The immune suppressive factors CD155 and PD-L1 show contrasting expression patterns and immune correlates in ovarian and other cancers

(1)

Citation for this paper:

Smazynski, J., Hamilton, P. T., Thornton, S., Milne, K., Wouters, M. C. A., Webb, J. R., &

Nelson, B. H. (2020). The immune suppressive factors CD155 and PD-L1 show contrasting

expression patterns and immune correlates in ovarian and other cancers. Gynecologic

Oncology, 158(1), 167-177. https://doi.org/10.1016/j.ygyno.2020.04.689.

UVicSPACE: Research & Learning Repository

_____________________________________________________________

Faculty of Science

Faculty Publications

_____________________________________________________________

The immune suppressive factors CD155 and PD-L1 show contrasting expression

patterns and immune correlates in ovarian and other cancers

Julian Smazynski, Phineas T. Hamilton, Shelby Thornton, Katy Milne, Maartje C. A.

Wouters, John R. Webb, & Brad H. Nelson

July 2020

© 2020 Julian Smazynski et al. This is an open access article distributed under the terms of the

Creative Commons Attribution License.

https://creativecommons.org/licenses/by-nc-nd/4.0/

This article was originally published at:

(2)

The immune suppressive factors CD155 and PD-L1 show contrasting

expression patterns and immune correlates in ovarian and other cancers

Julian Smazynski

a,b,1

, Phineas T. Hamilton

a,1

, Shelby Thornton

a

, Katy Milne

a

, Maartje C.A. Wouters

a

,

John R. Webb

a,b

, Brad H. Nelson

a,b,c,

⁎

a

Deeley Research Centre, BC Cancer, Victoria, BC V8R 6V5, Canada

b

Department of Biochemistry and Microbiology, University of Victoria, Victoria, BC V8P 3E6, Canada

c

Department of Medical Genetics, University of British Columbia, Vancouver, BC V6T 1Z3, Canada

H I G H L I G H T S

• Members of the CD155/TIGIT immune checkpoint pathway are commonly expressed in HGSC and other cancers. • In HGSC, expression of CD155 and TIGIT is substantially more frequent than expression of PD-L1 and PD-1. • In contrast to PD-L1, CD155 is commonly expressed by immunologically cold tumors.

• CD155 and PD-L1 appear to represent non-redundant immune checkpoints in HGSC.

a b s t r a c t

a r t i c l e i n f o

Article history:

Received 10 December 2019 Accepted 15 April 2020 Available online 20 May 2020 Keywords: Immunotherapy CD155 PD-L1 TIGIT Cold tumors Checkpoint blockade

Objective. We recently showed that tumors with an immunologically‘cold’ phenotype are enriched for ex-pression of stemness-associated genes and PVR/CD155, the ligand of the immunosuppressive molecule TIGIT. To explore the therapeutic implications of thisﬁnding, we investigated the relationship between PVR/CD155 ex-pression, tumor-inﬁltrating lymphocytes (TIL), and prognosis in high-grade serous ovarian cancer (HGSC) and other cancers.

Methods. Expression of CD155, TIGIT, PD-1, PD-L1, and other immune markers in HGSC was assessed by high-dimensionalflow cytometry, multi-color histological imaging, and/or gene expression profiling. The prognostic significance of PVR/CD155 and CD274/PD-L1 expression was assessed bioinformatically in HGSC and 32 other cancers in The Cancer Genome Atlas.

Results. T cells from HGSC frequently co-expressed TIGIT and PD-1, and the ratio of TIGIT to PD-1 expression increased markedly after in vitro expansion with a clinically relevant protocol. CD155 was commonly expressed on malignant epithelium in HGSC and showed a negative or non-signiﬁcant association with TIL. In contrast, PD-L1 was predominantly expressed by tumor-associated macrophages and positively associated with TIL. These contrasts between CD155 and PD-L1 were seen across HGSC patients, across metastatic sites within individual patients, and even within individual tumor deposits. PVR/CD155 and CD274/PD-L1 exhibited divergent prognos-tic associations across diverse cancer types in TCGA, including HGSC.

Conclusions. CD155 and PD-L1 exhibit contrasting expression patterns, TIL associations and prognostic signif-icance, suggesting they represent non-redundant immunosuppressive mechanisms. The CD155/TIGIT pathway represents a compelling immunotherapeutic target for HGSC and for immunologically cold tumors in general.

1. Introduction

Monoclonal antibodies targeting the PD-1/PD-L1 signaling axis have shown striking clinical success against multiple malignancies, leading to regulatory approvals for melanoma, non-small cell lung cancer, renal cell cancer, and many others [1]. However, despite this important advance, the majority of cancers show unacceptably low response rates to PD-1/PD-L1 blockade [2]. While the reasons for this remain

⁎ Corresponding author at: British Columbia Cancer Agency, 2410 Lee Avenue, Victoria, BC V8R 6V5, Canada.

E-mail address:bnelson@bccrc.ca(B.H. Nelson).

1

These authors contributed equally to this work.

https://doi.org/10.1016/j.ygyno.2020.04.689

Contents lists available atScienceDirect

Gynecologic Oncology

(3)

incompletely understood, the efficacy of PD-1/PD-L1 blockade is associ-ated with high tumor mutation burden [3] and the presence of tumor-in_{filtrating lymphocytes (TIL) prior to therapy [}3]. Thus, new immuno-therapeutic strategies are urgently needed for the many cancers with low/intermediate mutation burden and a TIL-deficient (immunologi-cally_{‘cold’) phenotype.}

High-grade serous ovarian cancer (HGSC) is a challenging disease that is largely resistant to today's immunotherapies, as evidenced by a 10–15% objective response rate to PD-1/PD-L1 blockade [4]. This may partly reflect the intermediate mutation load associated with HGSC and the corresponding paucity of mutation-derived antigens (neoantigens) [5]. Moreover, we and others have shown in HGSC that PD-L1 is expressed predominantly by tumor-infiltrating macro-phages rather than malignant cells [6], suggesting the PD-1/PD-L1 axis may represent a less significant barrier to TIL in HGSC compared to other malignancies.

Despite being resistant to PD-1/PD-L1 blockade, HGSC is clearly an immunologically active disease. In particular, a substantial proportion of HGSC cases present with vigorous TIL responses involving CD8 and CD4 T cells, B cells, plasma cells, and macrophages [7]. We recently re-ported evidence that TIL responses are productive in HGSC, as indicated by‘pruning’ of tumor clones and increased signs of immune editing at TIL-positive (‘hot’) tumor sites [8]. Indeed, the presence of TIL is strongly associated with patient survival in multiple HGSC cohorts [9]. These findings provide hope that immunological control of HGSC may be pos-sible, but this will require identification of the most significant immuno-logical barriers in this malignancy.

Since its initial discovery, CD155 has been shown to play a key role in controlling anti-viral and anti-tumor immune responses [10,11]. Encoded by the PVR (poliovirus receptor) gene, CD155 belongs to the nectin family of proteins [12] and, along with CD112 (encoded by PVRL2), participates in a signaling network that promotes cell adhesion, migration, proliferation, and contact inhibition [12–14]. Expression of CD155 and CD112 is induced by the Raf-Mek-ERK, Sonic Hedgehog, and ATM/ATR DNA damage response pathways [13,15,16]. Accordingly, heightened expression of CD155 and corresponding negative correla-tions with survival have been reported in multiple cancers including pancreatic cancer [17], osteosarcoma [18], non-small lung cancer [19] and breast cancer [20].

