121 bottom layer was taken for analysis. The final concentrations of histamine found naturally in tuna extracts after addition of ethanol ranged from 0 to 30 ppm, while the spiking levels of histamine in tuna meat were set to 0, 3, 30, and 90 ppm. Three independent samples were prepared at each concentration.   MIPs-PVC film: adsorption & desorption of histamine in tuna For the adsorption of histamine in tuna, we placed one piece of MIPs-PVC film into tuna extract (1 mL) and incubated it at 25 °C for 5 min. The film was removed and dipped into ethanol to remove components on the film that were not firmly attached. Once it was dried, a AuNPs colloidal solution (1 mL) was used to desorb histamine on the MIPs-PVC film by mixing for 2.5 min.   5.2.3 Preparation of gold colloid   Gold nanoparticles (AuNPs) were prepared following the Lee-Meisel method with modification [210, 211]. After boiling, HAuCl4 (1mM, 50 mL) was reduced by sodium citrate (38.8 mM, 5 mL). Boiling was continued for another 30 min and a wine red colloidal solution was obtained.  5.2.4 Determination of histamine by SERS and data analysis   A Renishaw Raman spectrometer (Gloucestershire, UK) equipped with a near IR excitation laser (785 nm, 25 mW), 1200 groove/mm grating, and a 578 × 385 pixel CCD array detector was used. The inelastic Raman scattered photons were generated, dispersed, and finally converted to readable signals. An optical microscope (Leica Biosystems, Germany) coupled with a 50× objective (NA = 0.75, WD = 0.37 mm) and a movable stage in the x-y-z dimension was   122 employed for sample scanning. The MIPs-PVC treated gold colloidal solution (0.5 µL) was deposited onto clean aluminum foil and dried. Raman spectral collection within the wavenumber region of 1082-1603 cm-1 (in 1 s or 10 s exposure time) was conducted. All the collected spectra were first baseline corrected to remove fluorescence background by Vancouver Raman Algorithm Software [78]. Principal component analysis (PCA) and partial least square regression (PLSR) were performed to investigate the variations in spectra collected from tuna samples with different spiking levels of histamine. The entire Raman shift range (1082-1603 cm-1) was selected to construct the PCA plot and PLSR model. Both PCA and PLSR were carried out using Matlab R2014a with the PLS toolbox (Eigenvector Research Inc., Manson, WA). For each spiking level, we chose a set of 18 spectra to serve as a training dataset (72 spectra in total) and selected another 6 spectra to form the testing dataset (24 spectra in total). To build the calibration model, we randomly split the training set spectra into 8 groups and performed leave-one group-out cross-validation. Three latent variables were selected according to root-mean-square error of cross-validation (RMSECV) and root-mean-square error of prediction (RMSEP) values [62]. The testing dataset was applied to the PCA plot and PLSR regression model to evaluate the performance. The predictive squared correlation coefficient (Q2) was calculated to assess PLSR model with respect to its predictive capability [62]. All of the related equations are shown as below:  RMSECV = !!!!! !!! !  (5.1)  RMSEP = !!!!! !!! !  (5.2)  Q2 = 1 − !!!!! !!! !!!!! !!!  (5.3)   123 in which, n denotes the number of spectra, 𝑐!  represents the concentration of histamine created in the sample by spiking (shown as the x axis in the PLSR plot), and 𝑐! is the predicted spiking value derived from spectral features (shown as the y axis in the PLSR plot). We determine 𝑐! as the arithmetic mean of the 24 𝑐! values in the test dataset.  5.3 Results and discussion  5.3.1 MIPs-PVC film: adsorption & desorption of histamine in tuna  Figure 5.1 gives a schematic illustration of the experimental design and working mechanism of the MIPs-PVC-SERS analytical approach.  Characterization of MIPs-PVC film Experiments characterized the analytical effectiveness of MIPs-PVC films by comparing their performance with NIPs-PVC films. Equations (4) and (5) define the adsorption capacity (Q) and imprinting factor (IF) [209]:  Q = (Ci – Cf) × V/W (5.4)  IF = QMIPs/QNIPs (5.5) in which, Ci and Cf refer to the initial and final concentration of the histamine standard solution. V represents the volume of the histamine solution (1 mL), and W is the weight of MIPs (or NIPs) used to fabricate the film (5 mg). Figure 5.3A shows that both QMIPs and QNIPs increase with incubation time, reaching adsorption equilibrium after 1 hour. Note the significant difference between the values of QMIPs and QNIPs. This adsorption test validates the superior performance of MIPs-PVC versus NIPs-  124 PVC with respect to histamine binding. We set the incubation time to one hour based on our adsorption equilibrium study. For this static adsorption test, we noticed that QNIPs increased slightly for higher levels of histamine, as measured by Ci . QMIPs increased significantly more than QNIPs (Figure 5.3B). Equation 5.5 yields an IF of 2.81 for QMIPs and QNIPs obtained at a Ci of 100 ppm. IF reflects the imprinting effect of the target substrate, histamine, on the polymers. Larger values of IF indicate a greater capacity of the MIP to retain a complementary binding site for the analyte. That NIPs adsorb histamine at all can be explained by non-specific interactions. The carboxylic group on the MIPs particles readily attracts the biogenic amine histamine. This reflects the fact that MIPs contain non-specific binding sites as well as molecularly imprinted ones, but the number of non-specific binding sites is limited. Therefore, we find much higher QMIPs compared with QNIPs.   Figure 5.3: (A) Kinetic plots of MIPs-PVC and NIPs-PVC towards histamine at 100 ppm. (B) Isotherm of 1 h static study on MIPs-PVC and NIPs-PVC to histamine. (Values are shown as mean ± standard error of the mean, n = 3.)   0 0.4 0.8 1.2 1.6 2 0 20 40 60 80 100 120 Adsorption capacity (mg/g) Concentration (ppm) MIP-PVC NIP-PVC 0 0.4 0.8 1.2 1.6 2 0 50 100 150 Adsorption capacity (mg/g) Time (min) MIP-PVC NIP-PVC A B   125 We used the structurally similar compound, L-histidine to test the selective interaction between MIPs and histamine (see Figure 5.4 and Figure 5.5). We obtained a negative result for the adsorption capacity for L-histidine, in keeping with previous work reported by Trikka [201]. This may reflect a swelling of the polymer by absorbing solvent, effectively increasing the concentration of L-histidine in the solution. In addition, we find that a smaller QMIPs than QNIPs for this case, indicating that NIPs express a higher retaining capacity to L-histidine than MIPs. Taken together, these studies indicate that the MIPs-PVC film was successfully imprinted by histamine and that histamine rebinding occurs selectively for histamine rather than its structural analogue (i.e. L-histidine).   Figure 5.4: (A) The ionization of histamine at different pH. (B) The structure of L-histidine. HNNHNH3NNHNH3HNN NH3NNHNH2HNN NH2I II III pKa2 = 9.40 (B) NNHOHONH2L-histidine (A)   126  Figure 5.5: The comparison of adsorption capacities of MIPs-PVC and NIPs-PVC towards histamine and L-histidine at 100 ppm after 1 h incubation. (Values are shown as mean ± standard error of the mean, n = 3.)  Preparation of tuna samples After homogenization and spiking of histamine, the current study used one-step solid-liquid extraction. We omitted trichloroacetic acid or HClO4, reagents commonly used for peptide precipitation, in order to avoid further neutralization and ion removal by ionic exchange resin, which would have necessitated a complicated pretreatment. As reported by previous studies [212-214], ethanol is widely used to purify and concentrate proteins and nucleic acids by precipitation. Because of its suitability as a solvent for histamine, ethanol was used in the current study to remove peptides and nucleic acids reduced interferents, simplifying the pretreatment required to reduce interferents.     