PROTON N M R STUDIES OF H U M A N L U M B A R INTERVERTEBRAL DISCS Teresa Rind B. Sc. (Physics) University of Victoria A THESIS SUBMITTED IN PARTIAL F U L F I L L M E N T OF T H E REQUIREMENTS FOR T H E D E G R E E OF MA S T E R OF SCIENCE in T H E FACULTY OF GRADUATE STUDIES DEPARTMENT OF PHYSICS We accept this thesis as conforming to the required standard T H E UNIVERSITY OF BRITISH COLUMBIA October 1990 © Teresa Rind In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Physics The University of British Columbia 1956 Main Mall Vancouver, Canada Date: Abstract Problems with the low back resulting from diseases of the lumbar disc account for much human suffering and medical expense. Through early identification of changes in the disc, Magnetic Resonance Imaging offers a potential method for recognition of those at risk of low back trouble. An understanding of the chemical and structural basis of the MR image will therefore be of great help in understanding the degenerative mechanism itself. With this in mind, NMR studies of human intervertebral discs were performed to extract data that could be correlated with different stages of degeneration. In vitro samples of anulus fibrosus and nucleus pulposus of grades II and IV discs were exam-ined. The Carr-Purcell-Meiboom-Gill (CPMG) pulse sequence was used to obtain T 2 relaxation data, which were analysed with a non-negative least squares algorithm and presented as a continuous spectrum of exponentials. An inversion-recovery sequence was used to obtain T i data which were analysed with the same algorithm, then pre-sented as a discrete exponential spectrum. The same procedures were used to produce relaxation times from samples of isolated disc collagen and proteoglycan. Also, T 2 re-laxation data were obtained with a twelve-echo CPMG imaging sequence from healthy volunteers, and these data analysed with the same procedure used for in vitro data. In vivo and in vitro results were compared and found to agree in terms of the T 2 relaxation values. Most change between grades occurred in the nucleus, where both Ti and T 2 values decreased from grade II to grade IV. The results of the proteoglycan and collagen studies support the theory that structure, not chemical composition, is responsible for the observed changes in relaxation times. Based on estimates of the percentages of protons in the nucleus due to water, proteoglycan and collagen, tentative ii assignments were made to the components of the T 2 spectra for grades II and IV nucleus, and these were supported by the results of all parts of the study. iii Acknowledgements I would first like to thank my supervisor Alex MacKay, who made this project possible. He provided guidance and encouragement, and shared his expert knowledge of the workings of the Room 100 magnet and spectrometer. He also set up and developed the partnership with Richard Pearce's group in the UBC Department of Pathology, found many valuable reference papers, offered advice on this thesis, and obtained all of the MRI CPMG data. I enjoyed working with Dick Pearce, Joyce Mathieson, and Gwen Bebault from the Department of Pathology at the UBC Hospital, and greatly appreciate all the time and thought they contributed. Specifically, I would like to thank Dick for sharing his extensive knowledge of the intervertebral disc and offering suggestions on the direction of the project, as well as providing all of the samples and many valuable reference papers. I would also like to thank Joyce who prepared in vitro disc samples and taught me how to prepare the collagen and proteoglycan samples; and Gwen who shared thoughtful insights on the study of degenerative discs. The crew in Room 100 have been very supportive and provided help in solving day-to-day problems with the equipment and computing system. Thanks to Clare Morrison, our system manager, the computing system in the lab has been running well and has enabled me to do experimental work and writing without interruption. Also, the relaxation results would not have been possible without Ken Whittall's NNLS program. Finally, I would like to thank Julia Wallace for showing me the ropes in Room 100. iv Table of Contents Abstract ii Acknowledgements iv List of Tables vii List of Figures viii 1 Introduction 1 2 The Intervertebral Disc 3 2.1 Composition of the Disc 5 2.2 Disc Degeneration 7 2.3 Disc Grading Scheme 9 3 Proton N M R Theory 12 3.1 Basic NMR and Relaxation Theory 12 3.2 NMR of Water in Biological Samples 17 4 Application of N M R to Intervertebral Discs 23 4.1 Advantages of Magnetic Resonance 23 4.2 Purpose of This Work 24 5 Experimental Methods 25 5.1 Experiments Performed 25 v 5.2 Sample Preparation 27 5.3 Data Collection 27 5.3.1 T 2 and the CPMG Sequence 28 5.3.2 Ti and the Inversion Recovery Sequence 29 5.3.3 In Vivo CPMG 30 5.4 Data Analysis 30 5.4.1 Analysis of CPMG Data 31 5.4.2 Analysis of Inversion-Recovery Data 32 6 Results 37 6.1 T 2 and Tx of Disc Samples 37 6.2 In Vivo MRI T 2 Values 46 6.3 Solid Signal in Disc Samples 51 6.4 Ti and T 2 of Collagen Samples 51 6.5 T x and T 2 of Proteoglycan Samples 53 7 Discussion 55 Bibliography 61 vi List of Tables 5.1 Disc Samples 26 5.2 Collagen Samples • • • 26 5.3 Proteoglycan Samples 26 6.1 T 2 Values for Nucleus and Anulus of Disc Samples 38 6.2 Ti Values for Nucleus and Anulus of Disc Samples 38 6.3 Average T 2 Values for Grades II and IV Anulus and Nucleus 40 6.4 Average Tj Values for Grades II and IV Anulus and Nucleus 41 6.5 T 2 Values of In Vivo MRI Samples 46 6.6 Average In Vivo T 2 Values for Each Lumbar Disc 47 6.7 Ti and T 2 Values for Collagen Samples 53 6.8 Ti and T 2 Values for Proteoglycan Samples 53 7.1 Tentative Assignments to the T 2 Spectrum 59 vn List of Figures 2.1 The Lumbar Spine and The Intervertebral Discs 4 2.2 Components of the Disc 4 2.3 Disc Grading System by Morphology and MRI 10 5.1 Examples of Discrete and Continuous NNLS Spectra 34 5.2 NNLS Spectra With Different Values of Facmu 35 5.3 Example of NNLS Fit to CPMG Data 36 6.1 Typical T2 Spectra for In Vitro Disc Samples 39 6.2 T2 Values for In Vitro Disc Samples — Nucleus 42 6.3 T 2 Values for In Vitro Disc Samples — Anulus 43 6.4 Ti Values for In Vitro Disc Samples — Nucleus 44 6.5 Ti Values for In Vitro Disc Samples — Anulus 45 6.6 T 2 Values for In Vivo MRI Samples 48 6.7 Graph of Mean MRI T 2 Values for Each Disc 49 6.8 Histogram of T 2 Values for MRI Samples 50 6.9 Solid Signal in Disc Samples 52 6.10 Proteoglycan Ti and T 2 Relaxation Rates Versus Concentration . . . . 54 viii Chapter 1 Introduction Problems with the spinal column in humans are widespread, costly, and serious. Sixty to eighty percent of all adults have experienced back pain at some time in their lives, with most of these symptoms occurring in the lower portion of the spine, the lumbar region. The most common disorders giving rise to the pain are instability and misalignment, disc degeneration, spinal stenosis, and facet disease. Accompanying the pain are sensory changes and weakness in the body. The social impact of these phenomena is striking when one considers that back trouble leads to many worker's compensation claims, time lost from work, and great hospital costs. In fact, degenerative disorders of the spine are the leading cause of disability in the working years; over any twelve month period approximately 25% of working men will have back trouble. In addition, these disorders are among the most common causes of impairment in general, and account for a major portion of medical expenses. [1,2] Most scientists believe that lower back pain is often due to disc degeneration, as alterations in disc structure may cause pain and impairment by affecting nerves and the spinal cord. Extensive research has been conducted in the epidemiology, anatomy, biomechanics, biochemistry and neuromechanics of the disc and spine, but the patho-physiology of disc degeneration and the role of biochemical changes are poorly under-stood, and complex. Autopsy results show that by the fifth decade of life, 85-95% of the population have degenerate discs. Possible contributing factors to disc degeneration are thought to be genetic makeup and biomechanics. Indeed, disc failure occurs in the 1 Chapter 1. Introduction 2 areas of the spine that undergo the most mechanical stress; the role of genetics is not so easy to observe or define. [1,2,3] What