In addition to their direct roles in tumorigenesis [10,14], CD155 and CD112 can modulate immune function by interacting with receptors expressed by immune cells, including the co-stimulatory molecule DNAX Accessory Molecule-1 (DNAM-1; encoded by CD226) [21] and the inhibitory molecules TIGIT (T cell immunoreceptor with Ig and ITIM domains) [22] and CD96 (Tactile; T cell activation, increased late expression) [22]. Given the complexity of this network, the net in flu-ence of CD155 and CD112 expression on immune responses depends on context. However, in the setting of cancer, TIGIT in particular is under active investigation as a target for checkpoint blockade owing to its clear inhibitory effects on T cell proliferation and effector function [23,24]. In preclinical models, TIGIT blockade has limited efficacy as a monotherapy but can significantly potentiate the efficacy of PD-1 and CD96 blockade [11,25].

In a recent pan-cancer analysis, we identified a pervasive association between a stem cell-like (‘stemness’) gene expression signature in tu-mors and a TIL-deficient phenotype [26]. This stemness signature was strongly associated with expression of PVR, suggesting that CD155 ex-pression is negatively associated with cancer cell differentiation and may contribute to the immunologically cold phenotype of high stemness tumors. To explore this possibility, we investigated the expression pat-terns and prognostic significance of members of the CD155/TIGIT path-way relative to the PD-1/PD-L1 pathpath-way in diverse HGSC cohorts, as well as other cancers from The Cancer Genome Atlas (TCGA). Our find-ings indicate that blockade of the TIGIT/CD155 axis merits clinical inves-tigation in HGSC and related malignancies and may be particularly relevant to immunologically cold tumors.

2. Materials and methods

2.1. Patient cohorts for multispectral_{ﬂow cytometry and histological} analysis

Patient specimens were accessed through BC Cancer's Tumor Tissue Repository (a member of the Canadian Tissue Repository Network). Specimens and clinical data were obtained with either written informed consent or a formal waiver of consent under protocols approved by the Research Ethics Board of the BC Cancer and the University of British Co-lumbia (H07-00463). All specimens were obtained from primary sur-geries prior to chemotherapy or any other treatment.

2.2. Multispectralﬂow cytometry and In vitro T-cell expansion See Supplementary methods.

2.3. Immunohistochemistry, immunoﬂuorescence and image analysis See Supplementary methods.

2.4. Transcriptomic, genomic and clinical datasets

Publicly available datasets used in this study included i) harmonized HGSC and other cancer RNA-seq and clinical data in the PanCanAtlas (https://gdc.cancer.gov/about-data/publications/pancanatlas); ii) HGSC gene expression and clinical data in the curatedOvarianData Bioconductor package [27] (which also includes TCGA samples); and iii) published gene expression (Nanostring Immune Panel) and matched IHC-based TIL data [8].

2.5. Statistical methods

We used R v.3.5.2 (the R Project for Statistical Computing) and GraphPad Prism (v8.2) for statistical analyses andﬁgure generation. As-sociations between checkpoint ligand expression and other measures (e.g., T cell inﬁltration) were tested using general linear models and log-transformed response variables (log10(x + 1)) where required to

meet linear model assumptions. For cohorts with hierarchical designs (e.g., multiple tumor sites per patient [8]), we used mixed effects models to test for associations between ligand expression and TIL markers. Cox proportional hazards models were used to evaluate associations be-tween checkpoint ligand expression and patient outcome. The metafor package [28] was used to conduct meta-analysis of outcomes, as outlined in the curatedOvarianData package. For survival analyses across different cancers, we stratified Cox models by cancer type to account for cancer-specific effects on hazard. We further evaluated cancer-specific hazards by estimating separate log hazard ratios for each cancer. Scripts to reproduce the analysis will be posted atwww.github.com/vicDRC/ ithCD155.

3. Results

3.1. T cells in HGSC co-express TIGIT and PD-1, and TIGIT expression is en-hanced during in vitro T cell expansion

To evaluate the potential influence of TIGIT signaling on anti-tumor immunity in HGSC, we used multispectralflow cytometry to profile the cell surface phenotype of lymphocytes from matched primary ascites and solid tumor samples from 11 patients. We will refer to lymphocytes from ascites as“tumor-associated lymphocytes” (TAL) and those from solid tumor as“tumor-infiltrating lymphocytes” (TIL). We evaluated 22 phenotypic markers, including canonical T cell exhaustion and/or ac-tivation surface markers (e.g. TIGIT, PD-1, CD96, CD39, GITR, DNAM-1, CD69, CD103, CD27, and CD28), as well as CD56 to identify NK and NKT cells (Supplementary Table S1 and Supplementary Fig. S1).

(4)

t-SNE analysis revealed diverse T cell sub-populations (Fig. 1A) that were largely shared across the 11 patients. TIGIT was frequently expressed by CD4+ and CD8+ TAL (mean 33.58% and 57.56%, respec-tively) and even more frequently by CD4+ and CD8+ TIL (59.28% and 72.74%, respectively; Pb 0.05;Fig. 1B). In comparison, the frequency of PD-1 expression was moderately lower on CD4+ and CD8+ TAL

(29.79% and 42.3%, respectively) and similar on TIL (64.03% and 69.61%, respectively; Pb 0.01;Fig. 1B).

Co-expression of TIGIT and PD-1 was seen on only a minority of TAL (17.35% and 28.9% of CD4+ and CD8+ TAL; Fig. S2A) but was signi ﬁ-cantly more frequent on TIL (50.82% and 50.49% of CD4+ and CD8+ TIL; P_{b 0.05; Fig. S2A). The majority of TIGIT+ CD8+ and CD4+ TIL}

Fig. 1. Expression of TIGIT, PD-1 and other immunoregulatory factors on CD4 and CD8 T cells in HGSC. A. t-SNE projections of combinedflow cytometry data depicting CD45+ cells from primary tumor samples (n = 11 patients). Heatmap shows the individual marker expression normalized to the medianfluorescent intensity. B. Expression of cell surface proteins from tumor-associated lymphocytes (TAL) and tumor-infiltrating lymphocytes (TIL) from the same 11 cases as panel A, as assessed by flow cytometry. Points indicate individual patients (n = 11). The percentages of CD8+ or CD4+ T cells (gated on CD3) expressing the indicated surface markers from ascites (green) and tumor (purple) are shown. P values were calculated using paired t-tests (*Pb 0.05; **P b 0.01; ***P b 0.001). An example of the gating strategy is shown in Fig. S1. C. Correlation between TIGIT and PD1 (PDCD1) gene expression in HGSC cases from TCGA (Pb 10−12_{; Spearman correlation). The residuals of linear models controlling from T cell infiltration are shown to account for overall infiltration of T cells in analysis. Colors indicate}

(5)

co-expressed PD-1 (78.87% and 87.21%, respectively; Fig. S2C) and con-versely, the majority of PD-1+ CD8+ and CD4+ TIL co-expressed TIGIT (83.58% and 79.65%, respectively; Fig. S2C). TIGIT+ CD4+ and CD8+ TIL showed lower expression of DNAM-1 (a co-stimulatory ligand for CD155) compared to TIGIT- TIL (Pb 0.001) (Supplementary Fig. S3A). CD4+ and CD8+ TIL and TAL that co-expressed PD-1 and TIGIT showed lower expression of CD45RA and CCR7, suggesting an effector memory

phenotype (Supplementary Fig. S3B and C). Compared to TAL, TIL showed elevated expression of CD39, CD69, and CD103 on the CD8+ subset (Fig. 1B; P_{b 0.01) and CD39, CD69, and GITR on the CD4+ subset} (Fig. 1B; Pb 0.01).

