MERTK Acts as a Costimulatory Receptor on Human CD8⁺ T Cells

The TAM receptor kinases—TYRO3, AXL, and MERTK—are negative regulators of inflammatory responses on antigen-presenting cells, like macrophages and dendritic cells (DC; ref. 1). TAM receptor signaling is induced by the ligands growth arrest–specific gene 6 (GAS6) and protein S (PROS1), which act as bridging molecules by binding to phosphatidylserine (PtdSer; ref. 2). The PtdSer–GAS6 or PtdSer–PROS1 complex subsequently activates the TAM receptors (3). TAM receptors are reported to be expressed on DCs, macrophages, human platelets, natural killer [NK(T)] cells, and B cells (4–8). On cells of the innate immune system, TAM signaling dampens activation and promotes silent engulfment of apoptotic cells (9, 10). As for T cells, the ligand PROS1 is expressed by activated human and mouse CD4⁺ T cells. The previous thought that T cells did not express TAM receptors was challenged with the report of MERTK being expressed by activated human CD4⁺ T cells (11). In contrast, another study reported that activated mouse T cells do not express MERTK (12).

The TAM receptors also act as oncogenes. Many solid and hematologic cancers express TAM receptors, PtdSer, and the ligands PROS1 or GAS6 (13). These cancer cells are capable of TAM autosignaling, which is associated with oncogenic traits such as survival, invasion, chemoresistance, and metastasis (13–15). As a consequence, overexpression of TAM receptors in cancer is associated with a poor prognosis (reviewed in ref. 13). A range of inhibitors of TAM receptor signaling are in development or clinical testing for treatment of cancers such as leukemia (16–18).

Cytotoxic CD8⁺ T cells establish and maintain antitumor immune responses. Some cancer immunotherapies are based on the capacity of T cells to recognize and kill cancer cells (19). The use of checkpoint antibodies that block inhibitory signaling on T cells in the tumor microenvironment (TME) is associated with tumor regression in a variety of cancers. Here we characterize another molecule that affects T-cell functionality.

We show that human CD8⁺ T cells expressed MERTK and its ligand PROS1 upon T-cell receptor (TCR)–mediated activation. PROS1-mediated MERTK activation delivered a costimulatory signal in cytotoxic CD8⁺ T cells. We show that PROS1–MERTK signaling in CD8⁺ T cells affected the function, gene expression, and metabolism of the cell. Finally, we show that PROS1–MERTK signaling benefited tumor-infiltrating lymphocyte (TIL) expansion and autologous tumor cell killing by TILs from metastatic melanoma patients. Such TAM receptor function on activated CD8⁺ T cells could have therapeutic implications.

All procedures were approved by the Scientific Ethics Committee for the Capital Region of Denmark. Written informed consent was obtained from all patients according to the Declaration of Helsinki. All cells were cultured in a humidified 37°C, 5% CO₂ incubator. Cell lines PC-3, FM82, and MDA-MB-231 were cultured in RPMI-1640 + 10% heat-inactivated fetal calf serum (FCS; Gibco). Cell lines were obtained from ATCC or ESTDAB in 2014 or later and frozen upon initial expansion. The cell lines used in experiments were cultured for a maximum of 20 passages. Cells were not reauthenticated and were tested as Mycoplasma negative. Peripheral blood mononuclear cells (PBMC) from healthy donor buffy coats were isolated by gradient centrifugation and used immediately or cryopreserved.

Biopsies from metastatic lesions of patients with American Joint Committee on Cancer (AJCC) 7th edition stage III or IV melanoma were collected from 2006 to 2013 and used for expansion of TILs. TILs and autologous tumor cell lines used for in vitro cell killing assays were generated as described previously (20, 21). Autologous tumor cell lines were cultured in RPMI medium supplemented with 10% FCS.

