The Inhibitory Signaling Receptor Protocadherin-18 Regulates Tumor-Infiltrating CD8⁺ T-cell Function

Effective CD8⁺ antitumor T-cell responses require several stages of development and activity: differentiation of antigen-specific cells, homing and entrance into tumor tissue, recognition of MHC-restricted cognate peptide on tumor cells, and cytolytic function. Each of these functional stages can be deficient or defective in cancer, but the presence of antigen-specific tumor-infiltrating lymphocytes (TIL) in many types of tumors argues that antitumor T cells develop and home (1). However, the tumor microenvironment is robustly immunosuppressive with potential inhibitory contributions from many cells in the tumor: B cells, cancer-associated fibroblasts, mast cells, myeloid-derived cells, pericytes, FoxP3⁺ regulatory T cells, or vascular endothelia (2). Reversing inhibition in endogenous TILs (and in adoptively transferred antitumor T cells) is now recognized as a major therapeutic objective (3).

T-cell activation results after T-cell receptor (TCR) interaction with cognate antigen. This generates a positive signal that is integrated with negative signals resulting from the trans interaction of T-cell surface inhibitory signaling receptors (ISR) with their cognate ligands. ISRs, also known as coinhibitory receptors or cell surface immune checkpoint molecules, are widely expressed and regulate T-cell activation, differentiation, homeostasis, and effector phase functions. In conditions of chronic antigen exposure and inflammation, hyporesponsive T cells develop, which characterize latent infection or tumor growth (4). Two ISRs, CTLA-4 (CD152) and PD-1 (CD274), have emerged as major inhibitors of antitumor T-cell function. mAb-based inhibition of CTLA-4 or PD-1 activation (targeting either PD-1 or its ligand PD-L1) elicits potent antitumor activity in various models and clinical trials, especially in treatment of human tumors having strong inherent antigenicity (5). Although clinical results have been dramatic (6), successes are limited to subsets of patients having certain tumor types, suggesting that additional ISRs may be involved in the functional regulation of TILs in human cancer (7). Supporting this notion, various other T-cell ISRs (8), such as BTLA, LAG-3, TIGIT, TIM-3 or VISTA, have different cellular patterns of expression and are being targeted in experimental mAb-based blockade mono- and combination therapy trials (9).

Analysis of multiple ISRs that are simultaneously expressed on hyporesponsive T cells shows that any individual ISR expressed can regulate cell function in vitro, implying that, collectively, the repertoire of expressed ISR may function in an additive or synergistic manner, especially in vivo (10). TILs can be nonlytic in vivo (1, 11, 12) and express a variety of different ISRs (8) that restrict TIL functionality, leading to tumor escape from immune killing. Because ISR expression in TILs is intrinsic to T-cell differentiation and activation status (13–15), the availability of ISR ligands on cells within the tumor determines whether a given ISR is engaged and thus contributes to TIL effector phase dysfunction (2). Multiple cell types within the tumor microenvironment can potentially express ISR ligands, including vascular endothelia, tumor cells, myeloid components of the stroma (DCs, macrophages, and myeloid cells), and TILs themselves, making regulation of TIL function by ISRs complex.

A role for protocadherin-18 (pcdh18) in T-cell function was identified through study of the nonlytic phenotype of CD8⁺ TILs in a murine model of colon cancer, in which it was discovered to interact with p56^lck, thereby blocking proximal TCR signaling and cytolysis (13, 16). Proximal TCR-mediated signaling (calcium flux) in purified TILs is intact if assessed by activation in vitro with anti-CD3 (17), and TIL lytic function is restored upon purification and brief culture in vitro (16), properties suggestive of ISR activity. As opposed to other ISRs, pcdh18 is expressed in activated CD8⁺ memory T cells (CD8⁺CD44⁺CD62L⁺CD127^hi), is not expressed in B cells, NK cells, naïve CD8⁺ T cells, or primary CD8⁺ effector cells, and the kinetics of its transcriptional regulation upon TCR ligation are that of an immediate early response gene (13). The only known ligand for pcdh18 is itself, and it mediates homophilic binding (18). Here, we show in a murine cancer model that pcdh18 is a key regulator of antitumor CD8⁺ T-cell effector-phase function.

