RESEARCH ARTICLE Open AccessProgrammed cell death ligand 1 cut-pointis associated with reduced disease specificsurvival in resected pancreatic ductaladenocarcinomaBasile Tessier-Cloutier1,2, Steve E. Kalloger2,5,7,9* , Mohammad Al-Kandari1, Katy Milne6, Dongxia Gao3,Brad H. Nelson6,8, Daniel J. Renouf3,5,7, Brandon S. Sheffield9 and David F. Schaeffer1,2,4,5AbstractBackground: Programmed cell death 1 (PD1) inhibitors have recently shown promising anti-cancer effects in anumber of solid tumor types. A predictive biomarker to this class of drugs has not been clearly identified; however,overexpression of the PD1 ligand (PD-L1) has shown particular promise in lung adenocarcinoma. In this study, weexplore the staining characteristics, prevalence, and clinico-molecular correlates of PD-L1 overexpression inpancreatic ductal adenocarcinoma (PDAC).Methods: A tissue microarray (TMA) was constructed from cases of resected PDAC. PD-L1 immunohistochemistry(IHC) was performed using the SP142 primary antibody. Immunohistochemical assessment for deficient mismatchrepair status (MMRd), CD3 and CD8 were performed. All biomarkers were assessed independently by twoanatomical pathologists and consensus achieved on all cases. Survival analysis was performed using three thresholds(> = 1%, >5% and >10%) for tumor cell membrane staining.Results: Two-hundred fifty-two cases were included in the TMA and evaluable by IHC. Thirty-one (12%), 17 (7%),12(5%) cases were positive at percentage cut offs of >0, >5, and >10% respectively. Increased PD-L1 expression wasassociated with inferior prognosis (p = 0.0367). No statistically significant association was identified between PD-L1status and MMR status or tumor infiltrating lymphocytes.Conclusions: This data suggests that there is an inverse relationship between PD-L1 expression and disease specificsurvival times in resected PDAC. Consequently, this association may represent a phenotype where increased PD-L1expression has an effect on tumor biology and could therefore identify a subgroup where PD1 blockade could haveenhanced effectiveness.Keywords: Pancreatic cancer, Programmed cell death 1 ligand, DNA mismatch repair, Tumor-infiltrating lymphocytes,Biomarkers, Immuno-oncology* Correspondence: skalloger@mac.comDaniel J. Renouf and David F. Schaeffer co-supervised this work.Basile Tessier-Cloutier and Steve E Kalloger contributed equally to this work.2Department of Pathology & Laboratory Medicine, University of BritishColumbia, Vancouver, British Columbia, Canada5Pancreas Centre BC, Vancouver, British Columbia, CanadaFull list of author information is available at the end of the article© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Tessier-Cloutier et al. BMC Cancer (2017) 17:618 DOI 10.1186/s12885-017-3634-5BackgroundPancreatic ductal adenocarcinoma (PDAC) ranks fourthfor overall cancer-related death with over forty-thousandestimated deaths in 2015 in the United States. The five-year survival rate is 26% in resectable disease and drops to2% if unresectable. Surgical resection is only attempted in20% of cases [1].Inhibitors of the programmed cell death 1 (PD1) sig-nalling axis have yielded improved survival benefits for anumber of solid tumor types. Large randomized clinicaltrials have been successful in treating melanoma, non-small cell lung cancer, and renal cell carcinoma [2, 3].Three phase 1/2 drug trials are ongoing involving treat-ment of PDACs with immunotherapy (NCT02583477,NCT02305186, NCT02452424). To date, no biomarkerhas been established to predict benefit from PD1-axis in-hibition for this disease [4].PD-1 is an inhibitory receptor expressed by T cells andother immune cell types. It plays an important role in im-mune suppression when activated by its ligand (PD-L1).The latter is physiologically expressed by normal tissueand can occasionally be aberrantly expressed by tumorcells as a means for evading immune destruction [4–8].Blockade of the PD-1/PD-L1 interaction promotes T-cellresponse against tumor cells [3, 9].The response to PD-1/PD-L1 inhibition has beenmixed in various malignancies such as colorectal, pros-tatic and pancreatic adenocarcinomas and is exemplifiedin the results of a study by Brahmer et al. which failed toshow an objective response to anti-PD-L1 therapy in 14patients with pancreatic cancer [3, 10]. In those cases,the use of biomarkers may have been useful in the iden-tification of patients who are more likely to respond toPD1-axis inhibition. Mismatch repair (MMR) status hasbeen shown to be predictive in colorectal carcinoma [11]and PD-L1 expression by immunohistochemistry (IHC)may be useful in lung and bladder carcinomas [12, 13].However, no cut-off has been uniformly defined for PD-L1 expression that would trigger the use of PD-L1 inhib-itors in PDAC. Current clinical trials often use 1% [14]but evidence suggests that