Human CD8⁺ T cells were isolated from human PBMCs by negative selection (Magnisort, Invitrogen). Purified CD8⁺ T cells or PBMCs were cultured with anti-CD3/anti-CD8–coated Dynabeads (Gibco) in X-VIVO 15 medium (Lonza) or CEF peptide pool (MabTech), supplemented with 5% heat-inactivated human AB serum (Sigma-Aldrich) and 50 U/mL hIL2 (Proleukin). Alternatively, cells were cultured in serum-free medium with 50 nmol/L PROS1 (Haematologic Technologies) or 250 nmol/L MERTK inhibitor UNC2025 (Sigma-Aldrich). For PROS1-blocking experiments or PtdSer-blocking experiments, cells were cultured in medium with 5% human serum with PROS1-blocking antibody (anti-PROS1; 10 μg/mL, clone PS7, SCBT) or unconjugated Annexin V (5 μg/mL, BD Biosciences), respectively.

For proliferation assays, cells were labeled with proliferation dye CellTrace Violet (Invitrogen). For surface staining, single-cell suspensions were stained with the following: anti-CD3 (clone UCHT1), anti-CD8 (RPA-T8), anti-CD45RO (UCHL-1, all BD Biosciences), anti-CD4 (SK3), anti-CD137 (VIC7), anti-CCR7 (G043H7, all BioLegend), anti-PROS1 (PS7, SCBT), anti-Mertk (125518), anti-Tyro3 (96201), and anti-Axl (108724, all R&D Systems). Sample acquisition was performed using a FACSCanto or LSR II (BD Biosciences), and data were analyzed using FlowJo v10.

Amounts of IFNγ and TNFα (Invitrogen) or PROS1 (Abcam) were measured in culture supernatants using enzyme-linked immunosorbent assays (ELISA) according to the manufacturer's instructions. Results were analyzed using Epoch plate reader (BioTek) and Gen5 Take3 software (v1.00.4, BioTek). Alternatively, culture supernatants were tested using the Bio-Plex Pro Human Cytokine 27-Plex Immunoassay (Bio-Rad). Samples were acquired on a Bio-Plex 200 system and analyzed with Bio-Plex Manager v.6 software. Samples analyzed with the Bio-Plex system that were below or above the standard curve range were excluded from analysis.

RNA was isolated with NucleoSpin RNA kit (Macherey-Nagel) and reverse transcribed using SuperScript VILO cDNA Synthesis Kit (Invitrogen). qPCR was performed in Agilent AriaMX System using the Brilliant III Ultra-Fast QPCR Master Mix (Agilent). Amplified products were checked by dissociation curves, and expression was normalized to a housekeeping gene. Primer sequences used are listed in Supplementary Table S1.

Western blotting was performed according to standard protocols. Briefly, cells were lysed using RIPA lysis buffer (Pierce) supplemented with protease and phosphatase inhibitor cocktails (Thermo Scientific). Proteins were quantified by BCA assay (Pierce) and separated using precast 4% to 12% Bolt Bis-Tris Plus SDS-PAGE gels (Invitrogen). Proteins were transferred to nitrocellulose membranes using the iBlot 2 system (Invitrogen). The following primary antibodies were used for protein detection: rabbit anti-human MERTK (D21F11), rabbit anti-human TYRO3 (D38C6, both Cell Signaling Technology), rat anti-human PROS1 (PS7), mouse anti-human AXL (B-2), and mouse anti-human actin (C4, all SCBT). Proteins were visualized using SuperSignal West ECL Kit (GE Healthcare) and Bio-Rad ChemiDoc Molecular Imager. Quantification of signal was done using Fiji ImageJ (v.1.49).

A set of three Stealth siRNA duplexes for targeted silencing of human MERTK were obtained from Invitrogen. siRNA duplex sequences are listed in Supplementary Table S1. For control experiments, three siRNAs with scrambled sequences possessing similar GC content (Invitrogen) were used. Following magnetic bead removal, 3-day stimulated CD8⁺ T cells were transfected with MERTK or mock siRNA with the ECM830 square wave electroporation system (BTX) using electroporation parameters as previously described (22). Knockdown of protein content was confirmed for every individual experiment.