is known is that degeneration begins early in life and occurs in everyone, to different extents among individuals [2,4]. A method of early detection of problem discs would help in preventing unnecessary injury and pain; a better understanding of the causes and mechanisms of disc degeneration would be the first step in finding an effective prevention and/or cure. Magnetic resonance may be the solution to both of these needs. MRI could be used to non-invasively detect early degenerate discs, and knowledge of the theory behind the images could lead to an understanding of the degenerative process. Chapter 2 The Intervertebral Disc Intervertebral discs play a remarkable role in stabilizing the spine by anchoring the vertebrae to one another, allowing a fair amount of movement of the spine, absorbing loads, and always regaining their original size and shape. Discs vary in size, from the smallest cervical discs in the neck region, to the largest lumbar discs in the lower back (Figure 2.1). They contain special types of cartilage which are noted for their strength and resilience. The tensile strength comes from collagen fibrils, and the resilience is due to large proteins called proteoglycans. In this type of tissue, it is the extracellular material that is important, rather than the cells themselves. In fact, the disc is mostly water and'extracellular matrix, including some of the largest proteins made by any cell in the body. [3,5] The body of a healthy disc is composed of two distinct regions: the inner gelatinous nucleus pulposu3 surrounded and contained by the fibrous rings of the anulus fibrosus. Outside the disc body are the cartilage endplates and vertebral bodies, as illustrated in Figure 2.2. The nucleus pulposus (NP) is a proteoglycan-rich matrix interspersed with a rela-tively small number of randomly oriented collagen fibres. Cell density in this region is only 4000 cells/mm3, being mostly chondrocytic and fibroblastic in type, with a smaller number of notochordal cells. [1,3] The anulus fibrosus (AF) is made up of 15-20 concentric layers of collagen fibres, organized so that the fibres of adjacent laminae are at 40-90° to one another. The cells 3 Chapter 2. The Intervertebral Disc 4 Intervertebral discs Intervertebral foramina Sacrum Lumbar curvature Figure 2.1: The lumbar spine and the intervertebral discs. This drawing of the lumbar spine shows the positioning of the intervertebral discs between the vertebrae. (From Human Anatomy by K. M. Van De Graaff, Wm. C. Brown Publishers, Iowa, 1988. pl71.) anulus End-plate Figure 2.2: The components of the healthy intervertbral disc. The drawing illustrates the anulus fibrosus, nucleus pulposus, endplates and surrounding vertebral bodies. Chapter 2. The Intervertebral Disc 5 of the outer anulus are fibrocytic while the inner cells are chondrocytic. [1,3] Nutrient transport in the disc is unique, due to the fact that only the outermost layers contain blood vessels, lymphatics and nerves. In fact, discs are the largest volumes of tissue in the body without nerves and vessels. Nutrients reach the cells in the disc by moving through the matrix from the blood vessels at the edges. Some transport is achieved by entrainment in moving fluid under changing loads, and another means is active transport by chondrocytes, but the main mode of nutrient travel through the matrix is diffusion. Thus, the maximum concentration of any nutrient or solute in the matrix may be controlled by the local proteoglycan content. [3,5,6] 2.1 Composition of the Disc Collagen Collagen is found throughout the disc, in various fibril concentration, di-ameter, and organization. Ten types of collagen have been found in the disc, with types I and II being the most abundant, followed by type VI. Type I, found mostly in the anulus, increases from 0-80% abundance towards the outer layers, while type II collagen increases 0-80% in the opposite direction, being found mostly in the nucleus. Type VI is found in both regions, 10% in the AF and 15-20% in the NP. In adult lum-bar discs, the dry mass in the outer anulus fibrosus is 60-70% collagen, and 10-20% collagen in the nucleus pulposus. As far as structure goes, the fibrils in the outer anulus are arranged in tight concentric layers at angles to one another; the inner anulus fibrils are less dense and less organized; and in the nucleus they form a loose network with no organization. [2,3] The general structure of a collagen fibril starts with procollagen, the precursor. This is a protein with about 1100 amino acids, every third one being glycine and one third of the others being proline or hydroxyproline. The procollagen molecule twists into C h a p t e r 2. T h e Intervertebral D i s c 6 a compact left-handed helix; three of these form a right-handed superhelix, a rod-like macromolecule approximately 1.4 nm in diameter and about 300 nm long. Groups of these superhelixes form the fibrils, which are from 10 to 200 nm in diameter, and of varyied length. They possess great tensile strength as a result of the multiple twisting and grouping. [5] Proteoglycan Like collagen, proteoglycans are present throughout the disc. They are similar to the proteoglycans of articular cartilage in many ways, but differ in several respects that may be important in the degenerative process. [1] Different regions of the disc contain one or more of five distinct species of pro-teoglycans, though the relationships amongst these species are not yet defined. One population of disc proteoglycans form large macromolecular aggregates while the rest do not, and the proportions of these two groups differ between anulus and nucleus. Their relative amounts also change with aging: some elaborate PG aggregates are found in newborns' discs but not in mature spines. Another grouping of the proteoglycans is in terms of their components. One family is rich in keratan sulfate while another contains large amounts of chondroitin-6 sulfate. These repeating disaccharides are two of the main constituents of proteoglycans in general, as described below. [3,7,8,9] The central molecule of a proteoglycan is the repeating disaccharide hyaluronic acid, to which are attached amino acid chains called core proteins. The attachment is achieved without chemical bonding, by use of a link protein and a good fit with a globular shape formed from the end of the core protein. Each core protein then has many polysaccharide chains attached to it. Three regions are defined on the core protein by virtue of what is attached: (1) the linkage region containing the globular end and a few small sugar chains; (2) a region having keratan sulfate chains and small o-linked oligosaccharides; and (3) a region with chondroitin sulfate chains. There are Chapter 2. The Intervertebral Disc 7 approximately 50 keratan sulfate chains and 100 chondroitin sulfate chains on a core protein, and there may be up to 100 core proteins on a central strand of hyaluronic acid when large aggregates form. The overall weight of such an aggregate may be many tens of millions of atomic units. [5,6] Water Acting with collagen and proteoglycans to give discs their unique proper-ties is water, which makes up about 70% of the healthy disc. The repeating disac-charides chondroitin sulfate and keratan sulfate contain sugars which carry negative charges, forming a large ordered array of electronegativity. Since water is a dipole, H 2 O molecules become organised in layers around these proteoglycan charges. When the disc is under pressure from loading, the water is forced away from the negative charges and they become closer together. Electrostatic repulsion prevents further com-pression; as soon as the load is reduced the H 20 returns to its former arrangement and the structure is restored. Thus water plays a major role alongside proteoglycans in providing the disc with resilience. The movement of fluids also helps to nourish the tissue as explained in Section 2. [3,5] While a proteoglycan molecule can structure an amount of water many times its own weight, this is not the only way H 20 is stored in the disc: some is trapped in the interstices of the proteoglycan and collagen matrix. The structured water, however, is reponsible for the important disc properties. [5] 2.2 Disc Degeneration As the disc ages or degenerates, structural and chemical changes occur affecting all components