higher cut-points mayoptimize patient stratification for PD-L1 therapies [15].PD-1 expression in tumor infiltrating immune cells, thedirect target of nivolumab, has shown, unlike tumor PD-L1 expression, only borderline association with clinicaloutcome to PD-1 blockade [16]. Other methods to pre-dict response to immune checkpoint inhibitors have alsobeen investigated, including immune cell infiltration,hypermutation signature, and gene expression linked tochemokine expression [17–19], but are yet to be vali-dated in prospective clinical trials.In this study, we explore the prevalence of PD-L1 ex-pression in PDAC using IHC and compare this to clin-ical characteristics, including MMR status and tumorinfiltrating lymphocytes and examine if an associationwith clinical outcome exists.MethodsEthical approval and a waiver of consent for researchon this retrospective cohort was obtained from theUniversity of British Columbia Clinical Research EthicsBoard (H12–03484).Sample identification and TMA constructionA tissue microarray was constructed using duplicate0.6 mm cores from the epithelial component of all avail-able, resected, pathologically confirmed pancreatic ductaladenocarcinomas derived from the archives of theVancouver Coastal Health Region between 1995 and 2014.All patients received primary surgery with a subset receiv-ing adjuvant chemotherapy with a pyrimidine nucleosideanalog. Cores for the tissue microarray were obtainedfrom areas of tumor as determined by routine microscopyon hematoxylin and eosin-stained sections. Cases were ex-cluded if they lacked clinical follow-up data or if clinico-pathologic variables were lacking.Immunohistochemical staining of PD-L1 and mismatchrepair markersImmunohistochemistry was performed on 4-μm-thickformalin-fixed paraffin-embedded sections of tissue mi-croarrays. PD-L1 immunohistochemistry was performedat the Deely Research Centre at the British ColumbiaCancer Agency using the Intellipath FLX autostainer(Biocare) platform. Mismatch repair, CD3 and CD8 im-munohistochemistry was performed in the clinical labora-tory of Vancouver General Hospital using the VentanaDiscovery XT and the Ventana Benchmark XT automatedsystem (Ventana Medical Systems, Tucson, AZ).For PD-L1, slides were incubated with the cloneSP142 (Spring Bioscience, Pleasanton, USA) at 1/100 di-lution in Da Vinci Green diluent at room temperaturefor 30 min. Slides were then washed and incubated withMach2 Rabbit-HRP polymer for 30 min at roomtemperature and detected with IP DAB chromogen for5 min. Nuclei were counterstained with a 1/10 dilutionof CAT hematoxylin then slides were again washed, airdried and coverslipped with Ecomount. The antibodyclone was selected based on its strong concordance tothree other PD-L1 clones and RNA in situ hybridization(ISH) in NSCLC [19].For MMR stains, slides were incubated with MLH1(mouse monoclonal antibody, 1:50 dilution, cat#: NCL-L- MLH1, clone ID:ES05; Leica Microsystems, New- cas-tle, UK), MSH2 (mouse monoclonal antibody, 1:1000 dilu-tion, cat#: 286 M-16, clone ID:G219–1129; Cellmarque,Rocklin, CA), MSH6 (rabbit monoclonal antibody, 1:200dilution, cat#: CLAC-0047, clone ID: EP49; CedarlaneTessier-Cloutier et al. BMC Cancer (2017) 17:618 Page 2 of 10Corporation, Burlington, ON, Canada), and PMS2 (rabbitmonoclonal antibody, 1:20 dilution, cat#: CLAC-0049,clone ID:EP51; Cedarlane Corporation) for 32 min atroom temperature. For the slides to be stained for PMS2were additionally prepped with the Epitomics DAB prepkit before antibody incubation.Antibodies were detected using the Ventana DABMapkit, counterstained with hematoxylin and treated with aproprietary bluing agent (Ventana). Positive and negativecontrols are performed as part of the routine clinical qualityassurance; in addition to the external quality control pro-gram (Canadian Immunohistochemistry Quality Control(cIQc), a provider of proficiency testing for Canadianclinical laboratories).Interpretation of Immunohistochemical stainsPD-L1 status was assessed independently by two anatom-ical pathologists (BSS and DG) and consensus achieved onall cases. Positivity was evaluated by H-Score, a combin-ation of staining intensity and percentage of tumor cellstaining. Staining intensity was scored as 0 (negative), 1(weak), 2 (moderate), or 3 (strong) based on membranouslocalization and each score multiplied by the percentageof cells (0% - 100%) staining. Therefore, H-scores rangedfrom 0 to 300. To account for potential intra-tumoral het-erogeneity, the mean of both cores was used to generatethe score for each case.Mismatch repair (MMR) was quantified as per Riazyet al. [20]. Briefly, protein expression for MLH1, MSH2,MSH6, and PMS2 was considered intact (normal) if anypercentage of definite positive nuclear staining of themalignant cells was detected on either TMA core. Incases where one or more mismatch