Sorted CD8⁺ T cells from three healthy donors were cultured in the presence or absence of anti-CD3/anti-CD8 beads or 10 μg/mL anti-PROS1 for 3 days. Subsequently, cells were incubated with Brefeldin A (BioLegend) for 4 hours. Final input for transcriptomic analysis was 1 × 10⁵ viable cells per condition. RNA and protein samples were split and processed separately. Transcriptomic analysis was performed using the nCounter Vantage 3D RNA:Protein Immune Cell Signaling Panel (NanoString Technologies). Samples were subsequently processed in the fully automated nCounter Prepstation (NanoString Technologies) and analyzed in the nCounter Digital Analyzer (NanoString Technologies). The nSolver4 software (NanoString Technologies) was used for data normalization and differential gene-expression analyses. Data analysis was performed according to NanoString gene-expression data analysis guidelines and the nCounter Advanced Analysis 2.0. Selected housekeeper mRNAs and proteins are listed in Supplementary Table S2. The significance of differential gene-expression between paired groups was estimated using a mixed module significance testing with the algorithm included in the nCounter Advanced Analysis. In this module, a negative binomial mixture model for low-expression probes or a simplified negative binomial model for high-expression probes is used. If both models fail, a log-linear model is applied. Differential expression is indicated as the log₂ fold change in gene or protein expression, and the obtained P values were adjusted for multiple testing by the Benjamini and Hochberg method (BH P value) to control the false discovery rate. Differentially expressed genes and proteins were depicted as volcano plot using R/RStudio v1.0.44.

The bioenergetics from CD8⁺ T cells were measured in the presence or absence of costimulatory MERTK signaling in real-time using an XF-96 Extracellular Flux Analyzer (Seahorse Bioscience, Agilent). Anti-CD3/anti-CD8–stimulated CD8⁺ T cells were grown in the presence or absence of 50 nmol/L PROS1 (serum-free medium) or anti-PROS1 (10 μg/mL, serum-containing medium) for 3 days prior to use. Cells were resuspended in Seahorse assay media (Seahorse Bioscience, Agilent), supplemented with 1 mmol/L pyruvate, 2 mmol/L glutamine, adjusted to pH 7.4, and subsequently seeded in a Seahorse 96-well plate using Cell-Tak adherent (Corning). Oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) were measured, and then wells were treated with 1 μmol/L oligomycin and 10 mmol/L 2-deoxy-d-glucose to measure ATP turnover and glycolytic capacity from the changes in OCR and ECAR, respectively, or with 0.4 μmol/L carbonyl cyanide p-(trifluoromethoxy) phenylhydrazone (FCCP) to determine reserve respiratory capacity from change in OCR. All wells received a final treatment with 2 μmol/L antimycin A.

Whole-cell ATP content was measured in 3-day anti-CD3/anti-CD8–stimulated CD8⁺ T cells grown in the presence or absence of 50 nmol/L PROS1 or anti-PROS1 (10 μg/mL) using a luciferase-based assay (ViaLight MDA Plus Detection Kit, Lonza). Luminescence was quantified in a MicroBeta2 Scintillation Counter (PerkinElmer).

For “young” TIL outgrowth, biopsy material was cut into small fragments (1–2 mm²) and cultured overnight. The following day, tumor fragments and cells were washed and used immediately or cryopreserved. TILs were expanded in RPMI-1640 supplemented with 10% heat-inactivated human AB serum, IL2 (6,000 IU/mL), penicillin, streptomycin, and fungizone. Expansion conditions included culture medium in the presence or absence of 50 nmol/L PROS1 or 10 μg/mL anti-PROS1. Outgrowth of “young” TILs was measured by manual, unblinded counting of live cells, and fold expansion was calculated.

TILs designated for in vitro killing assays were isolated and expanded in vitro from metastatic melanoma lesions with a two-step process as described previously (21). Expanded TILs with high specificity for the HLA-A2 restricted MART-1/MelanA peptide analogue ELAGIGILTV (>90% specific with peptide-MHC multimer staining) were obtained through electronic sorting of relevant CD8⁺ “young” TILs, using peptide-MHC multimers. TILs were subsequently subjected to the rapid expansion protocol as previously described (20).

