The Balance Between Cytotoxic T-cell Lymphocytes and Immune Checkpoint Expression in the Prognosis of Colon Tumors

The Balance Between Cytotoxic T-cell Lymphocytes and Immune Checkpoint Expression in the... Abstract Background Immune checkpoint (ICK) expression might represent a surrogate measure of tumor-infiltrating T cell (CTL) exhaustion and therefore be a more accurate prognostic biomarker for colorectal cancer (CRC) patients than CTL enumeration as measured by the Immunoscore. Methods The expression of ICKs, Th1, CTLs, cytotoxicity-related genes, and metagenes, including Immunoscore-like metagenes, were evaluated in three independent cohorts of CRC samples (260 microsatellite instable [MSI], 971 non-MSI). Their associations with patient survival were analyzed by Cox models, taking into account the microsatellite instability (MSI) status and affiliation with various Consensus Molecular Subgroups (CMS). PD-L1 and CD8 expression were examined on a subset of tumors with immunohistochemistry. All statistical tests were two-sided. Results The expression of Immunoscore-like metagenes was statistically significantly associated with improved outcome in non-MSI tumors displaying low levels of both CTLs and immune checkpoints (ICKs; CMS2 and CMS3; hazard ratio [HR] = 0.63, 95% confidence interval [CI] = 0.43 to 0.92, P = .02; and HR = 0.55, 95% CI = 0.34 to 0.90, P = .02, respectively), but clearly had no prognostic relevance in CRCs displaying higher levels of CTLs and ICKs (CMS1 and CMS4; HR = 0.46, 95% CI = 0.10 to 2.10, P = .32; and HR = 1.13, 95% CI = 0.79 to 1.63, P = .50, respectively), including MSI tumors. ICK metagene expression was statistically significantly associated with worse prognosis independent of tumor staging in MSI tumors (HR = 3.46, 95% CI = 1.41 to 8.49, P = .007). ICK expression had a negative impact on the proliferation of infiltrating CD8 T cells in MSI neoplasms (median = 0.56 in ICK low vs median = 0.34 in ICK high, P = .004). Conclusions ICK expression cancels the prognostic relevance of CTLs in highly immunogenic colon tumors and predicts a poor outcome in MSI CRC patients. The host immune response, and especially the T cell immune infiltrate, intersects the molecular and clinical characteristics of primary colorectal cancers (CRCs), including TNM staging. It is associated with prognosis of localized colon tumors. Expression levels for CTL/Th1/cytotoxicity markers have been demonstrated to constitute strong and independent prognostic indicators in CRC patients, with a high density of lymphocyte infiltration being consistently associated with prolonged survival (1). Based on multiple analyses, a scoring system was developed (termed Immunoscore) to quantify cytotoxic and memory T cells in the core of the tumor and in the tumor’s invasive margin (2). Immunoscore is a prognostic index in CRC (3). Several authors have suggested that the dense Th1/CTL lymphocytic infiltrate could explain the better prognosis of colon tumors displaying microsatellite instability (MSI), compared with non-MSI (MSS) CRC (4–6). MSI (7–12) leads to the synthesis of aberrant and potentially immunogenic neo-antigens by the tumor cells (for review, see [13,14]). A likely consequence of this is that MSI tumors are heavily infiltrated with activated cytotoxic T-cell lymphocytes (CTLs) and Th1 cells. A recent study reported that concomitant expression of multiple active immune checkpoint (ICK) markers (eg, CTLA-4, PD-1, PD-L1, LAG-3 targeted by immunotherapy) in CRC may counterbalance the antitumoral Th1/CTL immune response, notably in MSI CRC (15). One hypothesis that stems from these observations is that MSI tumors would be more responsive to ICK inhibitors because ICK expression is expected to protect highly immunogenic tumor cells from being destroyed. In line with this hypothesis, Le et al. (16) evaluated the clinical activity of an anti–PD-1 ICK inhibitor (pembrolizumab) in a cohort of metastatic CRC patients with MSS and MSI primary tumors. The results of this phase II study showed that MSI status was predictive of clinical benefit from ICK blockade therapy with pembrolizumab. The second hypothesis is that ICK expression should exert a negative impact on the prognosis of MSI tumors untreated by ICK inhibitors. To our knowledge, this second hypothesis has not yet been assessed thoroughly. It was reported in a recent study that the Immunoscore was prognostic in MSI CRC (17), but the authors did not integrate ICK markers in their survival analyses. Their results were obtained by investigating a unique series of CRC, and no further validation using publicly available clinical data was performed because the Immunoscore requires direct access to primary tumor samples. In other publications, only the role of PD1/PDL1 expression in CRC has been explored, leading to various conclusions and thus demonstrating the complexity involved with analysis of the CRC immune response (18–21). Rather than analyzing CRC immune response through the expression of PD1/PDL1 or T cell enumeration, our aim here was for a more holistic assessment of T cell immune response quality based on the analysis of both Th1/CTL and ICK markers. We also took into account the major molecular subtypes of CRC, that is, MSI/MSS status and consensus molecular subtypes (CMS) (22). Finally, the functional relevance of ICK overexpression was determined by pathway enrichment analysis and by quantification of immune infiltrates and CD8 T cell proliferation in MSI tumors. Methods Immune Genes Immune checkpoint and modulator genes were selected according to Llosa et al. (15) and a recent review (23). Markers for cytotoxic T lymphocytes, cytotoxicity, and T helper1 were selected as described earlier (see also the Supplementary Materials and Methods, available online) (15,24). Cohort Data Tissue samples from a large multisite cohort of CRC patients were collected as part of the “Cartes d’Identité des Tumeurs” (CIT) research program/enlever network, including tumors with or without microsatellite instability (MSI or MSS, respectively) and adjacent nontumoral tissue samples (NT). Samples from 146 MSI tumors, 444 MSS tumors, and 56 NT were analyzed for gene expression profiles on Affymetrix U133 plus 2 chips as described earlier (25). Data were normalized using the frozen RMA method (26) followed by Combat normalization (27) to remove technical batch effects (SVA R package). In addition, the CRC cohort from the The Cancer Genome Atlas (TCGA) consortium was also used (86 MSI tumors, 527 MSS tumors, 51 NT). Both data sets were centered for each gene by subtracting the median value of the non-tumoral sample and are provided as a Supplementary Data File (available online). An additional retrospective, multisite series of 28 stage IV primary MSI CRCs was analyzed separately for gene expression as part of an independent, ongoing study using the NanoString nCounter platform. A set of immune genes that comprised 14 of the 32 analyzed markers, including seven ICKs, was screened. All patients from this metastatic cohort (11 synchronous metastatic lesions, 17 metachronous metastatic lesions) received standard of care chemotherapy but not ICK blockade treatment. The NanoString data set also includes a subset of the CIT cohort samples for which sufficient RNA material was available for analysis (for protocol details, see the Supplementary Materials and Methods, available online). Supplementary Table 1 (available online) gives a description of all the cohorts investigated in this study, totaling 260 MSIs and 971 MSS CRCs. For non-metastatic MSI CRC patients, informed consent was recorded in each case; the standard hospital blanket consent was considered sufficient by French law. For patients included in the cohort of metastatic CRCs, ethical approval was provided by an institutional review board (institutional review board No. 00003835) on November 27, 2014. Statistical Analysis Metagene values were computed to summarize an immune gene category as the median value per sample of all the genes in the category. Comparisons of the mean expression between gene or metagene values across groups were assessed using the Student’s t-test or the analysis of variance test, as appropriate. Correlations were assessed by Pearson correlation coefficients. Associations between categorical variables were assessed using chi-square tests. MSigDB gene sets/pathways enrichment analysis was performed by comparing low- and high-ICK groups, using the hypergeometric test and based on the genes most statistically significantly differentially expressed between these two groups. Associations between gene /metagene expression values and survival in the different cohorts were assessed by univariate and bivariate Cox proportional-hazards regression analyses and adjusted by tumor series and TNM stage, using the R package “survival,” and the results were combined using a generic inverse variance meta-analysis approach (see also the Supplementary Materials and Methods, available online). The proportional-hazards assumption was tested prior to survival analysis to examine the models’ appropriateness using the cox.zph function. Survival curves were obtained using Kaplan-Meier estimates, and differences between curves were assessed using the log-rank test. Overall survival was used and defined as time from diagnosis to death from any cause. All statistical tests were two-sided. The cut point for statistical significance was .05. Results Prognostic Value of Immune Genes and Metagenes in Function of CMS Classification of Colorectal Cancer To test our working hypothesis, summarized in Figure 1A, we evaluated the prognostic significance of ICKs, Th1, CTLs, and cytotoxicity markers in the combined CIT (n = 590 CRC, comprising 146 MSI and 444 MSS) and TCGA (n = 613 CRC, comprising 86 MSI and 527 MSS) series. In both cohorts, MSS tumors were categorized into one of the four CMS subtypes of CRC (22). We investigated 32 immune markers classified into the above four immune marker groups (see the Supplementary Materials and Methods, available online, for a list of these markers) (for review, see [24]) as independent genes and as four metagenes. Four of the 19 immune checkpoints and modulators were not statistically significantly overexpressed in CRC as compared with nontumor colonic mucosa and were subsequently removed (Supplementary Figure 1, available online). Further analyses were thus carried out on the 28 remaining genes. Finally, to obtain Immunoscore surrogates, six Immunoscore-like metagenes were built based on the expression of Immunoscore-related markers (Supplementary Table 2, available online). Figure 1. View largeDownload slide Prognostic value of immune gene expression. A) Schematic representation of the working hypothesis on immune checkpoint prognosis. Arrows represent inhibition. B) Overall survival stratified by microsatellite instable (MSI)/microsatellite stable (MSS) status (left) and Consensus Molecular Subgroups (CMS) subtypes (right). Curves of overall survival (OS) rate (y-axis) according to time from diagnosis (in years; x-axis) were obtained by the method of Kaplan-Meier for both the Cartes d’Identité des Tumeurs (CIT) and The Cancer Genome Atlas (TCGA) series. Differences between survival distributions were assessed by the log-rank test using an end point of five years. C) Prognostic values of immune gene/metagene expression and of clinical factors in MSI tumors. Forest plot of overall survival (OS) hazard ratios (HRs) estimated by combining independent univariate Cox analyses on the CIT and TCGA series, adjusted for TNM stage. Hazard ratio, as well as related Wald test P value and 95% confidence intervals (CIs), are given for metagenes (which aggregate the gene expression values of a gene set related to the four immune categories [immune checkpoints {ICK}, cytotoxic T lymphocytes {CTL}, cytotoxicity, Th1 functional orientation]), individual immune genes, and clinical annotations. Diamonds represent the hazard ratios, and horizontal bars represent the 95% confidence intervals. Red indicates a hazard ratio greater than 1 with a P value of less than .1 (worse prognosis), blue a hazard ratio of less than 1 with a P value of less than .1 (better prognosis), and gray a hazard ratio with a Wald test P value of .1 or less. D) Overall survival stratified by overexpression of immune checkpoint genes within MSI (left), MSS (middle), and both CRCs (right). MSI tumors in the higher quartile of ICK metagene values, in CIT and TCGA series independently, were assigned ICK + (n = 55), and the other tumors ICK- (n = 139). The minimal ICK metagene value within MSI ICK+ tumors was used to divide MSS tumors into ICK + (n = 26) and ICK- (n = 765). Curves of overall survival rate (y-axis) according to time from diagnosis (in years; x-axis) were obtained by the method of Kaplan and Meier for both the CIT and TCGA series. Differences between survival distributions were assessed by the log-rank test using an end point of five years. E) Prognostic value of immune metagenes according to the type of CRC. Each column defines a specific type of CRC (all CRC, MSI CRC, MSS CRC, MSS CMS1 CRC, MSS CMS2 CRC, MSS CMS3 CRC, MSS CMS4 CRC) and the number of related samples with available survival data in the combined CIT + TCGA series. Each row defines a given metagene (ICK, CTL, CYTOX, TH1, IS-like v1, IS-like v2). Each cell shows the hazard ratio for overall survival of the relevant metagene expression for the specific CRC type (obtained via a Cox model). Each cell also contains the P value for the Wald test for equality of this hazard ratio to 1. The hazard ratio shown in each cell was calculated by combining the hazard ratios for each of the CIT and TCGA data sets using the meta-analysis approach of DerSimonian. The genes relevant to each metagene are indicated in (C) (ICK, CTL, CYTOX, TH1) or are given below the metagene name. F) Bivariate Cox models of OS, combining the ICK metagene with other metagenes, in MSI tumors. Forest plot of OS hazard ratio estimated by bivariate Cox analysis of ICK, CTL, CYTOX, Th1, and Immunoscore-like metagene expression in the CIT series, adjusted by TNM stage. All statistical tests were two-sided. CIT = Cartes d’Identité des Tumeurs; CRC = colorectal cancer; CMS = Consensus Molecular Subgroups; CTL = cytotoxic T lymphocytes; high = high expression level; ICK = immune checkpoint genes; MSI = microsatellite instable; MSS = microsatellite stable; OS = overall survival; TCGA = The Cancer Genome Atlas. Figure 1. View largeDownload slide Prognostic value of immune gene expression. A) Schematic representation of the working hypothesis on immune checkpoint prognosis. Arrows represent inhibition. B) Overall survival stratified by microsatellite instable (MSI)/microsatellite stable (MSS) status (left) and Consensus Molecular Subgroups (CMS) subtypes (right). Curves of overall survival (OS) rate (y-axis) according to time from diagnosis (in years; x-axis) were obtained by the method of Kaplan-Meier for both the Cartes d’Identité des Tumeurs (CIT) and The Cancer Genome Atlas (TCGA) series. Differences between survival distributions were assessed by the log-rank test using an end point of five years. C) Prognostic values of immune gene/metagene expression and of clinical factors in MSI tumors. Forest plot of overall survival (OS) hazard ratios (HRs) estimated by combining