repair proteins wereinterpreted as negative staining, examination utilizingimmunohistochemistry on whole sections was per-formed. Each protein was considered lost (abnormal) ifthere was complete loss of nuclear staining in the tumorcells and if there was a positive internal control (intactnuclear staining of stromal elements such as inflamma-tory cells and/or endothelial cells) on whole section.Cases showing a complete absence of nuclear stainingpattern of both tumor cells and stromal elements weredeemed uninterpretable and thus excluded from thestudy. Cases that demonstrated loss of any MMR markeron the TMA were subjected to confirmatory whole slidesection staining and were scored independently by twopathologists (BSS and DFS), who were blinded to clinicalcharacteristics and patient outcomes. Divergent assess-ments were reconciled by consensus conference. A casewas labeled as mismatch repair deficient (MMRd) if anyof the four mismatch repair proteins was completely ab-sent on immunohistochemistry. Cases were classified asmismatch repair proficient (MMRp) if all four proteinsstained positive to some degree.Individual tumor infiltrating lymphocytes were countedand typed in the epithelial and stromal compartmentsusing clinically validated IHC stains for CD3 and CD8.Scoring was performed independently by two anatomicalpathologists (BSS and MA-K) and consensus achievedon all cases. To account for potential intra-tumoralheterogeneity, the average of both cores were used togenerate the final score for CD8+ and CD3+ tumorinfiltrating lymphocytes.Clinico-pathologic variables and outcomeStandard treatment, clinical and pathologic parameterswere collected from the British Columbia Cancer Agencywhich included: age at surgery, sex, adjuvant chemother-apy agents used, lymphovascular invasion, perineural inva-sion, pathologic primary tumor (pT) stage, and pathologicregional lymph-node status (pN). The primary outcomemeasure was defined to be disease-specific survival, wheresurvival time was calculated as the difference between thedate of last follow-up and the date of surgery, expressed inyears. Patients were censored if they were alive at lastfollow-up or had died from a cause other than theirpancreatic malignancy. Deaths attributable to treatmentrelated toxicities or inter-current diseases were consideredcensored observations for this analysis.Statistical analysisTo determine if H-Score or the percent of positive cellsfor PD-L1 expression yield differential prognostic ability,each scoring method was subjected to an omnibus as-sessment utilizing the Cox-Proportional Hazards Modelto determine if the expression of PD-L1 was a significantprognostic marker in the context of the clinico-pathologic variables outlined previously with the excep-tion of pT-Stage due the fact that the vast majority ofthe cases in this cohort are pT3. The proportionality as-sumption for each variable was assessed through exam-ination of Cox-Snell residuals and continuous variableswere assessed for linearity. The PD-L1 scoring method-ology with the smallest P-value was determined to havethe strongest prognostic effect. Parametric survival ana-lysis was used in order to further elucidate the gradientdependent effect of PD-L1 expression on disease specificsurvival (DSS) for the scoring methodology with thegreatest prognostic effect. This procedure modelled thedisease specific survival data with 5 different distribu-tions which included: weibull, log-normal, exponential,frechet, and log-logistic. The best distribution to be usedfor parametric survival analysis was determined byselecting the one with the lowest Bayesian InformationCriterion from the model fits. This analysis produces aquantile plot with logSurvival Time plotted against PD-L1 expression which illustrates the gradient dependentrelationship between disease specific survival and PD-L1Tessier-Cloutier et al. BMC Cancer (2017) 17:618 Page 3 of 10expression. Based on these findings, a series of threecut-points were created starting with an H-Score or per-centage positive cells of 1- as these identify the equiva-lent cases. Subsequent cut-points were set at incrementsof 5 which correspond to the increments used for the as-sessment of percent positive cells. The resultant groupswere subjected to univariable survival analysis to quan-tify differences in disease specific survival using theKaplan-Meier method. A multivariable approach todisease specific survival, using the Cox ProportionalHazards Model, was used to determine if survivaldifferences between PD-L1 expression categories wereindependent of adjuvant chemotherapy. Assessment forheterogeneity of clinico-pathologic variables was per-formed with the following statistical approaches: con-tinuous variables were examined using the WilcoxonRank-Sum Test, categorical comparisons were computedusing Fisher’s exact