127 MIPs-PVC film: adsorption & desorption of histamine in tuna After 5 min incubation in 100 ppm histamine solution, the NIPs-PVC film retains very little histamine. The amount grows with time but MIPs-PVC film adsorbs much more histamine at all times up to saturation (Figure 5.3A). The NIPs-PVC film exhibits a kinetic of non-specific interaction with histamine. It has been concluded above and elsewhere [201, 209] that MIPs contain non-specific binding sites in addition to binding sites templated to the target molecule. The behavior of NIPs-PVC film in the current study appears to reflect a small level of non-specific interaction between the MIPs-PVC film and histamine. To maximally reduce the non-specific interaction and achieve a better selectivity of the MIPs-PVC film, we have chosen to sacrifice the adsorption quantity of histamine by MIPs-PVC film by setting the adsorption time to 5 min. Trikka et al. reported that pH and NaCl concentration (i.e. ionic strength) affect MAA-involved polymer binding [201]. Thus, ion-exchange may be responsible for the binding between polymers and histamine. In the current study, when the pH of histamine dissolved in 30 mM ammonium acetate (mediated by acetic acid) was lower than 4.87, no binding of histamine could be observed. This result indicated that carboxylic group on the polymers was protonated and histamine was in dicationic form (see Figure 5.4A); thus, no electrostatic interaction occurred. Increasing the pH of the histamine solution improved the binding of histamine to both MIPs-PVC and NIPs-PVC as expected. Trikka et al. also discussed the possibility of reducing non-specific binding by increasing salt concentration as they found that the binding for both MIPs and NIPs to histamine decreased in the same manner at lower NaCl concentration [201]. Here, we changed the medium to ammonium acetate and compared the binding of MIPs-PVC and NIPs-PVC to histamine as a function of the concentration of ammonium acetate (Figure 5.6). For   128 histamine dissolved in water, the difference between QMIPs and QNIPs was not significant (p = 0.34), which may be due to the non-specific binding. When the pH was between 7.61 and 7.86, form II of histamine predominates regardless of tautomer (Figure 5.4A). Both QMIPs and QNIPs values decreased, but the IF increased with an increase in the ionic strength. In fact, we observed an increase in QMIPs for solutions increased from 20 mM to 30 mM in ammonium acetate, and this may well reflect an increase in the number of deprotonated carboxylic groups at complementary binding sites caused by a higher pH value. This can either offset or overcome the ionic strength effect. Taken together, these results confirm ion-exchange as the main factor that determines the binding strength of histamine to polymer carboxylic groups.   Figure 5.6: The adsorption capacities of MIPs-PVC and NIPs-PVC toward histamine at 100 ppm as a function of different concentrations of ammonium acetate in the solution after 1 h incubation. (Values are shown as mean ± standard error of the mean, n = 3.)    129 Reasoning from this observation, we were able to use the ionic strength of gold colloid to disrupt the histamine-carboxylic binding and desorb histamine from the MIPs-PVC film. In addition, the surface charge on AuNPs is negative [215], contributing an electrostatic attraction term that further facilitated histamine desorption from MIPs-PVC film. An aliquot of 1 µL of AuNPs was deposited onto the aluminum foil at different time points (30, 60, 90, 120, 150 and 180 s) after placing the MIPs-PVC film into 1 mL AuNPs. The MIPs-PVC film (or NIPs-PVC film) was obtained after 5 min incubation with histamine-spiked tuna extract (final concentration of histamine: 30 ppm). SERS spectra were collected (exposure time: 1 s) in the Raman shift region from 1082 to 1603 cm-1 to determine the most suitable desorption time. In Figure 5.7, the averaged SERS spectra (n = 4) of histamine-spiked tuna extract are shown and stacked according to desorption time. We see no significant variation in the spectra collected between 30 s and 60 s of desorption, or between 90 s and 120 s of desorption. Non-specifically bonded histamine on the surface of MIPs-PVC film readily desorbed within 30 s. After 60s, the interior of the film began to release histamine. The higher SERS signals collected after desorption times between 90 s and 120 s reflect the histamine desorbed from the surface together with that non-specifically bound to the film interior. Starting with samples given 120 s or more desorption time, we begin to see the selectively bonded histamine slowly released. The intensity of SERS signals continuously increases. We settled on a desorption time of 150 s because longer desorption times overloaded the SERS signal, owing to inherent limitations of the Raman instrumentation and operating software.      130   Figure 5.7: Averaged SERS spectra (n = 4) of histamine desorbed by AuNPs from MIPs-PVC film that was incubated with tuna extract (final concentration of histamine in this tuna extract: 30 ppm) for 5 min at different time points. (Desorption time: 1: 30 s, 2: 60 s, 3: 90 s, 4: 120 s, 5: 150 s, 6: 180 s; Raman shift range: 1082-1603 cm-1; exposure time: 1 s.)  For comparison, Figure 5.8 shows desorption data from NIPs-PVC films treated in the same manner and desorbed by AuNPs for 2.5 min. The exposure time was set to be 10 s to increase the signal-to-noise ratio. The SERS spectra are shown in Figure 5.8. Besides the featured histamine SERS bands at 1304 cm-1 and 1576 cm-1, the bands at 1341 cm-1 and 1522 cm-1 may be attributed to L- histidine. All in all, there were numerous interfering bands that were associated with tuna components, resulting in difficulty in band identification and assignment. Moreover, the intensities of featured SERS bands at 1304 cm-1and 1576 cm-1 were much lower than the ones shown in Figure 5.9, indicating the difference in adsorption capacity of histamine between MIPs-PVC and NIPs-PVC. This result further validated that MIPs were successfully imprinted by histamine.  Raman shift (cm -1) Time (s) 1 2 3 4 6 5 1576 1304 1267 1317 Raman intensity (arbitrary units)   131   Figure 5.8: SERS spectra (4 samples) of histamine and interferents desorbed by AuNPs from NIPs-PVC film that was incubated with 30 ppm tuna extract for 5 min. (Desorption time: 2.5 min; Raman shift range: 1082-1603 cm-1; exposure time: 10 s.)  -5000 5000 15000 1000 1100 1200 1300 1400 1500 1600 1700 Raman intensity (arbitrary units) Raman shift (cm-1) 1304 1576 1341 1522   132  Figure 5.9: SERS spectra (n = 4 at each concentration) of histamine desorbed by AuNPs from MIPs-PVC film that was incubated with tuna extract at different spiking levels for 5 min. (Final concentration of histamine in tuna extract: A: 30 ppm, B: 10 ppm, C: 1 ppm, D: 0 ppm, E: blank MIPs-PVC film treated by pure ethanol; desorption time: 2.5 min; Raman shift range: 1082-1603 cm-1; exposure time: 10 s.)  5.3.2 SERS spectral collection and data analysis SERS spectral collection was conducted by 10 s exposure times to increase the signal-to-noise ratio. Figure 5.10 compares the normal Raman spectra of histamine (10 × 103 ppm) and SERS spectra of histamine (10 ppm). Featured bands at 1267, 1304, 1317 and 1576 cm-1 in the SERS spectra appear among those most enhanced, which correlates well with bands in the normal Raman spectra (peak assignment in Table 5.1 [187, 216]). The enhancement factor based on the Raman band of histamine at 1576 cm-1 was calculated to be ~104. As the structural analogue of -5000 5000 15000 25000 35000 45000 55000 1000 1100 1200 1300 1400 1500 1600 1700 Raman intensity (arbitrary units) Raman shift (cm-1) A D C B E 1576 1267 1304 1317   133 histamine, L-histidine did not show a high SERS activity (Figure 5.11) owing to the fact that in L-histidine (1 × 103 ppm) SERS spectra, the intensity of the highest band (1306 cm-1) is around 3000 counts, while the one (1576 cm-1) in HA (10 ppm) SERS spectra is about 30000 counts. Moreover, the band positions could further facilitate distinguishing L-histidine from histamine.   Figure 5.10: Raman spectra of histamine. [A: SERS spectra (n = 4) of histamine at 10 ppm, B: normal Raman spectra (n = 4) of histamine at 10 × 103 ppm.]       -5000 5000 15000 25000 35000 45000 1000 1100 1200 1300 1400 1500 1600 1700 Raman intensity (arbitrary units) Raman shift (cm-1) A B 1267 1317 1576 1304   134   Figure 5.11: SERS spectra (4 samples) of L-histidine at 1 × 103 ppm. (Raman shift range: 1082-1603 cm-1; exposure time: 10 s.)  Table 5.1: Peak assignment of histamine in normal Raman spectrum. Raman shift (cm-1) Assignments 1105 imidazole C-H in-plane bending 1164 imidazole N-H in-plane bending, NH2 rocking 1232 imidazole C-H in-plane bending 1267 imidazole ring stretching 1305 imidazole ring stretching 1320 methylene twisting 1356 imidazole ring stretching 1383 methylene wagging 1440 methylene scissoring 1481 imidazole ring stretching 1570 imidazole ring stretching  -1000 0 1000 2000 3000 4000 1000 1100 1200 1300 1400 1500 1600 1700 Raman intensity (arbitrary units) Raman shift (cm-1) 1341 1306 1522   135  Figure 5.9 shows the SERS spectra of histamine extracted by MIPs-PVC from tuna meat and then desorbed by AuNPs from MIPs-PVC film. Different spiking levels of histamine resulted in the variations of spectral intensities. A blank test was carried out by introducing MIPs-PVC film into 1 mL ethanol to confirm that there was no leaking of histamine from MIPs-PVC film (spectrum E in Figure 5.9). Rather than the disorganized SERS spectra in Figure 5.8 (by NIPs-PVC), the resultant SERS spectra from MIPs-PVC treatment were in a clear pattern and all the SERS histamine featured bands (1267, 1304, 1317, and 1576 cm-1) could be easily identified. For further spectral analysis, we constructed an unsupervised PCA plot to visualize the variations in spectra of tuna extracts with different spiking levels of histamine. Moreover, we constructed a PLSR model to verify the reliability of this MIPs-PVC-SERS approach to quantify histamine in tuna. In the PCA plot (Figure 5.12), the clusters of 0, 1, and 10 ppm histamine in tuna samples were relatively close to each other, whereas the cluster of tuna sample spiked with 30 ppm histamine was located on the other side of the coordinate. The difference in spectra of 0, 1 and 10 ppm histamine in tuna were not as large as that between 10 and 30 ppm histamine in tuna. Nevertheless, each group was tightly clustered and well segregated, with the testing dataset well fitted into the calibration plot. The first principal component (PC1) accounts for 53% of the total explained variance for the separation, while the PC2 (13%) and PC3 (6%) still attribute to some minor variations.      