independent univariate Cox analyses on the CIT and TCGA series, adjusted for TNM stage. Hazard ratio, as well as related Wald test P value and 95% confidence intervals (CIs), are given for metagenes (which aggregate the gene expression values of a gene set related to the four immune categories [immune checkpoints {ICK}, cytotoxic T lymphocytes {CTL}, cytotoxicity, Th1 functional orientation]), individual immune genes, and clinical annotations. Diamonds represent the hazard ratios, and horizontal bars represent the 95% confidence intervals. Red indicates a hazard ratio greater than 1 with a P value of less than .1 (worse prognosis), blue a hazard ratio of less than 1 with a P value of less than .1 (better prognosis), and gray a hazard ratio with a Wald test P value of .1 or less. D) Overall survival stratified by overexpression of immune checkpoint genes within MSI (left), MSS (middle), and both CRCs (right). MSI tumors in the higher quartile of ICK metagene values, in CIT and TCGA series independently, were assigned ICK + (n = 55), and the other tumors ICK- (n = 139). The minimal ICK metagene value within MSI ICK+ tumors was used to divide MSS tumors into ICK + (n = 26) and ICK- (n = 765). Curves of overall survival rate (y-axis) according to time from diagnosis (in years; x-axis) were obtained by the method of Kaplan and Meier for both the CIT and TCGA series. Differences between survival distributions were assessed by the log-rank test using an end point of five years. E) Prognostic value of immune metagenes according to the type of CRC. Each column defines a specific type of CRC (all CRC, MSI CRC, MSS CRC, MSS CMS1 CRC, MSS CMS2 CRC, MSS CMS3 CRC, MSS CMS4 CRC) and the number of related samples with available survival data in the combined CIT + TCGA series. Each row defines a given metagene (ICK, CTL, CYTOX, TH1, IS-like v1, IS-like v2). Each cell shows the hazard ratio for overall survival of the relevant metagene expression for the specific CRC type (obtained via a Cox model). Each cell also contains the P value for the Wald test for equality of this hazard ratio to 1. The hazard ratio shown in each cell was calculated by combining the hazard ratios for each of the CIT and TCGA data sets using the meta-analysis approach of DerSimonian. The genes relevant to each metagene are indicated in (C) (ICK, CTL, CYTOX, TH1) or are given below the metagene name. F) Bivariate Cox models of OS, combining the ICK metagene with other metagenes, in MSI tumors. Forest plot of OS hazard ratio estimated by bivariate Cox analysis of ICK, CTL, CYTOX, Th1, and Immunoscore-like metagene expression in the CIT series, adjusted by TNM stage. All statistical tests were two-sided. CIT = Cartes d’Identité des Tumeurs; CRC = colorectal cancer; CMS = Consensus Molecular Subgroups; CTL = cytotoxic T lymphocytes; high = high expression level; ICK = immune checkpoint genes; MSI = microsatellite instable; MSS = microsatellite stable; OS = overall survival; TCGA = The Cancer Genome Atlas. As a preliminary step, the prognostic values of MSI and CMS status were confirmed with, as expected, an improved prognosis for patients with MSI CRCs, compared with those with MSS CRCs, and a statistically significant prognostic value of the CMS classification (Figure 1B;Supplementary Figure 2, available online). The prognostic values of the 28 immune markers mentioned and the four metagenes were then analyzed by univariate Cox models. Most ICK genes were individually associated with poorer prognosis in MSI CRC, as reflected by the ICK metagene (HR =  3.46, 95% CI = 1.41 to 8.49, P = .007) (Figure 1, C–E; Supplementary Figure 3A, available online). This association remained statistically significant in non-metastatic MSI CRC (Supplementary Figure 3B, available online). No prognostic association was observed in patients with MSS CRCs, either for individual ICK markers or for the ICK metagene (Figure 1, D and E; Supplementary Figure 3C, available online). Subdividing MSS CRCs by CMS did not change this result, except in CMS3, where the ICK metagene was associated with better prognosis (HR = 0.46, 95% CI = 0.21 to 1.00, P = .05) (Figure 1E;Supplementary Table 2, available online). As expected, the expression of most Immunoscore-like metagenes was associated with improved outcome of CRC patients (Figure 1E;Supplementary Table 2, available online). However, when CMS and MSI/MSS status were taken into account, expression of these Immunoscore surrogates was predictive of better prognosis only in MSS tumors from CMS2 and CMS3 (HR = 0.63, 95% CI = 0.43 to 0.92, P = .02; and HR = 0.55, 95% CI = 0.34 to 0.90, P = .02, respectively). In contrast, they had no prognostic relevance in CRC from CMS1 and CMS4 (HR =  0.46, 95% CI = 0.10 to 2.10, P = .32; and HR = 1.13, 95% CI = 0.79 to 1.63, P = .50, respectively) (Figure 1E;Supplementary Table 2, available online). In univariate models, the overexpression of CTL/Th1/cytotoxicity/Immunoscore markers and metagenes was also associated with adverse prognosis in MSI CRC (Figure 1, C and E; Supplementary Figure 3, A and B, available online). We therefore hypothesized that high expression levels of ICKs in MSI CRC could counterbalance and mask the expected positive effect of CTL/Th1/cytotoxic cells or Immunoscore-related cells on prognosis. Bivariate Cox models at the metagene level were consistent with this assumption (Figure 1F). All together, these data underline that ICKs and Immunoscore biomarkers constitute an independent prognostic factor for overall survival in MSI and MSS tumors, respectively. Expression and Prognostic Value of Immune Checkpoints in an Independent Metastatic MSI CRC Patient Series The CIT and TCGA series included mostly non-metastatic MSI CRC patients (n = 220/232, 94.8%). Because ICK blockade was recently proposed as a promising new therapy for metastatic MSI CRC, we endeavored to further evaluate the prognostic relevance of ICK expression in an additional retrospective, multisite series of 28 stage IV primary MSI CRCs treated with standard care that were analyzed separately for gene expression as part of an independent, ongoing study using the NanoString nCounter platform. To do this, we analyzed the expression of the seven ICKs that were previously screened (CD274, PDCD1LG2, HAVCR2, LAG3, ICOS, CTLA4, PDCD1). As with non-metastatic MSI colon tumors, we observed statistically significant association of PD-L1 (CD274) expression with worse OS and worse survival after relapse (SAR) (Figure 2A). Associations of TIM-3 (HAVCR2) and LAG3 expression, although not statistically significant, were suggested. Again, the overexpression of metagenes corresponding to CTL/Th1/cytotoxicity/Immunoscore markers (CTL = CD3D, CD3G, CD8A, PTPRC; CYTOX = GNLY, GZMK; Th1= IFNG) tended to associate with adverse prognosis in univariate models (Figure 2B). However, in bivariate Cox models, the expression of CTL/Th1/cytotoxicity/Immunoscore-related metagenes was associated with good prognosis (hazard ratio [HR] < 1.00), as expected (Figure 2C). Several ICK markers, in particular the druggable PD-L1 (CD274) and TIM-3 (HAVCR2) molecules (23), are therefore likely to constitute biomarkers for poor prognosis in both metastatic and non-metastatic MSI CRC patients. Figure 2. View largeDownload slide Prognostic value of immune checkpoints in an independent metastatic microsatellite instable (MSI) colorectal cancer (CRC) patient series. A) Prognostic value of ICK gene expression in metastatic MSI tumors. Forest plot of hazard ratios (HRs) estimated by univariate Cox analysis of overall survival (OS; left panel) and survival after relapse (SAR; right panel) on all ICK genes available in the NanoString data. Diamonds represent hazard ratio estimates, and bars the related 95% confidence intervals (CIs). Red indicates a worse prognosis hazard ratio, blue a better prognosis, and gray a hazard ratio with a Wald test P value of .1 or less. B) Prognostic value of metagene expression in MSI metastatic tumors. Forest plot of hazard ratios estimated by univariate Cox analysis of survival after relapse for metagene expression. The ICK metagene aggregates the three most associated ICK genes in univariate analysis (HAVCR2, CD274, and LAG3). The other metagene aggregate genes available in the NanoString data set for the corresponding category were CTL = CD3D, CD3G, CD8A, PTPRC; CYTOX = GNLY, GZMK; Th1= IFNG. Metagene expression values were calculated by averaging the expression values obtained within the corresponding gene set. C) Bivariate Cox models of OS and SAR, combining the ICK metagene with other metagenes, in metastatic MSI tumors. Forest plot of SAR hazard ratios estimated by bivariate Cox analysis of ICK, CTL, CYTOX, Th1, and Immunoscore-like metagene expression in the independent MSI metastatic series. All statistical tests were two-sided. CI = confidence interval; CTL = cytotoxic T lymphocytes; HR = hazard ratio; ICK = immune checkpoint genes; OS = overall survival; SAR = survival after relapse. Figure 2. View largeDownload slide Prognostic value of immune checkpoints in an independent metastatic microsatellite instable (MSI) colorectal cancer (CRC) patient series. A) Prognostic value of ICK gene expression in metastatic MSI tumors. Forest plot of hazard ratios (HRs) estimated by univariate Cox analysis of overall survival (OS; left panel) and survival after relapse (SAR; right panel) on all ICK genes available in the NanoString data. Diamonds represent hazard ratio estimates, and bars the related 95% confidence intervals (CIs). Red indicates a worse prognosis hazard ratio, blue a better prognosis, and gray a hazard ratio with a Wald test P value of .1 or less. B) Prognostic value of metagene expression in MSI metastatic tumors. Forest plot of hazard ratios estimated by univariate Cox analysis of survival after relapse for metagene expression. The ICK metagene aggregates the three most associated ICK genes in univariate analysis (HAVCR2, CD274, and LAG3). The other metagene aggregate genes available in the NanoString data set for the corresponding category were CTL = CD3D, CD3G, CD8A, PTPRC; CYTOX = GNLY, GZMK; Th1= IFNG. Metagene expression values were calculated by averaging the expression values obtained within the corresponding gene set. C) Bivariate Cox models of OS and SAR, combining the ICK metagene with other metagenes, in metastatic MSI tumors. Forest plot of SAR hazard ratios estimated by bivariate Cox analysis of ICK, CTL, CYTOX, Th1, and Immunoscore-like metagene expression in the independent MSI metastatic series. All statistical tests were two-sided. CI = confidence interval; CTL = cytotoxic T lymphocytes; HR = hazard ratio; ICK = immune checkpoint genes; OS = overall survival; SAR = survival after relapse. Immune Checkpoint Gene Expression Distribution in Colorectal Cancer We next investigated the level of variation in ICK expression among CRC tumors in order to further assess their relevance as prognostic and theranostic markers. ICK expression was analyzed in stage I through IV CRC and in nontumor colonic mucosa from our CIT cohort (590 CRC, 56 NT) and in the TCGA cohort (613 CRC, 51 NT). In both cohorts, the metagenes corresponding to ICKs, CTL, cytotoxicity, and Th1 orientation were overexpressed in MSI and in MSS tumors belonging to CMS1 and CMS4, as compared with MSS CRCs from CMS2 and CMS3 (Figure 3A). Variable expression of ICKs relative to NT was noted in all CMS subtypes in both cohorts (Figure 3B). A high degree of heterogeneity was observed in CMS1 tumors (Figure 3B), particularly in MSI tumors, where high to very high expression levels of ICKs were observed in a large proportion of cases (Figure 3B). Figure 3. View largeDownload slide Immune gene expression distribution in colorectal cancer (CRC). A) Immune metagene expression relative to nontumoral colonic mucosa across CRC subtypes. Height panels show expression values for four metagenes (columns) in two series (rows): Cartes d’Identité des Tumeurs (CIT; top) and The Cancer Genome Atlas (TCGA; bottom). Each metagene aggregates the gene expression values of a gene set related to one of four immune categories (immune checkpoints [ICK], cytotoxic T lymphocytes [CTL], cytotoxicity, Th1 functional orientation). Gene (and metagene) expression values are relative to the median expression in nontumor samples. Within each panel, the metagene values are presented in six boxplots related to six sample types (nontumor tissue [NT], MSI tumors, MSS tumors subdivided into the four consensus molecular subtypes [CMS]). P values were calculated using a Student’s t test. P values were calculated using a Student’s t test to assess the statistical significance of differences between MSI and NT and between MSI and MSS/CMS (***P < .001, **P < .01, *P < .05). B) Heterogeneous expression of ICK genes among MSI and MSS CRCs. Heatmap shows the gene expression z-scores of ICK genes in MSI and MSS tumor samples in CIT (top) and TCGA (bottom) series. Central and dispersion parameters used to generate z-scores correspond to the median and median absolute deviation (mad) in NT samples. Samples are ordered according to MSI/MSS status, CMS within MSS, and median expression value of immune checkpoint genes. C) Correlation of expression of immune genes within MSI tumors. Heatmap of Pearson coefficients of correlation between expression values of immune genes within MSI tumors of CIT (top) and TCGA (bottom) series separately. Genes are ordered according to four immune categories (ICK, cytotoxic T lymphocytes [CTL], cytotoxicity, Th1 functional orientation). D) Heterogeneous expression of ICK genes among MSI metastatic tumors. Heatmaps show gene expression of ICK genes in primary tumor samples from an independent metastatic MSI CRC series (bottom), compared with a subset of the CIT MSI series (top). Expression values were obtained using NanoString technology, where only a subset of immune genes was evaluated. NT-relative gene expression ratios are displayed. All statistical tests were two-sided. CIT = Cartes d’Identité des Tumeurs; CMS = Consensus Molecular Subgroups; CTL = cytotoxic T lymphocytes; ICK = immune checkpoint genes; high = high expression level; MSI = microsatellite instable; MSS = microsatellite stable; NT = nontumoral tissue samples; TCGA = The Cancer Genome Atlas. Figure 3. View largeDownload slide Immune gene expression distribution in colorectal cancer (CRC). A) Immune metagene expression relative to nontumoral colonic mucosa across CRC subtypes. Height panels show expression values for four metagenes (columns) in two series (rows): Cartes d’Identité des Tumeurs (CIT; top) and The Cancer Genome Atlas (TCGA; bottom). Each metagene aggregates the gene expression values of a gene set related to one of four immune categories (immune checkpoints [ICK], cytotoxic T lymphocytes [CTL], cytotoxicity, Th1 functional orientation). Gene (and metagene) expression values are relative to the median expression in nontumor samples. Within each panel, the metagene values are presented in six boxplots related to six sample types (nontumor tissue [NT], MSI tumors, MSS tumors subdivided into the four consensus molecular subtypes [CMS]). P values were calculated using a Student’s t test. P values were calculated using a Student’s t test to assess the statistical significance of differences between MSI and NT and between MSI and MSS/CMS (***P < .001, **P < .01, *P < .05). B) Heterogeneous expression of ICK genes among MSI and MSS CRCs. Heatmap shows the gene expression z-scores of ICK genes in MSI and MSS tumor samples in CIT (top) and TCGA (bottom) series. Central and dispersion parameters used to generate z-scores correspond to the median and median absolute deviation (mad) in NT samples. Samples are ordered according to MSI/MSS status, CMS within MSS, and median expression value of immune checkpoint genes. C) Correlation of expression of immune genes within MSI tumors. Heatmap of Pearson coefficients of correlation between expression values of immune genes within MSI tumors of CIT (top) and TCGA (bottom) series separately. Genes are ordered according to four immune categories (ICK, cytotoxic T lymphocytes [CTL], cytotoxicity, Th1 functional orientation). D) Heterogeneous expression of ICK genes among MSI metastatic tumors. Heatmaps show gene expression of ICK genes in primary tumor samples from an independent metastatic MSI CRC series (bottom), compared with a subset of the CIT MSI series (top). Expression values were obtained using NanoString technology, where only a subset of immune genes was evaluated. NT-relative gene expression ratios are displayed. All statistical tests were two-sided. CIT = Cartes d’Identité des Tumeurs; CMS = Consensus Molecular Subgroups; CTL = cytotoxic T lymphocytes; ICK = immune checkpoint genes; high = high expression level; MSI = microsatellite instable; MSS = microsatellite stable; NT = nontumoral tissue samples; TCGA = The Cancer Genome Atlas. Expression levels for all of the 28 immune markers were highly correlated in the MSI CRCs from both cohorts (Figure 3C). The present results highlight the extent of heterogeneity of MSI CRC with respect to immunity and to the overexpression of ICK molecules. This was observed regardless of MSI CRC origin (inherited or sporadic) or of other clinical or molecular parameters such as sex, tumor location, tumor stage, CMS, or KRAS/BRAF mutations (Supplementary Figure 4, available online). Considerable variation in the expression of ICK markers was also observed in the independent metastatic MSI CRC series evaluated by NanoString (Figure 3D). Functional Relevance of Immune Checkpoint Expression in CRC We next addressed the possible physiological relevance of ICK overexpression in CRC. Tumor infiltration by immune cells was quantified using MCP counter software (28) in both the CIT and TCGA cohorts. A strong correlation was observed between ICK expression and infiltration by lymphoid (NK, T cells, cytotoxic cells) and myeloid cells. In contrast, B cells, fibroblasts, vessels, and granulocytes were less, or not at all, associated with ICK expression (Figure 4A). These results suggest that ICK expression occurs in response to an efficient in situ adaptive T cell immune response. Pathway enrichment analysis was performed to compare the expression profiles of MSI tumors with low vs high ICK expression levels. Statistically significant associations were observed between ICK expression and immune response gene sets, including positive activation of T cell response, negative regulation of T cell activation, T cell exhaustion, IL-10 response, and chronic viral infection (29) (Figure 4B). Hence, we conclude that there is a strong correlation between ICK expression and the presence of an exhausted T cell immune response in MSI CRC. Figure 4. View largeDownload slide Functional relevance of immune checkpoint expression in colorectal cancer (CRC). A) Transcriptome-based abundance estimates of microenvironment cell populations (MCPcounter) and relation to ICK gene expression. Heatmap showing the abundance estimates for nine immune (Lymphoid, natural killer [NK] or T, T derived, NK, Cytotoxic, B derived, myeloid, monocyte derived, granulocyte) and two nonimmune (vessels, fibroblasts) cell populations in microsatellite instable (MSI) tumors and microsatellite stable (MSS) Consensus Molecular Subgroups (CMS) in the Cartes d’Identité des Tumeurs (CIT; left) and The Cancer Genome Atlas (TCGA; right) series. For each of these 11 cell populations, abundance estimates were calculated using MCP counter software as the average NT-relative expression ratios of corresponding markers. ICK metagene expression and CMS classification are also shown. MSI tumor samples are divided into three groups according to ICK metagene expression (high, intermediate, and low). These three groups are compared with CMS subtypes; analysis of variance test P values are reported on the right of the heatmap. Associations between abundance estimates of immune cell populations and the ICK metagene expression within MSI tumors were measured by Pearson's correlations. Coefficients and associated P values are reported on the right of the heatmap. For MSS tumors, the median values within each CMS are displayed. Samples are ordered according to the ICK metagene values. B) Functional associations with overexpression of ICK genes. MSI samples are divided into three groups according to ICK metagene expression (high, intermediate, and low). Heatmap showing the expression score of a selection of immune gene sets statistically significantly enriched in genes that were differentially expressed between high- and low-ICK-expressing MSI tumor groups, in CIT (left) and TCGA (right) series. For each gene set, expression scores are calculated as the mean NT-relative expression ratios across all genes in the gene set. Pearson's correlation coefficients and associated P values are reported. C) Representative slide of PD-L1 and CD8 immunohistochemical (IHC) staining of MSI CRC with high expression of ICK genes on transcriptomic data. Black scale bar is 1 mm. Yellow Scale bar is 100 µm. D) CD8 cell density according to PD-L1 expression within tumors, stroma, and healthy mucosa. Upper panel: number of CD8 cells per mm² in intratumoral area, peritumoral stromal area, and healthy mucosa in patients with high PD-L1 expression vs low or no PD-L1 expression. Lower panel: correlation between semiquantitative PD-L1 expression with IHC staining and CD8 number per mm² in intratumoral area, peritumoral stromal area, and healthy mucosa. E) Representative slide of CD8 cell infiltration and PD-L1 expression. Left panel: representative slide of CD8, DAPI, and Ki67 staining with classical composite immunofluorescence image and phenotype map using Inform software to recognize tumor cells, proliferating and nonproliferating CD8 cells, and other cells. Scale bar is 100 µm. Right panel: quantification of proliferating CD8 or nonproliferating CD8 cells in PDL-L1 high vs low or negative tumor. All statistical tests were two-sided. Error bars represent standard deviation. CIT = Cartes d’Identité des Tumeurs; CMS = Consensus Molecular Subgroups; ICK = immune checkpoint genes; MSI = microsatellite instable; MSS = microsatellite stable; TCGA = The Cancer Genome Atlas. Figure 4. View largeDownload slide Functional relevance of immune checkpoint expression in colorectal cancer (CRC). A) Transcriptome-based abundance estimates of microenvironment cell populations (MCPcounter) and relation to ICK gene expression. Heatmap showing the abundance estimates for nine immune (Lymphoid, natural killer [NK] or T, T derived, NK, Cytotoxic, B derived, myeloid, monocyte derived, granulocyte) and two nonimmune (vessels, fibroblasts) cell populations in microsatellite instable (MSI) tumors and microsatellite stable (MSS) Consensus Molecular Subgroups (CMS) in the Cartes d’Identité des Tumeurs (CIT; left) and The Cancer Genome Atlas (TCGA; right) series. For each of these 11 cell populations, abundance estimates were calculated using MCP counter software as the average NT-relative expression ratios of corresponding markers. ICK metagene expression and CMS classification are also shown. MSI tumor samples are divided into three groups according to ICK metagene expression (high, intermediate, and low). These three groups are compared with CMS subtypes; analysis of variance test P values are reported on the right of the heatmap. Associations between abundance estimates of immune cell populations and the ICK metagene expression within MSI tumors were measured by Pearson's correlations. Coefficients and associated P values are reported on the right of the heatmap. For MSS tumors, the median values within each CMS are displayed. Samples are ordered according to the ICK metagene values. B) Functional associations with overexpression of ICK genes. MSI samples are divided into three groups according to ICK metagene expression (high, intermediate, and low). Heatmap showing the expression score of a selection of immune gene sets statistically significantly enriched in genes that were differentially expressed between high- and low-ICK-expressing MSI tumor groups, in CIT (left) and TCGA (right) series. For each gene set, expression scores are calculated as the mean NT-relative expression ratios across all genes in the gene set. Pearson's correlation coefficients and associated P values are reported. C) Representative slide of PD-L1 and CD8 immunohistochemical (IHC) staining of MSI CRC with high expression of ICK genes on transcriptomic data. Black scale bar is 1 mm. Yellow Scale bar is 100 µm. D) CD8 cell density according to PD-L1 expression within tumors, stroma, and healthy mucosa. Upper panel: number of CD8 cells per mm² in intratumoral area, peritumoral stromal area, and healthy mucosa in patients with high PD-L1 expression vs low or no PD-L1 expression. Lower panel: correlation between semiquantitative PD-L1 expression with IHC staining and CD8 number per mm² in intratumoral area, peritumoral stromal area, and healthy mucosa. E) Representative slide of CD8 cell infiltration and PD-L1 expression. Left panel: representative slide of CD8, DAPI, and Ki67 staining with classical composite immunofluorescence image and phenotype map using Inform software to recognize tumor cells, proliferating and nonproliferating CD8 cells, and other cells. Scale bar is 100 µm. Right panel: quantification of proliferating CD8 or nonproliferating CD8 cells in PDL-L1 high vs low or negative tumor. All statistical tests were two-sided. Error bars represent standard deviation. CIT = Cartes d’Identité des Tumeurs; CMS = Consensus Molecular Subgroups; ICK = immune checkpoint genes; MSI = microsatellite instable; MSS = microsatellite stable; TCGA = The Cancer Genome Atlas. To further investigate the functional relevance of ICKs in MSI tumors, we studied eight primary MSI tumors showing upregulation of ICKs and 12 without. PD-L1 and CD8 expression were examined using immunohistochemistry (IHC). PD-L1 expression was observed only in the tumor bed, whereas CD8 was present both in the tumor core and in stromal areas (Figure 4C). Moreover, PD-L1 expression correlated strongly with ICK expression (Supplementary Figure 5, available online), while we observed concomitant increases of CD8 infiltrates in both the tumor bed and in the peritumoral stroma with PD-L1 IHC staining (Figure 4D). Proliferation and functional activity of CD8 T cells were then determined using multiparametric immunofluorescence microscopy (Figure 4E, left panel). CD8 T cells that were close to or in contact with PD-L1-expressing tumors were less proliferative, as observed with Ki67 labeling (median = 0.56 in ICK low vs median = 0.34 in ICK high, P = .004) (Figure 4E, right panel). These results indicate that interactions between CD8 T cells and ICK ligands in MSI primary tumors can impede CD8 T cell function. Discussion During cancer progression, tumor-infiltrating T cells have been shown to display increased chronic expression of different antagonist ICKs, causing functional exhaustion and unresponsiveness of T cells (30). The exhausted CD8 T cells fail to proliferate in response to antigen and lack critical anticancer effector functions (31). Checkpoint inhibitors could boost the anticancer immune response, and the potential relevance of these inhibitors for the treatment of metastatic MSI CRC patients was highlighted (16). In the present study, we showed that ICK overexpression represents a more accurate prognostic biomarker for MSI CRC patients treated with standard care than the classical assessment of T cell number by Immunoscore (1). This may be explained by the presence of exhausted nonproliferative CD8 T cells in the core of these neoplasms. More generally, our data indicate that assessment of the prognostic significance of antitumor immunity in CRC needs to take into account ICK expression. This is particularly relevant for colon tumors displaying immunogenic profiles with both high Immunoscores and ICK expression, such as in MSI tumors and probably a substantial proportion of MSS CRCs. The current results were obtained using univariate Cox models for survival analysis and a transcriptome-based method to quantify both ICK and CTL/Th1/cytotoxicity (Immunoscore) markers. We validated our method by building Immunoscore-like surrogates that were associated with statistically significantly improved survival of CRC patients. Under the same conditions, the CTL/Th1/cytotoxicity and Immunoscore markers were both associated with worse prognostic in MSI CRC. These results are potentially in conflict with a recent publication that observed a statistically significant association between Immunoscore and improved outcome in a single series of 105 MSI CRC patients (17). Although the studies are not directly comparable, here we assessed three independent cohorts of CRC patients, totaling more than 1200 cases and including 260 MSI CRCs. However, it does not include classical Immunoscore evaluation by immunohistochemistry. Bivariate Cox analysis, combining ICK metagene with CTL or Th1 or cytotoxicity or Immunoscore-like metagene, revealed that expression of metagenes related to CTL/Th1/cytotoxicity and Immunoscore markers was associated with trends for better prognosis in MSI CRC from both the CIT and NanoString metastatic series, whereas the ICK metagene was statistically significantly associated with worse prognosis. However, this association was not observed in the TCGA cohort as it had too few events (n = 5) to get sufficient statistical power. Therefore, validation in another larger cohort is required. In contrast with the earlier study that focused only on the PD1/PDL1 couple (17), a more global assessment of ICK gene expression in the tumor core, as proposed in the present study, allows a more holistic view of the T cell immune response in CRC. Moreover, the transcriptome-based method reported here is easier to use and more amenable to standardization. It can be also used to test publicly available clinical data sets, whereas this is not possible with Immunoscore because of the need to assess primary tumor samples. Recent clinical trials have demonstrated that antibodies targeting PD-1 or PD-L1 can induce a major response in many types of cancers (32). The overall survival rate with more than five years’ follow-up for stage II and III MSI CRC patients is approximately 70% without adjuvant chemotherapy and 75% to 90% with standard care adjuvant chemotherapy (33–35), whereas it is less than 5% for stage IV MSI CRC patients. We report here for the first time the prognostic significance of ICK overexpression in both metastatic and nonmetastatic MSI CRC and in the absence of immunotherapy. These findings should help to better inform the prognosis of MSI CRC