Wild-type C57BL/6 male mice were from Taconic. The pcdh18 gene-deleted mouse (Pcdh18^{tm1(KOMP)Vlcg}) was obtained from the Knockout Mouse Project (KOMP) Repository (# 14494, via Regeneron Pharmaceuticals).

Doxycycline, puromycin, protein G-agarose, and polymyxin B-agarose were from Sigma-Aldrich. Magnetic immunobeads for isolation of human (130-096-495) or murine (130-049-401) T cells were from Miltenyi Biotec. Primary antibodies for flow cytometry were from eBioscience and secondary reagents (Alexa 647-, HRP-anti-rabbit Ig) from Jackson Immuno-Research (111-625-144 and 111-035-144). Anti-TCR H57-597 was purified from hybridoma-conditioned medium using protein G-agarose. Rabbit Ab reactive to the cytoplasmic domain of pcdh18 was produced using a Gst-fusion protein as immunogen as described previously (13). Rabbit and mouse Abs reactive to the extracellular domain of pcdh18 (amino acids 37, Glu- 690, Ser) were produced using a His-tag protein expressed in CHO cells as immunogen. The rabbit Ab was purified on Protein G-agarose and the murine serum used as serum in passive transfer (see below). SIINFEKL-tetramer was from MBL International. Anti-CD8 (clone 53.6.7) used for in vivo depletion was purified from hybridoma-conditioned medium using Protein G-agarose and was absorbed on Polymyxin B-agarose before dialysis versus PBS. Anti-PD-1 Ab 29F.1A12 (a gift from G. Freeman, Dana-Farber Cancer Institute, Boston, MA) or control Rat IgG2a (Bio X Cell) were similarly treated for potential LPS contamination.

MCA38 cells (obtained from N. Restifo, NIH, Bethesda, MD, circa 1990) and RMA-S cells (from M. Bevan, University of Washington, Seattle, WA) were routinely tested for mycoplasma contamination (MP0025-1KT, Sigma-Aldrich). The cell lines have not been authenticated by our laboratory and were cultured for <10 passages before new stocks were thawed. Listeria monocytogenes-^ova (L. monocytogenes-^OVA), a gift from Eric Pamer (Memorial Sloan Kettering Cancer Center, New York, NY), was prepared and used as described previously (13). Human Leukopaks were obtained from the New York Blood Center (Queens, NY).

The pTRIPZ vector (Dharmacon/GE Healthcare Life Sciences) was used to express candidate pcdh18 shRNAs. Candidate pcdh18 targets were: ATGTCCTGGCTAAGAATCTGAA = “a,” CACCAAGCCTCTCCTCAGTGAG = “b,” CGCCACTCCTGCTGTTGAGGTC = “c,” and scrambled control GACTAGTCTTACGATACATGCA. Recombinant vectors were sequenced to confirm insert sequences. Virus was produced in HEK293 cells, concentrated by centrifugation, and used for infection in the presence of Polybrene. Vectors containing TurboGFP (and separately lacking GFP) were produced and were titered on HEK293 cells. Knockdown in MCA38 cells used an MOI of 1.

MCA38 cells were fixed and permeabilized (eBioscience, 88-8824) before labeling with rabbit anti-pcdh18, or control rabbit Ig, which was detected using PE-conjugated goat-anti-rabbit IgG. Human CD8⁺ memory T cells were isolated after Ficoll purification of peripheral blood mononuclear cells (PBMC), followed by FACS isolation using antibodies to CD8, CD44, CD27, and CD45RO (eBioscience). TILs were isolated by magnetic immunobeading (Miltenyi Biotec).

Wild-type MCA38 cells or MCA38 cells transduced with either virus encoding shRNA “pcdh18b” or control virus, both lacking GFP, were injected intraperitoneally or subcutaneously into 6- to 8-week-old male mice and observed for 15 weeks or until sacrifice was required. Mice received doxycycline in drinking water (50 μg/mL) as indicated. Subcutaneous tumors were measured with calipers and volume was calculated as: [(W² × L)/2]. P values for group comparisons of tumor growth were calculated using the two-tailed nonparametric Mann–Whitney (GraphPad Prism 5.0).