The tumor-specific killing ability of TILs was assessed with an impedance-based cytotoxicity assay (23). Briefly, antigen-specific TILs were thawed and rested in IL2-free media (RPMI-1640 supplemented with 10% human serum, penicillin, and streptomycin) for 72 hours. Autologous tumor cells were seeded on E-plate 96 plates (ACEA Biosciences, Inc.) which were loaded onto RTCA SP real-time cell analyzer (ACEA Biosciences, Inc.). After 24 hours, TILs were added with 50 nmol/L PROS1 or a titration ranging from 0 to 100 nmol/L PROS1.

For coculture experiments, high TAM receptor–expressing MDA-MB-231 cells were cultured in serum-free X-VIVO medium for 1 week prior to coculture. Subsequently, MDA-MB-231 cells were plated in a flat-bottom 96-well plate and left to adhere for approximately 4 hours. Sorted allogeneic nonreactive CD8⁺ T cells and anti-CD3/anti-CD8 beads were added in a 1:10 tumor cell:T-cell ratio. A PROS1 titration was added in the range of 0 to 100 nmol/L PROS1. After 4 days of coculture, supernatants were harvested and analyzed by ELISA.

Data are plotted as mean ± SEM. Comparisons between groups were analyzed with two-tailed paired Student t tests or two-way ANOVA with Bonferroni multiple comparisons tests, as appropriate. Data analysis was performed with GraphPad Prism (v8.00) software unless specified otherwise. Used statistical tests and number of biological replicates are indicated in the figure legends. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

We analyzed the expression of the TAM receptors and ligand PROS1 upon CD8⁺ T-cell activation. Anti-CD3/anti-CD8–mediated activation of sorted human CD8⁺ T cells led to an increase in surface staining of PROS1 on CD8⁺ T cells from day 2 onward (Fig. 1A and B). This was correlated with an approximately 60-fold induction of PROS1 mRNA expression and an increase in endogenous protein expression (Fig. 1C and D). PROS1 surface staining was partly reversible by blockage of PtdSer (Supplementary Fig. S1A and S1B). Furthermore, activated CD8⁺ T cells significantly increased TAM receptor MERTK surface expression from day 2 onward, only on PROS1-positive cells (Fig. 1E and Supplementary Fig. S1). On day 3 postactivation, approximately 25% of CD8⁺ T cells were MERTK positive, whereas resting T cells remained MERTK negative (Fig. 1E and F). MERTK expression was confirmed by mRNA and protein expression on 3-day activated CD8⁺ T cells (Fig. 1G and H). To assess if MERTK expression was limited to a certain CD8⁺ T-cell subset, we analyzed MERTK expression on 3-day activated CD8⁺ T cells that were costained with subset markers CCR7 and CD45R0. These data show that MERTK expression was significantly higher on T_CM CD8⁺ T cells (Fig. 1I and J). Finally, to confirm that MERTK upregulation was not due to persistent stimulation by CD3/CD28, human PBMCs were activated with a pool of 23 peptides derived from cytomegalovirus, Epstein–Barr virus, and influenza. Using CD137, recently TCR activated, naturally occurring, CD8⁺ T cells can be identified (24). Only recently activated CD137⁺ CD8⁺ T cells expressed MERTK (Fig. 2A–E). Additionally, we found that resting or activated CD8⁺ T cells expressed little TYRO3 and did not express AXL (Supplementary Fig. S2A–S2D).