patients. They may be useful for guiding future immunotherapy involving antibody blockade of ICKs in nonmetastatic MSI CRC patients and for having predictive factors of immunotherapy efficacy for patients with metastatic disease. The limitations of our study are mainly related to the fact that it is not yet ready for routine clinical use. First, it will be necessary to confirm in large and prospective series of patients from randomized trials that the outcome of CRC patients is determined by the CTL/ICK balance. Second, it will be necessary to further assess the prognostic relevance of combined ICK and CTL marker expression in MSI CRC. This should allow patients who are at high risk of relapse to be identified and thus indicated for more intensive treatments linked to these molecular markers. Such work will help to confirm ICK molecules as clinically relevant theranostic factors for the treatment of MSI CRC. To conclude, our results highlight the extent of heterogeneity of CRC with respect to immunity, and the overexpression of ICK molecules in particular. They suggest that prediction of CRC patient outcomes through evaluation of immune components in the tumor microenvironment will likely be improved by the integration of ICK markers, the prognosis of colon tumors being determined by the CTL/ICK balance (Figure 5). More particularly, our results indicate that ICK expression impacts the prognosis of MSI tumors. To inform future immunotherapy involving antibody blockade of ICKs and resistance to these molecules in MSI CRC patients, additional studies on the molecular mechanisms underlying the immune reaction against MSI tumor cells are required, for example, the number and type of MSI-driven mutational events that drive the synthesis of aberrant, immunogenic peptides (36). Identifying these somatic events with respect to quantitative and qualitative antitumoral immunity may improve the personalized treatment of MSI CRC patients with ICK inhibitors, in both metastatic and nonmetastatic settings. Figure 5. View largeDownload slide Immune model for prognostication of colorectal cancer (CRC) patients. The diagram shows a model of decision tree based on the three criteria, fibroblast presence, immune infiltration, and immune checkpoint expression, to define the four distinct prognosis groups according to immunity observed in CRC patients. Expected prognosis, relevant markers, subtypes, and microsatellite instable status are given for each group. CMS = Consensus Molecular Subgroups; CTL = cytotoxic T lymphocytes; ICK = immune checkpoint genes; high = high expression level; MSI = microsatellite instable; MSS = microsatellite stable. Figure 5. View largeDownload slide Immune model for prognostication of colorectal cancer (CRC) patients. The diagram shows a model of decision tree based on the three criteria, fibroblast presence, immune infiltration, and immune checkpoint expression, to define the four distinct prognosis groups according to immunity observed in CRC patients. Expected prognosis, relevant markers, subtypes, and microsatellite instable status are given for each group. CMS = Consensus Molecular Subgroups; CTL = cytotoxic T lymphocytes; ICK = immune checkpoint genes; high = high expression level; MSI = microsatellite instable; MSS = microsatellite stable. Funding This work was supported by the Inserm Institute, UPMC, and the “Cartes d'Identité des Tumeurs” (CIT) research program, funded and directed by the “Ligue Nationale Contre le Cancer,” and by grants from the “Institut National du Cancer” (INCa to AD). AD group has the label “La Ligue Contre le Cancer.” Notes Laetitia Marisa, Magali Svrcek, Ada Collura, Etienne Becht, Pascale Cervera, Kristell Wanherdrick, Olivier Buhard, Anastasia Goloudina, Vincent Jonchère, Janick Selves, Gerard Milano, Dominique Guenot, Romain Cohen, Chrystelle Colas, Pierre Laurent-Puig, Sylviane Olschwang, Jérémie H. Lefèvre, Yann Parc, Valérie Boige, Côme Lepage, Thierry André, Jean-François Fléjou, Valentin Dérangère, François Ghiringhelli*, Aurélien de Reynies*, Alex Duval* *Authors contributed equally to this work. Affiliations of authors: Programme “Cartes d'Identité des Tumeurs,” Ligue Nationale Contre le Cancer, Paris, France (LM, EB, AdR); INSERM, UMRS 938 - Centre de Recherche Saint-Antoine, Equipe “Instabilité des Microsatellites et Cancers,” Equipe labellisée par la Ligue Nationale contre le Cancer, Paris, France (MS, AC, PC, KW, OB, AG, VJ, RC, CC, JHL, YP, TA, JFF, AD); Sorbonne Université, UPMC Univ Paris 06, Paris, France (MS, AC, PC, KW, OB, AG, VJ, RC, CC, JHL, YP, TA, JFF, AD); AP-HP, Hôpital Saint-Antoine, Service d’Anatomie et Cytologie Pathologiques, Paris, France (MS, PC, JFF); Centre de Recherche en Cancérologie de Toulouse, UMR 1037 INSERM - Université Toulouse III, Department of Pathology, CHU, Toulouse, France (JS); Laboratoire d'Oncopharmacologie, EA 3836, Centre Antoine Lacassagne, Nice, France (GM); INSERM, U682, Développement et Physiopathologie de l’Intestin et du Pancréas, Strasbourg, France (DG); AP-HP, Hôpital Saint-Antoine, Service d’Oncologie Médicale, Paris, France (RC, TA); AP-HP, Laboratoire d’oncogénétique et d’Angiogénétique, GH Pitié-Salpétrière, Paris, France (CC); INSERM, Unité Mixte de Recherche, Paris Sorbonne Cité, Université Paris Descartes, Paris, France (PLP); Aix Marseille Univ, INSERM, GMGF, Marseille, France and RGDS, HP Clairval, Marseille, France (SO); AP-HP, Service de Chirurgie Générale et Digestive, Hôpital Saint-Antoine, Paris, France (JHL, YP); Department of Oncologic Medicine, Gustave-Roussy, Villejuif, France (VB); Université Paris Descartes, Paris Sorbonne Cité INSERM UMR-S775, Paris, France (VB); INSERM, Burgundy Cancer Registry, U866, Burgundy University, Dijon University Hospital, BP 87900 21079 Dijon, France (CL); Department of Medical Oncology, Centre Georges-François Leclerc, Dijon, France (VD, FG); INSERM, UMR866, Burgundy University (VD, FG); Platform of transfer in oncology, Burgundy University, Centre Georges-François Leclerc, Dijon, France (VD, FG). The funder, the French League Against Cancer, participated in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; and the decision to submit the manuscript for publication. We thank Prof. W. H. Fridman for his scientific assistance. Study concept and design: L. Marisa, F. Ghiringhelli, A. de Reynies, A. Duval; acquisition of data: all other authors; analysis and interpretation of data: L. Marisa, V. Dérengère, E. Becht, F. Ghiringhelli, A. de Reynies, A. Duval; drafting of the manuscript: L. Marisa, M. Svrcek, A. Collura, E. Becht, P. Cervera, K. Wanherdrick, O. Buhard, A. Goloudina, V. Jonchère, J. Selves, G. Milano, D. Guenot, R. Cohen, C. Colas, P. Laurent-Puig, S. Olschwang, J. H Lefèvre, Y. Parc, V. Boige, C. Lepage, T. André, JF Fléjou, V. Dérangère, F. Ghiringhelli, A. de Reynies, A Duval; statistical analysis: L. Marisa, A. de Reynies; study supervision: A. Duval; biological material suppliers: all other authors. References 1 Galon J, Costes A, Sanchez-Cabo F, et al.   Type, density, and location of immune cells within human colorectal tumors predict clinical outcome. Science.  2006; 313( 5795): 1960– 1964. Google Scholar CrossRef Search ADS PubMed  2 Angell HK, Gray N, Womack C, Pritchard DI, Wilkinson RW, Cumberbatch M. Digital pattern recognition-based image analysis quantifies immune infiltrates in distinct tissue regions of colorectal cancer and identifies a metastatic phenotype. Br J Cancer.  2013; 109( 6): 1618– 1624. Google Scholar CrossRef Search ADS PubMed  3 Mlecnik B, Tosolini M, Kirilovsky A, et al.   Histopathologic-based prognostic factors of colorectal cancers are associated with the state of the local immune reaction. J Clin Oncol.  2011; 29( 6): 610– 618. Google Scholar CrossRef Search ADS PubMed  4 Lothe RA, Peltomaki P, Meling GI, et al.   Genomic instability in colorectal cancer: relationship to clinicopathological variables and family history. Cancer Res.  1993; 53( 24): 5849– 5852. Google Scholar PubMed  5 Gryfe R, Kim H, Hsieh ET, et al.   Tumor microsatellite instability and clinical outcome in young patients with colorectal cancer. N Engl J Med.  2000; 342( 2): 69– 77. Google Scholar CrossRef Search ADS PubMed  6 Bauer K, Nelius N, Reuschenbach M, et al.   T cell responses against microsatellite instability-induced frameshift peptides and influence of regulatory T cells in colorectal cancer. Cancer Immunol Immunother.  2013; 62( 1): 27– 37. Google Scholar CrossRef Search ADS PubMed  7 Aaltonen LA, Peltomaki P, Leach FS, et al.   Clues to the pathogenesis of familial colorectal cancer. Science.  1993; 260( 5109): 812– 816. Google Scholar CrossRef Search ADS PubMed  8 Leach FS, Nicolaides NC, Papadopoulos N, et al.   Mutations of a mutS homolog in hereditary nonpolyposis colorectal cancer. Cell.  1993; 75( 6): 1215– 1225. Google Scholar CrossRef Search ADS PubMed  9 Thibodeau SN, Bren G, Schaid D. Microsatellite instability in cancer of the proximal colon. Science.  1993; 260( 5109): 816– 819. Google Scholar CrossRef Search ADS PubMed  10 Ionov Y, Peinado MA, Malkhosyan S, Shibata D, Perucho M. Ubiquitous somatic mutations in simple repeated sequences reveal a new mechanism for colonic carcinogenesis. Nature.  1993; 363( 6429): 558– 561. Google Scholar CrossRef Search ADS PubMed  11 Fishel R, Lescoe MK, Rao MR, et al.   The human mutator gene homolog MSH2 and its association with hereditary nonpolyposis colon cancer. Cell.  1993; 75( 5): 1027– 1038. Google Scholar CrossRef Search ADS PubMed  12 Dorard C, de Thonel A, Collura A, et al.   Expression of a mutant HSP110 sensitizes colorectal cancer cells to chemotherapy and improves disease prognosis. Nat Med.  2011; 17: 1283– 1289. Google Scholar CrossRef Search ADS PubMed  13 Duval A, Hamelin R. Mutations at coding repeat sequences in mismatch repair-deficient human cancers: Toward a new concept of target genes for instability. Cancer Res.  2002; 62( 9): 2447– 2454. Google Scholar PubMed  14 Hamelin R, Chalastanis A, Colas C, et al.   [Clinical and molecular consequences of microsatellite instability in human cancers]. Bull Cancer.  2008; 95( 1): 121– 132. Google Scholar PubMed  15 Llosa NJ, Cruise M, Tam A, et al.   The vigorous immune microenvironment of microsatellite instable colon cancer is balanced by multiple counter-inhibitory checkpoints. Cancer Discov.  2015; 5( 1): 43– 51. 16 Le DT, Uram JN, Wang H, et al.   PD-1 blockade in tumors with mismatch-repair deficiency. N Engl J Med.  2015; 372( 26): 2509– 2520. Google Scholar CrossRef Search ADS PubMed  17 Mlecnik B, Bindea G, Angell HK, et al.   Integrative analyses of colorectal cancer show immunoscore is a stronger predictor of patient survival than microsatellite instability. Immunity.  2016; 44( 3): 698– 711. Google Scholar CrossRef Search ADS PubMed  18 Dunne PD, McArt DG, O'Reilly PG, et al.   Immune-derived PD-L1 gene expression defines a subgroup of stage II/III colorectal cancer patients with favorable prognosis who may be harmed by adjuvant chemotherapy. Cancer Immunol Res.  2016; 4( 7): 582– 591. 19 Masugi Y, Nishihara R, Yang J, et al.   Tumour CD274 (PD-L1) expression and T cells in colorectal cancer. Gut.  2016; in press. 20 Lee LH, Cavalcanti MS, Segal NH, et al.   Patterns and prognostic relevance of PD-1 and PD-L1 expression in colorectal carcinoma. Mod Pathol.  2016; in press. 21 Li Y, Liang L, Dai W, et al.   Prognostic impact of programed cell death-1 (PD-1) and PD-ligand 1 (PD-L1) expression in cancer cells and tumor infiltrating lymphocytes in colorectal cancer. Mol Cancer.  2016; 15( 1): 55. Google Scholar CrossRef Search ADS PubMed  22 Guinney J, Dienstmann R, Wang X, et al.   The consensus molecular subtypes of colorectal cancer. Nat Med.  2015; 21( 11): 1350– 1356. Google Scholar CrossRef Search ADS PubMed  23 Melero I, Berman DM, Aznar MA, Korman AJ, Perez Gracia JL, Haanen J. Evolving synergistic combinations of targeted immunotherapies to combat cancer. Nat Rev Cancer.  2015; 15( 8): 457– 472. Google Scholar CrossRef Search ADS PubMed  24 Fridman WH, Pages F, Sautes-Fridman C, Galon J. The immune contexture in human tumours: Impact on clinical outcome. Nat Rev Cancer.  2012; 12( 4): 298– 306. Google Scholar CrossRef Search ADS PubMed  25 Marisa L, de Reynies A, Duval A, et al.   Gene expression classification of colon cancer into molecular subtypes: Characterization, validation, and prognostic value. PLoS Med.  2013; 10( 5): e1001453. Google Scholar CrossRef Search ADS PubMed  26 McCall MN, Bolstad BM, Irizarry RA. Frozen robust multiarray analysis (fRMA). Biostatistics.  2010; 11( 2): 242– 253. Google Scholar CrossRef Search ADS PubMed  27 Johnson WE, Li C, Rabinovic A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics.  2007; 8( 1): 118– 127. Google Scholar CrossRef Search ADS PubMed  28 Becht E, Giraldo NA, Lacroix L, et al.   Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression. Genome Biol.  2016; 17( 1): 218. Google Scholar CrossRef Search ADS PubMed  29 Wherry EJ, Kurachi M. Molecular and cellular insights into T cell exhaustion. Nat Rev Immunol.  2015; 15( 8): 486– 499. Google Scholar CrossRef Search ADS PubMed  30 Blackburn SD, Shin H, Haining WN, et al.   Coregulation of CD8+ T cell exhaustion by multiple inhibitory receptors during chronic viral infection. Nat Immunol.  2009; 10( 1): 29– 37. Google Scholar CrossRef Search ADS PubMed  31 Ferris RL, Lu B, Kane LP. Too much of a good thing? Tim-3 and TCR signaling in T cell exhaustion. J Immunol.  2014; 193( 4): 1525– 1530. Google Scholar CrossRef Search ADS PubMed  32 Topalian SL, Drake CG, Pardoll DM. Immune checkpoint blockade: A common denominator approach to cancer therapy. Cancer Cell.  2015; 27( 4): 450– 461. Google Scholar CrossRef Search ADS PubMed  33 Sargent DJ, Marsoni S, Monges G, et al.   Defective mismatch repair as a predictive marker for lack of efficacy of fluorouracil-based adjuvant therapy in colon cancer. J Clin Oncol.  2010; 28( 20): 3219– 3226. Google Scholar CrossRef Search ADS PubMed  34 Venderbosch S, Nagtegaal ID, Maughan TS, et al.   Mismatch repair status and BRAF mutation status in metastatic colorectal cancer patients: A pooled analysis of the CAIRO, CAIRO2, COIN, and FOCUS studies. Clin Cancer Res.  2014; 20( 20): 5322– 5330. Google Scholar CrossRef Search ADS PubMed  35 Andre T, de Gramont A, Vernerey D, et al.   Adjuvant fluorouracil, leucovorin, and oxaliplatin in stage II to III colon cancer: Updated 10-year survival and outcomes according to BRAF mutation and mismatch repair status of the MOSAIC Study. J Clin Oncol.  2015; 33( 35): 4176– 4187. Google Scholar CrossRef Search ADS PubMed  36 Saeterdal I, Bjorheim J, Lislerud K, et al.   Frameshift-mutation-derived peptides as tumor-specific antigens in inherited and spontaneous colorectal cancer. Proc Natl Acad Sci U S A.  2001; 98( 23): 13255– 13260. Google Scholar CrossRef Search ADS PubMed  37 Becht E, de Reynies A, Giraldo NA, et al.   Immune and stromal classification of colorectal cancer is associated with molecular subtypes and relevant for precision immunotherapy. Clin Cancer Res.  2016; 22( 16): 4057– 4066. Google Scholar CrossRef Search ADS PubMed  Author notes See the Notes section for the full list of authors and affiliations. © The Author 2017. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png JNCI: Journal of the National Cancer Institute Oxford University Press

Loading next page...