Five or 10 days after seeding of tumor as indicated, mice received intraperitoneal injections of 0.2 mg of purified anti–PD-1 Ab 29F.1A12 or control Rat IgG2a (in PBS), or 0.05 mL mouse anti-pcdh18 sera (or control sera) twice per week. MCA38 tumors were initiated subcutaneously, and on day 10, mice received control Ig, anti-PD-1, anti-pcdh18, or both. Data are representative of two independent experiments (n = 5/group). For the experiment shown in Fig. 4, comparison of average tumor size at 4 weeks of growth for treatment starting at day 5 is P < 0.001 (wild-type average size was 0.78 cm³ and pcdh18^–/– is 0.21 cm³). Murine anti-pcdh18 serum was prepared by immunization of pcdh18^–/– mice with recombinant pcdh18 extracellular domain (amino acids 39-690). Immunized mice were bled and pooled sera were used for passive transfer.

MCA38 cells (2.5 × 10⁴) infected with “pcdh18b” or control lentivirus were plated in triplicate in 48-well plates in the presence of doxycycline (0.001 mg/mL) and were enumerated after Trypan blue staining.

TILs were prepared from either 5 (for use as effector cells in in vitro killing assays) or 10 to 15 pooled tumors [for RNA sequencing (RNA-seq) analyses] per experiment by magnetic immunobeading as described previously (13, 16). For Nanostring and RNA-seq analyses, after enrichment by magnetic immunobeading, TILs were stained with anti-CD8α and anti-TCRβ and purified by FACS before purification of RNA. For assay of IFNγ production, 10 tumors were pooled to purify TILs. Quadruplicate wells (2 × 10⁵ cells) were stimulated in vitro (for 36 hours using plate-bound anti-TCRβ) before assay of supernatants by ELISA (eBioscience, # 88-8314-88, minimum sensitivity = 0.7 pg/mL).

Target cell killing was assessed (using freshly isolated TILs or TILs that were cultured in complete RPMI1640 medium overnight) by MTT assay (Sigma-Aldrich) in quadruplicate wells for each E:T ratio. Targets were either MCA38 tumor or RMA-S cells (pulsed with SIINFEKL or control K^b binding peptide for 2 hours at 26°C). For determination of lytic efficiency, maximal target cell lysis was calculated after treating target cells with 1% Triton X-100 and used the formula: % cytotoxicity = (experimental − spontaneous release)/(maximal release − spontaneous release) × 100.

Wild-type mice were reconstituted with bone marrow from congenic pcdh18^–/– mice (Thy1.2) and MCA38 tumors developed. CD8⁺ TILs were combined from 5 individual mice, then FACS-purified (by Thy1 expression) before RNA isolation and analysis by NanoString PanCancer Immune Profiling Panel (NanoString Technologies) performed by the NYU School of Medicine's Genome Technology Center (New York, NY).

MCA38 TILs were purified (combining 10 pooled tumors grown in wild-type mice) by magnetic immunobeading and FACS, RNA was prepared, converted to cDNA, and used and analyzed by gene array (13) or RNA-seq. RNA-seq was performed on two independent biological replicates of TILs (n = 6 and 10 pooled tumors each) isolated by magnetic immunobeading followed by FACS.

Anonymized normal lymph node samples were deparafinized and reacted with rabbit anti–pcdh-18 (1:3,000) that was detected with HRP-conjugated donkey anti-rabbit and amplified with biotin/tyramide (Sigma-Aldrich) and Alexa 594 streptavidin essentially as described previously (19). Samples were also labeled with mouse anti-CD8 (1:300) that was detected with donkey anti-mouse Alexa 488. Two individual lymph nodes were analyzed and 5 microscopic fields were counted for each sample, which were scored by a blinded observer.

PBMCs were prepared by Ficoll gradient from a single donor (1.21 × 10⁹ total cells) and used to isolate CD8⁺ T cells by magnetic immunobeading. Total CD8⁺ cells were isolated by negative selection (“untouched”), then CD45RA⁺ cells by positive selection (naïve). CD45RO⁺ cells (memory) were then used to isolate Cm by positive selection (CD27⁺) and Em (CD27^–, “untouched”). Cells were plated in triplicate in round-bottom wells using 96-well plates (22 × 10³ cells/well) with or without blocking or control Ab (10 μg/mL) as indicated.

