Multifaceted Role of BTLA in the Control of CD8⁺ T-cell Fate after Antigen Encounter

Adoptive T-cell therapy (ACT) using the body's own expanded tumor-infiltrating lymphocytes (TIL) with prior lymphodepletion and followed by high-dose IL-2 administration has demonstrated an overall response rate of 38%–51% in multiple clinical trials for stage IIIC/IV metastatic melanoma (1–4). A comprehensive immunophenotyping of TIL infusion products has revealed an unexpected finding that a population of CD8⁺ TILs expressing a molecule known to attenuate T-cell response, the B and T lymphocyte associated (BTLA), was strongly associated with a positive clinical response (4).

BTLA is an inhibitory molecule expressed by T cells, B cells, dendritic cells, and NK cells (5). Herpes virus entry mediator (HVEM), a TNF receptor superfamily member 14 (TNFRSF14), is the known ligand for BTLA (6). The cytoplasmic domain of BTLA consists of three motifs; an immunoreceptor tyrosine-based inhibition motif (ITIM), an immunoreceptor tyrosine-based switch motif (ITSM), and a growth factor receptor–bound protein 2 motif (Grb2). Ligation of BTLA by HVEM has been shown to recruit Src homology 2 (SH2) domain-containing phosphatase 1 and 2 (SHP-1 and SHP-2) to the ITIM and ITSM motifs, resulting in suppression of T-cell receptor (TCR) activation (5, 7). Indeed, ITIM and ITSM motifs were required for the full function of BTLA to inhibit T-cell proliferation and cytokine production including IFNγ, IL-2, and IL-10 (8, 9). A recent report indicated that PD-1 ligation, which also contains ITIM and ITSM motifs, selectively inhibits both the Akt and Ras–MEK–ERK pathways (10). So far, it remains inconclusive whether BTLA utilizes mechanism similar to PD-1 as ITIM and ITSM motifs are commonly shared between these two receptors.

Unlike PD-1, BTLA also harbors a Grb2 motif. Although the function of the Grb2 motif remains unclear, some evidence suggests that it may actually transmit a positive signal. In fact, an in vitro binding assay demonstrated the potential interaction of the Grb2-binding motif with the Grb2 adaptor protein and the p85 subunit of phosphatidylinositol 3-kinase (p85 PI3K; refs. 7, 11). Interestingly, a gene expression analysis of mouse CD4⁺ T cells following activation by anti-CD3 and anti-BTLA demonstrated a highly overlapping transcription profile with that produced by anti-CD3 in combination with positive costimulators (CD28, ICOS, and CD80), but not with inhibitory molecules (PD-1 and CTLA-4; ref. 12).

Emerging evidence also demonstrates that BTLA serves as a T-cell differentiation marker in human T cells as BTLA expression is highly enriched in naïve T cells and central memory T cells (Tcm) and downregulated upon T-cell differentiation (13). Our recent work demonstrated that CD8⁺BTLA⁺TILs exhibited the molecular signature of less-differentiated T cells as compared with their CD8⁺BTLA⁻ counterpart and had increased persistence following adoptive transfer in treated patients (14). Several studies in both immunodeficient murine and non-human primate models also demonstrated that central memory–derived effector CD8⁺ T cells established a pool of in vivo persistent memory T cells (15–18). In addition, adoptive transfer of memory T cells with stem cell properties (T_SCM) was shown to confer in vivo persistence and better tumor control due to enhanced survival and antitumor properties (15).

Thus far, it remains understudied whether the intrinsic properties of less-differentiated TILs highly enriched in BTLA-expressing cells and/or BTLA signaling itself contribute to the favorable clinical outcome of TIL-treated patients. In this study, we have uncovered survival advantages of the BTLA⁺ subset that allows for serial killing of target tumor cells, which may explain our previous correlation between this subset and response to TIL ACT. In addition, our results unveiled a role for the BTLA-associated Grb2-binding motif in T-cell proliferation and IL-2 production following TCR engagement that was independent of the inhibitory function of ITIM/ITSM motifs. The use of a pmel BTLA knockout system demonstrates a weaker priming of T cells in response to the cognate antigen and the absence of a recall response. Overall, this study has uncovered a previously unappreciated role of the Grb2 motif of BTLA in providing positive costimulatory signal to T cells and the ability of CD8⁺BTLA⁺TIL to function as serial killers.