136  Figure 5.12: PCA plot of SERS spectra of histamine acquired from different tuna extract. (Raman shift range: 1082-1603 cm-1, exposure time: 10 s.)  Then, a PLSR model was constructed using the training spectral dataset and this calibration model was further evaluated using the testing spectral dataset (Figure 5.13). RMSEP and Q2 are the parameters to estimate the predictivity of the PLSR model. The lower the RMSEP and the higher the Q2 are, the better the PLSR model is. In general, the developed PLSR model could accurately quantify the spiking level of histamine in canned fish (R2=0.947 and RMSECV=3.526, Figure 5.13). We also compared the performance between PLSR and univariate linear regression models (Figure 5.14 and Table 5.2). The linear regression model constructed on the basis of the band height at 1576 cm-1 showed slightly better performance (lower RMSEP and higher Q2) than that of the PLSR model, while the other univariate linear ï20 ï10 0 10 20 30 40 50ï30ï20ï100102030ï20ï15ï10ï5051015202530PC2(13% PC1(53%) PC3(6%)0 ppmïCal1 ppmïCal10 ppmïCal30 ppmïCal0 ppmïTest1 ppmïTest10 ppmïTest30 ppmïTestPC1 (53%) PC3 (6%) PC2 (13%)   137 regression models could not accurately quantify the spiking levels of histamine in canned tuna samples.    Figure 5.13: PLSR model of SERS spectra of histamine acquired from different tuna extract. (Raman shift range: 1082-1603 cm-1, exposure time: 10 s.)        0 5 10 15 20 25 30ï50510152025303540Real concentration (ppm)Predicted concentration (ppm)  FitCalibrationTestR2 = 0.9473 Latent VariablesRMSECV = 3.526RMSEP = 3.4367Q2 = 0.932  138  Figure 5.14: Univariate linear regression models based on different peaks and their intensities. (A: 1267 cm-1, B: 1304 cm-1, C: 1317 cm-1, D: 1576 cm-1.)  Table 5.2: Comparison of univariate linear regression models constructed based on different peaks and their intensities. Raman shift (cm-1) 1267 1304 1317 1576 RMSECV 5.589 4.256 5.378 3.155 RMSEP 5.761 4.133 4.966 2.170 Q2 0.785 0.882 0.836 0.968   5.4 Conclusion We have developed a MIPs-PVC-SERS approach to determine histamine in canned tuna. A MIPs-PVC film was fabricated and verified in its selectivity towards histamine over L-histidine.   139 This film was applied to extract histamine from tuna samples and subsequently placed into AuNPs solution to release histamine to the SERS substrate (i.e. AuNPs) through an ion-exchange mechanism. AuNPs desorption maximally simplified the operation and facilitated the detection of histamine by SERS. PCA plot, PLSR model, and univariate linear regression models were constructed to verify the reliability of this approach.     140  Conclusion and outlook Chapter 6: In this thesis, I presented and discussed an innovative MIPs-SERS approach to identify and analyze small molecules in complicated food samples (i.e. chloramphenicol in milk and honey, Sudan I in paprika powder, and histamine in canned tuna). MIPs serve as the specific substrate that can selectively extract analytes and exclude interferants simultaneously.      As sorbent material, MIPs were used to developed different types of separation elements. Specifically, solid phase cartridge provided easy operation. Once MIPs particles were tightly packed into SPE cartridge, loading, washing and eluting of sample solutions could be carried out. MIPs-TLC plate was prepared by spreading MIPs onto glass slide. As only the analyte (i.e. Sudan I) was retained at the original sample spot and all other interferents were washed forward with developing solvent, separation was done less than a minute. In addition, MIPs-TLC plate enabled on-line detection of Sudan I spot, which further facilitated the process of determining Sudan I in paprika powder. Additionally, MIPs-PVC film as another separation element showed the characteristic of convenient separation. The flexible and thin film was easy to operate as MIPs particles were tightly immobilized. Dendritic silver and colloidal gold nanoparticles function as SERS active substrates and consequently the sensitive analysis can be yielded. Multivariate statistical analysis (PCA and PLSR) has been employed to realize the rapid classification and regression.      Chloramphenicol-imprinted MIPs were formed via precipitation polymerization. Static and kinetic studies validated the specific selectivity of MIPs toward chloramphenicol over non-imprinted polymers (imprinting factor > 4). Canadian penny-based silver nano-structure was synthesized as a SERS-active substrate for the determination of chloramphenicol in foods.   141 Collected spectra were analyzed by PCA to differentiate various concentrations of chloramphenicol in different foods. PLSR models showed good prediction values (R > 0.9) of actual spiked contents (0, 0.1, 0.5, 1, 5 ppm) of chloramphenicol in milk and honey. This developed approach is low cost, requires little sample pretreatment, and can provide reliable detection of trace level of chloramphenicol in foods within a total of 15 min, including food sample preparation.     In Chapter 4, the integrated MIPs-TLC-SERS method was developed and could determine the levels of Sudan I in paprika powder down to 1 ppm (or 2 ng/spot). Sudan I spiked paprika extracts (spiking levels: 0, 1, 5, 10, 40, 70 and 100 ppm) were prepared. Sudan I imprinted MIPs were used as a stationary phase for TLC and could selectively retain Sudan I at the original spot with little interferents left A homemade gold colloid SERS substrate could enhance Raman signal intensity for Sudan I separated by this MIP-TLC system. PLSR (R2 = 0.978) models were constructed and a linear regression model (R2 = 0.983) correlated spiking levels (5, 10, 40, 70 and 100 ppm) with the peak intensities (721 cm−1) of Sudan I SERS spectra. Both separation (30-40 s) and detection (1 s or 0.1 s) are extremely fast by using both commercial bench-top and custom made portable Raman spectrometers. This MIPs-TLC-SERS sensor can be applied as a rapid, low-cost, and reliable tool for screening Sudan I adulteration in foods.     Chapter 5 introduces a rapid, cost effective, and reliable approach to determine the level of histamine in canned tuna. PVC was used to immobilize MIPs, yielding a MIPs-PVC film that functions as a recognition element to selectively separate histamine from tuna extract. A gold colloid solution served both as an eluting solvent to extract histamine from MIPs-PVC film and conduct a SERS detection of histamine. PCA together with a PLSR model (R2 = 0.947, RMSECV = 3.526) confirmed the reliability of this MIPs-PVC-SERS approach for the detection   142 and spectral analysis of histamine. Linear regression models were also constructed to correlate the intensity of different histamine SERS bands with the corresponding spiking levels. One such model (using a band at 1576 cm-1) performed slightly better in predicting histamine content in tuna than the PLSR model. Taken together, this MIPs-PVC-SERS approach can rapidly and reliably determine histamine at levels from 3 to 90 ppm in canned tuna meat.     Future work can be focused on developing other rapid separation and sensitive detection techniques. MIPs have been well studied for a very long time. Various synthesis methods have been introduced to obtain MIPs with different physical and chemical properties. Then, the study of MIPs and corresponding separation techniques can be further extended. For example, the reported in situ polymerization in capillary tube can be another option without subsequent packing or fabrication [217]. Besides MIPs, other separation elements such as antibodies and aptamers can be good candidates due to their specific binding capacity. Integrating MIPs (or antibodies) and signal reporter (e.g. fluorophore or SERS substrate) can be another option. TERS has been validated as a sensitive and efficient technique to enhance the intensity of Raman scattering signals with a higher spectral reproducibility than SERS technique. Further, TERS can provide information about sample distribution on a substrate, such as the sample spot on TLC plate and MIPs-PVC film. Therefore, TERS may require no desorption of analytes and ultimately realize on-line detection. Development on other detection techniques should be explored as well.      