test. A P-value of <0.05 was consid-ered as statistically significant for all analyses. Allanalyses were computed with JMP v13.1 (SAS Institute,Cary, NC).ResultsAfter exclusion criteria were applied, two-hundred fifty-two cases remained (Fig. 1). The demographic informa-tion for the cohort were tabulated and are shown inTable 1. The Cox-Proportional Hazards Analysis for PD-L1 H-score and percent positive indicated that the latterwas a stronger prognostic indicator with P = 0.0466compared to H-Score with a P = 0.10 (Table 2).Subjecting the survival data to the 5 distributionmodels outlined in the methods and ranking those fitsby the Bayesian Information Criterion revealed thatthe log-logistic distribution fit best and was used asthe basis for parametric disease specific survivalanalysis. Parametric disease specific survival of PD-L1 per-cent positive and H-Score demonstrated an inverse rela-tionship between increased PD-L1 expression and survivaltime (Fig. 2). As indicated in the multivariable survivalanalysis, PD-L1 percent positive had a slightly strongerprognostic association compared to H-Score.Based on the prognostic non-inferiority of PD-L1percent positive, we elected to pursue this scoringmethod for the remainder of the study. Cut-points weredetermined according to our criteria outlined in themethods and resulted in: > = 1% (N = 31; 12.3% of thecohort), >5% (N = 17; 6.7% of the cohort), and >10%(N = 12; 4.8% of the cohort). Univariable survival ana-lysis using these three cut-points showed no disease spe-cific survival differences at the > = 1 cut-point (p = 0.51)or the >5 cut-point (0.52), but the >10 cut-point yieldedstatistically significant disease specific survival differ-ences of p = 0.027 (Fig. 3). Multivariable DSS analysis ofthe >10% positive PD-L1 expression cut-point alongwith the other clinico-pathologic covariates outlinedin Table 2, indicates that this subset of twelve caseshas a trend toward inferior prognosis with a Risk Ra-tio and 95%CI 1.90 [0.96–3.42] (P = 0.06). When wesequentially removed statistically insignificant variablesfrom the model (age, histopathologic grade, sex, and lypho-vascular Invasion, PD-L1 > 10% became statistically sig-nificant Risk Ratio and 95%CI 2.05 [1.03–3.66](P = 0.0410). The remaining statistically significant vari-ables included pN-Stage (P < 0.0001), adjuvant chemo-therapy (P = 0.0002), and perineurial invasion (P = 0.009)and resection status (P = 0.0263).Analysis for heterogeneity across clinico-pathologicparameters which included: age, sex, adjuvant chemo-therapy use, histopathological grade, lymphovascular in-vasion, perineural inavasion, pN-Stage, and resectionResected PDAC Cases from the Vancouver Coastal Health RegionN = 277PD-L1 IHC FailureN = 16Treated With Adjuvant ChemotherapyN = 74Missing Clinico-Pathologic InformationN = 9Post-Surgical ObservationN = 178Exclusion CrtiteriaRemaining CohortN = 252Fig. 1 Patient selection diagram illustrating inclusion and exclusion criteria for this study with final numbers for the cohorts who receivedadjuvant pyrimidine nucleoside analogs or subjected to post-surgical observation onlyTessier-Cloutier et al. BMC Cancer (2017) 17:618 Page 4 of 10status demonstrated a significant relationship betweenincreased PD-L1 expression and higher grade cases(Table 3). The remaining clinico-pathologic variables,mismatch repair, and the stromal or epithelial compart-ment specific prevalence of CD3+ or CD8+ tumor in-filtrating lymphocytes were not associated with thePD-L1 > 10% positive cells cut-point.DiscussionIn this study, we have found a gradient dependent asso-ciation between PD-L1 expression and inferior diseasespecific survival in resected pancreatic ductal adenocar-cinoma. This finding was independent of the improvedprognosis associated with the application of adjuvantchemotherapy with a pyrimidine nucleoside analog.We assessed multiple scoring methods, H-Score orpercent of positive cells, for the quantification of PD-L1expression and determined that the estimation of per-cent positive cells yields a stronger association with in-ferior survival than H-Score. This suggests that theaddition of a subjective intensity assessment to generatean H-score may represent an unnecessary step for thequantification of PD-L1 in this disease. Examination ofother clinico-pathologic parameters revealed no statisti-cally significant associations with PD-L1 expression atany cut-point, which indicates that PD-L1 expressiondoes not select for any known prognostic variable otherthan histo-pathologic grade. Due to the limited powerassociated with our cohort combined with the smallfraction of cases that express PD-L1 at a high level, wewere limited in our ability to perform multivariable dis-ease specific survival analyses with numerous variables.Exploratory multivariable disease specific survival mod-eling