In both tumor and ascites, NK cells expressed substantially more TIGIT than PD-1 (P_{b 0.0001; Supplementary Fig. S4A) and infrequently} co-expressed the two markers (Supplementary Fig. S4D). Expression of

Fig. 2. Assessment of CD155, CD112 and PD-L1 expression in HGSC. A. Representative IHC images of HGSC cores stained with antibodies to PD-L1, CD112 or CD155 and categorized as low versus high expression. B. Summary of IHC-based H-scores for PD-L1, CD112 and CD155 from a cohort of 51 HGSC cases. The dashed line indicates an H-score of 50, which was used to categorize samples as having high versus low expression. Each point represents the mean H-score for replicate TMA cores from an individual patient. C. Relationship between CD155, CD112 and PD-L1 expression based on the IHC data from panel B. For each patient, mean H scores for the indicated markers were compared using Spearman correlation. D. Expression of CD155, CD112 and other immunoregulatory factors on myeloid populations across HGSC patient tumors. t-SNE projection depicting CD45+ populations from primary patient tumor samples ex vivo (n = 11). Heatmap shows individual marker expression normalized to the medianﬂuorescent intensity.

(6)

both TIGIT and PD-1 was higher on NK cells from tumor compared to as-cites (Pb 0.05). NK cells also showed high expression of CD39, CD69, CD103, and GITR (Supplementary Fig. S4A).

In general, the results for NKT cells (CD3+ CD56+) were similar to T cells, including elevated frequencies of TIGIT+, PD-1+ and TIGIT+PD-1 + NKT cells derived from tumor compared to ascites (Supplementary Fig. S4B, C, and D), as well as frequent expression of CD39 and CD69 (52.23% and 93.00%, respectively; Supplementary Fig. S4B).

In contrast to TIL and TAL, T cells from healthy donor PBMC had much lower expression of TIGIT (Pb 0.05) and PD-1 (P b 0.001) (Sup-plementary Fig. S5A). This was also the case for NK cells (Supplemen-tary Fig. S5C). Likewise, in matched samples from an HGSC patient, expression of both TIGIT and PD-1 was higher on CD4+ and CD8+ TIL compared to PBMC (Supplementary Fig. S5B).

We evaluated changes in TIGIT expression after in vitro expansion of 5 patient ascites samples using a standard clinical protocol (see Supple-mentary Table S2 forﬂow cytometry panel). Following expansion, the frequency of TIGIT+ CD8+ T cells increased from 39.9% to 72.3% (P = 0.024; Supplementary Fig. S6A). A similar fold increase was seen on CD4+ T cells (from 24.1% to 43.3%), although this did not reach sig-niﬁcance (P = 0.055; Supplementary Fig. S6A). In contrast, expression of PD-1 on CD8+ T cells decreased dramatically after expansion (from 38.8% to 5.24%; paired t-test, Pb 0.001; Supplementary Fig. S6A), while remaining unchanged on CD4+ T cells (17.8% to 17.9%; P = 0.99; Supplementary Fig. S6A). Accordingly, after expansion, the pro-portion of T cells that co-expressed TIGIT and PD-1 decreased for CD8 + T cells (28.9% to 4.63%; P = 0.0131) but not CD4+ T cells (12.5% to 13.2%; P = 0.8994; Supplementary Fig. S6A). Expansion induced other changes that may also affect TIGIT/CD155 signaling. Of particular inter-est, CD96 (a co-inhibitory binding partner for CD155) dramatically in-creased on both CD4+ and CD8+ T cells (12.1% to 70.4% and 13.6% to 42.1%, respectively; P = 0.0003 and P = 0.0208).

To confirm the association between TIGIT and PD-1 expression on CD8+ T cells in a larger dataset, we assayed mRNA expression data from HGSC cases profiled by TCGA, focusing on cases with defined mo-lecular subtypes (n = 292). Based on recent approaches [29], we in-ferred enrichment of TIGIT and PD-1 using the residuals of log-linear models that controlled for CD8+ T cell levels (as the geometric mean of CD8A and CD8B to account for overall CD8+ T cell infiltration [29]). This residual TIGIT correlated with residual PD-1 (PDCD1) expression (ρ = 0.39, P b 10−11_;_{Fig. 1}_{C), corroborating their co-expression while}

controlling for overall CD8 levels. To exclude the possibility that this co-expression arose from NK cells, we alsoﬁt linear models that con-trolled for estimated NK inﬁltration [29], which did not substantively af-fect our estimates of the association between TIGIT and PD-1.

3.2. CD155 and CD112 are highly expressed on tumor cells, APCs and tumor vasculature in HGSC

To better understand the potential role of CD155 as an immune checkpoint in HGSC, we evaluated the expression of CD155 and the re-lated protein CD112 by immunohistochemical staining of a tissue mi-croarray (TMA) containing primary (untreated) tumor samples from 51 HGSC cases (Table 1). To mitigate the sampling issues inherent in the use of TMA cores and to ensure diverse representation of the im-mune microenvironment, the TMA was constructed to contain replicate tumor and stromal regions from each case. Both CD155 and CD112 were primarily expressed on the malignant epithelium, with relatively less expression on stromal or immune cells (Fig. 2A). High expression (H scoreN 50) of CD155 and CD112 was seen in 98.1% and 68.6% of cases, respectively (Fig. 2B). Expression of CD155 and CD112 was positively correlated across patients (Spearman'sρ = 0.45, P b 0.001;Fig. 2C), con-sistent with a common regulatory pathway underlying their expression [10]. We observed low CD155 expression in normal fallopian tube epi-thelium relative to the majority of HGSCs, suggesting a relationship be-tween epithelial CD155 expression and malignancy (Supplementary

Fig. S7B). Last, we examined tumor vasculature and found that CD31+ endothelial cells were frequently positive for CD155 (Supplementary Fig. S8), consistent with prior reports [30]. These data collectively suggest that TIGIT expressing T cells will commonly encounter the in-hibitory ligands CD155 and CD112 in the HGSC microenvironment. In-terestingly, neither CD155 nor CD112 showed a correlation with PD-L1 expression (P = 0.33 and 0.83, respectively;Fig. 2C). Moreover, in contrast to CD155/CD112, PD-L1 was primarily expressed on inﬁltrating immune cells (mostly macrophages) in stromal regions (Supplemen-tary Fig. S7A), as we have previously reported [6].

To further characterize other cell types expressing CD155 and CD112, we evaluated lymphoid and myeloid cell populations from matched pri-mary ascites and tumor byﬂow cytometry using a 21-antibody panel that allowed discrimination of myeloid subpopulations (Supplementary Table S3). t-SNE analysis revealed a distinct macrophage subpopulation (CD3_{− CD14+ HLA-DRhigh+) that co-expressed CD155, CD112,} CD276, and CD32 (Fig. 2D). We sub-divided macrophages into M1 (CD206−/CD163−) and M2 (CD206+/CD163+) populations (Supple-mentary Figs. S9 and S10). Intriguingly, M2 populations had substantially increased expression of CD155, CD112, PD-L1, and PD-L2 relative to M1 macrophages (P b 0.0001; Supplementary Fig. S10A). Tumor cells (EpCAM+/CD45−) expressed high levels of CD155, CD112, CD276, and HLA-DR (allN70% positivity; Supplementary Fig. S10B) but expressed PD-L1 and PD-L2 expression infrequently (14.7% and 15.1% respectively) (Supplementary Fig. S10B).