In conclusion, the introduced MIPs-SERS approach for the determination of small molecules in food samples has been validated to be rapid and sensitive, indicating wide application of this approach in food science and food analysis.    143 Bibliography  [1] Grunert, K.G., Food quality and safety: consumer perception and demand, European Review of Agricultural Economics 2005, 32, 369-391. [2] Krska, R., Becalski, A., Braekevelt, E., Koerner, T., Cao, X.L., Dabeka, R., Godefroy, S., Lau, B., Moisey, J., Rawn, D.F., Challenges and trends in the determination of selected chemical contaminants and allergens in food, Analytical and Bioanalytical Chemistry 2012, 402, 139-162. [3] Nigg, J.T., Lewis, K., Edinger, T., Falk, M., Meta-analysis of attention-deficit/hyperactivity disorder or attention-deficit/hyperactivity disorder symptoms, restriction diet, and synthetic food color additives, Journal of the American Academy of Child & Adolescent Psychiatry 2012, 51, 86-97. [4] Oliver, S.P., Jayarao, B.M., Almeida, R.A., Foodborne pathogens in milk and the dairy farm environment: food safety and public health implications, Foodbourne Pathogens & Disease 2005, 2, 115-129. [5] Kataoka, H., Kurisu, M., Shindoh, S., Determination of volatile N-nitrosamines in combustion smoke samples, Bulletin of Environmental Contamination and Toxicology 1997, 59, 570-576. [6] Tricker, A., Preussmann, R., Carcinogenic N-nitrosamines in the diet: occurrence, formation, mechanisms and carcinogenic potential, Mutation Research/Genetic Toxicology 1991, 259, 277-289. [7] McCann, D., Barrett, A., Cooper, A., Crumpler, D., Dalen, L., Grimshaw, K., Kitchin, E., Lok, K., Porteous, L., Prince, E., Food additives and hyperactive behaviour in 3-year-old and 8/9-year-old children in the community: a randomised, double-blinded, placebo-controlled trial, Lancet 2007, 370, 1560-1567. [8] Carvalho, F.P., Agriculture, pesticides, food security and food safety, Environmental Science & Policy 2006, 9, 685-692. [9] Kataoka, H., Lord, H.L., Pawliszyn, J., Applications of solid-phase microextraction in food analysis, Journal of Chromatography A 2000, 880, 35-62. [10] Pomeranz, Y., Food analysis: theory and practice, Springer Science & Business Media 2013. [11] Schenck, F.J., Lehotay, S.J., Does further clean-up reduce the matrix enhancement effect in gas chromatographic analysis of pesticide residues in food?, Journal of Chromatography A 2000, 868, 51-61. [12] Hajšlová, J., Zrostlıkova, J., Matrix effects in (ultra) trace analysis of pesticide residues in food and biotic matrices, Journal of Chromatography A 2003, 1000, 181-197. [13] Otles, S., Ozyurt, V.H., Instrumental Food Analysis 2015. [14] Lehotay, S.J., Hajšlová, J., Application of gas chromatography in food analysis, TrAC Trends in Analytical Chemistry 2002, 21, 686-697. [15] Bianchi, F., Careri, M., Mangia, A., Musci, M., Retention indices in the analysis of food aroma volatile compounds in temperature-programmed gas chromatography: database creation and evaluation of precision and robustness, Journal of Separation Science 2007, 30, 563-572. [16] El-Saeid, M., Selim, M., Multiresidue analysis of 86 pesticides using gas chromatography mass spectrometry: II-nonleafy vegetables, Journal of Chemistry 2012, 2013, 1-10.   144 [17] Xu, M.L., Liu, J.B., Lu, J., Determination and control of pesticide residues in beverages: a review of extraction techniques, chromatography, and rapid detection methods, Applied Spectroscopy Reviews 2014, 49, 97-120. [18] Matthews, G., Pesticide application methods, John Wiley & Sons 2008. [19] Grube, A., Donaldson, D., Kiely, T., Wu, L., Pesticides industry sales and usage, US Enviromment Protection Agency, Washington, DC 2011. [20] Eisert, R., Levsen, K., Solid-phase microextraction coupled to gas chromatography: a new method for the analysis of organics in water, Journal of Chromatography A 1996, 733, 143-157. [21] Eisert, R., Pawliszyn, J., Automated in-tube solid-phase microextraction coupled to high-performance liquid chromatography, Analytical Chemistry 1997, 69, 3140-3147. [22] Berijani, S., Assadi, Y., Anbia, M., Hosseini, M.-R.M., Aghaee, E., Dispersive liquid–liquid microextraction combined with gas chromatography-flame photometric detection: very simple, rapid and sensitive method for the determination of organophosphorus pesticides in water, Journal of Chromatography A 2006, 1123, 1-9. [23] Akasaka, K., Sasaki, I., Ohrui, H., Meguro, H., A simple fluorometry of hydroperoxides in oils and foods, Bioscience, Biotechnology, and Biochemistry 1992, 56, 605-607. [24] Osborne, B.G., Near-infrared spectroscopy in food analysis, Encyclopedia of Analytical Chemistry 2000. [25] Berrueta, L.A., Alonso-Salces, R.M., Héberger, K., Supervised pattern recognition in food analysis, Journal of Chromatography A 2007, 1158, 196-214. [26] Vigli, G., Philippidis, A., Spyros, A., Dais, P., Classification of edible oils by employing 31P and 1H NMR spectroscopy in combination with multivariate statistical analysis. A proposal for the detection of seed oil adulteration in virgin olive oils, Journal of Agricultural and Food Chemistry 2003, 51, 5715-5722. [27] Liu, G., Lin, Y., Biosensor based on self-assembling acetylcholinesterase on carbon nanotubes for flow injection/amperometric detection of organophosphate pesticides and nerve agents, Analytical Chemistry 2006, 78, 835-843. [28] Abad, J., Pariente, F., Hernandez, L., Abruna, H., Lorenzo, E., Determination of organophosphorus and carbamate pesticides using a piezoelectric biosensor, Analytical Chemistry 1998, 70, 2848-2855. [29] Kim, N., Park, I.S., Kim, D.K., High-sensitivity detection for model organophosphorus and carbamate pesticide with quartz crystal microbalance-precipitation sensor, Biosensors and Bioelectronics 2007, 22, 1593-1599. [30] Raman, C.V., Krishnan, K.S., A new type of secondary radiation, Nature 1928, 121, 501-502. [31] Boyaci, I.H., Temiz, H.T., Geniş, H.E., Soykut, E.A., Yazgan, N.N., Güven, B., Uysal, R.S., Bozkurt, A.G., İlaslan, K., Torun, O., Dispersive and FT-Raman spectroscopic methods in food analysis, RSC Advances 2015, 5, 56606-56624. [32] Long, D.A., Raman spectroscopy, New York 1977. [33] McCartney, E.J., Optics of the atmosphere: scattering by molecules and particles, New York, John Wiley and Sons, Inc. 1976. [34] Lord, R., Yu, N.T., Laser-excited Raman spectroscopy of biomolecules: I. Native lysozyme and its constituent amino acids, Journal of Molecular Biology 1970, 50, 509-524. [35] Wright, S.J., Wright, D.J., Introduction to confocal microscopy, Cell Biological Applications of Confocal Microscopy, in Methods in Cell Biology 2002, 70, 1-85.   145 [36] Hartschuh, A., Sánchez, E.J., Xie, X.S., Novotny, L., High-resolution near-field Raman microscopy of single-walled carbon nanotubes, Physical Review Letters 2003, 90, 095503. [37] Graf, D., Molitor, F., Ensslin, K., Stampfer, C., Jungen, A., Hierold, C., Wirtz, L., Spatially resolved Raman spectroscopy of single-and few-layer graphene, Nano Letters 2007, 7, 238-242. [38] Campbell, I., Fauchet, P.M., The effects of microcrystal size and shape on the one phonon Raman spectra of crystalline semiconductors, Solid State Communications 1986, 58, 739-741. [39] Manoharan, R., Shafer, K., Perelman, L., Wu, J., Chen, K., Deinum, G., Fitzmaurice, M., Myles, J., Crowe, J., Dasarl, R.R., Raman spectroscopy and fluorescence photon migration for breast cancer diagnosis and imaging, Photochemistry and Photobiology 1998, 67, 15-22. [40] Liu, Z., Davis, C., Cai, W., He, L., Chen, X., Dai, H., Circulation and long-term fate of functionalized, biocompatible single-walled carbon nanotubes in mice probed by Raman spectroscopy, Proceedings of the National Academy of Sciences 2008, 105, 1410-1415. [41] De Juan, A., Tauler, R., Dyson, R., Marcolli, C., Rault, M., Maeder, M., Spectroscopic imaging and chemometrics: a powerful combination for global and local sample analysis, TrAC Trends in Analytical Chemistry 2004, 23, 70-79. [42] Cheng, J.X., Xie, X.S., Coherent Raman scattering microscopy, CRC press 2012. [43] Hirschfeld, T., Chase, B., FT-Raman spectroscopy: development and justification, Applied Spectroscopy 1986, 40, 133-137. [44] Lu, X., Al-Qadiri, H., Lin, M., Rasco, B., Application of mid-infrared and Raman spectroscopy to the study of bacteria, Food and Bioprocess Technology 2011, 4, 919-935. [45] Chase, D.B., Fourier transform Raman spectroscopy, Journal of the American Chemical Society 1986, 108, 7485-7488. [46] Pettinger, B., Ren, B., Picardi, G., Schuster, R., Ertl, G., Nanoscale probing of adsorbed species by tip-enhanced Raman