suggested that our categorized PD-L1 expressionutilizing the cut-point of >10% of positive cells is inde-pendently associated with inferior disease specific sur-vival and was only surpassed by the presence of regionallymph node metastasis and perineural invasion in termsof negative prognostic variables.Recent studies have demonstrated that PD-L1 expres-sion is associated with tumor types known to havehigher somatic mutation load, as is the case for melano-mas, NSCLC and RCC [21, 22]. Considering that PDAChas a lower mutation burden, it is not surprising that wefound only 4 to 12% PD-L1 positive tumors compared tothe reported 83% in melanoma, 50% in NSCLC and 80%in RCC [10]. Nonetheless, PDAC is associated with to-bacco use and BRCA loss-of-function, and is predictedto, at least occasionally, show an increased mutationburden as a result of these [23]. Consequently, the lowerthan average rate of PD-L1 expression in PDAC com-pared to other malignancies may explain poor responseto checkpoint inhibitors in clinical trials since PD-L1was either not accounted for or the positivity thresholdswere only set between 1% and 5% [10, 11]. Althoughour patient cohort was mostly treatment naive, wewere able to identify differential outcomes based onhigher PD-L1 expression.The observed increased trend of lymphocyte tumor in-filtration (CD3+) in PD-L1 positive patients has been re-ported in previous studies [24]. Sanmamed et al. showedthat tumor infiltrating lymphocytes release IFN-Gammaas part of the host response to the tumor, which inducesupregulation, and expression of, PD-L1 by tumor cells[25]. Our results indicate that a cut-point > = 1% yieldsthe strongest association with CD3+ infiltrating T-cellsbut due to reduced power associated with increasing thePD-L1 cut-point, statistical significance is lost at higherthresholds.Table 1 Demographics of the entire cohortVariable Levels ValuesAge Median [IQR] 66.4 [13.3]Sex Male 139 (55.2%)Female 113 (44.8%)Adjuvant Chemotherapy Given 74 (29.4%)Observation 178 (70.6%)Histologic Grade 1 2 (0.8%)2 186 (73.8%)3 64 (25.4%)Lymphovascular Invasion Present 144 (57.1%)Absent 108 (42.9%)Perineurial Invasion Present 232 (92.1%)Absent 20 (7.9%)pT Stage 1 2 (0.8%)2 11 (4.4%)3 238 (94.4%)4 1 (0.4%)pN Stage 0 64 (25.4%)1 168 (74.6%)Resection Status R0 190 (75.4%)R1 62 (24.6%)CD3 Epithelial Median [IQR] 0 [0]CD3 Stromal Median [IQR] 50 [52]CD8 Epithelial Median [IQR] 0 [0]CD8 Stromal Median [IQR] 11 [36]MMR Status Proficient 211 (84.1%%)Deficient 40 (15.9%)Follow-up Time (Years) Median [IQR] 1.33 [1.59]Events Disease Specific Deaths 200 (79.4%)Censorings 52 (20.6%)Tessier-Cloutier et al. BMC Cancer (2017) 17:618 Page 5 of 10We found no significant association between MMRand PD-L1 status. Our results are somewhat differentfrom what was observed by Le et al. (2016) who re-ported that, in a series of 30 cases, PD-L1 was onlyexpressed in MMR deficient (MMRd) tumors, most ofwhich being colorectal carcinomas [11]. This inconsist-ency might be explained by the lower mutational burdenseen in PDAC compared to MMRd colon carcinoma,melanoma, NSCLC and RCC [22]. Tumors with lowmutational burden tend to be less immunogenic, makingthem less likely to develop immune silencing mechanismduring their evolution.There are several limitations to our study, one beingthe lack of consensus for PD-L1 IHC expression cut-offand gold standard, which our study has attempted to ex-plore. Our IHC protocol for PD-L1 previously showedfairly strong concordance when compared to three otherPD-L1 clones and RNA in situ hybridization (ISH),Sheffield et al., in NSCLC [26]. Our sample size is lim-ited given the small percentage of PD-L1 expression andmay have been underpowered to detect some moresubtle associations, especially in the higher PD-L1cut-points. Finally, since the IHC was performed on aTMA rather than full section, we might have under-represented the amount of PD-L1 positive PDAC dueto sampling error, although this method approximatesthe biopsy sampling error in encountered in clinicalpractice.The prevalence of PD-L1 positivity in PDAC has beenexamined in numerous other studies with the percentageof tumor cells staining positive ranging from 4% - 49%.Each of these previous studies utilized different cut-points that varied between 1% - 10% making their resultsnearly impossible to compare [27–29]. Of particularinterest, our results are somewhat different from whathas been reported by Nomi et al. who demonstrated afound a 39% PD-L1 positivity in pancreatic cancerusing a 10% positivity threshold [28]. Their cohort in-cluded 51 cases from Japan, which were stained usingAnti-Human CD274, clone MIH1. The difference inPD-L1 expression is notable and although the CD274is not commonly used in the clinical research settingthis result may indicate variability associated withethnicity.Table 2 Multivariable