3.3. CD155 is expressed on immunologically cold tumors in HGSC We and others have previously reported that PD-L1 expression is positively associated with the presence of TIL and improved prognosis in HGSC [6,9], presumably reﬂecting PD-L1 induction via T cell-derived IFN-_{γ [}31]. In contrast, CD155 and CD112 are upregulated as part of the Raf-Mek-ERK, SHH, and ATM/ATR DNA damage response pathways [10] and hence might be expected to show a different associ-ation with immune inﬁltration. Indeed, we observed a positive associa-tion between T cells (both CD3+ and CD8+ subsets) and PD-L1 expression in our 51-case HGSC cohort, whereas CD155 showed nega-tive correlations with these TIL subsets (Fig. 3A).

We expanded our analysis using other HGSC cohorts with IHC and/ or gene expression-based measurements of CD155/PVR and related markers. By Nanostring analysis, we again detected a negative correla-tion between PVR (CD155) expression and immune in_{ﬁltration in our} 51-case cohort (Fig. 3B). Furthermore, PVR (measured by NanoString) showed a striking negative association with CD8+ T cells (measured by IHC) in the multisite cohort of Zhang et al. [8] (Fig. 3C). These multi-site HGSC samples were also previously clustered into 3 immune subtypes: N-TIL (tumors with no/few TIL), S-TIL (TIL restricted to stroma), and ES-TIL (TIL in tumor epithelium and stroma) [8]. Consis-tent with the above analyses, N-TIL sites showed the highest expression of PVR (linear mixed effect model; P = 0.003) and the lowest expression of CD274 (PD-L1; Pb 0.001) (Fig. 3D). Analysis of these multi-site

Table 1

Clinical and pathological characteristics of patients in the study cohort.

Characteristic (n)

Age (years), mean ± SD 64.2 ± 11.3 Histologic subtype

High grade serous 51

Stage IC 1 IIB 2 IIIC 3 IIIA 1 IIIB 3 IIIC 35 IV 6

(7)

(8)

samples also revealed that PVR expression was significantly higher in ovarian and other disease sites compared to omental sites (Pb 0.05) (Supplemental Fig. S11), with the latter also showing a trend towards greater T cell infiltration (P = 0.055; data not shown). However, not all datasets showed these negative associations. For instance, in TCGA transcriptome datasets for HGSC, inferred CD8 T cells were not signi fi-cantly related to PVR expression (rho = 0.05, P = 0.39, N = 292) but did show the expected positive association with CD274 expression (rho = 0.50, Pb 10−15_{). Finally, analysis across cancer types}

repre-sented in TCGA revealed negative correlations between PVR gene ex-pression and cytolytic score (geometric mean of PRF1 and GZMA expression) in 12 of 33 cancers, with the remaining cancers showing neutral or even positive associations (Table 2).

3.4. Microregional analysis reveals greater co-localization of TIL with PD-L1 than CD155

The precedingfindings suggested that, in contrast to PD-L1, CD155 may be associated with lymphocyte exclusion in HGSC. To further ex-plore this possibility, we assessed at the microregional level whether T cell infiltrates were reduced or absent in CD155+ regions of tumors rel-ative to PD-L1+ regions. We developed a multi-color IF panel that allowed simultaneous detection of CD155, PD-L1, and T cell markers (CD3 and CD8). We included CD68 to detect macrophages (the predom-inant PD-L1+ cell type in HGSC) and pan-cytokeratin to detect tumor epithelium. We attempted to include TIGIT in this panel; however, none of the commercial anti-TIGIT antibodies we assessed gave reliable staining patterns on FFPE tissue (see Supplementary methods and Sup-plementary Table S6). Instead, we used PD-1 as a surrogate for TIGIT, given ourfinding that the majority of PD-1+ CD8+ and CD4+ TIL co-express TIGIT (83.58% and 79.65%, respectively) (Supplementary Fig. S2C). The IF panel was applied to our 51-case primary HGSC cohort (Fig. 4A, Supplementary Table S5). We evaluated co-localization of CD8, PD-1, CD155 and PD-L1 in all TMA cores that had at least 5 cells express-ing each of these markers (n = 46 evaluable patients with at least one core each). Similar to prior studies [8,32], we used the Getis-Ord Gi* sta-tistic [33] to identify‘hotspots’ of CD8, PD-1, CD155 and PD-L1 expres-sion in tumor epithelium. We then calculated a modified Fcscore for

each core, deﬁned as the proportion of CD155+ or PD-L1+ tumor hotspots that were also PD1+ or CD8+ immune cell hotspots. CD8+ T cell hotspots showed a substantially higher overlap with PD-L1+ hotspots than with CD155+ hotspots (Pb 10−5_;_{Fig. 4}_{B). Similarly,}

PD-1+ hotspots showed a substantially higher overlap with PD-L1+ hotspots than with CD155+ hotspots (Pb 10−5_;_{Fig. 4}_{B). Thus, even}

at the microregional level, CD8+ and PD-1+ TIL showed relative exclu-sion from CD155+ tumor regions compared to PD-L1+ regions. 3.5. Prognostic signiﬁcance of CD155 and PD-L1 expression across cancers

Recent studies have reported that CD155 (or PVR) expression is as-sociated with poor outcome in a variety of cancers [17,19,20]. As our 51-case cohort had an insufﬁcient sample size for robust outcomes anal-ysis, we instead explored this issue systematically across cancers by comparing PVR and CD274 (PD-L1) expression with patient survival using harmonized pan-cancer data provided by TCGA (https://gdc. cancer.gov/about-data/publications/pancanatlas). For most cancers, PVR and CD274 showed divergent associations with survival, as reﬂected by non-correlated hazard ratios across cancers (P = 0.8, Fig. 5A). PVR expression was negatively associated with survival in many cancers and demonstrated an overall negative association in

pan-cancer survival models stratiﬁed by cancer type (log hazard = 0.18 ± 0.026 (SE); Pb 10–12;Fig. 5A). In contrast, CD274 (PD-L1) did not show a signiﬁcant association with survival in this modeling frame-work (P = 0.261;Fig. 5A).

Within HGSC, we used meta-analysis of publicly available gene ex-pression datasets (using multivariable models to control for debulking status [27]). This revealed a non-signiﬁcant association between PVR ex-pression and survival (_{ﬁxed effects meta-analysis; P = 0.11; log HR =} 0.04,Fig. 5B). In contrast, CD274 expression was positively prognostic (Pb 0.001; log HR = −0.13;Fig. 5B), highlighting the distinct biological roles of these immunological checkpoints in HGSC.

4. Discussion

It is well established that immune cells face diverse inhibitory stim-uli within the tumor microenvironment, but understanding the key sig-naling events that limit effective anti-tumor immunity remains a challenge. While targeting PD-1 or its ligand PD-L1 has shown striking efﬁcacy in some cancers, it is ineffective in many others, including HGSC. Here, we showed that T cells from the ascites and solid tumor micro-environments of HGSC frequently co-express TIGIT and PD-1, and expression of TIGIT is exacerbated after expansion with a standard clinical protocol. In contrast to PD-L1, CD155 and CD112 were broadly

Table 2

Association between cytolytic gene score (geometric mean of PRF1 and GZMA) and PVR ex-pression in cancers in TCGA.