Platinum-E retroviral packaging cell line, MEL 526 tumor line, and primary melanoma tumor cell line #2549, as well as B16F10, and B16OVA were maintained in Roswell Park Memorial Institute (RPMI) medium supplemented with 10% FBS (Gemini bio product), 10 mmol/L HEPES (Gibco), 10 mmol/L penicillin–streptomycin (Gibco), and 10 mmol/L Glutamine (Gibco), selenium–transferrin–insulin (Gibco), and 0.05 mmol/L β-mercaptoethanol (Gibco) as described previously (2, 19). Platinum-E retro packing cell line was purchased from Cell Biolabs. MEL 526 tumor line was obtained from Dr. Steven A. Rosenberg at the National Cancer Institute. Autologous primary melanoma tumor cell line #2549 was generated at MD Anderson Cancer Center (Houston, TX) from a tumor sample of a patient enrolled on an ongoing adoptive T-cell therapy study. The cell line #2549 was last authenticated on 03/24/2015 by STR DNA fingerprinting using the Promega 16 High Sensitivity STR Kit (Catalog #DC2100). The STR profiles were compared to online search databases (DSMZ/ATCC/JCRB/RIKEN) of 2,455 known profiles; along with the MD Anderson Characterized Cell Line Core (CCLC) database of 2,556 known profiles. The STR profiles matched known DNA fingerprint of patient's PBMCs. No authentication was performed in all other cell lines.

Tumor samples were obtained from patients with stage IIIc and stage IV melanoma undergoing surgery at The University of Texas MD Anderson Cancer Center according to an Institutional Review Board–approved protocol and with patient consent (IRB# 2004-0069, LAB06-0755, NCT00338377). This study was carried out in compliance with the protocol and Good Clinical Practice concerning medical research in humans, as described in the Declaration of Helsinki.

Fragments from melanoma tumors were cut into 1 to 2 mm³. Each fragment was placed into a single wells in 24-well culture plates (Falcon) and maintained with RPMI supplemented with 10% heat-inactivated Human AB serum (Gemini bio product), IL-2 6,000 IU/mL (Proleukin, Novartis), 10 mmol/L HEPES, 10 mmol/L penicillin–streptomycin (Gibco), and 10 mmol/L Glutamine (Gibco) as described previously (20).

Murine BTLA was amplified from fully sequenced murine BTLA from Mammalian Gene Collection Clones (MGC, Open Biosystems) by PCR with primers mBTLA-F and R (Supplementary Table S1). The PCR products were cloned into pRVKM retroviral vectors. To generate BTLA mutants, we substituted tyrosine for phenyl alanine in either Grb2 motifs or ITIM and ITSM motifs of BTLA cytoplasmic tails (Supplementary Table S2). These included murine ΔGrb2 mutants (Y245F), and murine Δ ITIM and ITSM mutants (Y274F and Y299F). Sequences of all constructs were validated by DNA sequencing.

pRVKM retroviral vectors and pEco plasmids were cotransfected into Plate-E cells using using PolyJet (Signagen Laboratories). The supernatants were harvested 60 hours later and concentrated using Vivaspin-20 (Vivaproducts). Splenocytes from OT-1 BTLA KO mice were cultured with RPMI1640 with 10% FBS and hIL-2 at 300 IU/mL, and activated with anti-mouse CD3 at a concentration of 0.3 μg/mL (Clone 145-2C11, eBioscience) for 24 hours. The cells were then infected with a concentrated retrovirus and further expanded in RPMI1640 with 10% FBS and hIL-2 for 3 days. The cells were sorted on the basis of the expression of GFP using a FACSAria (BD Biosciences) and propagated with hIL-2 at 300 IU/mL for 5 days.