of the tissue. As the disc becomes more fibrous and disorganised, the distinction between anulus and nucleus becomes less clear. All collagen fibrils increase in diameter and more chondrocytic cells appear, while proteoglycan and water content Chapter 2. The Intervertebral Disc 8 decrease. Water loss runs to 30%, with the majority lost from the nucleus. Disc volume can shrink dramatically, in some cases from 15 cc to 1 cc, causing secondary effects from increased forces at the facet joint, such as misalignment, arthritic changes and potential fractures. [1,2,3,4] Some of the water loss is related to changes in-the proteoglycan molecules; a de-crease in the molecular "weight of proteoglycan aggregates in the nucleus leads to a decrease in water-structuring capacity. Not only do the aggregates lose size, but there are fewer of them, and the remaining proteoglycans become smaller. The small scale chemical structures of the proteoglycans do not change, but the relative amounts of the components do. For example, the ratio of chondroitin sulfate to keratan sulfate in-creases in some parts of the disc and decreases in others. Also, a diminished association with collagen results in some loss of tensile strength. [2,3] Data seem to indicate that proteoglycan is lost from all discs of a spine before any morphologically recognizable degeneration takes place. This loss of proteoglycan is thought by some to stem from a long term decrease in nutrient supply. This may be partly due to the fact that less vessels penetrate the anulus as the disc ages. It may also be related to an accumulation of large macromolecular fragments which the disc cannot remove, resulting in abnormal biomechanical and biophysical tissue properties. The cellular feedback mechanism would become confused and the matrix would not function properly or maintain itself. This is one possible explanation of degeneration. [1,3,4,8] The extracellular matrix does seem to have a limited ability for resynthesis and re-pair, though it is not known if the repair is good enough to restore long term function. Most restructuring occurs in the anulus with the collagenous material being resynthe-sized most easily; repair in the nucleus is minimal. [3] Chapter 2. The Intervertebral Disc 9 One of the main problems in finding an explanation for disc degeneration is that the relationships have not been well defined between age-related changes in the disc and the development of clinical syndromes or change in function. Normal aging and abnormal degeneration are difficult to separate. [3] 2.3 Disc Grading Scheme A disc grading scheme has been devised by J. P. Thompson and associates [3, page 193] [10,11] for assessment by gross morphology and Magnetic Resonance Imaging. The scheme delineates five stages of degeneration based upon the appearance of mid-sagittal sections of the lumbar disc. Features affecting the classification are the nucleus pulposus, anulus fibrosus, cartilaginous and bony end-plates, and the periphery of the vertebral body. Figure 2.3 illustrates the appearances of the five grades by morphology and MRI. Grade I discs (found in the young) are identified by a bulging gel nucleus, a distinct fibrous anulus with discrete layers of collagenous fibres, and a uniformly thick hyaline endplate. In the MR T2-weighted spin echo image with TE = 90 ms, the nucleus appears uniformly bright and the anulus uniformly dark grey, with the endplate appearing as a single dark line. Grade II discs (the healther discs examined in this project) still have a bulging nucleus with gel at its centre, but white fibrous tissue appears around the edges, and the anulus contains mucinous material between the fibrous laminae. The endplate is no longer uniform in thickness. MRI shows some loss of brightness in the central horizontal plane of the nucleus in the form of dark bands, and light areas in the anulus. The infiltration of fibrous tissue into the nucleus and the invasion of mucinous ma-terial into the anulus increase through grades III and IV, with the clear demarcation Chapter 2. The Intervertebral Disc 10 Figure 2.3: The morphologic and MRI grading system. Photographs of actual disc slices are shown at left, and MR images of corresponding grade are shown at right. (Courtesy of R. H. Pearce) Chapter 2. The Intervertebral Disc 11 between nucleus and anulus dissolving at the grade III level. In a grade IV nucleus there are also horizontal clefts parallel to the endplate, and the anulus contains fo-cal disruptions. In the MR image the nucleus and anulus are indistinguishable, with increasing regions of bright and dark signal and decreasing grey areas. Focal defects appear in the line of the endplate in the grade IV disc. Finally, the grade V disc is severely degenerate, with clefts extending through the remaining tissue (nucleus and anulus are indistinguishable), and diffuse sclerosis in the endplate. The MR image reveals defects and thick areas in the endplate, extreme loss of overall disc height, and dominant bright and dark signals in the disc body. Chapter 3 Proton N M R Theory A brief overview of basic NMR theory leading to the Ti and T 2 relaxation processes is contained in the following section, which has been summarized from C. P. Slichter's Principles of Magnetic Resonance [12]. For more details please refer to this source. The second part of this section describes briefly some relevant aspects of the NMR of water in biological tissues, including water structuring, proton transfer, cross-re-laxation, and compartmentation. 3.1 Basic N M R and Relaxation Theory A nucleus has total magnetic moment fl and total angular momentum J related by p = 7j (3.i) Define a dimensionless operator / such that J = %I and it follows that (Im\fj,xi\Im') = yh(Irn\Ixi\Im') where m and m' are spin quantum numbers. If one applies a field H, the interaction energy of the nucleus will be —jl-H. The Hamiltonian is thus H = -fi-H Taking the field to be H0 along the z-direction gives a Hamiltonian of Ti = —yhH0Iz 12 Chapter 3. Proton NMR Theory 13 The eigenvalues are thus multiples of the eigenvalues of Iz: for / = 1/2, we have E = -j%H0m, m = -1/2, +1/2 To detect these energy levels, one needs an interaction that causes transitions between these levels, i.e. one needs AE = hu where io is an angular frequency. In NMR one uses an alternating magnetic field applied perpendicular to a static field Ho. The perturbing term in the Hamiltonian is Hvtrt = — jhH®Ixcosujt where H® is the amplitude of the alternating field and u is the frequency of that field. Ix has matrix elements between states m and m' which disappear unless m' — m i l . Allowed transitions are between adjacent levels: huj = AE = ~/h,H0 so that CJ = -yHQ Thus we can find u needed to observe a resonance, if we know 7. To estimate 7, a classical picture is used, by computing the magnetic moment and angular momentum of a charged particle moving in a circular path. Then by comparing the expressions for fj. and J and considering equation 3.1, we get 7 = e/2mc, where e is the charge of the particle, m is the mass, and c is the speed of light. A general result of this is that large masses have low 7's. For magnetic fields around 2 Tesla, hydrogen nuclei (protons) resonate at about 90 MHz. Chapter 3. Proton NMR Theory 14 Spin-Lattice Relaxation: Consider nuclei of spin 1/2, each with two states, m = +1/2, —1/2, called |+) and |—) . Some nuclei will be in the |+) state, and some in the |—) state at a given time. Call the number in the |+) state N+, and the number in the |—) state JV_. To see that there must usually be a coupling of the spins to another system (the lattice), consider the magnetization of an unmagnetized piece of material: a net number of transitions from upper to lower state is required, so the spins give up energy, which is accepted by another system, a reservoir (the "lattice"). Let W | be the probability per second that the coupling will induce a spin transition from |+) to |—) (higher energy) and let W[ be the probability for the reverse process. W\ and W[ are unequal, seen by considering what is happening in the coupled system (Slichter, page 8). Then if n is defined as N+ — 7V_, the population difference, and N is the total population, using dN+ we get where and The solution is , = NJWl- N+W1 dt dn no — n It ~~ Tj W[ - W]\ (3.2) n0 = N W[ + WT n = n 0 + Ae~t,Tl where A is a constant of integration, no is the population difference at thermal equi-librium, and Ti is a characteristic time related to the approach to equilibrium. T a is called the "spin-lattice relaxation time". Chapter 3. Proton NMR Theory 15 Effect of Alternating Field Hi( t ) : An alternating field can be considered as two rotating components of equal amplitude, rotating in opposite directions. One rotates in the same sense as the precession of the spins, but the other, going the opposite way, will not be "seen" by the spins. A spin sees only Hi(t) — Hi(icostozt + jsinuzt). The equation of motion for a spin in fields Ho and H\ (t) is * = / ? x 7 p o + ii ( 0 ] One can eliminate the time dependence by working in a rotating coordinate system (rotating about the z-axis with frequency uz). Then both H0 and Hi will be time independent. If we set up the x-axis along Hi, we have | = / i X + iHo) + HHI Near resonance, u>z + jH0 m 0 (set LOz = —u = —jH0). The above equation can also be written as where Heff =k (H0-u/y) + iHi So in the rotating system, the moment will precess in a cone about Htg at angular frequency ^Heg. At exact resonance (u> = 7^0) we have Htg = Hit, and fl will precess in the y-z plane. Application of Hi then for a time tp (applying a "pulse") will cause the moment to precess through an angle e = 7H1 tp One can choose Hi and tp to produce any angle 6; for example, a "180° pulse" is when Chapter 3. Proton NMR Theory 16 Thus to put spins into the x-y laboratory plane, one must apply a ninety degree pulse (6 = 7r/2); in the lab frame, spins will then precess about the z-direction. This is detected through an induced emf in the same coil used to apply the pulse. Interactions of spins with each other and with the lattice cause decay of the signal (emf) detectable in the x-y plane. This decay is called the "Free Induction Decay" (free of H\) and may be exponential in nature, though this is not generally the case. Bloch Equations for Free Induction Decay: Knowing that Mz = 7?in/2 and using equation 3.2, one expects that dMz _ MQ-MZ dt ~ Tx where MQ is the magnetization at thermal equilibrium. Also, the x and y components of the magnetization are reduced over time due to spin-spin interactions, and this decay is represented by the following equations: dMx Mx ~dT = dMy _ My ~dT ~ ~ ~T^ where T 2 is introduced as a different relaxation time than T i . This is because while the longitudinal (Ti) decay involves a change of energy (transferred to the lattice), the transverse decay conserves energy. The exponential form of the transverse decay is arbitrary but is useful to describe some important phenomena. It is important to note that the processes causing T 2 relaxation are irreversible processes. Spin Echoes: For liquids, whose T 2 values are on the order of tenths of a second or seconds, inhomogeneities in the Ho field can dephase spins at a rate higher than the transverse relaxation rate, thus "covering up" the real T 2 value which one wants to observe. To eliminate the effect of the magnet inhomogeneity, spin echoes are used. Chapter 3. Proton NMR Theory 17 A spin echo is created by applying a 90° pulse to put the magnetization into the x-y plane, followed by a 180° pulse a time r afterwards. Before the 180° pulse, the spins will have dephased due to the inhomogeneities; some will be "ahead" and some will be "behind", by the angle 6 = fSHr where 8H = H — H0, the difference in local field from H0. After the 180° pulse, the spins that were ahead will now be behind, and vice versa, so that after another time r they will line up again. At this time, 2r from the initial 90° pulse, a spin echo will be observed as an isolated peak in the detected transverse magnetization amplitude. Note that during the period 2r the effects of T 2 relaxation will continue unaffected by the 180° pulse, so that the amplitude of the echo signal will be M(2r) = M0e~2T^2 for single exponential relaxation behaviour. 3.2 N M R of Water in Biological Samples Most or all of an NMR or MRI signal from a biological sample is due to protons of water, so understanding the behaviour of this water is vital to any analysis of relaxation results. Some signal may come from the proton-rich solutes which are a source of relaxation events, but this is fairly insignificant compared to the contribution of water protons. [15] States of Water in Tissues: Water in biological tissues behaves differently from bulk water, in that it is structured around macromolecules as described in Section 2.1. A small fraction of the H 20 molecules interact with slow-moving constituents such Chapter 3. Proton NMR Theory 18 as proteins and cell walls via electrostatic interactions and hydrogen bonding, form-ing a well-defined hydration layer in solutions, hydrated fibres such as collagen, and all biological samples such as intracellular H2O and tissues. The hydration layer is characterized by water molecules that are oriented and restricted in motion, thus with different properties from bulk water. [13,14] Most fat-free tissues can be considered to contain three water fractions, based on their location and behaviour: bound water consists of H 20 molecules hydrogen-bonded to fixed sites on macromolecules; structured water is motionally perturbed by the macromolecules but not bonded to them; and bulk water's motion is solely determined by the interactions between H 20 molecules. Each fraction has its own relaxation rates, governed by the motion of the molecules, proton transfer processes, and spin exchange. [15] Interactions between H 2 0 and macromolecules: The most important source of relaxation time shortening in tissues is due to changes in the motion of water molecules near the macromolecular surface. This restricted motion in the hydration layer is not uniform, leading to anisotropic diffusion of H 20 molecules. Macromolecules with dif-ferent protein content also have different water-perturbing capacities, so the behaviour of hydration water depends strongly on the nature of the hydrated species. [13,15] Magnetic dipole-dipole coupling is the important interaction in proton water studies, and this is affected by molecular motions such as rotation, which varies the orientation of local fields, and translation. Magnetic interactions are partially averaged by the rotations, and when water molecules diffuse through regions of different orientations. Averaging also occurs when proton transfer takes place, and this process is enhanced along a continuous hydrogen-bonded path in the hydration layer of a macromolecule. [13,15] Chapter 3. Proton NMR Theory 19 Relaxation rates in tissues: True T 2 decay is caused by random dephasing of the x-y magnetization due to rapid molecular tumbling motion and the resulting changing magnetic field. There is also a contribution to T 2 caused by dephasing from a "static" component of the magnetic coupling between proton pairs: large macromolecules rotate slowly on the NMR timescale, appearing to be motionless from the NMR point of view. The relatively short T 2 in tissues is partly a result of this static dephasing. [15] In general, tissue relaxation rates are affected by: (1) the strength of local mag-netic interactions between nuclei, (2) molecular motion (anisotropic rotation and trans-lation), (3) water perturbing capabilities of the macromolecules, (4) proton transfer rates, (5) diffusion rates, (6) water content, and (7) spin exchange or cross-relaxation rates [13,15]. The first three items affect intrinsic relaxation rates, while the remain-ing four characteristics affect the observed rates, which may differ from the intrinsic rates. Compartmentation is based on the assumption that sites of different relaxation characteristics exist and the protons at these sites interact with each other in various ways [14,15]. For example, the protons of different sites may be exchanged (proton transfer) or water molecules from one compartment may diffuse to other regions with different local relaxation rates. Water content affects the observed relaxation rates by determining the ratio of bulk water to structured water, both of which have different intrinsic rates. Restricted water motion near the macromolecules gives the structured water a smaller T 2 value [13]. When water content decreases and a larger proportion of water is structured, the observed T 2 value