Pcdh18 is expressed in dysfunctional MCA38 TILs, within which it interacts with p56^lck. It is postulated to play a functional role in the lack of cytolytic activity of MCA38-infiltrating T cells recovered from MCA38 tumor-bearing mice (13). The transiently blocked proximal TCR-mediated signaling in these TILs is coincident with the tumor-induced defect in cytotoxicity (20). Because pcdh18 binds to itself in a trans fashion (19), and the inhibition of TIL lytic function by contact with cognate tumor cells implies a reversible activation switch reminiscent of ISR interactions with their cognate ligands (16), we examined the activation of TIL pcdh18 when cognate tumor cells had expression of pcdh18 knocked down (21). We tested three shRNAs for their abilities to downregulate pcdh18 expression in the cognate tumor cells that were used as targets for in vitro cytolysis assays (Fig. 1). Expression of target sequence “b” in MCA38 reduced pcdh18 mRNA 94% compared with controls (Fig. 1A), and pcdh18 protein was undetectable in those tumor cells (Fig. 1B). As has been observed in almost every tumor model (2, 22), freshly isolated (“nonlytic”) TILs lacked lytic function against MCA38 that express pcdh18 (Fig. 1C, solid blue tracing), but after purification and brief culture of the TILs, cytolysis was restored (solid red tracing). In contrast, cognate MCA38 with pcdh18 knocked down due to expression of shRNA were efficiently killed, even by freshly isolated TILs (broken blue tracing). The robust lytic function of “fresh” TILs was more pronounced than for TILs after brief in vitro culture (“recovery”), possibly because many TILs are effete immediately after the in vitro “recovery” period, as suggested by higher Annexin-V binding compared with freshly isolated TILs (23). The expression of pcdh18 on tumor cells was thus necessary to engage pcdh18 on TILs and initiate its inhibitory function.

In the L. monocytogenes-^OVA infection model, antigen-specific CD8⁺ T cells accumulate in nonlymphoid tissue long after the infection is cleared, and these cells resemble TILs in terms of cell surface expression of memory markers (24). However, tissue-resident anti–L. monocytogenes^OVA T cells are cytolytic immediately upon isolation (as assessed by lytic activity toward OVA-expressing or SIINFEKL-pulsed RMA-S cells; refs. 24, 25), whereas TILs are not (17). Thus, we asked whether the tumor microenvironment might induce defective effector-phase function in anti–L. monocytogenes T cells that infiltrate tumor tissue. Mice were infected with L. monocytogenes-^OVA at different times relative to MCA38 tumor seeding (10 days prior, coincident, or 10 days after) and the lytic function of CD8⁺ T cells isolated from tumor tissue after 20 days of growth was determined (Fig. 1D). The time of infection relative to tumor seeding did not affect the anti–L. monocytogenes-^OVA response. CD8⁺ T cells had no lytic activity toward cognate MCA38 tumor cells (red tracings). The same population of T cells, however, had efficient lytic activity toward RMA-S cells pulsed with SIINFEKL (solid black tracings), but not target RMA-S cells pulsed with control peptide.This experiments shows that residence in tumor tissue is not sufficient to induce either pcdh18 expression or lytic dysfunction. We found that mice infected with L. monocytogenes-^OVA developed antigen-specific T cells that both expressed pcdh18 protein and were maintained more than 1 year after infection (Supplementary Fig. S1).

Knockdown of pcdh18 in tumor cells did not affect either the in vitro growth of MCA38 cells (Fig. 2A) or the incidence or growth rate of tumors in CD8^–/– mice (Fig. 2B). However, tumor growth in wild-type mice was significantly delayed (Fig. 2C, P < 0.001), indicating that antitumor growth control mediated by T cells was enhanced in the absence of pcdh18 in the tumor. To evaluate the effects on tumor growth of T cells that lack pcdh18, MCA38 tumors with normal expression of pcdh18 were grown in pcdh18^–/– mice (Fig. 2D). Similar to delay in tumor growth of tumor cells lacking pcdh18 in wild-type mice (Fig. 2C), MCA38 tumor growth in pcdh18^–/– mice was significantly delayed compared with wild-type mice, implying that TILs in pcdh18^–/– mice have enhanced effector-phase function.