Next, we asked whether PROS1 acts as a costimulatory signal for CD8⁺ T cells, as previously shown for CD4⁺ T cells (11). We confirmed that PtdSer would not be a limiting factor of TAM signaling in our culture conditions, as the presence of PtdSer is essential for optimal TAM receptor activation. We excluded apoptosis as the sole PtdSer source by using a caspase inhibitor (Supplementary Fig. S3A–S3D). To study the effect of PROS1-mediated MERTK signaling on CD8⁺ T cells, we activated CD8⁺ T cells in the presence of 50 nmol/L PROS1. PROS1 acted as a costimulatory molecule by significantly increasing proliferation, but only when CD8⁺ T cells were activated (Fig. 3A and B). PROS1-mediated costimulation also increased secretion of effector cytokines IFNγ, TNFα, and chemokine CXCL10 (Fig. 3C). In addition, PROS1-costimulated CD8⁺ T cells secreted more memory cytokine IL7 (Fig. 3C). To substantiate these findings, we studied the effect of a PROS1-blocking antibody (anti-PROS1), which led to a significant inhibition of CD8⁺ T-cell proliferation (Fig. 3D and E).

To verify if the functional changes related to PROS1 were due to signaling through MERTK, we inhibited MERTK signaling in activated CD8⁺ T cells. We established an siRNA-mediated knockdown of MERTK. As our earlier results have shown that resting T cells do not express MERTK, CD8⁺ T cells were activated for 3 days prior to siRNA electroporation. We confirmed that siRNA knockdown resulted in a 70% reduction in MERTK protein expression compared with control (Fig. 4A and B). When reactivated, MERTK-knockdown cells produced less IFNγ (Fig. 4C). Moreover, IL7, but not IL15, secretion was significantly decreased (Fig. 4C). We verified these results using UNC2025, a MERTK inhibitor that is in development for the treatment of leukemia (18). MERTK inhibition significantly decreased anti-CD3/anti-CD8–mediated CD8⁺ T-cell proliferation with no decrease in cell viability (Fig. 4D and E). Correspondingly, MERTK inhibition could reverse the positive effects of PROS1 on IFNγ secretion (Fig. 4F).

To investigate the intracellular effects of this PROS1–MERTK axis on CD8⁺ T cells, we analyzed the transcriptome of 3-day anti-CD3/anti-CD8–activated CD8⁺ T cells in the presence or absence of anti-PROS1. An overview of the resulting differential regulation of approximately 800 genes and 30 proteins is depicted in Fig 5A. The most differentially upregulated genes and proteins in PROS1-blocked cells versus control were LTA, TNFSRF9, IL2, and IFNγ, whereas the most differentially downregulated genes were IL4R, DUSP4, CD99, ITGAL, and CCL5 (Fig. 5B–F). These results, along with the observation that activation-associated MERTK expression was more pronounced on T_CM cells, led us to hypothesize whether PROS1–MERTK signaling could influence differentiation of “long-lived” memory cells.

The gene encoding the transcription factor IRF4 was downregulated in PROS1-blocked cells versus control (Fig. 6A). IRF4 has been correlated with metabolic programming of CD8⁺ T cells, where it induces a metabolic shift essential for antigen-specific responses and effector differentiation and function (25). Subsequently, we studied the metabolism of activated CD8⁺ T cells in the presence or absence of PROS1–MERTK signaling. Bioenergetic phenotypes are shown to be predictive for CD8⁺ T-cell differentiation into the various memory subsets (26). The basal respiration rate of PROS1-blocked cells was significantly decreased to 35% of control-stimulated cells (Fig. 6B). Accordingly, the ATP turnover of PROS1-blocked cells was reduced to 31% (Fig. 6C). Finally, the spare respiratory capacity (SRC) of PROS1-blocked cells was significantly decreased (Fig. 6D). This contrasts with activated CD8⁺ T cells supplemented with PROS1, where no significant changes were found (Supplementary Fig. S4A–S4F). For both PROS1-blocked and PROS1-supplemented cells, no significant change of glycolytic reserve capacity was discovered (Fig. 6E and Supplementary Fig. S4D). To test whether this effect was due to a lack of overall energy, we measured the whole-cell content of ATP, which increased by 140% in PROS1-blocked CD8⁺ T cells (Fig. 6F). Representative plots of metabolic experiments are depicted in Fig. 6G. This demonstrates that the decreased activity of oxidative phosphorylation and mitochondrial respiration in PROS1-blocked cells was not a result of starvation of ATP. Taken together, these results indicate that when PROS1–MERTK signaling is absent in activated CD8⁺ T cells, the mitochondrial respiration capacity, necessary for long-lived T_CM cells, is significantly decreased.