spectroscopy, Physical Review Letters 2004, 92, 096101. [47] Clark, R.J.H., Wang, Q., Correia, A., Can the Raman spectrum of anatase in artwork and archaeology be used for dating purposes? Identification by Raman microscopy of anatase in decorative coatings on Neolithic (Yangshao) pottery from Henan, China, Journal of Archaeological Science 2007, 34, 1787-1793. [48] Efremov, E.V., Ariese, F., Gooijer, C., Achievements in resonance Raman spectroscopy: Review of a technique with a distinct analytical chemistry potential, Analytica Chimica Acta 2008, 606, 119-134. [49] Impens, S., Reybroeck, W., Vercammen, J., Courtheyn, D., Ooghe, S., De Wasch, K., Smedts, W., De Brabander, H., Screening and confirmation of chloramphenicol in shrimp tissue using ELISA in combination with GC–MS 2 and LC–MS 2, Analytica Chimica Acta 2003, 483, 153-163. [50] Matousek, P., Towrie, M., Parker, A., Fluorescence background suppression in Raman spectroscopy using combined Kerr gated and shifted excitation Raman difference techniques, Journal of Raman Spectroscopy 2002, 33, 238-242. [51] Fujino, T., Tahara, T., Picosecond time-resolved Raman study of trans-azobenzene, The Journal of Physical Chemistry A 2000, 104, 4203-4210. [52] Meyer, S.A., Ru, E.C.L., Etchegoin, P.G., Quantifying resonant Raman cross sections with SERS, Journal of Physical Chemistry A 2010, 114, 5515-5519. [53] Yoshizawa, M., Kurosawa, M., Femtosecond time-resolved Raman spectroscopy using stimulated Raman scattering, Physical Review A 1999, 61, 013808.   146 [54] Willets, K.A., Van Duyne, R.P., Localized surface plasmon resonance spectroscopy and sensing, Annual Review of Physical Chemistry 2007, 58, 267-297. [55] Krenn, J.R., Nanoparticle waveguides: watching energy transfer, Nature Materials 2003, 2, 210-211. [56] Albrecht, A.C., On the theory of Raman intensities, Journal of Chemical Physics 1961, 34, 1476-1484. [57] Kneipp, K., Wang, Y., Kneipp, H., Perelman, L.T., Itzkan, I., Dasari, R.R., Feld, M.S., Single molecule detection using surface-enhanced Raman scattering (SERS), Physical Review Letters 1997, 78, 1667. [58] Kneipp, K., Kneipp, H., Itzkan, I., Dasari, R.R., Feld, M.S., Surface-enhanced Raman scattering and biophysics, Journal of Physics: Condensed Matter 2002, 14, 597. [59] Zheng, J., He, L., Surface-enhanced Raman spectroscopy for the chemical analysis of food, Comprehensive Reviews in Food Science and Food Safety 2014, 13, 317-328. [60] Craig, A.P., Franca, A.S., Irudayaraj, J., Surface-enhanced Raman spectroscopy applied to food safety, Annual Review of Food Science and Technology 2013, 4, 369-380. [61] Lin, M., He, L., Awika, J., Yang, L., Ledoux, D., Li, H.A., Mustapha, A., Detection of melamine in gluten, chicken feed, and processed foods using surface enhanced Raman spectroscopy and HPLC, Journal of Food Science 2008, 73, 129-134. [62] Cheung, W., Shadi, I.T., Xu, Y., Goodacre, R., Quantitative analysis of the banned food dye Sudan I using surface enhanced Raman scattering with multivariate chemometrics, Journal of Physical Chemistry C 2010, 114, 7285-7290. [63] Farquharson, S., Maksymiuk, P., Simultaneous chemical separation and surface-enhanced Raman spectral detection using silver-doped sol-gels, Applied Spectroscopy 2003, 57, 479-482. [64] He, L., Haynes, C.L., Diez-Gonzalez, F., Labuza, T.P., Rapid detection of a foreign protein in milk using IMS–SERS, Journal of Raman Spectroscopy 2011, 42, 1428-1434. [65] Pan, G., Zhang, Y., Ma, Y., Li, C., Zhang, H., Efficient one-pot synthesis of water‐compatible molecularly imprinted polymer microspheres by facile RAFT precipitation polymerization, Angewandte Chemie International Edition 2011, 50, 11731-11734. [66] Feng, S., Gao, F., Chen, Z., Grant, E., Kitts, D.D., Wang, S., Lu, X., Determination of α-tocopherol in vegetable oils using a molecularly imprinted polymers-surface-enhanced Raman spectroscopic biosensor, Journal of Agricultural and Food Chemistry 2013, 61, 10467-10475. [67] Hu, Y., Feng, S., Gao, F., Li-Chan, E.C., Grant, E., Lu, X., Detection of melamine in milk using molecularly imprinted polymers-surface enhanced Raman spectroscopy, Food Chemistry 2015, 176, 123-129. [68] Liu, P., Liu, R., Guan, G., Jiang, C., Wang, S., Zhang, Z., Surface-enhanced Raman scattering sensor for theophylline determination by molecular imprinting on silver nanoparticles, Analyst 2011, 136, 4152-4158. [69] Zhao, Y., Liu, X., Lei, D.Y., Chai, Y., Effects of surface roughness of Ag thin films on surface-enhanced Raman spectroscopy of graphene: spatial nonlocality and physisorption strain, Nanoscale 2014, 6, 1311-1317. [70] Stöckle, R.M., Suh, Y.D., Deckert, V., Zenobi, R., Nanoscale chemical analysis by tip-enhanced Raman spectroscopy, Chemical Physics Letters 2000, 318, 131-136. [71] Cheng, J.X., Xie, X.S., Coherent snti-Stokes Raman scattering microscopy:  instrumentation, theory, and applications, Journal of Physical Chemistry B 2004, 108, 827-840.   147 [72] Haupt, K., Mosbach, K., Molecularly imprinted polymers and their use in biomimetic sensors, Chemical Reviews 2000, 100, 2495-2504. [73] Evans, C.L., Xie, X.S., Coherent anti-Stokes Raman scattering microscopy: chemical imaging for biology and medicine, Annual Review of Analytical Chemistry. 2008, 1, 883-909. [74] Kneipp, K., Moskovits, M., Kneipp, H., Surface-enhanced Raman scattering: physics and applications, Springer Science & Business Media 2006. [75] Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E., Tatham, R.L., Multivariate data analysis, Pearson Prentice Hall Upper Saddle River, NJ 2006. [76] Wise, B.M., Gallagher, N., Bro, R., Shaver, J., Windig, W., Koch, R.S., PLS Toolbox 4.0, Eigenvector Research Incorporated 2007, 3905. [77] Mahadevan-Jansen, A., Richards-Kortum, R.R., Raman spectroscopy for the detection of cancers and precancers, Journal of Biomedical Optics 1996, 1, 31-70. [78] Zhao, J., Lui, H., McLean, D.I., Zeng, H., Automated autofluorescence background subtraction algorithm for biomedical Raman spectroscopy, Applied Spectroscopy 2007, 61, 1225-1232. [79] Brault, J., White, O., The analysis and restoration of astronomical data via the fast Fourier transform, Astronomy and Astrophysics 1971, 13, 169. [80] Welling, M., Fisher linear discriminant analysis, Department of Computer Science, University of Toronto 2005. [81] Tobias, R.D., An introduction to partial least squares regression,  Processing Annual SAS Users Group Internetional Conference, 20th, Orlando, FL, Citeseer, 1995, 2-5. [82] Abdi, H., Multivariate analysis, Encyclopedia for research methods for the social sciences. Thousand Oaks: Sage 2003, 699-702. [83] Wold, S., Sjöström, M., Eriksson, L., PLS-regression: a basic tool of chemometrics, Chemometrics and intelligent laboratory systems 2001, 58, 109-130. [84] Wulff, G., Enzyme-like catalysis by molecularly imprinted polymers, Chemical Reviews 2002, 102, 1-28. [85] Shea, K.J., Sasaki, D.Y., An analysis of small-molecule binding to functionalized synthetic polymers by 13C CP/MAS NMR and FT-IR spectroscopy, Journal of the American Chemical Society 1991, 113, 4109-4120. [86] Sellergren, B., Hall, A.J., Molecularly imprinted polymers, Supramolecular Chemistry: from Molecules to Nanomaterials 2012. [87] Mayes, A., Whitcombe, M., Synthetic strategies for the generation of molecularly imprinted organic polymers, Advanced Drug Delivery Reviews 2005, 57, 1742-1778. [88] Haupt, K., Peer Reviewed: Molecularly imprinted polymers: the next generation, Analytical Chemistry 2003, 75, 376-383. [89] Kempe, M., Mosbach, K., Direct resolution of naproxen on a non-covalently molecularly imprinted chiral stationary phase, Journal of Chromatography A 1994, 664, 276-279. [90] O’Mahony, J., Molinelli, A., Nolan, K., Smyth, M., Mizaikoff, B., Towards the rational development of molecularly imprinted polymers: 1 H NMR studies on hydrophobicity and ion-pair interactions as driving forces for selectivity, Biosensors and Bioelectronics 2005, 20, 1884-1893. [91] Karim, K., Breton, F., Rouillon, R., Piletska, E.V., Guerreiro, A., Chianella, I., Piletsky, S.A., How to find effective functional monomers for effective molecularly imprinted polymers?, Advanced Drug Delivery Reviews 2005, 57, 1795-1808.   