Cancer type PVR CD274

ρ Padjvalue ρ Padjvalue

Cholangiocarcinoma −0.42 6.19E-02 0.34 8.87E-02 Colon adenocarcinoma −0.29 1.32E-07* 0.73 Pb E-10* Pancreatic adenocarcinoma −0.27 1.37E-03* 0.41 2.36E-08* Prostate adenocarcinoma −0.26 5.61E-07* 0.51 Pb E-10* Adrenocortical carcinoma −0.26 5.42E-02 0.49 7.17E-06* Rectum adenocarcinoma −0.23 1.47E-02* 0.59 Pb E-10* Skin Cutaneous Melanoma −0.19 2.31E-04* 0.71 Pb E-10* Thyroid carcinoma −0.19 2.31E-04* 0.36 Pb E-10* Liver hepatocellular carcinoma −0.17 1.06E-02* 0.4 Pb E-10* Stomach adenocarcinoma −0.16 8.66E-03* 0.65 Pb E-10* Kidney renal clear cell carcinoma −0.13 1.45E-02* 0.12 1.21E-02* Head and neck squamous cell carcinoma −0.11 4.06E-02* 0.53 Pb E-10* Acute Myeloid Leukemia −0.1 3.30E-01 0.44 2.15E-09* Lung adenocarcinoma −0.087 1.26E-01 0.53 Pb E-10* Esophageal carcinoma −0.058 6.60E-01 0.48 Pb E-10* Sarcoma −0.054 6.13E-01 0.39 Pb E-10* Thymoma −0.012 9.79E-01 0.26 4.95E-03* Lung squamous cell carcinoma −0.0087 9.79E-01 0.41 Pb E-10* Glioblastoma multiforme −0.00096 9.90E-01 0.28 6.48E-04* Kidney renal papillary cell carcinoma 0.0037 9.80E-01 0.17 8.42E-03* Uterine Corpus Endometrial Carcinoma 0.0088 9.79E-01 0.47 Pb E-10* Uterine Carcinosarcoma 0.014 9.79E-01 0.49 1.52E-04* Ovarian serous cystadenocarcinoma 0.015 9.79E-01 0.62 Pb E-10* Kidney Chromophobe 0.019 9.79E-01 0.046 7.70E-01 Breast invasive carcinoma 0.024 6.60E-01 0.57 Pb E-10* Mesothelioma 0.025 9.79E-01 0.48 5.25E-06* Pheochromocytoma and Paraganglioma 0.043 7.98E-01 0.16 4.26E-02* Cervical squamous cell carcinoma and

endocervical adenocarcinoma

0.088 2.38E-01 0.47 Pb E-10* Uveal Melanoma 0.15 3.30E-01 0.46 2.89E-05* Brain Lower Grade Glioma 0.19 1.73E-04* 0.32 Pb E-10* Testicular Germ Cell Tumors 0.23 2.75E-02* 0.7 Pb E-10* Bladder Urothelial Carcinoma 0.28 3.30E-07* 0.64 Pb E-10* Lymphoid Neoplasm Diffuse Large

B-cell Lymphoma

0.32 5.94E-02 0.54 1.27E-04*

* Padjb 0.05.

Fig. 3. Expression of CD155/PVR and PD-L1/CD274 relative to TIL in HGSC. A. Relationship between CD8+ T cell density and PD-L1 or CD155 as assessed by IHC of a 51-case cohort (from

Fig. 2). Each point represents an individual patient. B. Relationship between IHC-based CD8+ T cell density and NanoString-based CD274 or PVR gene expression in the same cohort as panel A. Each point represents an individual patient. C. Relationship between IHC-based CD8+ T cell density and NanoString-based CD274 or PVR gene expression in a multi-site HGSC cohort [8]. Colors represent individual patients, and points represent individual tumor sites. D. Expression of CD274 and PVR (assessed by NanoString) in multi-site HGSC with TIL patterns previously classified [8] as non-infiltrated (N-TIL), infiltrated in stroma only (S-TIL), or infiltrated in both stroma and epithelium (ES-TIL) (linear mixed effects model; P b 0.01).

(9)

expressed on malignant epithelium in HGSC and correlated with one another. Whereas PD-L1 was positively associated with TIL, CD155 was commonly expressed in TIL-negative tumors, and this effect was seen across patients, across metastatic sites within individual patients, and even at the microregional level in individual tumor specimens. As expected, PD-L1 expression correlated with increased survival in HGSC, whereas CD155 expression did not. The dramatically greater ex-pression of TIGIT versus PD-1 on TIL, and of CD155 versus PD-L1 on tumor epithelium, provides strong support for clinical investigation of TIGIT blockade in the setting of HGSC, and more generally, against tu-mors with an immunologically cold phenotype.

We found that in both ascites and solid tumor samples, TIGIT was expressed on the majority of CD4+ T cells, CD8+ T cells, and NK cells. In addition, a large proportion of TIGIT+ T cells co-expressed PD-1 and other co-stimulatory/co-inhibitory related proteins, including CD69, CD103, GITR, and CD39. CD8+ T cells that co-expressed TIGIT and PD-1 were enriched for a CD45RA-CCR7- effector memory (TEM)

phenotype, consistent with a previous report that TIGIT+PD-1+

tumor-reactive CD8+ T cells in melanoma exhibited a TEMphenotype

[23]. We found that expression of the co-stimulatory receptor DNAM-1 was drastically reduced on TIGIT+ T cells, both the PD-DNAM-1+ and PD-DNAM-1 − subpopulations, supporting previous reports that TIGIT may play a role in regulating DNAM-1 expression [11,23]. Given that DNAM-1 has been shown to participate in anti-tumor responses in the context of PD-1 blockade [11,34], our results suggest that high TIGIT and low DNAM-1 expression on TIL may contribute to the limited efﬁcacy of PD-1 blockade in HGSC and other cancers. Indeed, in several preclinical tumor models, TIGIT and PD-1 blockade have shown therapeutic syn-ergy [11,25,34].

A priority for improving adoptive cell therapies is to deﬁne which checkpoint markers are expressed on T cells after in vitro expansion. We found that, in contrast to PD-1, TIGIT expression is markedly in-creased on both CD8+ and CD4+ TIL after expansion with a clinically relevant protocol. Moreover, expanded TIL showed increased expres-sion of CD96, a second inhibitory receptor for CD155 and CD112. CD96 has been linked to diminished anti-tumor and anti-viral immunity

Fig. 4. Expression of CD155 and PD-L1 in relation to TIL in HGSC. A. Representative multi-color IF image from an HGSC TMA core, stained to detect the indicated markers. Dashed box in the merged image indicates the region shown in the smaller images. B. Summary of results from Getis-Ord Gi* analysis showing the extent of colocalization of PD-1+ (left) and CD8+ (right) hotspots with PD-L1+ and CD155+ hotspots in the tumor epithelium, as fractional co-localization (Fc). Each line indicates an individual patient (Pb 10−5; n = 46 evaluable patients).

(10)

[22,25,35], and in a recent study, CD96 blockade synergized with TIGIT and PD-1 blockade to enhance anti-tumor CD8+ T cell responses in multiple murine models [25]. We also observed signi_{ﬁcantly increased} expression of the co-stimulatory receptor GITR on CD4+ T cells and, to a lesser extent, CD8+ T cells after expansion. Ligation of GITR en-hances T cell proliferation and cytokine production and synergized with PD-1 blockade in a murine ovarian cancer model [36]. Further-more, in a murine colorectal carcinoma model, PD-1 blockade in combi-nation with GITR agonism rescued DNAM-1 signaling and reduced TIGIT expression on CD8+ TIL [34]. Collectively, the altered patterns of co-inhibitory and co-stimulatory receptor expression we observed after TIL expansion provide strong rationale for investigating TIGIT pathway blockade in the setting of adoptive cell therapy.