TAM receptors are expressed on cancer cells (13, 16). Cytotoxic CD8⁺ T cells have been thought to be negative for these TAM receptors. Our identification of potential joint expression of TAM receptors on T cells and cancer cells could indicate a situation of ligand competition. As a proof of principle, we measured PROS1 consumption from culture medium by T cells and cancer cells. As cancer cell lines express TAM receptors at a higher level, their PROS1 consumption was higher than that of activated T cells (Supplementary Fig. S5A and S5B). We aimed to study the ligand competition in a T-cell/cancer cell coculture using activated CD8⁺ T cells and the MDA-MB-231 cancer cell line (highly expressing TAM receptors; Fig. 7A and B). Scarcity of PROS1 in low concentrations resulted in an increased inhibition by cancer cells on CD8⁺ T-cell activation. Correspondingly, once an excess of PROS1 was present, this PROS1-competive inhibitory effect was abrogated.

Next, we sought to examine the clinical relevance of TAM signaling in T cells in relation to adoptive T-cell therapy, where T-cell–mediated antitumor immunity is essential. We studied if PROS1 signaling had an impact on the primary expansion of “young” TILs from metastatic melanoma patients. Treatment of TILs with anti-PROS1 during the outgrowth phase led to a significant decrease in fold expansion rate (Fig. 7C). Although PROS1-blocked conditions had a reduced total number of cells, analysis of T-cell subsets revealed that no specific subset was depleted (Fig. 7D). Finally, we asked whether joint TAM receptor expression on TILs and cancer cells could affect antitumor immunity. For this study, we screened cancer cells from 3 metastatic melanoma patients for TAM receptor expression (Fig. 7E). Next, rapidly expanded (“REP”) antigen-selected TILs from the highest MERTK-expressing patient were cocultured with their autologous TAM receptor–expressing tumor cells. Using xCELLigence technology, we followed real-time in vitro tumor cell killing by autologous TILs (Fig. 7F and Supplementary Fig. S5C and S5D). PROS1 alone had no effect on cancer cell growth. PROS1 supplementation had no significant effect on TIL-mediated killing (Fig. 7G).

TAM receptors play a role in dampening immune responses by negative feedback mechanisms in cells of the innate immune system (27). TAM receptor signaling thereby safeguards termination of inflammation and promotes tissue repair (13). A variety of cells are reported to express TAM receptors, but T cells have been reported negative (28). Our results demonstrate that CD8⁺ T cells express MERTK upon activation, and that MERTK signaling has proproliferative, costimulatory effects on human CD8⁺ T cells.

It was shown more than 20 years ago that cross-linking of PROS1 inhibited T-cell proliferation, but PROS1′s function as a TAM receptor ligand was yet unknown, and T-cell inhibition effect was attributed to PROS1′s anticoagulant role (29). In contrast, soluble PROS1 has proproliferative effects on human CD4⁺ T cells, an effect that was reversible with blocking antibodies for PROS1 (11). T cells express PROS1 upon activation (7, 30, 31), and T cell–derived PROS1 plays a role by negative feedback signaling to DCs (7). In this regard, low plasma concentration of TAM ligands is associated with a range of autoimmune disorders, suggesting that unbalanced TAM signaling could play a role in the pathogenesis of autoimmunity (27).