disease specific survival analysis for PD-L1 expression quantified by percent positive & H-scoreVariable Comparison Risk ratio 95%CI p-valuePD-L1 H-ScorePD-L1 H-Score Per unit change 1.01 0.997–1.02 0.10Age at Surgery Per unit change 1.01 0.991–1.02 0.44Sex Male v Female 1.14 0.85–1.53 0.37Adjuvant Chemotherapy Given v Observation 0.59 0.42–0.81 0.0011Histopathologic Grade 1 v 2 0.63 0.10–2.10 0.501 v 3 0.49 0.08–1.67 0.282 v 3 0.77 0.55–1.09 0.13Lymphovascular Invasion Present v Absent 1.27 0.93–1.74 0.13Perineural Invasion Present v Absent 1.66 0.96–3.09 0.07pN-Stage 1 v 0 1.83 1.27–2.68 0.0010Resection Status R0 v R1 0.67 0.49–0.94 0.0202PD-L1 Percent PositivePD-L1 Percent Positive Per unit change 1.03 1.0005–1.05 0.0466Age at Surgery Per unit change 1.0006 0.99–1.02 0.42Sex Male v Female 1.16 0.86–1.55 0.33Adjuvant Chemotherapy Given v Observation 0.58 0.42–0.80 0.0008Histopathologic Grade 1 v 2 0.64 0.10–2.13 0.511 v 3 0.50 0.08–1.71 0.302 v 3 0.78 0.56–1.10 0.15Lymphovascular Invasion Present v Absent 1.28 0.94–1.76 0.12Perineural Invasion Present v Absent 1.68 0.98–3.13 0.06pN-Stage 1 v 0 1.85 1.28–2.71 0.0009Rescetion Status R0 v R1 0.68 0.49–0.95 0.0189Tessier-Cloutier et al. BMC Cancer (2017) 17:618 Page 6 of 10ConclusionsIn summary, this is the first study to systematically in-vestigate the association between clinical outcome andabcFig. 3 Binarization cut-points for percent positive (a-c) show thatonly the highest cut-point (>10) yields statistically significantsurvival differencesabcFig. 2 Parametric disease specific survival analysis using the log-logisticdistribution to model disease specific survival in the entire cohort (a).Modeling of disease specific survival against PD-L1 expression assessedby percent positive (b) and H-Score (c) survival using a log-logisticdistribution demonstrates a substantial association between reducedsurvival and increased PD-L1 expression. Curves shown are fitted as afunction of the regressor representing the 0.9, 0.5, and 0.1 quantilesTessier-Cloutier et al. BMC Cancer (2017) 17:618 Page 7 of 10biomarker expression across differing scoring method-ologies and cut-points for PD-L1 immunohistochemistryin this disease. We have demonstrated a gradientdependent association between PD-L1 expression andinferior survival that is independent of the prognosticadvantage conferred by adjuvant chemotherapy. We pos-tulate that the association presented here may indicatethat higher PD-L1 protein expression levels represent aphenotype where PD-1 inhibition may be more effective.However, this hypothesis would have to be tested inthe context of a randomized clinical trial. With stud-ies in other diseases also indicating that deficientMMR (MMRd) status has been shown to be a pre-dictive biomarker for immunotherapy, it is entirelyplausible that PD-L1 immunohistochemistry is an im-perfect biomarker for sensitivity to anti-PD-1 therapy.Interestingly, we found no association between MMRdstatus and PD-L1 expression in this cohort. Moredata on the role of PD-1-axis inhibition in PDAC isneeded, specifically examining the use of predictivebiomarkers in the context of patients treated with im-munotherapy. Future studies should endeavor to buildpredictive models based on multi-marker expressionthat will serve as tools to triage the PDAC patientpopulation to immunotherapy or other treatmentregimens.AbbreviationsMMRd: Mismatch Repair Deficient; MMRp: Mismatch Repair Proficient;PD1: Programmed Cell Death 1; PDAC: Pancreatic Ductal Adenocarcinoma;PD-L1: Programmed Cell Death Ligand 1AcknowledgementsWe, the authors, would like to recognize the patients and their families fortheir direct and indirect contributions toward fighting this disease.FundingThis work was supported through unrestricted research funds provided bythe VGH and UBC Hospital Foundation and the BC Cancer Foundation whichwere administered through the Pancreas Centre BC. The above funders ofthis research had no influence upon the design of the study, collection,analysis nor interpretation of the data or writing of the manuscript.Table 3 Assessment for heterogeneity using a percent positive binarization cut-point of >10Variable Levels PD-L1%Positive < 10 PD-L1%Positive > 10 P-ValueAge Median [IQR] 66.5 [13.1] 65.6 [20.5] 0.96Sex Male 132 (55.0%) 7 (58.3%) 1.0Female 108 (45.0%) 4 (41.8%)Adjuvant Chemotherapy Given 71 (29.6%) 3 (25.0%) 1.0Observation 169 (70.4%) 9 (75.0%)Histologic Grade 1 2 (0.8%) 0 (0.0%) 0.0292 181 (75.4%) 5 (41.7%)3 57 (23.8%) 7 (58.3%)Lymphovascular Invasion Present 136 (56.7%) 8 (66.7%) 0.56Absent 104 (43.3%) 4 (33.3%)Perineurial Invasion Present 221 (92.1%) 11 (91.7%) 1.0Absent 19 (7.9%) 1 (9.3%)pT Stage 1 2 (0.8%) 0 (0.0%) 0.882 10 (4.2%) 1 (8.3%)3 227 (94.6%) 11 (91.7%)4 1 (0.4%) 0 (0.0%)pN Stage 0 62 (25.8%) 2 (16.7%) 0.741 178 (74.2%) 10 (83.3%)Resection Status R0 184 (76.7%) 