Why tumor development in pcdh18^–/– mice was only delayed and not obviated was investigated by analysis of gene expression in TILs from pcdh18^–/– mice. Bone marrow from wild-type and pcdh18^–/– mice was used to generate bone marrow chimeras in Thy1 congenic mice. Wild-type and pcdh18^–/– MCA38 TILs isolated from the same tumors were analyzed by Nanostring Immune Cell profiling containing 562 targets (Fig. 3). The data are displayed as the ratio of expression of a given gene in pcdh18^–/– TILs compared with wild-type TILs and showed pcdh18^–/– TILs had significant upregulation of many immune function genes, including those involved in T-cell activation and signaling (e.g., CD2, CD3, CD8, CD122, CD278, p56^lck, Zap70, TCRζ), the effector phase (GrzA and B, Prf), and chemokines and cytokine receptors (Cxcr6, Ccr7 and CD122, CD212, CD218, IL27Rα). In addition, pcdh18^–/– TILs robustly expressed genes associated with negative functional regulation (PD-1, NKG2a, CTLA-4, CD94, TIGIT, Sh2d1, CD5), most of which were previously identified by gene array as expressed in nonlytic wild-type TILs (Supplementary Fig. S2).

Because expression of PD-1 is both highly expressed in wild-type TILs (13) and significantly upregulated in pcdh18^–/– TILs, analysis of the major PD-1 ligand (PD-L1) in MCA38 tumor cells was assessed by flow cytometry (Fig. 4A, top). Cultured MCA38 cells uniformly expressed little PD-L1, but exposure to IFNγ induced expression. In vivo, more host stromal cells expressed PD-L1 compared with tumor cells early in tumor growth, but at later times, when the percentage of tumor cells within the tumor tissue increased, tumor cell expression became dominant (Fig. 4A, bottom). PD-L1 expression following inflammation in MCA38 tumor has been observed (26).

The finding that primary MCA38 tumors upregulate PD-L1 expression as a function of time of growth prompted analysis of tumor growth in mice treated with PD-1 blockade (Fig. 4B). Ab-mediated PD-1 blockade in wild-type mice significantly delayed growth of early-stage tumors (P < 0.001 on day 5, Fig. 4B, top, solid lines), an effect that was diminished if treatment was initiated after 10 days of tumor growth (dashed lines). Anti–PD-1 blockade in pcdh18^–/– mice inhibited tumor growth, even if treatment was delayed, and this effect was more pronounced than in wild-type mice (Fig. 4B, bottom). The average time to reach half-maximal tumor size (∼0.6 cm³) in wild-type mice treated with anti–PD-1 was 3 to 4 weeks but in pcdh18^–/– mice was 5 to 6 weeks. A similar effect was noted in mice deleted for the LAG-3 ISR (27).

An essential role for CD8⁺ T cells in mediating the delay of tumor growth following anti–PD-1 blockade was assessed by depleting T cells in both wild-type and pcdh18^–/– mice (Fig. 4C). Tumor volume was measured at 25 days of growth and showed both an essential role for CD8⁺ T cells and an additive benefit of PD-1 blockade in pcdh18^–/– mice compared with PD-1 blockade in wild-type mice (average tumor volume ∼0.1 cm³ versus ∼0.28 cm³, respectively).

The phenotype of TILs isolated from MCA38 tumors grown in wild-type and pcdh18^–/– mice was assessed by determination of IFNγ production upon isolation (at 26 days of growth) and activation in vitro (Fig. 4D). Following PD-1 blockade, TILs from wild-type mice produced IFNγ immediately after isolation and pcdh18^–/– TILs produced slightly more IFNγ. The highest IFNγ secretion was seen in TILs from pcdh18^–/– mice that were also treated with anti–PD-1, an observation that is in keeping with those mice having the smallest tumors (Fig. 4C).

Ab-mediated PD-1 blockade in wild-type mice delayed tumor growth (Fig. 4B, top), an effect that was enhanced in pcdh18^–/– mice (Fig. 4B, bottom), which we interpret to mean that activity of both PD-1 and pcdh18 restrains CD8⁺ TIL function. To test the effect of simultaneous blockade of PD-1 and pcdh18 on tumor growth, murine anti-pcdh18 serum reactive with recombinant extracellular domain of pcdh18 was developed and was tested for effect on tumor growth by passive transfer of sera (Supplementary Figs. S4 and S5). Control mice developed tumors with characteristic kinetics (detectable at 6–7 days), treatment with anti-pcdh18 slightly but consistently delayed growth, and anti–PD-1 was more effective at tumor inhibition than anti-pcdh18. Control mice developed tumors with characteristic kinetics (detectable at 6–7 days), treatment with anti-pcdh18 slightly but consistently delayed growth, and anti–PD-1 was more effective at tumor inhibition than anti-pcdh18. Consistent with the observations that anti–PD-1 treatment is more effective in pcdh18^–/– mice compared with wild-type mice (Fig. 4B), combined antibody therapy was dramatically more effective than either monotherapy.