It remains unclear why some tumors (or tumor regions) express CD155 while others do not, even within the same patient. As previously mentioned, we found in a pan-cancer analysis that PVR expression was associated with a stem cell-like gene expression program [26], suggest-ing that CD155 may be a natural component of the immune-privileged stem cell niche. Others have shown that CD155 expression is regulated by pathways associated with genotoxic stress and oncogenic signaling,

such as ATM/ATR, SHH and RAS pathways [13,15,16]. In theory, such mechanisms could promote CD155 expression in HGSC, which exhibits among the highest levels of copy number variation of all cancers and a correspondingly high degree of genotoxic stress [37]. An improved un-derstanding of the mechanisms that drive CD155 and CD112 expression in HGSC and other cancers may lead to new strategies to enhance tumor-speciﬁc T cell responses.

The identification of immune checkpoints in cold tumor microenvi-ronments remains an important goal of immunotherapy research. We found that CD155 was commonly expressed in TIL-negative tumors across multiple HGSC cohorts and many other cancers profiled by TCGA. This relationship was evident across patients, across metastatic sites within patients, and at the micro-regional level within individual tumor deposits. In contrast, and consistent with prior reports [6], we ob-served a positive association between PD-L1 and TIL. Accordingly, in a pan-cancer TCGA dataset, PVR/CD155 and CD274/PD-L1 showed diver-gent associations with survival. Several mechanisms have been identi-fied that can foster cold tumor microenvironments, including lack of tumor antigens or antigen presentation; impaired lymphocyte traf fick-ing; elevated oncogenic signalfick-ing; and biochemical barriers to immune

Fig. 5. CD155 (PVR) and PD-L1 (CD274) show contrasting associations with patient outcome in HGSC and diverse cancers. A. Relationship between PVR and CD274 gene expression and patient survival across cancers in TCGA. Log hazard ratios ±95% CI shown for each cancer type. Log hazard ratios for PVR and CD274 are uncorrelated across cancers (rho = 0.04; PN 0.05). B. Relationship between PVR and CD274 gene expression and patient survival across publicly available HGSC datasets (identiﬁed by GSE accession). Log hazard ratios ±95% CIs for each data set shown, with point size proportional to sample size. Overall log hazard estimates are based onﬁxed effects meta-analysis of hazard ratios, estimated while controlling for patient debulking status.

(11)

cell infiltration [38]. CD155 expression on tumor endothelium and/or epithelium could play an active role in lymphocyte exclusion through TIGIT-mediated inhibition of T cell signaling. Moreover, CD155 has mul-tiple intrinsic signaling functions that promote tumor cell migration and proliferation [40] and, in theory, could also stimulate the formation of an immunologically hostile extracellular matrix, cytokine milieu, or meta-bolic microenvironment. Furthermore, CD155 is commonly over-expressed in a soluble isoform which is associated with increased tumor burden [40]; whether soluble CD155 can have paracrine or sys-temic inhibitory effects on TIGIT+ T cells and NK cells is currently un-known. Notably, the immunological effects of CD155 may also depend on context: for example, in a murine melanoma model, CD155 and PD-L1 were co-expressed on tumor cells and positively associated with T cell infiltration [39]. Thus, further work is required to define whether and how expression of CD155 promotes lymphocyte exclusion in the tumor microenvironment.

Ourfindings suggest several new avenues for immunotherapy re-search. First, our phenotypic analysis of patient-derived TIL in HGSC suggests TIGIT as a priority target for clinical trials of new immune checkpoint blockade strategies. Second, the increased expression of TIGIT, CD96, DNAM-1 and GITR on TIL following in vitro expansion sug-gests new opportunities to enhance the efficacy of adoptive cell thera-pies. Third, the elevated expression of CD155 on malignant epithelium in HGSC suggests this malignancy may be susceptible to therapeutic strategies targeting CD155, such as oncolytic poliovirus, which is show-ing promisshow-ing results in phase I trials against malignant glioma [10,40]. Finally, ourfinding that CD155/PVR is commonly expressed in TIL-negative tumors suggests that targeting of the CD155/TIGIT pathway might prove complementary to PD-1/PD-L1-directed approaches.

Supplementary data to this article can be found online athttps://doi. org/10.1016/j.ygyno.2020.04.689.

Declarations

Ethics approval and consent to participate

Patient specimens were accessed through BC Cancer's Tumor Tissue Repository (a member of the Canadian Tissue Repository Network). Specimens and clinical data were obtained with either written informed consent or a formal waiver of consent under protocols approved by the Research Ethics Board of the BC Cancer and the University of British Co-lumbia (H07-00463). All specimens were obtained from primary sur-geries, prior to any other treatment.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Funding

Funding was provided by the BC Cancer Foundation (B.H.N), Ca-nadian Cancer Society (B.H.N), Canada's Networks of Centres of Ex-cellence (BioCanRx FY16/CORE7; B.H.N and M.C.A.W), Canadian Institutes of Health Research (MOP 142436 and MFE 158087; B.H. N), Terry Fox Research Institute (TFRI#1060; B.H.N.), Cancer search Society (B.H.N.), Michael Smith Foundation for Health Re-search (16631; M.C.A.W), Carraresi Foundation OVCARE ReRe-search Grants supported by the VGH & UBC Hospital Foundation (P.T.H), and Canadian Institutes of Health Research Postdoctoral Fellowships (P.T.H. and M.C.A.W).

Credit authorship contribution statement

Julian Smazynski: Conceptualization, Investigation, Formal analysis, Writing - original draft, Methodology. Phineas T.

Hamilton: Conceptualization, Formal analysis, Writing - original draft, Methodology. Shelby Thornton: Investigation. Katy Milne: Inves-tigation, Methodology, Formal analysis, Writing - original draft. Maartje C.A. Wouters: Methodology, Formal analysis, Writing - original draft. John R. Webb: Methodology, Formal analysis, Writing - original draft. Brad H. Nelson: Methodology, Formal analysis, Writing - original draft. Declaration of competing interest

No potential conﬂicts of interest to disclose. Acknowledgments

We thank Bronwyn Gibson-Wright, Stacey Ledoux, Kerri Olsen, Alicia Parker, Daniel Kos, and Donald Stevens for optimization and workup of immunohistochemistry and immunoﬂuorescence staining (Molecular and Cell Immunology Core, Deeley Research Centre, BC Can-cer). We thank Dr. Allen Zhang (University of British Columbia) for spa-tial analysis scripts. We thank Megan Fuller (BC Cancer) and Dr. Peter Watson (BC Cancer) for helpful discussions.

References

[1] P. Darvin, S.M. Toor, V. Sasidharan Nair, E. Elkord, Immune checkpoint inhibitors: re-cent progress and potential biomarkers, Exp. Mol. Med. 50 (2018) 1–11,https://doi. org/10.1038/s12276-018-0191-1.

[2] J.S. O’Donnell, G.V. Long, R.A. Scolyer, M.W.L. Teng, M.J. Smyth, Resistance to PD1/ PDL1 checkpoint inhibition, Cancer Treat. Rev. 52 (2017) 71–81,https://doi.org/ 10.1016/J.CTRV.2016.11.007.

[3] S. Maleki Vareki, High and low mutational burden tumors versus immunologically hot and cold tumors and response to immune checkpoint inhibitors, J. Immunother. Cancer 6 (2018) 4–8,https://doi.org/10.1186/s40425-018-0479-7.

[4] S.L. Gaillard, A.A. Secord, B. Monk, The role of immune checkpoint inhibition in the treatment of ovarian cancer, Gynecol. Oncol. Res. Pract. 3 (2016) 11,https://doi.org/ 10.1186/s40661-016-0033-6.