6 (50%) 0.08R1 56 (23.3%) 6 (50%)CD3 Epithelial Median [IQR] 0 [0] 0 [4] 0.16CD3 Stromal Median [IQR] 50 [52] 63 [76] 0.52CD8 Epithelial Median [IQR] 0 [0] 0 [0] 0.51CD8 Stromal Median [IQR] 11 [36] 14 [39] 0.72MMR Status Proficient 200 (83.7%) 11 (91.7%) 0.70Deficient 39 (16.3%) 1 (8.3%)Tessier-Cloutier et al. BMC Cancer (2017) 17:618 Page 8 of 10Availability of data and materialsThe datasets used and/or analysed during the current study are availablefrom the corresponding author on reasonable request.Authors’ contributionsBTC: Conceived the study, wrote the first draft of the manuscript and hadinput on revisions after internal review by co-authors. SEK: Conceived thestudy, performed all statistical analyses and participated in the writing ofthe manuscript. MA: Developed the method and provided all scoring forimmune infiltrates for epithelial and stromal components of each PDAC case.KM: Developed and optimized the immunohistochemical staining procedurefor the PD-L1 antibody and was responsible for applying this procedureto the TMA. Contributed to the methods section of the manuscript. DG:Developed the scoring system used for this study and performed thescoring for the PD-L1 antibody. Contributed to the methods section of themanuscript. BHN: Advised on the design of the study and served as aninternal reviewer for the manuscript. DJR: Acquired the clinical follow-up forpatients included in this TMA case series. Served as an internal reviewer forthe manuscript and served in a co-supervisory capacity for the project.BSS: Performed consensus IHC scoring for PD-L1, CD3, CD8, and MMRimmunohistochemistry. Advised on study design and helped to improve themanuscript though critical review. DFS: Advised on the study design andaided with the interpretation of the results. Served as an internal reviewerfor the manuscript and served in a co-supervisory capacity for the project.Performed original scoring assessment for MMR status. All authors havegiven final approval of the manuscript in its current form and agree to takeresponsibility for the accuracy and integrity of it’s content. Any errors andomissions are our own.Ethics approval and consent to participateEthical approval for research on this retrospective cohort was obtained fromthe University of British Columbia Clinical Research Ethics Board (H12- 03484). Awaiver of consent was provided to account for the large proportion of patientsthat were deceased at the time of the assembly of this cohort.Consent for publicationNot ApplicableCompeting interestsThe authors declare that they have no competing interests.Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims inpublished maps and institutional affiliations.Author details1Division of Anatomical Pathology, Vancouver General Hospital, Vancouver,British Columbia, Canada. 2Department of Pathology & Laboratory Medicine,University of British Columbia, Vancouver, British Columbia, Canada. 3Divisionof Medical Oncology, University of British Columbia , Vancouver, BritishColumbia, Canada. 4Genetic Pathology Evaluation Centre, University of BritishColumbia, Vancouver, British Columbia, Canada. 5Pancreas Centre BC,Vancouver, British Columbia, Canada. 6Deeley Research Centre, BritishColumbia Cancer Agency, Victoria, British Columbia, Canada. 7Division ofMedical Oncology, British Columbia Cancer Agency, Vancouver, BritishColumbia, Canada. 8Department of Biochemistry and Microbiology,University of Victoria, Victoria, British Columbia, Canada. 9Department ofAnatomical Pathology, Abbotosford Regional Hospital and Cancer Centre,Abbotsford, British Columbia, Canada.Received: 22 May 2017 Accepted: 28 August 2017References1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2016. CA Cancer J Clin.2016;66:7–30.2. Topalian SL, Sznol M, McDermott DF, Kluger HM, Carvajal RD, Sharfman WH,et al. Survival, durable tumor remission, and long-term safety in patients withadvanced melanoma receiving nivolumab. J Clin Oncol. 2014;32:1020–30.3. Brahmer JR, Tykodi SS, Chow LQ, Hwu W-JJ, Topalian SL, Hwu P, et al. safetyand activity of anti-PD-L1 antibody in patients with advanced cancer. NEngl J Med. 2012;366:2455–65.4. De Guillebon E, Roussille P, Frouin E, Tougeron D. Anti program death-1/anti program death-ligand 1 in digestive cancers. World J GastrointestOncol. 2015;7:95–101.5. Dong H, Strome SE, Salomao DR, Tamura H, Hirano F, Flies DB, et al. Tumor-associated B7-H1 promotes T-cell apoptosis: a potential mechanism ofimmune evasion. Nat Med. 2002;8:793–800.6. Dong H, Zhu G, Tamada K, Chen L. B7-H1, a third member of the B7 family,co-stimulates T-cell proliferation and interleukin-10 secretion. Nat Med.1999;5:1365–9.7. Freeman GJ, Long AJ, Iwai Y, Bourque K, Chernova T, Nishimura H, et al.Engagement of the PD-1 immunoinhibitory receptor by a novel B7 familymember leads to negative regulation of lymphocyte activation. J Exp Med.2000;192:1027–34.8. Topalian SL, Drake CG, Pardoll DM. Targeting the PD-1/B7-H1(PD-L1) pathwayto activate anti-tumor immunity. Curr Opin Immunol. 