The expression of pcdh18 in human T cells was examined by immunocytochemistry analysis of human lymph nodes (Fig. 5A). Regions containing CD8⁺ or CD4⁺ T cells showed less abundant, but coincident, pcdh18 staining in a subset of T cells, approximately 5% of each single positive cell type compared with single staining of CD8 or CD4 cells. We further analyzed pcdh18 expression in freshly isolated CD8⁺ memory T cells FACS-purified from normal donor PBMCs (Fig. 5B), wherein effector memory CD8⁺ cells (CD44⁺CD45RO⁺CD27^–) were shown to contain pcdh18 protein. Expression of pcdh18 in the Em₁ (CD44⁺CD27^–CD45RO^–) subpopulation of CD8⁺ memory cells was especially strong, given that the immunoblot signal was obtained from only 1.4 × 10⁵ cells, whereas it is a minor fraction of total CD8⁺ T cells in the peripheral blood. These analyses corroborated our previous findings in mice that show CD8⁺ effector memory T cells, but not naïve cells, express pcdh18 (13) and extend those findings, in that pcdh18 was also found in memory CD4⁺ T cells. Similar to mouse (13), human CD8⁺ human central memory T cells (CD27⁺, “Cm”) expressed pcdh18 protein after in vitro activation coincident with conversion to CD27^– effector memory cells (Supplementary Fig. S6).

The role of pcdh18 in effector-phase functions of human memory T cells was examined (Fig. 5C). Similar to the phenotype of TILs (23) and primary murine lytic effector cells transfected to express pcdh18 (13), in vitro activation of purified pcdh18⁺ effector memory CD8⁺ T cells resulted in AICD, but inclusion of anti-pcdh18 enhanced viability and cell recovery and had a slightly additive effect in conjunction with blocking anti–PD-1. Checkpoint inhibitor blockade also enhanced effector-phase function of activated primary human CD8⁺CD27^– memory T cells in that IFNγ secretion was dramatically increased (Fig. 5D).

Activation of T cells results from integration of a positive signal, generated by TCR recognition of cognate antigen, with a negative signal, generated by interaction of ISR with cognate ligands that are expressed on the antigen-expressing cell with which the T cell interacts (15). The relative balance of signals from these two opposing systems results from the number of receptors interacting with their cognate ligands and influences the T-cell activation threshold, thus determining the functional outcome. The lytic dysfunction of the effector phase that typifies CD8⁺ TILs (1) is mediated by intrinsic expression of the ISRs characteristic of effector memory cells (13). Engagement by the ISR ligands expressed in the tumor environment results in abrogation of proximal TCR signaling. Analysis of the basis for defective TIL lytic function in the murine adenocarcinoma MCA38 model showed that the lytic defect is manifested by a failure to mobilize TIL lytic granules upon conjugation with cognate tumor cells (17, 28), is transient in nature, being rapidly reversed upon purification (13, 17, 20), and is induced by TILs' contact with cognate tumor cells (20, 29). Biochemical assessment of TIL signal transduction in the MCA38 model revealed that although proximal TCR-mediated signaling is intact when assessed with a surrogate antigen (by anti-CD3 TCR ligation; ref. 17), when cognate tumor cells are used to stimulate TILs in vitro, Zap70 is not activated, implicating defective p56^lck activity (20). In nonlytic nonsignaling TILs, pcdh18 interacts with p56^lck and expression of pcdh18 in signaling-competent primary lytic effector CD8⁺ T cells (that do not endogenously express pcdh18) induces the TIL phenotype: Proximal TCR signaling is blocked at Zap70 activation coincident with abrogation of lytic function (13). These characteristics strongly supported pcdh18 being an inhibitory signaling receptor in TILs, one whose expression is restricted to memory T cells (13), and motivated the experiments reported in this work.