is shortened. Cross-relaxation (spin exchange): Cross-relaxation between protein and solvent protons also makes an important contribution to the relaxation rates of hydration water. Spin exchange occurs when the motion of protons from the hydration layer and protons from macromolecules is correlated. Under this condition, there will be a Chapter 3. Proton NMR Theory 20 mutual exchange of spin magnetization between these protons. When T 2 , e x p ( - i / T 2 t ) t=i where S(t) is the fitted signal amplitude at time t, N is the number of exponentials used in the fit (120 in this project), A{ is the weighting factor for the iih component, and T2l-is the ith T 2 value in the spectrum. A discrete or continuous fit could be chosen. A set of N T 2 times was created prior to running the program, logarithmically covering a range wide enough to include all possible transverse relaxation times for the materials being studied. For the disc samples, a scale from 10~3s to 10°s was adequate, while a wider range of times (to 10s) was needed for the proteoglycan and collagen samples. Examples of both discrete and continuous fits are shown in Figure 5.1. The continu-ous fit was chosen for all T 2 analysis as this gave unambiguous results. For this type of fit, an adjustable parameter called "facmu" controlled the amount of curvature allowed in the spectrum by weighting the first derivative. The smaller the value of facmu, the less defined were the spectrum peaks, and vice versa. Figure 5.2 illustrates the effect Chapter 5. Experimental Methods 32 of changing facmu from 100 to 500. In the study, facmu was chosen so that each peak in the spectrum was distinct, if not separated, from its neighbours. An indication of this was the symmetry of a peak; i.e., how close the weighted average T 2 value was to the value with the largest amplitude (at the top of the peak). T 2 results quoted in this report are the weighted average values, though as stated above these were very close to the maximum amplitude values. The goodness of fit was indicated by the x 2 value calculated by the NNLS program. A good fit should have x 2 close to the number of degrees of freedom (n/); i.e. in this case the number of points being fit. This number depended on the number of echoes in the top 99% of the signal, and this number fell within the range 90-190. The value of x 2 depended greatly on the estimated noise, or baseline standard deviation. In most cases, x2/nf ^ e n between 0.9 and 1.2, while sometimes just one or two errant points in the early data could be enough to blow up x 2 • Fit to the data was checked by plotting the fit along with the data points, and this always showed an excellent match (Figure 5.3). Changing facmu did not alter x 2 dramatically. 5.4.2 Analysis of Inversion-Recovery Data The raw Ti data was transformed into an array of values (M(0) — M(r)) versus r, by a program written by A. MacKay. For each of the eleven different r value data sets, the program subtracted the baseline and output the average intensity for a data window following the 90° pulse. A noise estimator was also calculated from the baselines of all eleven data sets, by finding their standard deviation. With the inversion-recovery data represented as a decreasing function of time (in the form 2M2(r=0)e"T/Tl) they could be analysed with the NNLS program in a similar procedure to the T 2 analysis. For the Ti data, however, the relaxation was nearly single exponential, and the discrete fit was chosen with no ambiguity of results. This Chapter 5. Experimental Methods 33 allowed for easy extraction of the Ti value by looking at the spectrum, and without the bother of adjusting facmu. The Ti time scale for the 120 exponential fit covered the range 10-3s to 101s. The \ 2 goodness of fit was expected to be near 11 (the number of degrees of freedom) but was often much lower due to overestimation of noise. Plotting the calculated data with the actual data showed an excellent fit as in the case of T 2 analysis. Chapter 5. Experimental Methods 34 Figure 5.1: Examples of N N L S spectra, fitted to C P M G data. (a) A discrete spectrum, (b) A continuous spectrum. The continuous spectrum has less ambiguity for multiexponential relaxation, and does not differ between data sets from the same sample. Chapter 5. Experimental Methods 35 Figure 5.2: The effect of changing facmu on an N N L S continuous spectrum. (a) Facmu=100. (b) Facmu=500. Facmu is a variable pa-rameter that controls the amount of curvature allowed in the spectrum, by weighting the first derivative. More peaks are visible and each one is narrower when facmu is larger. Chapter 5. Experimental Methods 36 Figure 5.3: A plot of an NNLS fit to C P M G data. The actual data are shown as "o" symbols and the calculated signal is a solid line; the two are a close match. Chapter 6 Results 6.1 T 2 and T a of Disc Samples Table 6.1 lists all the T 2 values and their percent contributions to the continous spectra obtained for the nucleus pulposus and anulus fibrosus samples of grades II and IV discs. With the exception of HD8N, results for r = 200/JS and r = AOOus were very close, so both tau value results are shown for HD8N. Some typical T 2 spectra are also shown in Figure 6.1 to illustrate the origin of the results in Table 6.1. All T-i results are summarized in Table 6.2, which lists the components of the discrete spectra and their percent contributions. Figures 6.2-6.5 are plots of the results in Tables 6.1 and 6.2, with the dominant contributions (above 20%) shown as solid symbols and the lesser components as empty symbols. For the case of HD8N, the tau value whose fitted x2/nf w a s closest to one was chosen for plotting in Figure 6.2. T 2 — Nucleus Pulposus: Figure 6.2 clearly shows a difference in T 2 values between grade II and grade IV nucleus samples. The dominant contributions (solid symbols) for grade II mostly lie above 100 ms while the T 2 values for grade IV fall in the range 20-70 ms. The only overlap is from sample 5 where there are two components over 20%, the lesser one (31%) at 55 ms and the larger one (49%) at 156 ms; and sample 9, where there are two components over 20% also, the smaller at 220 ms and the larger at 69 ms. Sample 9 was also the one for which the r = 200/15 and r = AOOus results 37 Chapter 6. Results Table 6.1: T 2 Values for Nucleus and Anulus of Disc Samples Sample AF/NP Grade T2 Values (ms) and % Contributions 1. HD1N NP II 5.5 (2%) 36 (8%) 105 (23%) 255 (67%) 2. HD3N NP II 6.4 (4%) 38 (12%) 101 (84%) 3. HD5N NP II 4.6 (2%) 20 (5%) 101 (93%) 4. HD7N NP II 2.5 (1%) 15 (4%) 156 (95%) 5. HD9N NP II 4.5 (5%) 15 (15%) 55 (31%) 156 (49%) 6. HD2N NP IV 3.6 (8%) 23 (92%) 7. HD4N NP IV 2.4 (2%) 12 (7%) 47 (90%) 498 (1%) 8. HD6N NP IV 9 (2%) 42 (96%) 247 (1%) 9. HD8Na NP IV 1.7 (2%) 18 (9%) 69 (52%) 220 (37%) 9. HD8Na NP IV 9 (5%) 104 (95%) 10. HD10N NP IV 5.4 (2%) 49 (98%) 11. HD12N NP IV 4.8 (2%) 49 (98%) 1. HD1A AF II 2.9 (14%) 13 (51%) 43 (33%) 392 (2%) 2. HD3A AF II 4.0 (12%) 15 (41%) 38 (47%) 3. HD5A AF II 2.9 (22%) 10 (52%) 29 (26%) 4. HD2A AF IV 1.7 (1%) 11 (6%) 64 (91%) 562 (2%) 5. HD4A AF IV 3.5 (22%) 12 (57%) 39 (21%) 6. HD6A AF IV 4.8 (10%) 22 (90%) "HD8N is the only sample for which CPMG results from the two different tau values did not agree, so both are shown here, in order of increasing r. Table 6.2: T i Values for Nucleus and Anulus of Disc Samples Tx Values (s) #. Sample AF/NP Grade and % Contributions 1. HD1N NP II 0.62 (2%) 1.4 (98%) 2. HD3N NP II 0.69 (34%) 1.5 (66%) 3. HD5N NP II 0.25 (5%) 1.6 (95%) 4. HD2N NP rv 0.64 (100%) 5. HD4N NP TV 0.83 (89%) 1.2(11%) 6. HD6N NP IV 0.75 (98%) 3.5 (2%) 1. HD1A AF II 0.69 (83%) 1.3 (17%) 2. HD3A AF II 0.47 (51%) 1.2(49%) 3. HD5A AF II 0.70 (93%) 1.4 (7%) 4. HD2A AF IV 1.10 (92%) 1.9 (8%) 5. HD4A AF IV 0.78 (95%) 1.5 (5%) 6. HD6A AF IV 0.69 (100%) Chapter 6. Results 39 Figure 6.1: Typical T 2 spectra for in vitro disc samples. (a) Grade II nucleus, (b) Grade IV nucleus, (c) Grade II anulus. (d) Grade IV anulus. Facmu=500 in all cases. Chapter 6. Results 40 did not agree. Thus this sample may be suspect (due to mis-grading or an undetected problem with the data collection that day) and the overlap between grades II and IV only occurs for sample 5. Sample 5 had the largest contributing component in the range above 100 ms, so the problem may originate with the arbitrary