 
/lp/ou_press/the-balance-between-cytotoxic-t-cell-lymphocytes-and-immune-checkpoint-Nd5wwLqCQU
Publisher
Oxford University Press
Copyright
© The Author 2017. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com
ISSN
0027-8874
eISSN
1460-2105
D.O.I.
10.1093/jnci/djx136
Publisher site
See Article on Publisher Site

Abstract

Abstract Background Immune checkpoint (ICK) expression might represent a surrogate measure of tumor-infiltrating T cell (CTL) exhaustion and therefore be a more accurate prognostic biomarker for colorectal cancer (CRC) patients than CTL enumeration as measured by the Immunoscore. Methods The expression of ICKs, Th1, CTLs, cytotoxicity-related genes, and metagenes, including Immunoscore-like metagenes, were evaluated in three independent cohorts of CRC samples (260 microsatellite instable [MSI], 971 non-MSI). Their associations with patient survival were analyzed by Cox models, taking into account the microsatellite instability (MSI) status and affiliation with various Consensus Molecular Subgroups (CMS). PD-L1 and CD8 expression were examined on a subset of tumors with immunohistochemistry. All statistical tests were two-sided. Results The expression of Immunoscore-like metagenes was statistically significantly associated with improved outcome in non-MSI tumors displaying low levels of both CTLs and immune checkpoints (ICKs; CMS2 and CMS3; hazard ratio [HR] = 0.63, 95% confidence interval [CI] = 0.43 to 0.92, P = .02; and HR = 0.55, 95% CI = 0.34 to 0.90, P = .02, respectively), but clearly had no prognostic relevance in CRCs displaying higher levels of CTLs and ICKs (CMS1 and CMS4; HR = 0.46, 95% CI = 0.10 to 2.10, P = .32; and HR = 1.13, 95% CI = 0.79 to 1.63, P = .50, respectively), including MSI tumors. ICK metagene expression was statistically significantly associated with worse prognosis independent of tumor staging in MSI tumors (HR = 3.46, 95% CI = 1.41 to 8.49, P = .007). ICK expression had a negative impact on the proliferation of infiltrating CD8 T cells in MSI neoplasms (median = 0.56 in ICK low vs median = 0.34 in ICK high, P = .004). Conclusions ICK expression cancels the prognostic relevance of CTLs in highly immunogenic colon tumors and predicts a poor outcome in MSI CRC patients. The host immune response, and especially the T cell immune infiltrate, intersects the molecular and clinical characteristics of primary colorectal cancers (CRCs), including TNM staging. It is associated with prognosis of localized colon tumors. Expression levels for CTL/Th1/cytotoxicity markers have been demonstrated to constitute strong and independent prognostic indicators in CRC patients, with a high density of lymphocyte infiltration being consistently associated with prolonged survival (1). Based on multiple analyses, a scoring system was developed (termed Immunoscore) to quantify cytotoxic and memory T cells in the core of the tumor and in the tumor’s invasive margin (2). Immunoscore is a prognostic index in CRC (3). Several authors have suggested that the dense Th1/CTL lymphocytic infiltrate could explain the better prognosis of colon tumors displaying microsatellite instability (MSI), compared with non-MSI (MSS) CRC (4–6). MSI (7–12) leads to the synthesis of aberrant and potentially immunogenic neo-antigens by the tumor cells (for review, see [13,14]). A likely consequence of this is that MSI tumors are heavily infiltrated with activated cytotoxic T-cell lymphocytes (CTLs) and Th1 cells. A recent study reported that concomitant expression of multiple active immune checkpoint (ICK) markers (eg, CTLA-4, PD-1, PD-L1, LAG-3 targeted by immunotherapy) in CRC may counterbalance the antitumoral Th1/CTL immune response, notably in MSI CRC (15). One hypothesis that stems from these observations is that MSI tumors would be more responsive to ICK inhibitors because ICK expression is expected to protect highly immunogenic tumor cells from being destroyed. In line with this hypothesis, Le et al. (16) evaluated the clinical activity of an anti–PD-1 ICK inhibitor (pembrolizumab) in a cohort of metastatic CRC patients with MSS and MSI primary tumors. The results of this phase II study showed that MSI status was predictive of clinical benefit from ICK blockade therapy with pembrolizumab. The second hypothesis is that ICK expression should exert a negative impact on the prognosis of MSI tumors untreated by ICK inhibitors. To our knowledge, this second hypothesis has not yet been assessed thoroughly. It was reported in a recent study that the Immunoscore was prognostic in MSI CRC (17), but the authors did not integrate ICK markers in their survival analyses. Their results were obtained by investigating a unique series of CRC, and no further validation using publicly available clinical data was performed because the Immunoscore requires direct access to primary tumor samples. In other publications, only the role of PD1/PDL1 expression in CRC has been explored, leading to various conclusions and thus demonstrating the complexity involved with analysis of the CRC immune response (18–21). Rather than analyzing CRC immune response through the expression of PD1/PDL1 or T cell enumeration, our aim here was for a more holistic assessment of T cell immune response quality based on the analysis of both Th1/CTL and ICK markers. We also took into account the major molecular subtypes of CRC, that is, MSI/MSS status and consensus molecular subtypes (CMS) (22). Finally, the functional relevance of ICK overexpression was determined by pathway enrichment analysis and by quantification of immune infiltrates and CD8 T cell proliferation in MSI tumors. Methods Immune Genes Immune checkpoint and modulator genes were selected according to Llosa et al. (15) and a recent review (23). Markers for cytotoxic T lymphocytes, cytotoxicity, and T helper1 were selected as described earlier (see also the Supplementary Materials and Methods, available online) (15,24). Cohort Data Tissue samples from a large multisite cohort of CRC patients were collected as part of the “Cartes d’Identité des Tumeurs” (CIT) research program/enlever network, including tumors with or without microsatellite instability (MSI or MSS, respectively) and adjacent nontumoral tissue samples (NT). Samples from 146 MSI tumors, 444 MSS tumors, and 56 NT were analyzed for gene expression profiles on Affymetrix U133 plus 2 chips as described earlier (25). Data were normalized using the frozen RMA method (26) followed by Combat normalization (27) to remove technical batch effects (SVA R package). In addition, the CRC cohort from the The Cancer Genome Atlas (TCGA) consortium was also used (86 MSI tumors, 527 MSS tumors, 51 NT). Both data sets were centered for each gene by subtracting the median value of the non-tumoral sample and are provided as a Supplementary Data File (available online). An additional retrospective, multisite series of 28 stage IV primary MSI CRCs was analyzed separately for gene expression as part of an independent, ongoing study using the NanoString nCounter platform. A set of immune genes that comprised 14 of the 32 analyzed markers, including seven ICKs, was screened. All patients from this metastatic cohort (11 synchronous metastatic lesions, 17 metachronous metastatic lesions) received standard of care chemotherapy but not ICK blockade treatment. The NanoString data set also includes a subset of the CIT cohort samples for which sufficient RNA material was available for analysis (for protocol details, see the Supplementary Materials and Methods, available online). Supplementary Table 1 (available online) gives a description of all the cohorts investigated in this study, totaling 260 MSIs and 971 MSS CRCs. For non-metastatic MSI CRC patients, informed consent was recorded in each case; the standard hospital blanket consent was considered sufficient by French law. For patients included in the cohort of metastatic CRCs, ethical approval was provided by an institutional review board (institutional review board No. 00003835) on November 27, 2014. Statistical Analysis Metagene values were computed to summarize an immune gene category as the median value per sample of all the genes in the category. Comparisons of the mean expression between gene or metagene values across groups were assessed using the Student’s t-test or the analysis of variance test, as appropriate. Correlations were assessed by Pearson correlation coefficients. Associations between categorical variables were assessed using chi-square tests. MSigDB gene sets/pathways enrichment analysis was performed by comparing low- and high-ICK groups, using the hypergeometric test and based on the genes most statistically significantly differentially expressed between these two groups. Associations between gene /metagene expression values and survival in the different cohorts were assessed by univariate and bivariate Cox proportional-hazards regression analyses and adjusted by tumor series and TNM stage, using the R package “survival,” and the results were combined using a generic inverse variance meta-analysis approach (see also the Supplementary Materials and Methods, available online). The proportional-hazards assumption was tested prior to survival analysis to examine the models’ appropriateness using the cox.zph function. Survival curves were obtained using Kaplan-Meier estimates, and differences between curves were assessed using the log-rank test. Overall survival was used and defined as time from diagnosis to death from any cause. All statistical tests were two-sided. The cut point for statistical significance was .05. Results Prognostic Value of Immune Genes and Metagenes in Function of CMS Classification of Colorectal Cancer To test our working hypothesis, summarized in Figure 1A, we evaluated the prognostic significance of ICKs, Th1, CTLs, and cytotoxicity markers in the combined CIT (n = 590 CRC, comprising 146 MSI and 444 MSS) and TCGA (n = 613 CRC, comprising 86 MSI and 527 MSS) series. In both cohorts, MSS tumors were categorized into one of the four CMS subtypes of CRC (22). We investigated 32 immune markers classified into the above four immune marker groups (see the Supplementary Materials and Methods, available online, for a list of these markers) (for review, see [24]) as independent genes and as four metagenes. Four of the 19 immune checkpoints and modulators were not statistically significantly overexpressed in CRC as compared with nontumor colonic mucosa and were subsequently removed (Supplementary Figure 1, available online). Further analyses were thus carried out on the 28 remaining genes. Finally, to obtain Immunoscore surrogates, six Immunoscore-like metagenes were built based on the expression of Immunoscore-related markers (Supplementary Table 2, available online). Figure 1. View largeDownload slide Prognostic value of immune gene expression. A) Schematic representation of the working hypothesis on immune checkpoint prognosis. Arrows represent inhibition. B) Overall survival stratified by microsatellite instable (MSI)/microsatellite stable (MSS) status (left) and Consensus Molecular Subgroups (CMS) subtypes (right). Curves of overall survival (OS) rate (y-axis) according to time from diagnosis (in years; x-axis) were obtained by the method of Kaplan-Meier for both the Cartes d’Identité des Tumeurs (CIT) and The Cancer Genome Atlas (TCGA) series. Differences between survival distributions were assessed by the log-rank test using an end point of five years. C) Prognostic values of immune gene/metagene expression and of clinical factors in MSI tumors. Forest plot of overall survival (OS) hazard ratios (HRs) estimated by combining independent univariate Cox analyses on the CIT and TCGA series, adjusted for TNM stage. Hazard ratio, as well as related Wald test P value and 95% confidence intervals (CIs), are given for metagenes (which aggregate the gene expression values of a gene set related to the four immune categories [immune checkpoints {ICK}, cytotoxic T lymphocytes {CTL}, cytotoxicity, Th1 functional orientation]), individual immune genes, and clinical annotations. Diamonds represent the hazard ratios, and horizontal bars represent the 95% confidence intervals. Red indicates a hazard ratio greater than 1 with a P value of less than .1 (worse prognosis), blue a hazard ratio of less than 1 with a P value of less than .1 (better prognosis), and gray a hazard ratio with a Wald test P value of .1 or less. D) Overall survival stratified by overexpression of immune checkpoint genes within MSI (left), MSS (middle), and both CRCs (right). MSI tumors in the higher quartile of ICK metagene values, in CIT and TCGA series independently, were assigned ICK + (n = 55), and the other tumors ICK- (n = 139). The minimal ICK metagene value within MSI ICK+ tumors was used to divide MSS tumors into ICK + (n = 26) and ICK- (n = 765). Curves of overall survival rate (y-axis) according to time from diagnosis (in years; x-axis) were obtained by the method of Kaplan and Meier for both the CIT and TCGA series. Differences between survival distributions were assessed by the log-rank test using an end point of five years. E) Prognostic value of immune metagenes according to the type of CRC. Each column defines a specific type of CRC (all CRC, MSI CRC, MSS CRC, MSS CMS1 CRC, MSS CMS2 CRC, MSS CMS3 CRC, MSS CMS4 CRC) and the number of related samples with available survival data in the combined CIT + TCGA series. Each row defines a given metagene (ICK, CTL, CYTOX, TH1, IS-like v1, IS-like v2). Each cell shows the hazard ratio for overall survival of the relevant metagene expression for the specific CRC type (obtained via a Cox model). Each cell also contains the P value for the Wald test for equality of this hazard ratio to 1. The hazard ratio shown in each cell was calculated by combining the hazard ratios for each of the CIT and TCGA data sets using the meta-analysis approach of DerSimonian. The genes relevant to each metagene are indicated in (C) (ICK, CTL, CYTOX, TH1) or are given below the metagene name. F) Bivariate Cox models of OS, combining the ICK metagene with other metagenes, in MSI tumors. Forest plot of OS hazard ratio estimated by bivariate Cox analysis of ICK, CTL, CYTOX, Th1, and Immunoscore-like metagene expression in the CIT series, adjusted by TNM stage. All statistical tests were two-sided. CIT = Cartes d’Identité des Tumeurs; CRC = colorectal cancer; CMS = Consensus Molecular Subgroups; CTL = cytotoxic T lymphocytes; high = high expression level; ICK = immune checkpoint genes; MSI = microsatellite instable; MSS = microsatellite stable; OS = overall survival; TCGA = The Cancer Genome Atlas. Figure 1. View largeDownload slide Prognostic value of immune gene expression. A) Schematic representation of the working hypothesis on immune checkpoint prognosis. Arrows represent inhibition. B) Overall survival stratified by microsatellite instable (MSI)/microsatellite stable (MSS) status (left) and Consensus Molecular Subgroups (CMS) subtypes (right). Curves of overall survival (OS) rate (y-axis) according to time from diagnosis (in years; x-axis) were obtained by the method of Kaplan-Meier for both the Cartes d’Identité des Tumeurs (CIT) and The Cancer Genome Atlas (TCGA) series. Differences between survival distributions were assessed by the log-rank test using an end point of five years. C) Prognostic values of immune gene/metagene expression and of clinical factors in MSI tumors. Forest plot