148 [92] Pietrzyk, A., Kutner, W., Chitta, R., Zandler, M.E., D’Souza, F., Sannicolò, F., Mussini, P.R., Melamine acoustic chemosensor based on molecularly imprinted polymer film, Analytical Chemistry 2009, 81, 10061-10070. [93] Yoshimatsu, K., Yamazaki, T., Chronakis, I.S., Ye, L., Influence of template/functional monomer/cross-linking monomer ratio on particle size and binding properties of molecularly imprinted nanoparticles, Journal of Applied Polymer Science 2012, 124, 1249-1255. [94] Masqué, N., Marcé, R.M., Borrull, F., Molecularly imprinted polymers: new tailor-made materials for selective solid-phase extraction, TrAC Trends in Analytical Chemistry 2001, 20, 477-486. [95] Cormack, P.A.G., Elorza, A.Z., Molecularly imprinted polymers: synthesis and characterisation, Journal of Chromatography B 2004, 804, 173-182. [96] Mijangos, I., Navarro-Villoslada, F., Guerreiro, A., Piletska, E., Chianella, I., Karim, K., Turner, A., Piletsky, S., Influence of initiator and different polymerisation conditions on performance of molecularly imprinted polymers, Biosensors and Bioelectronics 2006, 22, 381-387. [97] Wang, J., Cormack, P.A., Sherrington, D.C., Khoshdel, E., Monodisperse, molecularly imprinted polymer microspheres prepared by precipitation polymerization for affinity separation applications, Angewandte Chemie International Edition 2003, 42, 5336-5338. [98] Pérez, N., Whitcombe, M.J., Vulfson, E.N., Molecularly imprinted nanoparticles prepared by core-shell emulsion polymerization, Journal of Applied Polymer Science 2000, 77, 1851-1859. [99] Mayes, A.G., Mosbach, K., Molecularly imprinted polymer beads: suspension polymerization using a liquid perfluorocarbon as the dispersing phase, Analytical Chemistry 1996, 68, 3769-3774. [100] Titirici, M.-M., Sellergren, B., Thin molecularly imprinted polymer films via reversible addition-fragmentation chain transfer polymerization, Chemistry of Materials 2006, 18, 1773-1779. [101] Chen, Z., Ye, L., Controlling size and uniformity of molecularly imprinted nanoparticles using auxiliary template, Journal of Molecular Recognition 2012, 25, 370-376. [102] Theodoridis, G., Kantifes, A., Manesiotis, P., Raikos, N., Tsoukali-Papadopoulou, H., Preparation of a molecularly imprinted polymer for the solid-phase extraction of scopolamine with hyoscyamine as a dummy template molecule, Journal of Chromatography A 2003, 987, 103-109. [103] Kugimiya, A., Takeuchi, T., Surface plasmon resonance sensor using molecularly imprinted polymer for detection of sialic acid, Biosensors and Bioelectronics 2001, 16, 1059-1062. [104] Ye, L., Ramström, O., Mosbach, K., Molecularly imprinted polymeric adsorbents for byproduct removal, Analytical Chemistry 1998, 70, 2789-2795. [105] Zhang, H., Piacham, T., Drew, M., Patek, M., Mosbach, K., Ye, L., Molecularly imprinted nanoreactors for regioselective huisgen 1, 3-dipolar cycloaddition reaction, Journal of the American Chemical Society 2006, 128, 4178-4179. [106] Alexander, C., Smith, C.R., Whitcombe, M.J., Vulfson, E.N., Imprinted polymers as protecting groups for regioselective modification of polyfunctional substrates, Journal of the American Chemical Society 1999, 121, 6640-6651.   149 [107] Matsui, J., Nicholls, I.A., Karube, I., Mosbach, K., Carbon-carbon bond formation using substrate selective catalytic polymers prepared by molecular imprinting: An artificial class II aldolase, Journal of Organic Chemistry 1996, 61, 5414-5417. [108] Sergeyeva, T.A., Matuschewski, H., Piletsky, S.A., Bendig, J., Schedler, U., Ulbricht, M., Molecularly imprinted polymer membranes for substance-selective solid-phase extraction from water by surface photo-grafting polymerization, Journal of Chromatography A 2001, 907, 89-99. [109] Sellergren, B., Lepistoe, M., Mosbach, K., Highly enantioselective and substrate-selective polymers obtained by molecular imprinting utilizing noncovalent interactions. NMR and chromatographic studies on the nature of recognition, Journal of the American Chemical Society 1988, 110, 5853-5860. [110] Hilal, N., Kochkodan, V., Al‐Khatib, L., Busca, G., Characterization of molecularly imprinted composite membranes using an atomic force microscope, Surface and Interface Analysis 2002, 33, 672-675. [111] Sánchez-Barragán, I., Karim, K., Costa-Fernández, J.M., Piletsky, S.A., Sanz-Medel, A., A molecularly imprinted polymer for carbaryl determination in water, Sensors and Actuators B: Chemical 2007, 123, 798-804. [112] Jenkins, A.L., Yin, R., Jensen, J.L., Molecularly imprinted polymer sensors for pesticide and insecticide detection in water, Analyst 2001, 126, 798-802. [113] Ju, J.Y., Shin, C., Whitcombe, M., Vulfson, E., Binding properties of an aminostyrene-based polymer imprinted with glutamylated monascus pigments, Biotechnology Techniques 1999, 13, 665-669. [114] Vivekanandan, K., Swamy, M.G., Prasad, S., Mukherjee, R., A simple method of isolation of chloramphenicol in honey and its estimation by liquid chromatography coupled to electrospray ionization tandem mass spectrometry, Rapid Communications in Mass Spectrometry 2005, 19, 3025-3030. [115] Shi, X., Wu, A., Zheng, S., Li, R., Zhang, D., Molecularly imprinted polymer microspheres for solid-phase extraction of chloramphenicol residues in foods, Journal of Chromatography B 2007, 850, 24-30. [116] Singer, C., Katz, S., Microbiological assay for chloramphenicol residues, Journal-Association of Official Analytical Chemists 1984, 68, 1037-1041. [117] Ferguson, J., Baxter, A., Young, P., Kennedy, G., Elliott, C., Weigel, S., Gatermann, R., Ashwin, H., Stead, S., Sharman, M., Detection of chloramphenicol and chloramphenicol glucuronide residues in poultry muscle, honey, prawn and milk using a surface plasmon resonance biosensor and Qflex® kit chloramphenicol, Analytica Chimica Acta 2005, 529, 109-113. [118] Posyniak, A., Zmudzki, J., Niedzielska, J., Evaluation of sample preparation for control of chloramphenicol residues in porcine tissues by enzyme-linked immunosorbent assay and liquid chromatography, Analytica Chimica Acta 2003, 483, 307-311. [119] Mottier, P., Parisod, V., Gremaud, E., Guy, P.A., Stadler, R.H., Determination of the antibiotic chloramphenicol in meat and seafood products by liquid chromatography–electrospray ionization tandem mass spectrometry, Journal of Chromatography A 2003, 994, 75-84. [120] Boyd, B., Björk, H., Billing, J., Shimelis, O., Axelsson, S., Leonora, M., Yilmaz, E., Development of an improved method for trace analysis of chloramphenicol using molecularly imprinted polymers, Journal of Chromatography A 2007, 1174, 63-71.   150 [121] Tang, Y., Fang, G., Wang, S., Sun, J., Qian, K., Rapid determination of metolcarb residues in foods using a biomimetic enzyme-linked immunosorbent assay employing a novel molecularly imprinted polymer film as artificial antibody, Journal of AOAC International 2013, 96, 453-458. [122] Lu, Q., Chen, X., Nie, L., Luo, J., Jiang, H., Chen, L., Hu, Q., Du, S., Zhang, Z., Tuning of the vinyl groups’ spacing at surface of modified silica in preparation of high density imprinted layer-coated silica nanoparticles: A dispersive solid-phase extraction materials for chlorpyrifos, Talanta 2010, 81, 959-966. [123] He, J., Fang, G., Deng, Q., Wang, S., Preparation, characterization and application of organic-inorganic hybrid ractopamine multi-template molecularly imprinted capillary monolithic column, Analytica Chimica Acta 2011, 692, 57-62. [124] Dieringer, J.A., McFarland, A.D., Shah, N.C., Stuart, D.A., Whitney, A.V., Yonzon, C.R., Young, M.A., Zhang, X., Van Duyne, R.P., Introductory lecture surface enhanced Raman spectroscopy: new materials, concepts, characterization tools, and applications, Faraday Discussions 2006, 132, 9-26. [125] Xu, H., Aizpurua, J., Käll, M., Apell, P., Electromagnetic contributions to single-molecule sensitivity in surface-enhanced Raman scattering, Physical Review E 2000, 62, 4318. [126] Steidtner, J., Pettinger, B., Tip-enhanced Raman spectroscopy and microscopy on single dye molecules with 15 nm resolution, Physical Review Letters 2008, 100, 236101. [127] Zhang, R., Zhang, Y., Dong, Z.C., Jiang, S., Zhang, C., Chen, L.G., Zhang, L., Liao, Y., Aizpurua, J., Luo, Y., Yang, J.L., Hou, J.G., Chemical mapping of a single molecule by plasmon-enhanced Raman scattering, Nature 2013, 498, 82-86. [128] Lu, X., Samuelson, D.R., Xu, Y., Zhang, H., Wang, S., Rasco, B.A., Xu, J., Konkel, M.E., Detecting and tracking nosocomial methicillin-resistant staphylococcus aureus using a microfluidic SERS biosensor, Analytical Chemistry 2013, 85, 2320-2327. [129] He, L., Kim, N.J., Li, H., Hu, Z., Lin, M., Use of a fractal-like gold nanostructure in surface-enhanced Raman spectroscopy for detection of selected food contaminants, Journal of Agricultural and Food Chemistry 2008, 56, 9843-9847. [130] Jackson, L.S., Chemical food safety issues in the United States: past, present, and future, Journal of Agricultural and Food Chemistry 2009, 57, 8161-8170. [131] Zhang, X.F., Zou, M.Q., Qi, X.H., Liu, F., Zhu, X.H., Zhao, B.H., Detection of melamine in liquid milk using surface-enhanced Raman scattering spectroscopy, Journal of Raman Spectroscopy 2010, 41, 1655-1660. [132] Shi, Y.E., Li, L., Yang, M., Jiang, X., Zhao, Q., Zhan, J., A disordered silver nanowires membrane for extraction and surface-enhanced Raman spectroscopy detection, Analyst 2014, 139, 2525-2530. [133] Zhang, Y., Huang, Y., Zhai, F., Du, R., Liu, Y., Lai, K., Analyses of enrofloxacin, furazolidone and malachite green in fish products with surface-enhanced Raman spectroscopy, Food Chemistry 2012, 135, 845-850. [134] Mabbott, S., Eckmann, A., Casiraghi, C., Goodacre, R., 2p or not 2p: tuppence-based SERS for the detection of illicit materials, Analyst 2013, 138, 118-122. [135] Mayes, A.G., Whitcombe, M.J., Synthetic strategies for the generation of molecularly imprinted organic polymers, Advanced Drug Delivery Reviews 2005, 57, 1742-1778.   