cut-off point of 20% for a "dominant" T 2 component. Every sample had a short T 2 component with less than 20% contribution to the signal, regardless of grade, though the grade II samples have a consistent second small component between 15 and 40 ms that is only present in two of the grade IV samples. Table 6.3 lists the average T 2 values for grades II and IV, for the dominant components and the lesser components. Table 6.3: Average T 2 Values for Grades II and IV Anulus and Nucleus AF/NP Grade Contribution 4f of Samples Mean T 2 (ms) Std. Dev. (ms) NP II > 20% 5 138 49 NP IV > 20% 5 42 11 NP II shortest 5 4.7 1.5 NP IV shortest 5 4.5 2.6 AF II > 20% 3 23 6 AF IV > 20% " 3 35 25 T 2 — Anulus Fibrosus: Unlike the nucleus samples, there was no significant dif-ference between the dominant contributions to T 2 relaxation of grades II and IV, as seen in Figure 6.3 and Table 6.3. Due to this observation, further studies of the anulus fibrosus were discontinued in order to devote more time to the nucleus pulposus, where most change occurs with degeneration. The nucleus is also the region examined most closely in MRI grading procedures as it usually dominated the image. As with the nucleus pulposus, all anulus samples had a short T 2 component of small amplitude, with the relaxation time falling between 1 and 5 ms. Generally, the Chapter 6. Results 41 variation between samples is greater for grade IV than for grade II, though one must take into account the small number of samples when evaluating such trends. Ti — Nucleus Pulposus: The T] values follow a similar pattern to the T 2 results, with a large decrease from grade II to grade IV. Table 6.4 lists the average T a times for the dominant contributions to the spectrum; note the factor of two difference between grades. The division is clearer between grades than it is for T 2 , and no overlap occurs between the two groups. The Tj relaxation is simpler than the T 2 , having only one or two components in the spectra. This is expected for biological systems, as discussed in section 3.2. Table 6.4: Average Tj Values for Grades II and IV Anulus and Nucleus AF/NP Grade Contribution # of Samples Mean T x (s) Std. Dev. (s) NP II > 20% 3 1.41 0.20 NP IV >20% 3 0.74 0.10 NP II <20% 2 0.44 0.26 NP IV <20% 2 2.40 1.60 AF II >20% 3 0.74 0.08 AF IV >20% 3 0.86 0.21 AF II < 20% 2 1.35 0.07 AF IV <20% 2 1.70 0.30 Ti — Anulus Fibrosus: The Ti values for the anulus are the same as those of the nucleus, except that the relative proportions of contributions to the spectrum are reversed: the dominant components have shorter Tj values, and the small components have longer relaxation times. Table 6.4 shows the average Ti values for large and small components of the spectra for grades II and IV disc anuli. Not all of the samples were found to have two components, or a large and small component — some had two contributions above 20%, for example. Chapter 6. Results 42 300 250 200 H T , 2 150H (ms) 100 n 50 0 o o LEGEND Gr. II Gr. IV O <20% A <20% • 2^0% A 2^0% ^ O O ? ? ? • A 2 3 4 5 Sample A 6 7 8 9 10 11 - Nucleus Figure 6.2: Plots of T 2 values for in vitro disc nucleus samples. Grade II and grade IV nucleus are displayed together (see leg-end). All components greater than or equal to 20% are shown as solid symbols, and the rest as empty symbols. Chapter 6. Results 43 T 2 (ms) 70 60 50 H 40 30 -2 0 -1 0 -0 i LEGEND Gr. II Gr. IV • <20% 0 <20% • ^20% • ^20% • • 1 2 3 Sample • • 4 5 6 Anu lus Figure 6.3: Plots of T 2 values for in vitro disc anulus samples. Grade II and grade IV anulus are displayed together (see leg-end). A l l components greater than or equal to 20% are shown as solid symbols, and the rest as empty symbols. Note that the vertical scale is much smaller than for the nucleus samples. Chapter 6. Results 44 (s) 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 i LEGEND Gr. II Gr. IV O <20% A <20% • 2^0% A 2^0% 1 2 3 Sample A • 4 5 6 Nucleus Figure 6.4: Plots of T i values for in vitro disc nucleus samples. Grades II and IV nucleus are displayed together (see legend). Symbols have the same meaning as in the corresponding T 2 plot. Chapter 6. Results 45 T . 2.5 2.0 1.5 H (s) 1.0-0 .5-i L E G E N D G r . II G r . IV • <20% 0 <20% • 2^0% • £20% • • 0.0 1 2 3 4 5 6 Sample — Anu lus Figure 6.5: Plots of T i values for in vitro disc anulus samples. Grades II and IV anulus are displayed together (see legend). Symbols have the same meaning as in the corresponding T 2 plot. Note the scale difference from the nucleus T i plots. Chapter 6. Results 46 Table 6.5: T 2 Values (in ms) of In Vivo MRI Samples Volunteer # Lumbar Disc Names: L1-L2 L2-L3 L3-L4 L4-L5 L5-S1 1 134 131 126 110 143 2 43 50 57 44 27 3 90 90 81 40 83 4 118 120 56 35 47 5 128 108 129 148 57 6 25, 97 85 122 119 157 7 76 58 46 58, 181 86 8 44 44 49 52 57 9 94 151 51, 159 60 144 10 136 135 178 129 111 11 129 152 137 33, 168 41 12 93 40, 136 48 83, 255 68 13 124 124 131 35, 175 104 14 87 62, 234 65 78 49 15 69, 197 55, 139 62 55, 313 55, 147 16 93 36,117 104 98 46 17 116 87 52 52 41 18 104 69 62 73 104 19 175 104, 395 117 35, 156 69 6.2 In Vivo MRI T 2 Values A l l of the in vivo M R I T 2 values for the lumbar disc nuclei of the 19 volunteers stud-ied are displayed in Table 6.5. In most cases where two components existed, they contributed fairly equally to the spectrum, so the actual percentages are not shown. Spectra were less complicated than those for in vitro samples because only twelve data points were available for fitting with the NNLS program. Several of the volunteers were found to have self-consistent T 2 values among discs, such as volunteer #1 (average T 2 was 129 ms with a standard deviation of 12 ms), #2 (average 44 ms, standard deviation 11 ms), #8 (49 ms; 6 ms) and #13 (118 ms; Chapter 6. Results 47 Table 6.6: Average In Vivo T 2 Values for Each Lumbar Disc Disc: L1-L2 L2-L3 L3-L4 L4-L5 L5-S1 Mean T 2 & Std. Error (ms) 104 ± 8 101 ± 8 91 ± 9 95 ± 1 0 81 ± 9 12 ms). This phenomenon can be seen in the first plot of Figure 6.6, which is a series of plots displaying all T 2 components found for each type of lumbar disc. The mean T 2 values for each disc type are shown in Table 6.6. The weighted average T 2 values for two-component spectra were used in the calculations. Figure 6.7 provides a graphic illustration of those results, showing a trend towards shorter T 2 values from L1-L2 to L5-S1. The error bars represent the standard error of the mean and are slightly larger for the lower lumbar discs. Figure 6.8 is a histogram plot of all the T 2 components found among the M R I volunteers, and shows clearly that there is a division of the values into two groups. The first group is centered around 50 ms, while the second is approximately centered on 120 ms. These numbers are not in disagreement with the average dominant T 2 values of in vitro nucleus samples found in Table 6.3. Chapter 6. Results 48 400H 300-J I I L L 1 - L 2 (•) L 2 - L 3 (*) (ms) 200-* * • 100- 1 D • * a • a o * *1 * • * 0-400- _i i_ T n i i i r i i i i i J 1 l I I l 1_ L 4 - L 5 A * A * ^ A ^ L 3 - L 4 300-(ms) 200-100- o o 0 ° ° o o o o ° ° o 0 —i 1 1 1 1 1 1 1 r-2 4 6 8 10 12 14 16 18 S a m p l e (disc) -i 1 1 1 1 1 1 1 r i i i i i i i i i L 5 - S 10 0 « * 0 o o 0 o 0 1 1 I 1 I I I I I 2 4 6 8 10 12 14 16 18 S a m p l e (d isc) Figure 6.6: T 2 values for MRI samples. Results for each disc are shown, with L1-L2 and L 2 - L 3 overlaid to show similarity between discs of the same spine. The same vertical scale is used for all plots. Chapter 6. Results i 1 1 r L1-L2 L2-L3 L3-L4 L4-L5 L5-S1 Disc Figure 6.7: Graph of mean MRI T 2 values for each disc. This plot demonstrates the shortening of T 2 values from L1-L2 to L5-S1. Error bars represent the standard errors of the means. Chapter 6. Results 50 0 50 100 150 200 250 300 350 4 0 0 4 5 0 T 2 ( m s ) Figure 6.8: Histogram of MRI T 2 values. A l l discs are included in this histogram, which demonstrates the grouping of T 2 values into two ranges. Chapter 6. Results 51 6.3 Solid Signal in Disc Samples The Ti inversion-recovery data show the signal decay at times immediately after the 90° pulse. For very short tau values one may sometimes observe a rapidly relaxing component which can be attributed to protons of solids in the sample. Figure 6.9 shows inversion-recovery-data from samples of grade II and IV nucleus, with r = 1 ms. The grade II sample showed very little solid component (3.8%), while 8.5% of the signal from the grade IV sample was due to solid protons. This was calculated by separating the two decay slopes and interpolating back to t = 0. Although not shown, inversion-recovery data from anulus samples had a solid component making up 12% of the signal at t — 0, for both grades of disc. 6.4 T i and T 2 of Collagen Samples Only one collagen sample from each of grades II and IV discs was examined, and these results are listed in Table 6.7. There did not appear to be any definite pattern to the relaxation times in this small sample group, but a few general observations were made. The Ti values for the collagen were very similar for all samples (grades II and IV anulus and nucleus), but the grade IV samples had a longer T x component that was not present with the grade II samples. The Ti values for grade II anulus and nucleus were identical. As far as T 2 went, the grade IV relaxation times were longer than those for grade II samples. A longer T 2 timescale was used for NNLS fits to accomodate the slow rate of relaxation. Chapter 6. Results 52 co • i-l cd 40 30 H 20 10 0 cd 35 0 30-25-£ 20 B 1 5 < 10 0 (a) ~50 100 150 200 250 300 ^ ( b ) f 0 50 100 150 200 250 300 Time (/xs) Figure 6.9: Inversion-recovery data showing solid signal for in vitro disc samples. (a) Grade II nucleus pulposus. (b) Grade IV nucleus pulposus. The first 50/is is baseline. The solid portion of the signal shows up as the steeper slope immediately after the pulse. Chapter 6. Results 53 Table 6.7: T x and T 2 Values for Collagen Samples Sample AF/NP Grade Ta Values (s) and % Contributions T 2 Values (ms) and % Contributions C1N NP II 1.5(100%) 76 (37%) 183 (63%) C2N NP IV 1.3 (81%) 3.1(19%) 327 (100%) CIA AF II 1.5(100%) 9 (1%) 99 (99%) C2A AF IV 1.0 (64%) 2.1(36%) 80 (72%) 1180(28%) 6.5 T i and T 2 of Proteoglycan Samples Both Ti and T 2 relaxation for proteoglycan samples in saline were single or almost single exponential. Table 6.8 lists the relaxation times. As proteoglycan concentration decreased, relaxation times increased for both anulus and nucleus proteoglycan, with the Ti and T 2 values shorter for anulus than nucleus at similar concentrations. Fig-ure 6.10 illustrates the changes of relaxation rates with proteoglycan concentration; the relationship is roughly linear for the three concentrations studied. In all cases, the T 2 values are much longer than those of the disc samples. Table 6.8: T i and T 2 Values for Proteoglycan Samples Concentration T2 Sample AF/NP (mg/ml) 00 (ms) PGN1 NP 155.3 2.00 666 PGN2 NP 74.0 2.42 1100 PGN3 NP 49.8 3.20 1340 PGA1 AF 143.6 1.42 440 PGA2 AF 71.6 1.74 666 PGA3 AF 50.1 — 750 Chapter 6. Results 54 0.9 1 0.2 • Nucleus PG A Anulus PG • (a) . • • • 30 50 70 90 110 130 150 170 PG Concentration (mg/ml) 3.0 2.6 2.2 1.8 1.4 1.0 0.6 H 0.2 • Nucleus PG A Anulus PG ( b ) 30 50 70 90 110 130 150 170 PG Concentration (mg/ml) Figure 6.10: Graphs of proteoglycan relaxation rates versus concen-tration in saline solution. (a) 1 /Ti versus concentration for anulus and nucleus P G . (b) I/T2 versus concentration for anulus and nucleus P G . Chapter 7 Discussion Correlation of relaxation times with grade: The results of the in vitro disc studies show that relaxation times correlate with disc grade. T 2 and T i values of the nucleus change significantly with grade, decreasing by approximately a factor of two between grades II and IV. Furthermore, the MRI T 2 values decrease with distance down the lumbar spine, which would be expected if T 2 were inversely proportional to disc grade: the lower discs of the lumbar region are generally the most degenerate. Another fact supporting the notion that T 2 is a sensitive indicator of degeneration is that the results for the grade IV nucleus look the same as those of both grades of anulus. This ties in with the fact that as discs degenerate, the nucleus becomes more and more like the anulus until they are indistinguishable, as noted in Section 2.2. Comparison of in vitro and in vivo relaxation times: The in vivo T 2 values agree with those found in vitro, being grouped into two ranges of relaxation times that match the grades II and IV in vitro T 2 values. These results agree despite the differ-ences in sample temperature and NMR frequency during data collection. The in vitro sample data were collected at approximately 23°C while the in vivo samples remained at 37°C. An increased temperature would be associated with faster thermally activated processes and increased diffusion rates. Correlation times decrease with increased ther-mal activity, but this should not affect T 2 unless slow motions are present. The lack of dependence of T 2 on CPMG tau value suggests that slow motions do not play a large 55 Chapter 7. Discussion 56 role in relaxation. The change in diffusion rate would shorten T 2 if the relaxation were diffusion-limited, but here that does not seem to be the case, as the T 2 times agree for room temperature and body temperature. As for frequency differences, the in vitro ex-periments were carried out at 90 MHz, while the MRI data were collected at 6.4 MHz. This would not be expected to affect T 2 values since the fundamental mechanism for transverse relaxation is dipole-dipole, which is independent of field strength. Ti on the other hand should be much shorter at the MRI frequencies compared to 90 MHz, and this is the case: the shortest in vitro T x value is 740 ms, while in vivo values are typically 370 ms [19]. Distribution of in vivo T 2 values: The two peaks in the distribution of in vivo relaxation times are relatively wide (Figure 6.8, indicating the disparity among indi-viduals' T 2 values. Separate consideration of the data for each subject suggests that relative, not absolute, evaluations of T 2 data would be most useful from a clinical point of view — often one disc in a spine will have a much lower T 2 value than any of the others, while another spine will have these "low" T 2 values for every disc, even though no other evidence of degeneration appears. For an example of this, compare samples 5 and 8 in Table 6.5. The L5-S1 disc of sample 5 has a T 2 value of 57 ms which is half as long as the values for the other discs, yet the discs of sample 8 all have T 2 values equal to or lower than 57 ms. Proteoglycan and collagen studies: The changes in T 2 with grade are probably related more to structure than to chemical composition, as noted in section 4.1. The results of the proteoglycan studies support this theory: though proteoglycan appears to be lost from the disc during degeneration, simply decreasing the concentration of isolated proteoglycan does not decrease the T 2 value; rather, it increases, contrary to Chapter 7. Discussion 57 the behaviour in discs when pg content is lowered. This is not unexpected, however, because when less proteoglycan is in solution, the water behaves more like bulk water, and the T 2 times should increase. These results are in agreement with a recent study by Weidenbaum et. al [20]. Collagen studies also demonstrate this contradictory be-haviour, with T 2 values being longer for grade IV collagen extract than for grade II. As suspected, studies of these macromolecules in isolation cannot be directly applied to disc studies. Instead, one must consider the structure of the cartilage as a whole. If proteoglycan aggregates and their subspecies are fragmenting with degeneration, this may shorten T 2 by reducing the volume of bulk water between molecules, i.e. decreasing the mean diffusion path for a water molecule. The decreasing water content would add to this effect. In addition, collagen increases in the nucleus with degeneration, reducing bulk water and increasing the amount of structured water. Details of T 2 spectra: The components of the T 2 spectra are at different T 2 values for grades II and IV, rather than having peaks at the same T 2 values with different relative contributions to the total area. This rules out a simple model of compartments relaxing separately, with relative volumes changing during degeneration. The only common component of the spectrum is the very short (less than 10 ms) T 2 peak, which contributes about 5%. The following calculation roughly estimates the percentages of protons in the water, proteoglycan and collagen of the grade II