[5] S.D. Martin, S.D. Brown, D.A. Wick, J.S. Nielsen, D.R. Kroeger, K. Twumasi-Boateng, R.A. Holt, B.H. Nelson, Low mutation burden in ovarian cancer may limit the utility of neoantigen-targeted vaccines, PLoS One 11 (2016), e0155189.https://doi.org/10. 1371/journal.pone.0155189.

[6] J.R. Webb, K. Milne, D.R. Kroeger, B.H. Nelson, PD-L1 expression is associated with tumor-inﬁltrating T cells and favorable prognosis in high-grade serous ovarian cancer, Gynecol. Oncol. 141 (2016) 293–302,https://doi.org/10.1016/J.YGYNO.2016.03.008. [7] D.R. Kroeger, K. Milne, B.H. Nelson, Tumor-inﬁltrating plasma cells are associated

with tertiary lymphoid structures, cytolytic T-cell responses, and superior prognosis in ovarian cancer, Clin. Cancer Res. 22 (2016) 3005–3015,https://doi.org/10.1158/ 1078-0432.CCR-15-2762.

[8] A.W. Zhang, A. McPherson, K. Milne, D.R. Kroeger, P.T. Hamilton, A. Miranda, T. Funnell, N. Little, C.P.E. de Souza, S. Laan, S. LeDoux, D.R. Cochrane, J.L.P. Lim, W. Yang, A. Roth, M.A. Smith, J. Ho, K. Tse, T. Zeng, I. Shlafman, M.R. Mayo, R. Moore, H. Failmezger, A. Heindl, Y.K. Wang, A. Bashashati, D.S. Grewal, S.D. Brown, D. Lai, A.N.C. Wan, C.B. Nielsen, C. Huebner, B. Tessier-Cloutier, M.S. Anglesio, A. Bouchard-Côté, Y. Yuan, W.W. Wasserman, C.B. Gilks, A.N. Karnezis, S. Aparicio, J.N. McAlpine, D.G. Huntsman, R.A. Holt, B.H. Nelson, S.P. Shah, Interfaces of malig-nant and immunologic clonal dynamics in ovarian cancer, Cell (2018)https://doi. org/10.1016/J.CELL.2018.03.073.

[9] W.-T. Hwang, S.F. Adams, E. Tahirovic, I.S. Hagemann, G. Coukos, Prognostic signifi-cance of tumor-infiltrating T cells in ovarian signifi-cancer: a meta-analysis, Gynecol. Oncol. 124 (2012) 192–198,https://doi.org/10.1016/J.YGYNO.2011.09.039.

[10] P. Kučan Brlić, T. Lenac Roviš, G. Cinamon, P. Tsukerman, O. Mandelboim, S. Jonjić, Targeting PVR (CD155) and its receptors in anti-tumor therapy, Cell. Mol. Immunol. 16 (2019) 51–63,https://doi.org/10.1038/s41423-018-0168-y.

[11] R.J. Johnston, L. Comps-Agrar, J. Hackney, X. Yu, M. Huseni, Y. Yang, S. Park, V. Javinal, H. Chiu, B. Irving, D.L. Eaton, J.L. Grogan, The immunoreceptor TIGIT regulates anti-tumor and antiviral CD8+ T cell effector function, Cancer Cell 26 (2014) 923–937,

https://doi.org/10.1016/J.CCELL.2014.10.018.

[12] Y. Takai, J. Miyoshi, W. Ikeda, H. Ogita, Nectins and nectin-like molecules: roles in contact inhibition of cell movement and proliferation, Nat. Rev. Mol. Cell Biol. 9 (2008) 603–615,https://doi.org/10.1038/nrm2457.

[13] T. Hirota, K. Irie, R. Okamoto, W. Ikeda, Y. Takai, Transcriptional activation of the mouse Necl-5/Tage 4/PVR/CD155 gene byﬁbroblast growth factor or oncogenic Ras through the Raf–MEK–ERK–AP-1 pathway, Oncogene 24 (2005) 2229–2235,

https://doi.org/10.1038/sj.onc.1208409.

[14] K.E. Sloan, B.K. Eustace, J.K. Stewart, C. Zehetmeier, C. Torella, M. Simeone, J.E. Roy, C. Unger, D.N. Louis, L.L. Ilag, D.G. Jay, CD155/PVR plays a key role in cell motility dur-ing tumor cell invasion and migration, BMC Cancer 4 (2004) 73,https://doi.org/10. 1186/1471-2407-4-73.

[15] D.J. Solecki, M. Gromeier, S. Mueller, G. Bernhardt, E. Wimmer, Expression of the human poliovirus receptor/CD155 gene is activated by sonic hedgehog, J. Biol. Chem. 277 (2002) 25697–25702,https://doi.org/10.1074/jbc.M201378200.

(12)

[16] M. Cippitelli, A. Santoni, M.L. Iannitto, V. Di Gialleonardo, A. Guarini, M.T. Petrucci, M.R. Ricciardi, A. Zingoni, C. Fionda, R. Foa, A. Soriani, C. Cerboni, ATM-ATR-depen-dent up-regulation of DNAM-1 and NKG2D ligands on multiple myeloma cells by therapeutic agents results in enhanced NK-cell susceptibility and is associated with a senescent phenotype, Blood 113 (2008) 3503–3511,https://doi.org/10. 1182/blood-2008-08-173914.

[17] S. Nishiwada, M. Sho, S. Yasuda, K. Shimada, I. Yamato, T. Akahori, S. Kinoshita, M. Nagai, N. Konishi, Y. Nakajima, Clinical signiﬁcance of CD155 expression in human pancreatic cancer, Anticancer Res. 35 (2015) 2287–2297http://www.ncbi.nlm.nih. gov/pubmed/25862891. (Accessed 28 May 2019).

[18] B. Zhuo, Y. Li, F. Gu, Z. Li, Q. Sun, Y. Shi, Y. Shen, F. Zhang, R. Wang, X. Wang, Over-expression of CD155 relates to metastasis and invasion in osteosarcoma, Oncol. Lett. 15 (2018) 7312–7318,https://doi.org/10.3892/ol.2018.8228.

[19] Y. Xu, G. Cui, Z. Jiang, N. Li, X. Zhang, Survival analysis with regard to PD-L1 and CD155 expression in human small cell lung cancer and a comparison with associated receptors, Oncol. Lett. 17 (2019) 2960–2968,https://doi.org/10.3892/ol.2019.9910. [20] H. Yong, R. Cheng, X. Li, G. Gao, X. Jiang, H. Cheng, X. Zhou, W. Zhao, CD155 expres-sion and its prognostic value in postoperative patients with breast cancer, Biomed. Pharmacother. 115 (2019) 108884,https://doi.org/10.1016/J.BIOPHA.2019.108884. [21] C.J. Chan, D.M. Andrews, N.M. McLaughlin, H. Yagita, S. Gilﬁllan, M. Colonna, M.J. Smyth, DNAM-1/CD155 interactions promote cytokine and NK cell-mediated sup-pression of poorly immunogenic melanoma metastases, J. Immunol. 184 (2010) 902–911,https://doi.org/10.4049/jimmunol.0903225.

[22] W.C. Dougall, S. Kurtulus, M.J. Smyth, A.C. Anderson, TIGIT and CD96: new check-point receptor targets for cancer immunotherapy, Immunol. Rev. 276 (2017) 112–120,https://doi.org/10.1111/imr.12518.