151 [136] Qian, K., Fang, G., He, J., Pan, M., Wang, S., Preparation and application of a molecularly imprinted polymer for the determination of trace metolcarb in food matrices by high performance liquid chromatography, Journal of Separation Science 2010, 33, 2079-2085. [137] Zhu, Q.-Z., Haupt, K., Knopp, D., Niessner, R., Molecularly imprinted polymer for metsulfuron-methyl and its binding characteristics for sulfonylurea herbicides, Analytica Chimica Acta 2002, 468, 217-227. [138] Levi, R., McNiven, S., Piletsky, S.A., Cheong, S.H., Yano, K., Karube, I., Optical detection of chloramphenicol using molecularly imprinted polymers, Analytical Chemistry 1997, 69, 2017-2021. [139] Schirmer, C., Meisel, H., Synthesis of a molecularly imprinted polymer for the selective solid-phase extraction of chloramphenicol from honey, Journal of Chromatography A 2006, 1132, 325-328. [140] Schirmer, C., Meisel, H., Chromatographic evaluation of polymers imprinted with analogs of chloramphenicol and application to selective solid-phase extraction, Analytical and Bioanalytical Chemistry 2009, 394, 2249-2255. [141] Kujawski, M.W., Namieśnik, J., Challenges in preparing honey samples for chromatographic determination of contaminants and trace residues, TrAC Trends in Analytical Chemistry 2008, 27, 785-793. [142] Faulkner, H., Hussein, A., Foran, M., Szijarto, L., A survey of vitamin A and D contents of fortified fluid milk in Ontario, Journal of Dairy Science 2000, 83, 1210-1216. [143] He, L., Lin, M., Li, H., Kim, N.J., Surface‐enhanced Raman spectroscopy coupled with dendritic silver nanosubstrate for detection of restricted antibiotics, Journal of Raman Spectroscopy 2010, 41, 739-744. [144] Xu, H., Shao, M., Chen, T., Zhao, Y., Lee, S.T., Surface‐enhanced Raman scattering on silver dendrite with different growth directions, Journal of Raman Spectroscopy 2012, 43, 396-404. [145] Zuo, C., Jagodzinski, P.W., Surface-enhanced Raman scattering of pyridine using different metals:  differences and explanation based on the selective formation of α-pyridyl on metal surfaces, Journal of Physical Chemistry B 2005, 109, 1788-1793. [146] Gutés, A., Carraro, C., Maboudian, R., Silver dendrites from galvanic displacement on commercial aluminum foil as an effective SERS substrate, Journal of the American Chemical Society 2010, 132, 1476-1477. [147] Lin, H., Mock, J., Smith, D., Gao, T., Sailor, M.J., Surface-enhanced Raman scattering from silver-plated porous silicon, Journal of Physical Chemistry B 2004, 108, 11654-11659. [148] Shen, X., Ho, C.M., Wong, T.S., Minimal size of coffee ring structure, Journal of Physical Chemistry B 2010, 114, 5269-5274. [149] Lai, K., Zhang, Y., Du, R., Zhai, F., Rasco, B., Huang, Y., Determination of chloramphenicol and crystal violet with surface enhanced Raman spectroscopy, Sensing and Instrumentation Food Quality and Safety 2011, 5, 19-24. [150] Movasaghi, Z., Rehman, S., Rehman, I.U., Raman spectroscopy of biological tissues, Applied Spectroscopy Reviews 2007, 42, 493-541. [151] Rejtharová, M., Rejthar, L., Determination of chloramphenicol in urine, feed water, milk and honey samples using molecular imprinted polymer clean-up, Journal of Chromatography A 2009, 1216, 8246-8253.   152 [152] Xue, J.Q., Li, D.W., Qu, L.L., Long, Y.T., Surface-imprinted core–shell Au nanoparticles for selective detection of bisphenol A based on surface-enhanced Raman scattering, Analytica Chimica Acta 2013, 777, 57-62. [153] Ahmed Refat, N.A., Ibrahim, Z.S., Moustafa, G.G., Sakamoto, K.Q., Ishizuka, M., Fujita, S., The induction of cytochrome P450 1A1 by Sudan dyes, Journal of Biochemical and Molecular Toxicology 2008, 22, 77-84. [154] Di Donna, L., Maiuolo, L., Mazzotti, F., De Luca, D., Sindona, G., Assay of Sudan I contamination of foodstuff by atmospheric pressure chemical ionization tandem mass spectrometry and isotope dilution, Analytical Chemistry 2004, 76, 5104-5108. [155] Chiye, T., Takashi, O., Noriko, O., Hiroki, K., Kyoko, S., Hiroshi, A., Yoko, K., Extraction method and determination of Sudan I present in sunset yellow FCF by isocratic high-performance liquid chromatography, American Journal of Analytical Chemistry 2012, 2012. [156] Prosek, M., Golc-Wondra, A., Vovk, I., Andrensek, S., Quantification of caffeine by off-line TLC-MS, JPC-Journal of Planar Chromatography-Modern TLC 2000, 13, 452-456. [157] He, L., Su, Y., Shen, X., Zeng, Z., Liu, Y., Determination of Sudan dye residues in eggs by liquid chromatography and gas chromatography-mass spectrometry, Analytica Chimica Acta 2007, 594, 139-146. [158] Nierengarten, J.F., Dietrich-Buchecker, C.O., Sauvage, J.P., Synthesis of a doubly interlocked [2]-catenane, Journal of the American Chemical Society 1994, 116, 375-376. [159] Gusev, A.I., Vasseur, O.J., Proctor, A., Sharkey, A.G., Hercules, D.M., Imaging of thin-layer chromatograms using matrix-assisted laser desorption/ionization mass spectrometry, Analytical Chemistry 1995, 67, 4565-4570. [160] Kandler, H., Bleisch, M., Widmer, V., Reich, E., A validated HPTLC method for the determination of illegal dyes in spices and spice mixtures, Journal of Liquid Chromatography & Related Technologies® 2009, 32, 1273-1288. [161] Wang, S., Xu, Z., Fang, G., Duan, Z., Zhang, Y., Chen, S., Synthesis and characterization of a molecularly imprinted silica gel sorbent for the on-line determination of trace Sudan I in chilli powder through high-performance liquid chromatography, Journal of Agricultural and Food Chemistry 2007, 55, 3869-3876. [162] Xu, X.Y., Tian, X.G., Cai, L.G., Xu, Z.L., Lei, H.T., Wang, H., Sun, Y.M., Molecularly imprinted polymer based surface plasmon resonance sensors for detection of Sudan dyes, Analytical Methods 2014, 6, 3751-3757. [163] Ren, D., He, J., Zhang, H., Synthesis, characterization and evaluation of hollow molecularly imprinted polymers for Sudan I, Analytical Methods 2014, 6, 3079-3085. [164] Pichon, V., Selective sample treatment using molecularly imprinted polymers, Journal of Chromatography. A 2007, 1152, 41-53. [165] Kindschy, L.M., Alocilja, E.C., A molecularly imprinted polymer on indium tin oxide and silicon, Biosensors and Bioelectronics 2005, 20, 2163-2167. [166] Baggiani, C., Anfossi, L., Giovannoli, C., Solid phase extraction of food contaminants using molecular imprinted polymers, Analytica Chimina Acta 2007, 591, 29-39. [167] Allender, C., Heard, C., Brain, K., Mobile phase effects on enantiomer resolution using molecularly imprinted polymers, Chirality 1997, 9, 238-242. [168] Lopez, M.I., Ruisanchez, I., Callao, M.P., Figures of merit of a SERS method for Sudan I determination at traces levels, Spectrochimica acta. Part A, Molecular and Biomolecular Spectroscopy 2013, 111, 237-241.   153 [169] Di Anibal, C.V., Marsal, L.F., Callao, M.P., Ruisanchez, I., Surface enhanced Raman spectroscopy (SERS) and multivariate analysis as a screening tool for detecting Sudan I dye in culinary spices, Spectrochimica acta. Part A, Molecular and Biomolecular Spectroscopy 2012, 87, 135-141. [170] Prabakaran, E., Pandian, K., Amperometric detection of Sudan I in red chili powder samples using Ag nanoparticles decorated graphene oxide modified glassy carbon electrode, Food Chemistry 2015, 166, 198-205. [171] Yang, D., Zhu, L., Jiang, X., Electrochemical reaction mechanism and determination of Sudan I at a multi wall carbon nanotubes modified glassy carbon electrode, Journal of Electroanalytical Chemistry 2010, 640, 17-22. [172] Mo, Z., Zhang, Y., Zhao, F., Xiao, F., Guo, G., Zeng, B., Sensitive voltammetric determination of Sudan I in food samples by using gemini surfactant-ionic liquid-multiwalled carbon nanotube composite film modified glassy carbon electrodes, Food Chemistry 2010, 121, 233-237. [173] Zhou, M., Chen, X., Xu, Y., Qu, J., Jiao, L., Zhang, H., Chen, H., Chen, X., Sensitive determination of Sudan dyes in foodstuffs by Mn–ZnS quantum dots, Dyes and Pigments 2013, 99, 120-126. [174] Chen, J., Abell, J., Huang, Y.W., Zhao, Y., On-chip ultra-thin layer chromatography and surface enhanced Raman spectroscopy, Lab on a Chip 2012, 12, 3096-3102. [175] Feng, S., Gao, F., Chen, Z., Grant, E., Kitts, D.D., Wang, S., Lu, X., Determination of alpha-tocopherol in vegetable oils using a molecularly imprinted polymers-surface-enhanced Raman spectroscopic biosensor, Journal of Agricultural and Food Chemistry 2013, 61, 10467-10475. [176] Hu, Y., Feng, S., Gao, F., Li-Chan, E.C., Grant, E., Lu, X., Detection of melamine in milk using molecularly imprinted polymers-surface enhanced Raman spectroscopy, Food Chemistry 2015, 176, 