2012;24:207–12.9. Iwai Y, Ishida M, Tanaka Y, Okazaki T, Honjo T, Minato N. Involvement ofPD-L1 on tumor cells in the escape from host immune system and tumorimmunotherapy by PD-L1 blockade. Proc Natl Acad Sci U S A. 2002;99:12293–7.10. Topalian SL, Hodi SF, Brahmer JR, Gettinger SN, Smith DC, McDermott DF, etal. Safety, activity, and immune correlates of anti-PD-1 antibody in cancer. NEngl J Med. 2012;366:2443–54.11. Le DT, Uram JN, Wang H, Bartlett BR, Kemberling H, Eyring AD, et al. PD-1blockade in tumors with mismatch-repair deficiency. N Engl J Med. 2015;372:2509–20.12. Herbst RS, Soria J-CC, Kowanetz M, Fine GD, Hamid O, Gordon MS, et al.predictive correlates of response to the anti-PD-L1 antibody MPDL3280A incancer patients. Nature. 2014;515:563–7.13. Powles T, Eder J, Fine GD, Braiteh FS, Loriot Y, Cruz C, et al. MPDL3280A(anti-PD-L1) treatment leads to clinical activity in metastatic bladder cancer.Nature. 2014;515:558–62.14. Herbst RS, Baas P, Kim D-WW, Felip E, Pérez-Gracia JL, Han J-YY, et al.Pembrolizumab versus docetaxel for previously treated, PD-L1-positive,advanced non-small-cell lung cancer (KEYNOTE-010): a randomisedcontrolled trial. Lancet. 2016;387:1540–50.15. Meng X, Huang Z, Teng F, Xing L, Yu J. Predictive biomarkers in PD-1/PD-L1checkpoint blockade immunotherapy. Cancer Treat Rev. 2015;41:868–76.16. Taube JM, Klein A, Brahmer JR, Xu H, Pan X, Kim JH, et al. Association ofPD-1, PD-1 ligands, and other features of the tumor immune microenvironmentwith response to anti-PD-1 therapy. Clin Cancer Res. 2014;20:5064–74.17. Ascierto PA, Capone M, Urba WJ, Bifulco CB, Botti G, Lugli A, et al. Theadditional facet of immunoscore: immunoprofiling as a possible predictivetool for cancer treatment. J Transl Med. 2013;11:54.18. Messina JL, Fenstermacher DA, Eschrich S, Qu X, Berglund AE, Lloyd MC, etal. 12-chemokine gene signature identifies lymph node-like structures inmelanoma: potential for patient selection for immunotherapy? Sci Rep.2012;2:765.19. Rizvi NA, Hellmann MD, Snyder A, Kvistborg P, Makarov V, Havel JJ, et al.Cancer immunology. Mutational landscape determines sensitivity to PD-1blockade in non-small cell lung cancer. Science. 2015;348:124–8.20. Riazy M, Kalloger SE, Sheffield BS, Peixoto RD, Li-Chang HH, Scudamore CH,et al. Mismatch repair status may predict response to adjuvant chemotherapyin resectable pancreatic ductal adenocarcinoma. Mod Pathol. 2015;28:1383–9.21. Grosso J, Inzunza D, Wu Q, Simon J, Singh P, Zhang X, et al. Programmeddeath-ligand 1 (PD-L1) expression in various tumor types. Journal forimmunotherapy of cancer. 2013;1:P53.22. Lawrence MS, Stojanov P, Polak P, Kryukov GV, Cibulskis K, Sivachenko A, etal. Mutational heterogeneity in cancer and the search for new cancer-associated genes. Nature. 2013;499:214–8.23. Alexandrov LB, Nik-Zainal S, Wedge DC, Aparicio SA, Behjati S, Biankin AV, et al.Signatures of mutational processes in human cancer. Nature. 2013;500:415–21.24. Furuta J, Inozume T, Harada K, Shimada S. CD271 on melanoma cell is anIFN-γ-inducible immunosuppressive factor that mediates downregulation ofmelanoma antigens. J Invest Dermatol. 2014;134:1369–77.25. Sanmamed MF, Chen L. Inducible expression of B7-H1 (PD-L1) and itsselective role in tumor site immune modulation. Cancer J. 2014;20:256–61.26. Sheffield BS, Fulton R, Kalloger SE, Milne K, Geller G, Jones M, et al. Investigationof PD-L1 biomarker testing methods for PD-1 Axis inhibition in non-squamousnon-small cell lung cancer. J Histochem Cytochem. 2016;64:587–600.Tessier-Cloutier et al. BMC Cancer (2017) 17:618 Page 9 of 1027. Diana A, Wang L, D’Costa Z, Allen P, Azad A, Silva MA, et al. Prognosticvalue, localization and correlation of PD-1/PD-L1, CD8 and FOXP3 with thedesmoplastic stroma in pancreatic ductal adenocarcinoma. Oncotarget.2016;7:40992–1004.28. Nomi T, Sho M, Akahori T, Hamada K, Kubo A, Kanehiro H, et al. Clinicalsignificance and therapeutic potential of the programmed death-1 ligand/programmed death-1 pathway in human pancreatic cancer. Clin CancerRes. 2007;13:2151–7.29. Wang L, Ma Q, Chen X, Guo K, Li J, Zhang M. Clinical significance of B7-H1and B7-1 expressions in pancreatic carcinoma. World J Surg. 2010;34:1059–65.• We accept pre-submission inquiries • Our selector tool helps you to find the most relevant journal• We provide round the clock customer support • Convenient online submission• Thorough peer review• Inclusion in PubMed and all major indexing services • Maximum visibility for your researchSubmit your manuscript atwww.biomedcentral.com/submitSubmit your next manuscript to BioMed Central and we will help you at every step:Tessier-Cloutier et al. BMC Cancer (2017) 17:618 Page 10 of 10