of overall survival (OS) hazard ratios (HRs) estimated by combining independent univariate Cox analyses on the CIT and TCGA series, adjusted for TNM stage. Hazard ratio, as well as related Wald test P value and 95% confidence intervals (CIs), are given for metagenes (which aggregate the gene expression values of a gene set related to the four immune categories [immune checkpoints {ICK}, cytotoxic T lymphocytes {CTL}, cytotoxicity, Th1 functional orientation]), individual immune genes, and clinical annotations. Diamonds represent the hazard ratios, and horizontal bars represent the 95% confidence intervals. Red indicates a hazard ratio greater than 1 with a P value of less than .1 (worse prognosis), blue a hazard ratio of less than 1 with a P value of less than .1 (better prognosis), and gray a hazard ratio with a Wald test P value of .1 or less. D) Overall survival stratified by overexpression of immune checkpoint genes within MSI (left), MSS (middle), and both CRCs (right). MSI tumors in the higher quartile of ICK metagene values, in CIT and TCGA series independently, were assigned ICK + (n = 55), and the other tumors ICK- (n = 139). The minimal ICK metagene value within MSI ICK+ tumors was used to divide MSS tumors into ICK + (n = 26) and ICK- (n = 765). Curves of overall survival rate (y-axis) according to time from diagnosis (in years; x-axis) were obtained by the method of Kaplan and Meier for both the CIT and TCGA series. Differences between survival distributions were assessed by the log-rank test using an end point of five years. E) Prognostic value of immune metagenes according to the type of CRC. Each column defines a specific type of CRC (all CRC, MSI CRC, MSS CRC, MSS CMS1 CRC, MSS CMS2 CRC, MSS CMS3 CRC, MSS CMS4 CRC) and the number of related samples with available survival data in the combined CIT + TCGA series. Each row defines a given metagene (ICK, CTL, CYTOX, TH1, IS-like v1, IS-like v2). Each cell shows the hazard ratio for overall survival of the relevant metagene expression for the specific CRC type (obtained via a Cox model). Each cell also contains the P value for the Wald test for equality of this hazard ratio to 1. The hazard ratio shown in each cell was calculated by combining the hazard ratios for each of the CIT and TCGA data sets using the meta-analysis approach of DerSimonian. The genes relevant to each metagene are indicated in (C) (ICK, CTL, CYTOX, TH1) or are given below the metagene name. F) Bivariate Cox models of OS, combining the ICK metagene with other metagenes, in MSI tumors. Forest plot of OS hazard ratio estimated by bivariate Cox analysis of ICK, CTL, CYTOX, Th1, and Immunoscore-like metagene expression in the CIT series, adjusted by TNM stage. All statistical tests were two-sided. CIT = Cartes d’Identité des Tumeurs; CRC = colorectal cancer; CMS = Consensus Molecular Subgroups; CTL = cytotoxic T lymphocytes; high = high expression level; ICK = immune checkpoint genes; MSI = microsatellite instable; MSS = microsatellite stable; OS = overall survival; TCGA = The Cancer Genome Atlas. As a preliminary step, the prognostic values of MSI and CMS status were confirmed with, as expected, an improved prognosis for patients with MSI CRCs, compared with those with MSS CRCs, and a statistically significant prognostic value of the CMS classification (Figure 1B;Supplementary Figure 2, available online). The prognostic values of the 28 immune markers mentioned and the four metagenes were then analyzed by univariate Cox models. Most ICK genes were individually associated with poorer prognosis in MSI CRC, as reflected by the ICK metagene (HR =  3.46, 95% CI = 1.41 to 8.49, P = .007) (Figure 1, C–E; Supplementary Figure 3A, available online). This association remained statistically significant in non-metastatic MSI CRC (Supplementary Figure 3B, available online). No prognostic association was observed in patients with MSS CRCs, either for individual ICK markers or for the ICK metagene (Figure 1, D and E; Supplementary Figure 3C, available online). Subdividing MSS CRCs by CMS did not change this result, except in CMS3, where the ICK metagene was associated with better prognosis (HR = 0.46, 95% CI = 0.21 to 1.00, P = .05) (Figure 1E;Supplementary Table 2, available online). As expected, the expression of most Immunoscore-like metagenes was associated with improved outcome of CRC patients (Figure 1E;Supplementary Table 2, available online). However, when CMS and MSI/MSS status were taken into account, expression of these Immunoscore surrogates was predictive of better prognosis only in MSS tumors from CMS2 and CMS3 (HR = 0.63, 95% CI = 0.43 to 0.92, P = .02; and HR = 0.55, 95% CI = 0.34 to 0.90, P = .02, respectively). In contrast, they had no prognostic relevance in CRC from CMS1 and CMS4 (HR =  0.46, 95% CI = 0.10 to 2.10, P = .32; and HR = 1.13, 95% CI = 0.79 to 1.63, P = .50, respectively) (Figure 1E;Supplementary Table 2, available online). In univariate models, the overexpression of CTL/Th1/cytotoxicity/Immunoscore markers and metagenes was also associated with adverse prognosis in MSI CRC (Figure 1, C and E; Supplementary Figure 3, A and B, available online). We therefore hypothesized that high expression levels of ICKs in MSI CRC could counterbalance and mask the expected positive effect of CTL/Th1/cytotoxic cells or Immunoscore-related cells on prognosis. Bivariate Cox models at the metagene level were consistent with this assumption (Figure 1F). All together, these data underline that ICKs and Immunoscore biomarkers constitute an independent prognostic factor for overall survival in MSI and MSS tumors, respectively. Expression and Prognostic Value of Immune Checkpoints in an Independent Metastatic MSI CRC Patient Series The CIT and TCGA series included mostly non-metastatic MSI CRC patients (n = 220/232, 94.8%). Because ICK blockade was recently proposed as a promising new therapy for metastatic MSI CRC, we endeavored to further evaluate the prognostic relevance of ICK expression in an additional retrospective, multisite series of 28 stage IV primary MSI CRCs treated with standard care that were analyzed separately for gene expression as part of an independent, ongoing study using the NanoString nCounter platform. To do this, we analyzed the expression of the seven ICKs that were previously screened (CD274, PDCD1LG2, HAVCR2, LAG3, ICOS, CTLA4, PDCD1). As with non-metastatic MSI colon tumors, we observed statistically significant association of PD-L1 (CD274) expression with worse OS and worse survival after relapse (SAR) (Figure 2A). Associations of TIM-3 (HAVCR2) and LAG3 expression, although not statistically significant, were suggested. Again, the overexpression of metagenes corresponding to CTL/Th1/cytotoxicity/Immunoscore markers (CTL = CD3D, CD3G, CD8A, PTPRC; CYTOX = GNLY, GZMK; Th1= IFNG) tended to associate with adverse prognosis in univariate models (Figure 2B). However, in bivariate Cox models, the expression of CTL/Th1/cytotoxicity/Immunoscore-related metagenes was associated with good prognosis (hazard ratio [HR] < 1.00), as expected (Figure 2C). Several ICK markers, in particular the druggable PD-L1 (CD274) and TIM-3 (HAVCR2) molecules (23), are therefore likely to constitute biomarkers for poor prognosis in both metastatic and non-metastatic MSI CRC patients. Figure 2. View largeDownload slide Prognostic value of immune checkpoints in an independent metastatic microsatellite instable (MSI) colorectal cancer (CRC) patient series. A) Prognostic value of ICK gene expression in metastatic MSI tumors. Forest plot of hazard ratios (HRs) estimated by univariate Cox analysis of overall survival (OS; left panel) and survival after relapse (SAR; right panel) on all ICK genes available in the NanoString data. Diamonds represent hazard ratio estimates, and bars the related 95% confidence intervals (CIs). Red indicates a worse prognosis hazard ratio, blue a better prognosis, and gray a hazard ratio with a Wald test P value of .1 or less. B) Prognostic value of metagene expression in MSI metastatic tumors. Forest plot of hazard ratios estimated by univariate Cox analysis of survival after relapse for metagene expression. The ICK metagene aggregates the three most associated ICK genes in univariate analysis (HAVCR2, CD274, and LAG3). The other metagene aggregate genes available in the NanoString data set for the corresponding category were CTL = CD3D, CD3G, CD8A, PTPRC; CYTOX = GNLY, GZMK; Th1= IFNG. Metagene expression values were calculated by averaging the expression values obtained within the corresponding gene set. C) Bivariate Cox models of OS and SAR, combining the ICK metagene with other metagenes, in metastatic MSI tumors. Forest plot of SAR hazard ratios estimated by bivariate Cox analysis of ICK, CTL, CYTOX, Th1, and Immunoscore-like metagene expression in the independent MSI metastatic series. All statistical tests were two-sided. CI = confidence interval; CTL = cytotoxic T lymphocytes; HR = hazard ratio; ICK = immune checkpoint genes; OS = overall survival; SAR = survival after relapse. Figure 2. View largeDownload slide Prognostic value of immune checkpoints in an independent metastatic microsatellite instable (MSI) colorectal cancer (CRC) patient series. A) Prognostic value of ICK gene expression in metastatic MSI tumors. Forest plot of hazard ratios (HRs) estimated by univariate Cox analysis of overall survival (OS; left panel) and survival after relapse (SAR; right panel) on all ICK genes available in the NanoString data. Diamonds represent hazard ratio estimates, and bars the related 95% confidence intervals (CIs). Red indicates a worse prognosis hazard ratio, blue a better prognosis, and gray a hazard ratio with a Wald test P value of .1 or less. B) Prognostic value of metagene expression in MSI metastatic tumors. Forest plot of hazard ratios estimated by univariate Cox analysis of survival after relapse for metagene expression. The ICK metagene aggregates the three most associated ICK genes in univariate analysis (HAVCR2, CD274, and LAG3). The other metagene aggregate genes available in the NanoString data set for the corresponding category were CTL = CD3D, CD3G, CD8A, PTPRC; CYTOX = GNLY, GZMK; Th1= IFNG. Metagene expression values were calculated by averaging the expression values obtained within the corresponding gene set. C) Bivariate Cox models of OS and SAR, combining the ICK metagene with other metagenes, in metastatic MSI tumors. Forest plot of SAR hazard ratios estimated by bivariate Cox analysis of ICK, CTL, CYTOX, Th1, and Immunoscore-like metagene expression in the independent MSI metastatic series. All statistical tests were two-sided. CI = confidence interval; CTL = cytotoxic T lymphocytes; HR = hazard ratio; ICK = immune checkpoint genes; OS = overall survival; SAR = survival after relapse. Immune Checkpoint Gene Expression Distribution in Colorectal Cancer We next investigated the level of variation in ICK expression among CRC tumors in order to further assess their relevance as prognostic and theranostic markers. ICK expression was analyzed in stage I through IV CRC and in nontumor colonic mucosa from our CIT cohort (590 CRC, 56 NT) and in the TCGA cohort (613 CRC, 51 NT). In both cohorts, the metagenes corresponding to ICKs, CTL, cytotoxicity, and Th1 orientation were overexpressed in MSI and in MSS tumors belonging to CMS1 and CMS4, as compared with MSS CRCs from CMS2 and CMS3 (Figure 3A). Variable expression of ICKs relative to NT was noted in all CMS subtypes in both cohorts (Figure 3B). A high degree of heterogeneity was observed in CMS1 tumors (Figure 3B), particularly in MSI tumors, where high to very high expression levels of ICKs were observed in a large proportion of cases (Figure 3B). Figure 3. View largeDownload slide Immune gene expression distribution in colorectal cancer (CRC). A) Immune metagene expression relative to nontumoral colonic mucosa across CRC subtypes. Height panels show expression values for four metagenes (columns) in two series (rows): Cartes d’Identité des Tumeurs (CIT; top) and The Cancer Genome Atlas (TCGA; bottom). Each metagene aggregates the gene expression values of a gene set related to one of four immune categories (immune checkpoints [ICK], cytotoxic T lymphocytes [CTL], cytotoxicity, Th1 functional orientation). Gene (and metagene) expression values are relative to the median expression in nontumor samples. Within each panel, the metagene values are presented in six boxplots related to six sample types (nontumor tissue [NT], MSI tumors, MSS tumors subdivided into the four consensus molecular subtypes [CMS]). P values were calculated using a Student’s t test. P values were calculated using a Student’s t test to assess the statistical significance of differences between MSI and NT and between MSI and MSS/CMS (***P < .001, **P < .01, *P < .05). B) Heterogeneous expression of ICK genes among MSI and MSS CRCs. Heatmap shows the gene expression z-scores of ICK genes in MSI and MSS tumor samples in CIT (top) and TCGA (bottom) series. Central and dispersion parameters used to generate z-scores correspond to the median and median absolute deviation (mad) in NT samples. Samples are ordered according to MSI/MSS status, CMS within MSS, and median expression value of immune checkpoint genes. C) Correlation of expression of immune genes within MSI tumors. Heatmap of Pearson coefficients of correlation between expression values of immune genes within MSI tumors of CIT (top) and TCGA (bottom) series separately. Genes are ordered according to four immune categories (ICK, cytotoxic T lymphocytes [CTL], cytotoxicity, Th1 functional orientation). D) Heterogeneous expression of ICK genes among MSI metastatic tumors. Heatmaps show gene expression of ICK genes in primary tumor samples from an independent metastatic MSI CRC series (bottom), compared with a subset of the CIT MSI series (top). Expression values were obtained using NanoString technology, where only a subset of immune genes was evaluated. NT-relative gene expression ratios are displayed. All statistical tests were two-sided. CIT = Cartes d’Identité des Tumeurs; CMS = Consensus Molecular Subgroups; CTL = cytotoxic T lymphocytes; ICK = immune checkpoint genes; high = high expression level; MSI = microsatellite instable; MSS = microsatellite stable; NT = nontumoral tissue samples; TCGA = The Cancer Genome Atlas. Figure 3. View largeDownload slide Immune gene expression distribution in colorectal cancer (CRC). A) Immune metagene expression relative to nontumoral colonic mucosa across CRC subtypes. Height panels show expression values for four metagenes (columns) in two series (rows): Cartes d’Identité des Tumeurs (CIT; top) and The Cancer Genome Atlas (TCGA; bottom). Each metagene aggregates the gene expression values of a gene set related to one of four immune categories (immune checkpoints [ICK], cytotoxic T lymphocytes [CTL], cytotoxicity, Th1 functional orientation). Gene (and metagene) expression values are relative to the median expression in nontumor samples. Within each panel, the metagene values are presented in six boxplots related to six sample types (nontumor tissue [NT], MSI tumors, MSS tumors subdivided into the four consensus molecular subtypes [CMS]). P values were calculated using a Student’s t test. P values were calculated using a Student’s t test to assess the statistical significance of differences between MSI and NT and between MSI and MSS/CMS (***P < .001, **P < .01, *P < .05). B) Heterogeneous expression of ICK genes among MSI and MSS CRCs. Heatmap shows the gene expression z-scores of ICK genes in MSI and MSS tumor samples in CIT (top) and TCGA (bottom) series. Central and dispersion parameters used to generate z-scores correspond to the median and median absolute deviation (mad) in NT samples. Samples are ordered according to MSI/MSS status, CMS within MSS, and median expression value of immune checkpoint genes. C) Correlation of expression of immune genes within MSI tumors. Heatmap of Pearson coefficients of correlation between expression values of immune genes within MSI tumors of CIT (top) and TCGA (bottom) series separately. Genes are ordered according to four immune categories (ICK, cytotoxic T lymphocytes [CTL], cytotoxicity, Th1 functional orientation). D) Heterogeneous expression of ICK genes among MSI metastatic tumors. Heatmaps show gene expression of ICK genes in primary tumor samples from an independent metastatic MSI CRC series (bottom), compared with a subset of the CIT MSI series (top). Expression values were obtained using NanoString technology, where only a subset of immune genes was evaluated. NT-relative gene expression ratios are displayed. All statistical tests were two-sided. CIT = Cartes d’Identité des Tumeurs; CMS = Consensus Molecular Subgroups; CTL = cytotoxic T lymphocytes; ICK = immune checkpoint genes; high = high expression level; MSI = microsatellite instable; MSS = microsatellite stable; NT = nontumoral tissue samples; TCGA = The Cancer Genome Atlas. Expression levels for all of the 28 immune markers were highly correlated in the MSI CRCs from both cohorts (Figure 3C). The present results highlight the extent of heterogeneity of MSI CRC with respect to immunity and to the overexpression of ICK molecules. This was observed regardless of MSI CRC origin (inherited or sporadic) or of other clinical or molecular parameters such as sex, tumor location, tumor stage, CMS, or KRAS/BRAF mutations (Supplementary Figure 4, available online). Considerable variation in the expression of ICK markers was also observed in the independent metastatic MSI CRC series evaluated by NanoString (Figure 3D). Functional Relevance of Immune Checkpoint Expression in CRC We next addressed the possible physiological relevance of ICK overexpression in CRC. Tumor infiltration by immune cells was quantified using MCP counter software (28) in both the CIT and TCGA cohorts. A strong correlation was observed between ICK expression and infiltration by lymphoid (NK, T cells, cytotoxic cells) and myeloid cells. In contrast, B cells, fibroblasts, vessels, and granulocytes were less, or not at all, associated with ICK expression (Figure 4A). These results suggest that ICK expression occurs in response to an efficient in situ adaptive T cell immune response. Pathway enrichment analysis was performed to compare the expression profiles of MSI tumors with low vs high ICK expression levels. Statistically significant associations were observed between ICK expression and immune response gene sets, including positive activation of T cell response, negative regulation of T cell activation, T cell exhaustion, IL-10 response, and chronic viral infection (29) (Figure 4B). Hence, we conclude that there is a strong correlation between ICK expression and the presence of an exhausted T cell immune response in MSI CRC. Figure 4. View largeDownload slide Functional relevance of immune checkpoint expression in colorectal cancer (CRC). A) Transcriptome-based abundance estimates of microenvironment cell populations (MCPcounter) and relation to ICK gene expression. Heatmap showing the abundance estimates for nine immune (Lymphoid, natural killer [NK] or T, T derived, NK, Cytotoxic, B derived, myeloid, monocyte derived, granulocyte) and two nonimmune (vessels, fibroblasts) cell populations in microsatellite instable (MSI) tumors and microsatellite stable (MSS) Consensus Molecular Subgroups (CMS) in the Cartes d’Identité des Tumeurs (CIT; left) and The Cancer Genome Atlas (TCGA; right) series. For each of these 11 cell populations, abundance estimates were calculated using MCP counter software as the average NT-relative expression ratios of corresponding markers. ICK metagene expression and CMS classification are also shown. MSI tumor samples are divided into three groups according to ICK metagene expression (high, intermediate, and low). These three groups are compared with CMS subtypes; analysis of variance test P values are reported on the right of the heatmap. Associations between abundance estimates of immune cell populations and the ICK metagene expression within MSI tumors were measured by Pearson's correlations. Coefficients and associated P values are reported on the right of the heatmap. For MSS tumors, the median values within each CMS are displayed. Samples are ordered according to the ICK metagene values. B) Functional associations with overexpression of ICK genes. MSI samples are divided into three groups according to ICK metagene expression (high, intermediate, and low). Heatmap showing the expression score of a selection of immune gene sets statistically significantly enriched in genes that were differentially expressed between high- and low-ICK-expressing MSI tumor groups, in CIT (left) and TCGA (right) series. For each gene set, expression scores are calculated as the mean NT-relative expression ratios across all genes in the gene set. Pearson's correlation coefficients and associated P values are reported. C) Representative slide of PD-L1 and CD8 immunohistochemical (IHC) staining of MSI CRC with high expression of ICK genes on transcriptomic data. Black scale bar is 1 mm. Yellow Scale bar is 100 µm. D) CD8 cell density according to PD-L1 expression within tumors, stroma, and healthy mucosa. Upper panel: number of CD8 cells per mm² in intratumoral area, peritumoral stromal area, and healthy mucosa in patients with high PD-L1 expression vs low or no PD-L1 expression. Lower panel: correlation between semiquantitative PD-L1 expression with IHC staining and CD8 number per mm² in intratumoral area, peritumoral stromal area, and healthy mucosa. E) Representative slide of CD8 cell infiltration and PD-L1 expression. Left panel: representative slide of CD8, DAPI, and Ki67 staining with classical composite immunofluorescence image and phenotype map using Inform software to recognize tumor cells, proliferating and nonproliferating CD8 cells, and other cells. Scale bar is 100 µm. Right panel: quantification of proliferating CD8 or nonproliferating CD8 cells in PDL-L1 high vs low or negative tumor. All statistical tests were two-sided. Error bars represent standard deviation. CIT = Cartes d’Identité des Tumeurs; CMS = Consensus Molecular Subgroups; ICK = immune checkpoint genes; MSI = microsatellite instable; MSS = microsatellite stable; TCGA = The Cancer Genome Atlas. Figure 4. View largeDownload slide Functional relevance of immune checkpoint expression in colorectal cancer (CRC). A) Transcriptome-based abundance estimates of microenvironment cell populations (MCPcounter) and relation to ICK gene expression. Heatmap showing the abundance estimates for nine immune (Lymphoid, natural killer [NK] or T, T derived, NK, Cytotoxic, B derived, myeloid, monocyte derived, granulocyte) and two nonimmune (vessels, fibroblasts) cell populations in microsatellite instable (MSI) tumors and microsatellite stable (MSS) Consensus Molecular Subgroups (CMS) in the Cartes d’Identité des Tumeurs (CIT; left) and The Cancer Genome Atlas (TCGA; right) series. For each of these 11 cell populations, abundance estimates were calculated using MCP counter software as the average NT-relative expression ratios of corresponding markers. ICK metagene expression and CMS classification are also shown. MSI tumor samples are divided into three groups according to ICK metagene expression (high, intermediate, and low). These three groups are compared with CMS subtypes; analysis of variance test P values are reported on the right of the heatmap. Associations between abundance estimates of immune cell populations and the ICK metagene expression within MSI tumors were measured by Pearson's correlations. Coefficients and associated P values are reported on the right of the heatmap. For MSS tumors, the median values within each CMS are displayed. Samples are ordered according to the ICK metagene values. B) Functional associations with overexpression of ICK genes. MSI samples are divided into three groups according to ICK metagene expression (high, intermediate, and low). Heatmap showing the expression score of a selection of immune gene sets statistically significantly enriched in genes that were differentially expressed between high- and low-ICK-expressing MSI tumor groups, in CIT (left) and TCGA (right) series. For each gene set, expression scores are calculated as the mean NT-relative expression ratios across all genes in the gene set. Pearson's correlation coefficients and associated P values are reported. C) Representative slide of PD-L1 and CD8 immunohistochemical (IHC) staining of MSI CRC with high expression of ICK genes on transcriptomic data. Black scale bar is 1 mm. Yellow Scale bar is 100 µm. D) CD8 cell density according to PD-L1 expression within tumors, stroma, and healthy mucosa. Upper panel: number of CD8 cells per mm² in intratumoral area, peritumoral stromal area, and healthy mucosa in patients with high PD-L1 expression vs low or no PD-L1 expression. Lower panel: correlation between semiquantitative PD-L1 expression with IHC staining and CD8 number per mm² in intratumoral area, peritumoral stromal area, and healthy mucosa. E) Representative slide of CD8 cell infiltration and PD-L1 expression. Left panel: representative slide of CD8, DAPI, and Ki67 staining with classical composite immunofluorescence image and phenotype map using Inform software to recognize tumor cells, proliferating and nonproliferating CD8 cells, and other cells. Scale bar is 100 µm. Right panel: quantification of proliferating CD8 or nonproliferating CD8 cells in PDL-L1 high vs low or negative tumor. All statistical tests were two-sided. Error bars represent standard deviation. CIT = Cartes d’Identité des Tumeurs; CMS = Consensus Molecular Subgroups; ICK = immune checkpoint genes; MSI = microsatellite instable; MSS = microsatellite stable; TCGA = The Cancer Genome Atlas. To further investigate the functional relevance of ICKs in MSI tumors, we studied eight primary MSI tumors showing upregulation of ICKs and 12 without. PD-L1 and CD8 expression were examined using immunohistochemistry (IHC). PD-L1 expression was observed only in the tumor bed, whereas CD8 was present both in the tumor core and in stromal areas (Figure 4C). Moreover, PD-L1 expression correlated strongly with ICK expression (Supplementary Figure 5, available online), while we observed concomitant increases of CD8 infiltrates in both the tumor bed and in the peritumoral stroma with PD-L1 IHC staining (Figure 4D). Proliferation and functional activity of CD8 T cells were then determined using multiparametric immunofluorescence microscopy (Figure 4E, left panel). CD8 T cells that were close to or in contact with PD-L1-expressing tumors were less proliferative, as observed with Ki67 labeling (median = 0.56 in ICK low vs median = 0.34 in ICK high, P = .004) (Figure 4E, right panel). These results indicate that interactions between CD8 T cells and ICK ligands in MSI primary tumors can impede CD8 T cell function. Discussion During cancer progression, tumor-infiltrating T cells have been shown to display increased chronic expression of different antagonist ICKs, causing functional exhaustion and unresponsiveness of T cells (30). The exhausted CD8 T cells fail to proliferate in response to antigen and lack critical anticancer effector functions (31). Checkpoint inhibitors could boost the anticancer immune response, and the potential relevance of these inhibitors for the treatment of metastatic MSI CRC patients was highlighted (16). In the present study, we showed that ICK overexpression represents a more accurate prognostic biomarker for MSI CRC patients treated with standard care than the classical assessment of T cell number by Immunoscore (1). This may be explained by the presence of exhausted nonproliferative CD8 T cells in the core of these neoplasms. More generally, our data indicate that assessment of the prognostic significance of antitumor immunity in CRC needs to take into account ICK expression. This is particularly relevant for colon tumors displaying immunogenic profiles with both high Immunoscores and ICK expression, such as in MSI tumors and probably a substantial proportion of MSS CRCs. The current results were obtained using univariate Cox models for survival analysis and a transcriptome-based method to quantify both ICK and CTL/Th1/cytotoxicity (Immunoscore) markers. We validated our method by building Immunoscore-like surrogates that were associated with statistically significantly improved survival of CRC patients. Under the same conditions, the CTL/Th1/cytotoxicity and Immunoscore markers were both associated with worse prognostic in MSI CRC. These results are potentially in conflict with a recent publication that observed a statistically significant association between Immunoscore and improved outcome in a single series of 105 MSI CRC patients (17). Although the studies are not directly comparable, here we assessed three independent cohorts of CRC patients, totaling more than 1200 cases and including 260 MSI CRCs. However, it does not include classical Immunoscore evaluation by immunohistochemistry. Bivariate Cox analysis, combining ICK metagene with CTL or Th1 or cytotoxicity or Immunoscore-like metagene, revealed that expression of metagenes related to CTL/Th1/cytotoxicity and Immunoscore markers was associated with trends for better prognosis in MSI CRC from both the CIT and NanoString metastatic series, whereas the ICK metagene was statistically significantly associated with worse prognosis. However, this association was not observed in the TCGA cohort as it had too few events (n = 5) to get sufficient statistical power. Therefore, validation in another larger cohort is required. In contrast with the earlier study that focused only on the PD1/PDL1 couple (17), a more global assessment of ICK gene expression in the tumor core, as proposed in the present study, allows a more holistic view of the T cell immune response in CRC. Moreover, the transcriptome-based method reported here is easier to use and more amenable to standardization. It can be also used to test publicly available clinical data sets, whereas this is not possible with Immunoscore because of the need to assess primary tumor samples. Recent clinical trials have demonstrated that antibodies targeting PD-1 or PD-L1 can induce a major response in many types of cancers (32). The overall survival rate with more than five years’ follow-up for stage II and III MSI CRC patients is approximately 70% without adjuvant chemotherapy and 75% to 90% with standard care adjuvant chemotherapy (33–35), whereas it is less than 5% for stage IV MSI CRC patients. We report here for the first time the prognostic significance of ICK overexpression in both metastatic and nonmetastatic MSI CRC and in the absence of immunotherapy. These findings should help to better inform the prognosis of MSI CRC patients. They may be