nucleus, for comparison with the components of the T 2 spectra. First of all, the healthy disc is about 70% water. Collagen makes up 10-20% of the dry weight in the nucleus (Section 2.2), so the remaining 80-90% would be proteoglycan. This means that of the total wet weight, there would be about 70% water, 4.5% collagen, and 25.5% proteoglycan. Water has 2 protons per 18 atomic units, or 0.11 protons per unit weight. Rough estimates of the number of protons Chapter 7. Discussion 58 per unit weight for the proteoglycan and collagen are 0.12 and 0.067, respectively. So, multiplying these fractions by the weight percentages of water, pg and collagen in the nucleus gives the following percentages of protons: 78% from water, 17% from proteoglycan, and 5% from collagen. For a grade IV nucleus, the percentage from water would decrease, as water content drops by 30%; the percentage from proteoglycan would be increased by water loss and decreased due to proteoglycan loss; and the percentage from collagen would increase both from decreased water content and increased collagen content. The solid component of the spectrum as shown in Figure 6.9 accounts for 3-4% of the signal in the grade II nucleus, so it seems likely to originate from the protons of the collagen. This theory is supported by two results: the solid component increases to 8-10% in grade IV nucleus, and collagen is known to increase with degeneration; also, the solid signal in the anulus was the largest at 12%, which also agrees with the known chemistry of the disc. Further evidence comes from the slope of the solid signal, which increases from grade II to grade IV, suggesting a more rigid structure is responsible for the signal; collagen fibrils are known to increase in diameter with degeneration. The shortest T 2 component in the grade II and grade IV nucleus spectra con-tributes 1-7% and thus could be attributed to bound water or proteoglycan, with cross-relaxation possibly playing a role. The second short component in the grade II nucleus spectra ranges from 4-15% contribution, with T 2 values from 15 to 38 ms, and this might be due to the protons of the proteoglycan, which are estimated at 17% abundance. The large components of the grade II spectra most likely originate from the bulk and hydration water. For the grade IV nucleus, the largest component of the spectra has T 2 values in the range 23-50 ms and no third T 2 component, so it appears that for grade IV, the proteoglycan and bulk water protons are indistinguishable in terms of T 2 values. Table 7.1 lists tentative assignments of all components of the T 2 Chapter 7. Discussion 59 Table 7.1: Tentative Assignments to the T 2 Spectrum for Disc Nucleus Constituent Assignment Grade II Proportion of Protons Grade IV Proportion of Protons Collagen TJ, 20-40/zs (solid signal) 3-4% 8-10% Bound water or proteoglycan T 2 , 1-6 ms 1-5% 2-7% Proteoglycan or water T 2 , 15-50 ms 4-14% 82-89% Water T 2 , >100 ms 48-92% Total=100%, including solid spectra for the disc nucleus, grades II and IV. Cross-relaxation between some or all of these constituents is probable and leaves several models open for investigation. Details of Tj spectra: The Tx spectra differ from the T 2 spectra in that they do have common values that occur in different relative contributions. Different processes seem to be at work for Ti relaxation, probably because cross-relaxation processes would have more time to act on a T x timescale. The large difference between Tj and T 2 indicates that fast motions with correlation times less than 1/90 MHz are of importance to the relaxation processes. The complex T 2 spectra are probably more useful for studying details of the degenerative mechanism, though the Ti results are necessary to complete the picture. Future work: More accurate assignment of the components of the T 2 spectrum to specific molecules and processes in the disc will require further experiments with in vitro and in vivo disc samples. First of all, more disc samples should be measured, both in vitro and in vivo, to gain a larger sample of the population from which robust statistics can be extracted. As well, the same samples should undergo chemical analysis Chapter 7. Discussion 60 so that the exact amounts of water, proteoglycan and collagen are known for each sample. This would allow precise comparisons between chemistry and NMR data, and contrast between samples of different grades. Histological analysis would also be valuable in correlating disc structure with the Ti and T 2 results. Meanwhile, in vitro data collection should be attempted at 37°C on samples that have not been previously frozen, to eliminate the possibility of structural differences between in vivo and in vitro tissues. Conclusions: This study has formed the framework for continuing research regarding the application of magnetic resonance to the intervertebral disc. The results of this investigation indicate that relaxation measurements, in particular T 2 , are sensitive to disc degeneration and could be used as the main tool for studying the details of the degenerative mechanism. The T i and T 2 spectra, the in vivo T 2 values, and the solid signal measurements strongly support the suggested roles of collagen, proteoglycan and water in NMR measurements of the intervertebral disc. Future work will be based on verifying and expanding the model proposed here. Bibliography [1] R. H. Pearce, J. P. Thompson, B. Flak, A. MacKay. "Biochemical and Radiologic Studies of Degeneration in the Human Intervertebral Disc". A research proposal to the Arthritis Society, December 1989. [2] M. T. Modic in Magnetic Resonance Imaging of the Spine. Chapter 3: "Degener-ative Disorders of the Spine", pp75-119. Yearbook Medical Publishers, Chicago, London, Boca Raton, 1989. [3] D. Eyre, P. Benya, J. Buckwalter, B. Caterson, D. Heinegard, T. Oegema, R. Pearce, M. Pope, J. Urban. New Perspectives on Low Back Pain. Part B: Basic Science Perspectives. Eds. J. W. Frymoyer, S. L. Gordon. Park Ridge, IL. American Academy of Orthopaedic Surgeons, ppl47-207, 1989. [4] R. H. Pearce, B. J. Grimmer, M. E. Adams. "Degeneration and the Chemical Composition of the Human Lumbar Intervertebral Disc". J. Orth. Research, 5:198-205, 1987. [5] A. I. Caplan. "Cartilage". Sci. Am., 251:84-94, 1984. [6] E. D. Harris, Jr. Scientific Bases of Rheumatology Chapter 16: "Biology of the Joint." (Futher information unavailable) [7] J. J. DiFabio, R. H. Pearce, B. Caterson, H. Hughes. Biochem. J., 224:27-33, 1987. 61 B i b l i o g r a p h y 62 [8] R. H. Pearce, B. J. Grimmer. "The Concentration of Proteoglycan Fractions in Healthy Young Human Lumbar Intervertebral Discs". Preprint, 1989. [9] G. M. Bebault, R. H. Pearce. "The Assay of Keratan Sulfate as Anion-Exchanger-Bound Hexose". Preprint, October 1989. [10] J. P. Thompson. "Classifications of Gross Morphologic and Magnetic Resonance Images of Human Intervertebral Discs". M.Sc. Thesis, UBC, 1987. [11] J. P. Thompson, R. H. Pearce, M. T. Schechter, M. E. Adams, I. K. Y. Tsang, P. B. Bishop. "Preliminary Evaluation of a Scheme for Grading the Gross Morphology of the Human Intervertebral Disc". Submitted to Spine, October 1989. [12] C. P. Slichter. Principles of Magnetic Resonance ppl-42. Springer-Verlag, Berlin, Heidelberg, New York, 1980. [13] R. Mathur-De Vre. "The NMR Studies of Water in Biological Systems" Proc. Biophys. Mol. Biol, 35:103-134, 1979. [14] P. S. Belton, R. G. Ratcliffe. "NMR and compartmentation in Biological Tissues" Progress in NMR Spectroscopy, 17:241-279, 1985. [15] G. D. Fullerton in Magnetic Resonance Imaging. Chapter 3: "Physiologic Basis of Magnetic Relaxation", pp36-55. C. V. Mosby Company, St. Louis, Washington, D. C. Toronto, 1988. [16] K. R. Brownstein, C. E. Tarr. "Importance of classical diffusion in NMR studies of water in biological cells" Phys. Rev. A., 19:2446-2453, 1979. [17] E. Sternin. Ph.D Thesis, UBC, 1988. pp34^5. [18] K. P. Whittall, A. L. MacKay. J. Magn. Resonance, 84:64, 1989. Bibliography 63 [19] A. MacKay, private communication. [20] M . Weidenbaum, R. J . Foster, B. A. Best, E . L. Nickoloff, A. Ratcliffe, R. Jelsma, J . H . Newhouse, V . C. Mow. "Magnetic resonance imaging studies of water and proteoglycan content of the intervertebral disc". p338, 36th Annual Meeting, Or-thopaedic Research Society, New Orleans, Louisiana, February 5-8, 1990.