123-129. [177] Gao, F., Feng, S., Chen, Z., Li-Chan, E.C., Grant, E., Lu, X., Detection and quantification of chloramphenicol in milk and honey using molecularly imprinted polymers: Canadian penny-based SERS nano-biosensor, Journal of Food Science 2014, 79, 2542-2549. [178] Freye, C.E., Crane, N.A., Kirchner, T.B., Sepaniak, M.J., Surface enhanced Raman scattering imaging of developed thin-layer chromatography plates, Analytical Chemistry 2013, 85, 3991-3998. [179] Brosseau, C.L., Gambardella, A., Casadio, F., Grzywacz, C.M., Wouters, J., Van Duyne, R.P., Ad-hoc surface-enhanced Raman spectroscopy methodologies for the detection of artist dyestuffs: thin layer chromatography-surface enhanced Raman spectroscopy and in situ on the fiber analysis, Analytical Chemistry 2009, 81, 3056-3062. [180] István, K., Keresztury, G., Szép, A., Normal Raman and surface enhanced Raman spectroscopic experiments with thin layer chromatography spots of essential amino acids using different laser excitation sources, Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 2003, 59, 1709-1723. [181] Li, D., Qu, L., Zhai, W., Xue, J., Fossey, J.S., Long, Y., Facile on-site detection of substituted aromatic pollutants in water using thin layer chromatography combined with surface-enhanced Raman spectroscopy, Environmental Science & Technology 2011, 45, 4046-4052.   154 [182] Lu, X., Rasco, B.A., Kang, D.H., Jabal, J.M.F., Aston, D.E., Konkel, M.E., Infrared and Raman spectroscopic studies of the antimicrobial effects of garlic concentrates and diallyl constituents on foodborne pathogens, Analytical Chemistry 2011, 83, 4137-4146. [183] Zhang, T., Liu, F., Chen, W., Wang, J., Li, K., Influence of intramolecular hydrogen bond of templates on molecular recognition of molecularly imprinted polymers, Analytica Chimica Acta 2001, 450, 53-61. [184] McGhee, J.D., von Hippel, P.H., Theoretical aspects of DNA-protein interactions: co-operative and non-co-operative binding of large ligands to a one-dimensional homogeneous lattice, Journal of Molecular Biology 1974, 86, 469-489. [185] Wang, C., Cheng, F., Wang, Y., Gong, Z., Fan, M., Hu, J., Single point calibration for semi-quantitative screening based on an internal reference in thin layer chromatography-SERS: the case of Rhodamine B in chili oil, Analytical Methods 2014, 6, 7218-7223. [186] Black, J.W., Duncan, W., Durant, C.J., Ganellin, C.R., Parsons, E., Definition and antagonism of histamine H2-receptors, Nature 1972, 236, 385-390. [187] Collado, J., Tunon, I., Silla, E., Ramírez, F., Vibrational dynamics of histamine monocation in solution: an experimental (FT-IR, FT-Raman) and theoretical (SCRF-DFT) study, Journal of Physical Chemistry A 2000, 104, 2120-2131. [188] Taylor, S.L., Eitenmiller, R.R., Histamine food poisoning: toxicology and clinical aspects, CRC Critical Reviews in Toxicology 1986, 17, 91-128. [189] Silva, C.C., Ponte, D.J., Dapkevicius, M.L.E., Storage temperature effect on histamine formation in big eye tuna and skipjack, Journal of Food Science 1998, 63, 644-647. [190] Taylor, S.L., Speckhard, M.W., Isolation of histamine-producing bacteria from frozen tuna, Marine Fisheries Review 1983, 45, 6. [191] Guizani, N., Al-Busaidy, M.A., Al-Belushi, I.M., Mothershaw, A., Rahman, M.S., The effect of storage temperature on histamine production and the freshness of yellowfin tuna (Thunnus albacares), Food Research International 2005, 38, 215-222. [192] Feng, C., Teuber, S., Gershwin, M.E., Histamine (scombroid) fish poisoning: a comprehensive review, Clinical Reviews in Allergy & Immunology 2015, 1-6. [193] Pessatti, T.L., Fontana, J.D., Pessatti, M.L., Spectrophotometric determination of histamine in fisheries using an enzyme immunoassay method,  Public Health Microbiology, Springer 2004. [194] Cinquina, A., Longo, F., Calı, A., De Santis, L., Baccelliere, R., Cozzani, R., Validation and comparison of analytical methods for the determination of histamine in tuna fish samples, Journal of Chromatography A 2004, 1032, 79-85. [195] Tahmouzi, S., Khaksar, R., Ghasemlou, M., Development and validation of an HPLC-FLD method for rapid determination of histamine in skipjack tuna fish (Katsuwonus pelamis), Food Chemistry 2011, 126, 756-761. [196] Rogers, P.L., Staruszkiewicz, W., Gas chromatographic method for putrescine and cadaverine in canned tuna and mahimahi and fluorometric method for histamine (minor modification of AOAC Official Method 977.13): collaborative study, Journal of AOAC International 1996, 80, 591-602. [197] Sequaris, J.M.L., Koglin, E., Direct analysis of high-performance thin-layer chromatography spots of nucleic purine derivatives by surface-enhanced Raman scattering spectrometry, Analytical Chemistry 1987, 59, 525-527.   155 [198] Frattini, V., Lionetti, C., Histamine and histidine determination in tuna fish samples using high-performance liquid chromatography: derivatization with o-phthalaldehyde and fluorescence detection or UV detection of “free” species, Journal of Chromatography A 1998, 809, 241-245. [199] Cohen, G., Rudnik, D.D., Laloush, M., Yakir, D., Karpas, Z., A novel method for determination of histamine in tuna fish by ion mobility spectrometry, Food Analytical Methods 2015, 8, 2376-2382. [200] Tong, A., Dong, H., Li, L., Molecular imprinting-based fluorescent chemosensor for histamine using zinc (II)-protoporphyrin as a functional monomer, Analytica Chimica Acta 2002, 466, 31-37. [201] Trikka, F.A., Yoshimatsu, K., Ye, L., Kyriakidis, D.A., Molecularly imprinted polymers for histamine recognition in aqueous environment, Amino Acids 2012, 43, 2113-2124. [202] Trikka, F., Yoshimatsu, K., Ye, L., Kyriakidis, D., Molecularly imprinted polymers that selectively recognize histamine, The FEBS Journal 2008, 275, 373-373. [203] Dai, J., Zhang, Y., Pan, M., Kong, L., Wang, S., Development and application of quartz crystal microbalance sensor based on novel molecularly imprinted sol–gel polymer for rapid detection of histamine in foods, Journal of Agricultural and Food Chemistry 2014, 62, 5269-5274. [204] Horemans, F., Alenus, J., Bongaers, E., Weustenraed, A., Thoelen, R., Duchateau, J., Lutsen, L., Vanderzande, D., Wagner, P., Cleij, T., MIP-based sensor platforms for the detection of histamine in the nano-and micromolar range in aqueous media, Sensors and Actuators B: Chemical 2010, 148, 392-398. [205] Mayes, A.G., Mosbach, K., Molecularly imprinted polymers: useful materials for analytical chemistry?, TrAC Trends in Analytical Chemistry 1997, 16, 321-332. [206] Stiles, P.L., Dieringer, J.A., Shah, N.C., Van Duyne, R.P., Surface-enhanced Raman spectroscopy, Annual Review of Analytical Chemistry 2008, 1, 601-626. [207] Camden, J.P., Dieringer, J.A., Wang, Y., Masiello, D.J., Marks, L.D., Schatz, G.C., Van Duyne, R.P., Probing the structure of single-molecule surface-enhanced Raman scattering hot spots, Journal of the American Chemical Society 2008, 130, 12616-12617. [208] Gao, F., Feng, S., Chen, Z., Li-Chan, E.C., Grant, E., Lu, X., Detection and quantification of chloramphenicol in milk and honey using molecularly imprinted polymers: Canadian penny‐based SERS nano-biosensor, Journal of Food Science 2014, 79, 2542-2549. [209] Gao, F., Hu, Y., Chen, D., Li-Chan, E.C., Grant, E., Lu, X., Determination of Sudan I in paprika powder by molecularly imprinted polymers-thin layer chromatography-surface enhanced Raman spectroscopic biosensor, Talanta 2015, 143, 344-352. [210] Lee, P., Meisel, D., Adsorption and surface-enhanced Raman of dyes on silver and gold sols, Journal of Physical Chemistry 1982, 86, 3391-3395. [211] Kim, K.B., Han, J.H., Choi, H., Kim, H.C., Chung, T.D., Dynamic preconcentration of gold nanoparticles for surface-enhanced Raman scattering in a microfluidic system, Small 2012, 8, 378-383. [212] Gaillard, C., Strauss, F.o., Ethanol precipitation of DNA with linear polyacrylamide as carrier, Nucleic Acids Research 1990, 18, 378. [213] Michaels, S., John, M., Amasino, R., Removal of polysaccharides from plant DNA by ethanol precipitation, Biotechniques 1994, 17, 274-276. [214] Feinberg, A.P., Vogelstein, B., A technique for radiolabeling DNA restriction endonuclease fragments to high specific activity, Analytical Biochemistry 1983, 132, 6-13.   156 [215] Alvarez-Puebla, R.A., Arceo, E., Goulet, P.J., Garrido, J.J., Aroca, R.F., Role of nanoparticle surface charge in surface-enhanced Raman scattering, Journal of Physical Chemistry B 2005, 109, 3787-3792. [216] Davis, K.L., McGlashen, M.L., Morris, M.D., Surface-enhanced Raman scattering of histamine at silver electrodes, Langmuir 1992, 8, 1654-1658. [217] Turiel, E., Martin-Esteban, A., Molecularly imprinted polymers: towards highly selective stationary phases in liquid chromatography and capillary electrophoresis, Analytical and Bioanalytical Chemistry 2004, 378, 1876-1886.   

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0223172/manifest

Comment

Related Items