ISRs initiate inhibitory signaling after binding to cognate ligand in trans (30), so we determined the effect of silencing pcdh18 ligand expression in cognate tumor cells upon TIL lytic response and found that the poor lytic ability of freshly isolated TILs could be overcome with the use of tumor targets lacking pcdh18, showing that pcdh18 acted like an ISR. The growth of MCA38 tumors lacking pcdh18 was significantly delayed, as was the growth of wild-type MCA38 in pcdh18^–/– mice, which corroborated the in vitro data. These experiments show that pcdh18 interaction with ligand initiated TIL functional deficiency.

Comparison of pcdh18^–/– with wild-type TILs showed changes in the expression of genes involved in the antitumor T-cell immune response, including genes encoding inhibitory signaling receptors (e.g., PD-1). Nonetheless, effector-phase functions were also increased in pcdh18^–/– TILs as shown by enhanced tumor clearance, which was correlated with increased expression of GrzB and IFNγ.

MCA38 TILs in wild-type mice are CD62^lo (23), defining them as effector memory T cells, and clearly contain pcdh18 protein (which was the basis for its identification; refs. 13, 16); thus, it was interesting to note that RNA encoding pcdh18 was not robustly expressed in freshly isolated pcdh18^–/– TILs. CD8⁺ memory T cells express pcdh18 RNA (13) and upon activation, as CD62L^hi central memory cells convert to CD62L^lo effector memory cells, similar to the kinetics of activation of immediate early response genes, pcdh18 mRNA is rapidly upregulated following activation, but is quickly terminated and degraded (13). This is likely the reason why prior RNA analysis of memory T cells failed to detect pcdh18 expression (31). Rapid loss of pcdh18 RNA after activation shows that dependency on RNA expression data for evaluation of involvement of a given candidate ISR in T-cell function needs be corroborated with protein expression analyses.

Pcdh18 is highly expressed in embryonic brain, where it functions as a patterning receptor (32), localizes to the neuronal synapse (33), and interacts with the brain src homolog p59^fyn (34). It is also expressed in various adult tissues (35), but until our initial report (13) was not known to be expressed in the immune system. In the hematopoietic system, expression was restricted to the T-cell lineage, including CD4⁺ T cells, but only following differentiation to the memory state. As pcdh18 was expressed only in memory cells (and dendritic cells), pcdh18 appears to play an adjunct role in regulation of the effector phase, in contrast to other IRS (e.g., PD-1) which are also expressed in naïve cells. Thus, we consider the function of pcdh18 to modify or sculpt the effector phase, depending upon the expression of its ligand (itself) in tumor cells. As multiple ISRs are coexpressed in antitumor T cells (36), it seems reasonable to propose that inhibition of pcdh18 in conjunction with other target ISRs like PD-1 may find utility in experimental therapy of cancer. In this regard, we note that inspection of the cBioPortal database shows amplification of pcdh18 ligand in a few cancer types (breast, prostate, desmoplastic small round-cell tumors), wherein it seems reasonable to predict anti-pcdh18 intervention may be impactful (37).

ISRs, typified by PD-1, often function to raise the threshold of immune cell activation by recruiting inhibitory phosphatases Shp-1 or Shp-2 (encoded by Ptpn6 and Ptpn11, respectively) from the cytoplasm into proximity with components of the antigen receptor wherein molecules important in signal transduction are inactivated by dephosphorylation (38). Thus, because pcdh18 binds directly to p56^lck (13), it differs mechanistically from most ISRs and may represent a novel class of ISR that functions by direct interaction with key enzymes in the proximal TCR signaling pathway, a class that may include other src-binding proteins LIME, Sit, and TSAd. In this regard, it is of interest that mRNA for all three of these src-binding proteins are expressed at high levels in nonlytic TILs (13).

No potential conflicts of interest were disclosed.

The author is indebted to the many people who materially contributed to this project: Adam Blasidell, Rachel Brody, Jeremy Burns, Luis Chiriboga, Devon Columbus, Adrian Erlebacher, Adriana Heguy, Tim Hemesath, Mythili Koneru, Ngozi Monu, Sasa Radoja, Mohini Ragasagi, David Schaer, Sergio Trombetta, Alejandro Ulloa, Claire Vanpouille-Box, and Edwin Vazquez-Cintron.

Research in the author's laboratory was supported by NIH R01CA108573 and Pfizer Inc. (through the CTI program; both to A.B. Frey), and to NYULSM for core facilities support (P30CA016087 for Cytometry and Cell Sorting, Histopathology, and Genome Technology).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.