useful for guiding future immunotherapy involving antibody blockade of ICKs in nonmetastatic MSI CRC patients and for having predictive factors of immunotherapy efficacy for patients with metastatic disease. The limitations of our study are mainly related to the fact that it is not yet ready for routine clinical use. First, it will be necessary to confirm in large and prospective series of patients from randomized trials that the outcome of CRC patients is determined by the CTL/ICK balance. Second, it will be necessary to further assess the prognostic relevance of combined ICK and CTL marker expression in MSI CRC. This should allow patients who are at high risk of relapse to be identified and thus indicated for more intensive treatments linked to these molecular markers. Such work will help to confirm ICK molecules as clinically relevant theranostic factors for the treatment of MSI CRC. To conclude, our results highlight the extent of heterogeneity of CRC with respect to immunity, and the overexpression of ICK molecules in particular. They suggest that prediction of CRC patient outcomes through evaluation of immune components in the tumor microenvironment will likely be improved by the integration of ICK markers, the prognosis of colon tumors being determined by the CTL/ICK balance (Figure 5). More particularly, our results indicate that ICK expression impacts the prognosis of MSI tumors. To inform future immunotherapy involving antibody blockade of ICKs and resistance to these molecules in MSI CRC patients, additional studies on the molecular mechanisms underlying the immune reaction against MSI tumor cells are required, for example, the number and type of MSI-driven mutational events that drive the synthesis of aberrant, immunogenic peptides (36). Identifying these somatic events with respect to quantitative and qualitative antitumoral immunity may improve the personalized treatment of MSI CRC patients with ICK inhibitors, in both metastatic and nonmetastatic settings. Figure 5. View largeDownload slide Immune model for prognostication of colorectal cancer (CRC) patients. The diagram shows a model of decision tree based on the three criteria, fibroblast presence, immune infiltration, and immune checkpoint expression, to define the four distinct prognosis groups according to immunity observed in CRC patients. Expected prognosis, relevant markers, subtypes, and microsatellite instable status are given for each group. CMS = Consensus Molecular Subgroups; CTL = cytotoxic T lymphocytes; ICK = immune checkpoint genes; high = high expression level; MSI = microsatellite instable; MSS = microsatellite stable. Figure 5. View largeDownload slide Immune model for prognostication of colorectal cancer (CRC) patients. The diagram shows a model of decision tree based on the three criteria, fibroblast presence, immune infiltration, and immune checkpoint expression, to define the four distinct prognosis groups according to immunity observed in CRC patients. Expected prognosis, relevant markers, subtypes, and microsatellite instable status are given for each group. CMS = Consensus Molecular Subgroups; CTL = cytotoxic T lymphocytes; ICK = immune checkpoint genes; high = high expression level; MSI = microsatellite instable; MSS = microsatellite stable. Funding This work was supported by the Inserm Institute, UPMC, and the “Cartes d'Identité des Tumeurs” (CIT) research program, funded and directed by the “Ligue Nationale Contre le Cancer,” and by grants from the “Institut National du Cancer” (INCa to AD). AD group has the label “La Ligue Contre le Cancer.” Notes Laetitia Marisa, Magali Svrcek, Ada Collura, Etienne Becht, Pascale Cervera, Kristell Wanherdrick, Olivier Buhard, Anastasia Goloudina, Vincent Jonchère, Janick Selves, Gerard Milano, Dominique Guenot, Romain Cohen, Chrystelle Colas, Pierre Laurent-Puig, Sylviane Olschwang, Jérémie H. Lefèvre, Yann Parc, Valérie Boige, Côme Lepage, Thierry André, Jean-François Fléjou, Valentin Dérangère, François Ghiringhelli*, Aurélien de Reynies*, Alex Duval* *Authors contributed equally to this work. Affiliations of authors: Programme “Cartes d'Identité des Tumeurs,” Ligue Nationale Contre le Cancer, Paris, France (LM, EB, AdR); INSERM, UMRS 938 - Centre de Recherche Saint-Antoine, Equipe “Instabilité des Microsatellites et Cancers,” Equipe labellisée par la Ligue Nationale contre le Cancer, Paris, France (MS, AC, PC, KW, OB, AG, VJ, RC, CC, JHL, YP, TA, JFF, AD); Sorbonne Université, UPMC Univ Paris 06, Paris, France (MS, AC, PC, KW, OB, AG, VJ, RC, CC, JHL, YP, TA, JFF, AD); AP-HP, Hôpital Saint-Antoine, Service d’Anatomie et Cytologie Pathologiques, Paris, France (MS, PC, JFF); Centre de Recherche en Cancérologie de Toulouse, UMR 1037 INSERM - Université Toulouse III, Department of Pathology, CHU, Toulouse, France (JS); Laboratoire d'Oncopharmacologie, EA 3836, Centre Antoine Lacassagne, Nice, France (GM); INSERM, U682, Développement et Physiopathologie de l’Intestin et du Pancréas, Strasbourg, France (DG); AP-HP, Hôpital Saint-Antoine, Service d’Oncologie Médicale, Paris, France (RC, TA); AP-HP, Laboratoire d’oncogénétique et d’Angiogénétique, GH Pitié-Salpétrière, Paris, France (CC); INSERM, Unité Mixte de Recherche, Paris Sorbonne Cité, Université Paris Descartes, Paris, France (PLP); Aix Marseille Univ, INSERM, GMGF, Marseille, France and RGDS, HP Clairval, Marseille, France (SO); AP-HP, Service de Chirurgie Générale et Digestive, Hôpital Saint-Antoine, Paris, France (JHL, YP); Department of Oncologic Medicine, Gustave-Roussy, Villejuif, France (VB); Université Paris Descartes, Paris Sorbonne Cité INSERM UMR-S775, Paris, France (VB); INSERM, Burgundy Cancer Registry, U866, Burgundy University, Dijon University Hospital, BP 87900 21079 Dijon, France (CL); Department of Medical Oncology, Centre Georges-François Leclerc, Dijon, France (VD, FG); INSERM, UMR866, Burgundy University (VD, FG); Platform of transfer in oncology, Burgundy University, Centre Georges-François Leclerc, Dijon, France (VD, FG). The funder, the French League Against Cancer, participated in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; and the decision to submit the manuscript for publication. We thank Prof. W. H. Fridman for his scientific assistance. Study concept and design: L. Marisa, F. Ghiringhelli, A. de Reynies, A. Duval; acquisition of data: all other authors; analysis and interpretation of data: L. Marisa, V. Dérengère, E. Becht, F. Ghiringhelli, A. de Reynies, A. Duval; drafting of the manuscript: L. Marisa, M. Svrcek, A. Collura, E. Becht, P. Cervera, K. Wanherdrick, O. Buhard, A. Goloudina, V. Jonchère, J. Selves, G. Milano, D. Guenot, R. Cohen, C. Colas, P. Laurent-Puig, S. Olschwang, J. H Lefèvre, Y. Parc, V. Boige, C. Lepage, T. André, JF Fléjou, V. Dérangère, F. Ghiringhelli, A. de Reynies, A Duval; statistical analysis: L. Marisa, A. de Reynies; study supervision: A. Duval; biological material suppliers: all other authors. References 1 Galon J, Costes A, Sanchez-Cabo F, et al.   Type, density, and location of immune cells within human colorectal tumors predict clinical outcome. Science.  2006; 313( 5795): 1960– 1964. Google Scholar CrossRef Search ADS PubMed  2 Angell HK, Gray N, Womack C, Pritchard DI, Wilkinson RW, Cumberbatch M. Digital pattern recognition-based image analysis quantifies immune infiltrates in distinct tissue regions of colorectal cancer and identifies a metastatic phenotype. Br J Cancer.  2013; 109( 6): 1618– 1624. Google Scholar CrossRef Search ADS PubMed  3 Mlecnik B, Tosolini M, Kirilovsky A, et al.   Histopathologic-based prognostic factors of colorectal cancers are associated with the state of the local immune reaction. J Clin Oncol.  2011; 29( 6): 610– 618. Google Scholar CrossRef Search ADS PubMed  4 Lothe RA, Peltomaki P, Meling GI, et al.   Genomic instability in colorectal cancer: relationship to clinicopathological variables and family history. Cancer Res.  1993; 53( 24): 5849– 5852. Google Scholar PubMed  5 Gryfe R, Kim H, Hsieh ET, et al.   Tumor microsatellite instability and clinical outcome in young patients with colorectal cancer. N Engl J Med.  2000; 342( 2): 69– 77. Google Scholar CrossRef Search ADS PubMed  6 Bauer K, Nelius N, Reuschenbach M, et al.   T cell responses against microsatellite instability-induced frameshift peptides and influence of regulatory T cells in colorectal cancer. Cancer Immunol Immunother.  2013; 62( 1): 27– 37. Google Scholar CrossRef Search ADS PubMed  7 Aaltonen LA, Peltomaki P, Leach FS, et al.   Clues to the pathogenesis of familial colorectal cancer. Science.  1993; 260( 5109): 812– 816. Google Scholar CrossRef Search ADS PubMed  8 Leach FS, Nicolaides NC, Papadopoulos N, et al.   Mutations of a mutS homolog in hereditary nonpolyposis colorectal cancer. Cell.  1993; 75( 6): 1215– 1225. Google Scholar CrossRef Search ADS PubMed  9 Thibodeau SN, Bren G, Schaid D. Microsatellite instability in cancer of the proximal colon. Science.  1993; 260( 5109): 816– 819. Google Scholar CrossRef Search ADS PubMed  10 Ionov Y, Peinado MA, Malkhosyan S, Shibata D, Perucho M. Ubiquitous somatic mutations in simple repeated sequences reveal a new mechanism for colonic carcinogenesis. Nature.  1993; 363( 6429): 558– 561. Google Scholar CrossRef Search ADS PubMed  11 Fishel R, Lescoe MK, Rao MR, et al.   The human mutator gene homolog MSH2 and its association with hereditary nonpolyposis colon cancer. Cell.  1993; 75( 5): 1027– 1038. Google Scholar CrossRef Search ADS PubMed  12 Dorard C, de Thonel A, Collura A, et al.   Expression of a mutant HSP110 sensitizes colorectal cancer cells to chemotherapy and improves disease prognosis. Nat Med.  2011; 17: 1283– 1289. Google Scholar CrossRef Search ADS PubMed  13 Duval A, Hamelin R. Mutations at coding repeat sequences in mismatch repair-deficient human cancers: Toward a new concept of target genes for instability. Cancer Res.  2002; 62( 9): 2447– 2454. Google Scholar PubMed  14 Hamelin R, Chalastanis A, Colas C, et al.   [Clinical and molecular consequences of microsatellite instability in human cancers]. Bull Cancer.  2008; 95( 1): 121– 132. Google Scholar PubMed  15 Llosa NJ, Cruise M, Tam A, et al.   The vigorous immune microenvironment of microsatellite instable colon cancer is balanced by multiple counter-inhibitory checkpoints. Cancer Discov.  2015; 5( 1): 43– 51. 16 Le DT, Uram JN, Wang H, et al.   PD-1 blockade in tumors with mismatch-repair deficiency. N Engl J Med.  2015; 372( 26): 2509– 2520. Google Scholar CrossRef Search ADS PubMed  17 Mlecnik B, Bindea G, Angell HK, et al.   Integrative analyses of colorectal cancer show immunoscore is a stronger predictor of patient survival than microsatellite instability. Immunity.  2016; 44( 3): 698– 711. Google Scholar CrossRef Search ADS PubMed  18 Dunne PD, McArt DG, O'Reilly PG, et al.   Immune-derived PD-L1 gene expression defines a subgroup of stage II/III colorectal cancer patients with favorable prognosis who may be harmed by adjuvant chemotherapy. Cancer Immunol Res.  2016; 4( 7): 582– 591. 19 Masugi Y, Nishihara R, Yang J, et al.   Tumour CD274 (PD-L1) expression and T cells in colorectal cancer. Gut.  2016; in press. 20 Lee LH, Cavalcanti MS, Segal NH, et al.   Patterns and prognostic relevance of PD-1 and PD-L1 expression in colorectal carcinoma. Mod Pathol.  2016; in press. 21 Li Y, Liang L, Dai W, et al.   Prognostic impact of programed cell death-1 (PD-1) and PD-ligand 1 (PD-L1) expression in cancer cells and tumor infiltrating lymphocytes in colorectal cancer. Mol Cancer.  2016; 15( 1): 55. Google Scholar CrossRef Search ADS PubMed  22 Guinney J, Dienstmann R, Wang X, et al.   The consensus molecular subtypes of colorectal cancer. Nat Med.  2015; 21( 11): 1350– 1356. Google Scholar CrossRef Search ADS PubMed  23 Melero I, Berman DM, Aznar MA, Korman AJ, Perez Gracia JL, Haanen J. Evolving synergistic combinations of targeted immunotherapies to combat cancer. Nat Rev Cancer.  2015; 15( 8): 457– 472. Google Scholar CrossRef Search ADS PubMed  24 Fridman WH, Pages F, Sautes-Fridman C, Galon J. The immune contexture in human tumours: Impact on clinical outcome. Nat Rev Cancer.  2012; 12( 4): 298– 306. Google Scholar CrossRef Search ADS PubMed  25 Marisa L, de Reynies A, Duval A, et al.   Gene expression classification of colon cancer into molecular subtypes: Characterization, validation, and prognostic value. PLoS Med.  2013; 10( 5): e1001453. Google Scholar CrossRef Search ADS PubMed  26 McCall MN, Bolstad BM, Irizarry RA. Frozen robust multiarray analysis (fRMA). Biostatistics.  2010; 11( 2): 242– 253. Google Scholar CrossRef Search ADS PubMed  27 Johnson WE, Li C, Rabinovic A. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics.  2007; 8( 1): 118– 127. Google Scholar CrossRef Search ADS PubMed  28 Becht E, Giraldo NA, Lacroix L, et al.   Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression. Genome Biol.  2016; 17( 1): 218. Google Scholar CrossRef Search ADS PubMed  29 Wherry EJ, Kurachi M. Molecular and cellular insights into T cell exhaustion. Nat Rev Immunol.  2015; 15( 8): 486– 499. Google Scholar CrossRef Search ADS PubMed  30 Blackburn SD, Shin H, Haining WN, et al.   Coregulation of CD8+ T cell exhaustion by multiple inhibitory receptors during chronic viral infection. Nat Immunol.  2009; 10( 1): 29– 37. Google Scholar CrossRef Search ADS PubMed  31 Ferris RL, Lu B, Kane LP. Too much of a good thing? Tim-3 and TCR signaling in T cell exhaustion. J Immunol.  2014; 193( 4): 1525– 1530. Google Scholar CrossRef Search ADS PubMed  32 Topalian SL, Drake CG, Pardoll DM. Immune checkpoint blockade: A common denominator approach to cancer therapy. Cancer Cell.  2015; 27( 4): 450– 461. Google Scholar CrossRef Search ADS PubMed  33 Sargent DJ, Marsoni S, Monges G, et al.   Defective mismatch repair as a predictive marker for lack of efficacy of fluorouracil-based adjuvant therapy in colon cancer. J Clin Oncol.  2010; 28( 20): 3219– 3226. Google Scholar CrossRef Search ADS PubMed  34 Venderbosch S, Nagtegaal ID, Maughan TS, et al.   Mismatch repair status and BRAF mutation status in metastatic colorectal cancer patients: A pooled analysis of the CAIRO, CAIRO2, COIN, and FOCUS studies. Clin Cancer Res.  2014; 20( 20): 5322– 5330. Google Scholar CrossRef Search ADS PubMed  35 Andre T, de Gramont A, Vernerey D, et al.   Adjuvant fluorouracil, leucovorin, and oxaliplatin in stage II to III colon cancer: Updated 10-year survival and outcomes according to BRAF mutation and mismatch repair status of the MOSAIC Study. J Clin Oncol.  2015; 33( 35): 4176– 4187. Google Scholar CrossRef Search ADS PubMed  36 Saeterdal I, Bjorheim J, Lislerud K, et al.   Frameshift-mutation-derived peptides as tumor-specific antigens in inherited and spontaneous colorectal cancer. Proc Natl Acad Sci U S A.  2001; 98( 23): 13255– 13260. Google Scholar CrossRef Search ADS PubMed  37 Becht E, de Reynies A, Giraldo NA, et al.   Immune and stromal classification of colorectal cancer is associated with molecular subtypes and relevant for precision immunotherapy. Clin Cancer Res.  2016; 22( 16): 4057– 4066. Google Scholar CrossRef Search ADS PubMed  Author notes See the Notes section for the full list of authors and affiliations. © The Author 2017. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com

Journal

JNCI: Journal of the National Cancer InstituteOxford University Press

Published: Aug 25, 2017

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off