TAM receptors were described as absent on both human and mouse lymphocytes in both resting and PMA/ionomycin-activated cells (12, 28, 32). Cabezon and colleagues later showed that human CD4⁺ T cells upregulate MERTK upon TCR-mediated activation after 3 days (11). In line with these results, we demonstrate that human CD8⁺ T cells expressed not only PROS1 but also MERTK from 2 days postactivation. MERTK upregulation was indeed induced by CD3/CD28 stimulation but also by physiologically relevant activation of naturally circulating CD8⁺ T cells specific for common viruses such as cytomegalovirus and influenza. PROS1 costimulation was subsequently mediated via MERTK. PtdSer was not a limiting factor for TAM signaling in the CD8⁺ T-cell cultures. PtdSer has various nonapoptotic signaling functions on activated T cells (33, 34). Indeed, PtdSer can be “downregulated” again on nonapoptotic cells, although why transient PtdSer-exposing viable cells are not phagocytosed remains controversial (35, 36).

Memory CD8⁺ T cells form a heterogenic population of cells (37, 38). The metabolism of cells is linked to their function (39). Our data suggest a role for IRF4-mediated MERTK signaling in CD8⁺ T-cell mitochondrial respiration, which is necessary for T_CM differentiation and longevity (39, 40). We have previously shown that the bioenergetics of even close differentiation stages can differ in their activity of glycolysis and oxidative phosphorylation (41). The downregulation of IRF4 and the change in bioenergetic phenotype demonstrated after anti-PROS1 treatment indicates a shift in differentiation stage of CD8⁺ T cells. This supports our hypothesis that costimulatory PROS1–MERTK signaling is needed for CD8⁺ T-cell differentiation. Along those lines, the transcriptomic data showed that IL2 mRNA were increased in the absence of MERTK signaling, which is needed for differentiation of effector T cells (37). The expression data also suggest a role for the IL4 receptor. This coincides with earlier findings showing that IL4 amplified PROS1 expression upon TCR stimulation in mouse T cells, and that IL4-deficient CD4⁺ T cells are not able to induce PROS1 expression (30, 31). IL4, as well as being a cytokine for CD4⁺ T cells, stimulates CD8⁺ T-cell memory differentiation (42, 43).

TAM receptor signaling on cancer cells has been implicated in proliferation, epithelial-to-mesenchymal transition, survival, and migration (13–15, 44). Next to this, AXL and MERTK have been suggested to play a role in therapy resistance. TAM receptor expression by cancer cells—and any cell in the TME—could thus set the stage for ligand competition. PROS1 secreted by T cells could be exploited by cancer cells for oncogenic TAM receptor signaling. This is corroborated by our coculture experiments, suggesting that cytokine secretion by T cells is inhibited at low concentrations of PROS1, supposedly due to ligand competition. TAM receptor signaling in cancer has been shown to lead to upregulation of PD-L1 (14, 45, 46). Therefore, oncogenic TAM receptor signaling could not only jeopardize T-cell costimulation but also inhibit PD-1–expressing T cells. This could restrain natural as well as treatment-induced anticancer T-cell responses, for instance, upon adoptive cell transfer (ACT; ref. 47). A hurdle for TIL-based ACT is outgrowth of T cells during the early phases of culture. We showed that melanoma TILs show a decreased expansion if PROS1 is cleared during culture. Together, our data on ex vivo and in vitro killing by TILs in an autologous setting support the notion that MERTK signaling may be inflicted by cancer cells, which may jeopardize the antitumor T-cell response.

Several agonistic antibodies directed against costimulatory molecules are in development or clinical testing as cancer treatments (48). Due to the widespread expression of TAM receptors, the therapeutic use of agonists is less straightforward. However, delivery approaches such as those using bispecific antibodies could open possibilities for targeting to T cells (49). Another therapeutic approach is the development of MERTK inhibitors (13). Due to their expression on various cancer types, TAM receptor small-molecule inhibitors have been, and are currently being, developed (12, 18; reviewed in 16). TAM receptor inhibitors have been suggested to be combined with immunotherapy, as T cells are assumed not to express TAM receptors (45, 50). Lee-Sherick and colleagues showed that MERTK inhibition in mice indirectly lowered PD-1 expression on T cells via effects on DCs and macrophages (12). As mouse T cells are assumed negative, there are, correspondingly, no expected direct effects of MERTK inhibition on (mouse) T cells. Data on human T cells, however, might paint a different picture. Highlighting this difference between mice and humans, Cabezon and colleagues showed that antibody-mediated MERTK inhibition decreased human CD4⁺ T-cell proliferation (11), whereas our study found inhibition of both CD8⁺ T-cell proliferation and IFNγ secretion using MERTK inhibitor UNC2025. Moreover, we confirmed these effects using MERTK knockdown. Our data indicate that clinical development should be halted or at least carefully monitored due to the expression of MERTK on activated human CD8⁺ T cells. Direct targeting of cancer cells could be positive, whereas inhibiting MERTK expressed on DCs implies more potent induction of T cells. However, targeting of MERTK could jeopardize both regular and antitumor T-cell responses and differentiation. Further studies are needed to scrutinize pros and cons of MERTK inhibition in cancer.