[23] J.-M. Chauvin, O. Pagliano, J. Fourcade, Z. Sun, H. Wang, C. Sander, J.M. Kirkwood, T.T. Chen, M. Maurer, A.J. Korman, H.M. Zarour, TIGIT and PD-1 impair tumor antigen– speciﬁc CD8 + T cells in melanoma patients, J. Clin. Invest. 125 (2015) 1–13,https:// doi.org/10.1172/JCI80445DS1.

[24] K.E. Pauken, E.J. Wherry, TIGIT and CD226: tipping the balance between costimula-tory and coinhibicostimula-tory molecules to augment the cancer immunotherapy toolkit, Cancer Cell 26 (2014) 785–787,https://doi.org/10.1016/J.CCELL.2014.11.016. [25] D. Mittal, A. Lepletier, J. Madore, A.R. Aguilera, K. Stannard, S.J. Blake, V.L.J. Whitehall,

C. Liu, M.L. Bettington, K. Takeda, G.V. Long, R.A. Scolyer, R. Lan, N. Siemers, A. Korman, M.W.L. Teng, R.J. Johnston, W.C. Dougall, M.J. Smyth, CD96 is an immune checkpoint that regulates CD8+ T-cell antitumor function, Cancer Immunol. Res. 7 (2019) 559–571,https://doi.org/10.1158/2326-6066.CIR-18-0637.

[26] A. Miranda, P.T. Hamilton, A.W. Zhang, S. Pattnaik, E. Becht, A. Mezheyeuski, J. Bruun, P. Micke, A. de Reynies, B.H. Nelson, Cancer stemness, intratumoral hetero-geneity, and immune response across cancers, Proc. Natl. Acad. Sci. 116 (2019) 9020–9029,https://doi.org/10.1073/PNAS.1818210116.

[27] B.F. Ganzfried, M. Riester, B. Haibe-Kains, T. Risch, S. Tyekucheva, I. Jazic, X.V. Wang, M. Ahmadifar, M.J. Birrer, G. Parmigiani, C. Huttenhower, L. Waldron, curated OvarianData: clinically annotated data for the ovarian cancer transcriptome, Data-base 2013 (2013)https://doi.org/10.1093/DATABASE/BAT013.

[28] W. Viechtbauer, Conducting meta-analyses in R with the metafor package, J. Stat. Softw. 36 (2010) 1–48,https://doi.org/10.18637/jss.v036.i03.

[29] P. Danaher, S. Warren, L. Dennis, L. D’Amico, A. White, M.L. Disis, M.A. Geller, K. Odunsi, J. Beechem, S.P. Fling, Gene expression markers of tumor inﬁltrating

leukocytes, J. Immunother. Cancer 5 (2017) 18, https://doi.org/10.1186/s40425-017-0215-8.

[30] N.K. Escalante, A. von Rossum, M. Lee, J.C. Choy, CD155 on human vascular endothe-lial cells attenuates the acquisition of effector functions in CD8 T cells, Arterioscler. Thromb. Vasc. Biol. 31 (2011) 1177–1184,https://doi.org/10.1161/ATVBAHA.111. 224162.

[31] F. Castro, A.P. Cardoso, R.M. Gonçalves, K. Serre, M.J. Oliveira, Interferon-gamma at the crossroads of tumor immune surveillance or evasion, Front. Immunol. 9 (2018) 847,https://doi.org/10.3389/ﬁmmu.2018.00847.

[32] S. Nawaz, A. Heindl, K. Koelble, Y. Yuan, Erratum: beyond immune density: critical role of spatial heterogeneity in estrogen receptor-negative breast cancer, Mod. Pathol. 28 (2015) 1621,https://doi.org/10.1038/modpathol.2015.133.

[33] A. Getis, J.K. Ord, The analysis of spatial association by use of distance statistics, Geogr. Anal. 24 (2010) 189–206, https://doi.org/10.1111/j.1538-4632.1992. tb00261.x.

[34] B. Wang, W. Zhang, V. Jankovic, J. Golubov, P. Poon, E.M. Oswald, C. Gurer, J. Wei, I. Ramos, Q. Wu, J. Waite, M. Ni, C. Adler, Y. Wei, L. Macdonald, T. Rowlands, S. Brydges, J. Siao, W. Poueymirou, D. MacDonald, G.D. Yancopoulos, M.A. Sleeman, A.J. Murphy, D. Skokos, Combination cancer immunotherapy targeting PD-1 and GITR can rescue CD8 + T cell dysfunction and maintain memory phenotype, Sci. Immunol. 3 (2018), eaat7061.https://doi.org/10.1126/sciimmunol.aat7061. [35] H. Harjunpää, S.J. Blake, E. Ahern, S. Allen, J. Liu, J. Yan, V. Lutzky, K. Takeda, A.R.

Aguilera, C. Guillerey, D. Mittal, X.Y. Li, W.C. Dougall, M.J. Smyth, M.W.L.L. Teng, De-ﬁciency of host CD96 and PD-1 or TIGIT enhances tumor immunity without signif-icantly compromising immune homeostasis, Oncoimmunology 7 (2018) 1–11,

https://doi.org/10.1080/2162402X.2018.1445949.

[36] L. Lu, X. Xu, B. Zhang, R. Zhang, H. Ji, X. Wang, Combined PD-1 blockade and GITR triggering induce a potent antitumor immunity in murine cancer models and synergizes with chemotherapeutic drugs, J. Transl. Med. 12 (2014) 36,https://doi. org/10.1186/1479-5876-12-36.

[37] D.D. Bowtell, S. Böhm, A.A. Ahmed, P.-J. Aspuria, R.C. Bast, V. Beral, J.S. Berek, M.J. Birrer, S. Blagden, M.A. Bookman, J.D. Brenton, K.B. Chiappinelli, F.C. Martins, G. Coukos, R. Drapkin, R. Edmondson, C. Fotopoulou, H. Gabra, J. Galon, C. Gourley, V. Heong, D.G. Huntsman, M. Iwanicki, B.Y. Karlan, A. Kaye, E. Lengyel, D.A. Levine, K.H. Lu, I.A. McNeish, U. Menon, S.A. Narod, B.H. Nelson, K.P. Nephew, P. Pharoah, D.J. Powell, P. Ramos, I.L. Romero, C.L. Scott, A.K. Sood, E.A. Stronach, F.R. Balkwill, Rethinking ovarian cancer II: reducing mortality from high-grade serous ovarian cancer, Nat. Rev. Cancer 15 (2015) 668–679,https://doi.org/10.1038/nrc4019.

[38] T.F. Gajewski, The next hurdle in cancer immunotherapy: overcoming the non–T-cell–inﬂamed tumor microenvironment, Semin. Oncol. 42 (2015) 663–671,

https://doi.org/10.1053/J.SEMINONCOL.2015.05.011.

[39] X.-Y. Li, I. Das, A. Lepletier, V. Addala, T. Bald, K. Stannard, D. Barkauskas, J. Liu, A.R. Aguilera, K. Takeda, M. Braun, K. Nakamura, S. Jacquelin, S.W. Lane, M.W.L. Teng, W.C. Dougall, M.J. Smyth, CD155 loss enhances tumor suppression via combined host and tumor-intrinsic mechanisms, J. Clin. Invest. 128 (2018) 2613–2625,

https://doi.org/10.1172/JCI98769.

[40] J.S. O’Donnell, J. Madore, X.Y. Li, M.J. Smyth, Tumor intrinsic and extrinsic immune functions of CD155, Semin. Cancer Biol. (2020)https://doi.org/10.1016/j.semcancer. 2019.11.013.