Despite our findings, several questions remain. Although T_CM cells are prone to express more MERTK, the identity of signals other than TCR signaling that lead to expression of MERTK is unknown. Furthermore, TAM receptor expression would be expected to influence T-cell activity and T-cell numbers in the TME, which again could affect overall survival. Due to expression of three TAM receptors and at least two ligands, these relationships are not trivial to study.

Our results reveal a role for TAM receptor MERTK in providing late costimulatory signaling to human CD8⁺ T cells. The mechanisms and implications of TAM receptor expression and signaling in T cells and effects on antitumor immunity remain unclear. Future studies are needed to determine whether oncogenic TAM signaling in cancer cells or costimulatory TAM signaling in cytotoxic T cells will tip the scale toward antitumor immunity.

M.J.W. Peeters has ownership interest in a patent application. J.C. Becker has received honoraria from the speakers bureau of Amgen, Sanofi, Merck, and Pfizer, and is a consultant/advisory board member for Merck, Amgen, eTheRNA, 4SC, Pfizer, and Sanofi. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.J.W. Peeters, A. Draghi, L.J. Rasmussen, Ö. Met, J.C. Becker, C. Desler, P. thor Straten

Development of methodology: M.J.W. Peeters, D. Dulkeviciute, S.K. Skadborg, A.M. Carnaz Simões

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.J.W. Peeters, D. Dulkeviciute, A. Draghi, A. Rahbech, S.K. Skadborg, T. Seremet, A.M. Carnaz Simões, E. Martinenaite, H.R. Halldórsdóttir, M.H. Andersen, G.H. Olofsson, I.M. Svane, J.C. Becker, M. Donia, P. thor Straten

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.J.W. Peeters, D. Dulkeviciute, C. Ritter, A. Rahbech, E. Martinenaite, H.R. Halldórsdóttir, M.H. Andersen, G.H. Olofsson, J.C. Becker, M. Donia, C. Desler, P. thor Straten

Writing, review, and/or revision of the manuscript: M.J.W. Peeters, A. Draghi, C. Ritter, E. Martinenaite, G.H. Olofsson, I.M. Svane, L.J. Rasmussen, J.C. Becker, M. Donia, C. Desler, P. thor Straten

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Martinenaite, P. thor Straten

Study supervision: J.C. Becker, P. thor Straten

The authors would like to thank Center for Cancer Immune Therapy staff, in particular M. Idorn, for technical assistance. The authors would also like to thank R. van Eijsden (NanoString Technologies) and S. Schwengberg (ACEA Biosciences) for valuable help with gene-expression studies and xCELLigence analysis, respectively. This study was supported by the Danish Council for Independent Research (grant no. DFF-1331-00095B), the Danish Cancer Society (grant no. R72-A4396-13-S2), the Training Network for the Immunotherapy of Cancer funded by the EU (IMMUTRAIN; H2020 grant no. 641549 to M.J.W. Peeters and P. thor Straten), The Danielsen Foundation, Axel Musfeldts fond, Dagmar Marshalls Fond, Else og Mogens Wedell-Wedellsborg Fond, AP Møller Fonden, and Den Bøhmske Fond.

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.