Murine OT-1 BTLA KO cells overexpressing BTLA WT or mutants constructs were restimulated with either 10 ng/mL anti-mouse CD3 (Clone 145-2C11, BD Pharmingen) alone or with recombinant mouse HVEM Fc (R&D Systems) plate-bound for 8 hours prior to harvest with the cell lysis buffer (kindly provided by RPPA core facility at The University of Texas MD Anderson Cancer Center). The cell lysates were centrifuged at 14,000 rpm for 10 minutes at 4°C. The protein supernatant was quantified using protein assay kit (Thermo Scientific). Reverse-phase protein array (RPPA) was processed and normalized as described previously (21). Differential fold expression of protein was analyzed using Linear models and empirical Bayes methods (22). Volcano plots were generated using R system. For human TILs, four TIL lines were stained with anti-CD8 (clone RPA-T8, BD Pharmingen), anti-BTLA (clone J168, BD Pharmingen), and Sytox blue (Molecular Probe) under aseptic condition. The cells were sorted on the basis of expression of CD8⁺BTLA⁺ using FACSAria (BD Biosciences). On the next day, sorted TILs were restimulated with anti-human CD3 (clone OKT-3, eBioscience) with or without recombinant human HVEM-Fc (R&D Systems) plate-bound for 8 hours prior to harvest with the cell lysis buffer. The protein samples were processed, normalized, and analyzed as similar to the mouse experiment described above.

OT-1 BTLA KO T cells overexpressing BTLA WT or mutants were re-activated with either dendritic cells pulsed with OVA peptide (SIINFEKL; American peptide) or dendritic cells alone at ratio of 1:40 in the presence of BD GolgiStop (according to the manufacturer's instructions; BD Pharmingen). After 4 hours, the cells were fixed and permeabilized using BD Cytofix/Cytoperm kits (according to the manufacturer's instructions; BD Pharmingen), and subsequently stained with anti-mouse IFNγ (clone XMG1.2; BD Pharmingen) and anti-mouse TNFα (clone MP6-XT22; BD Pharmingen).

Murine OT-1 BTLA KO cells overexpressing BTLA WT or mutants were restimulated with either 10 ng/mL anti-mouse CD3 (Clone 145-2C11, BD Pharmingen) alone or with recombinant mouse HVEM-Fc (R&D Systems) plate-bound for 24 hours. Supernatants were collected to quantify the secreted cytokines using a MILLIPLEX MAP Mouse CD8⁺ T Cell Magnetic Bead Panel (Millipore).

T cells were cocultured with tumor cells labeled with eFluor670 at ratio of 1:1, 1:3, and 1:10. After 3 hours, the cells were fixed and permeabilized using BD Cytofix/Cytoperm (according to manufacturer's protocol; BD Pharmingen), then stained with anti-cleaved caspase-3 (clone CPP32; BD Pharmingen), and analyzed by a BD FACSCanto II (BD Biosciences).

Murine OT-1 BTLA KO expressing BTLA WT or mutants were labeled with eFluor670 and restimulated with dendritic cells alone or dendritic cells pulsed with OVA peptide (SIINFEKL; American peptide) at a ratio of 1:40 for 48 hours prior to being analyzed using a BD FACSCanto II (BD Biosciences).

TILs and tumor cells were labeled with 1 μmol/L of red fluorescence dye (PKH26, Sigma) and 1 μL of green fluorescence dye (PKH68, Sigma), respectively. The cells were loaded onto nanowells at concentration of 1 × 10⁶/mL. Target cell cytolysis mediated by TILs was monitored using a Carl Zeiss Axio Observer fitted with Hamamatsu EM-CCD camera using 10 × 0.3 NA objective. Apoptotic cells became green when stained with Annexin V conjugated with Alexa 647 as described previously (23).

NOD-scid IL2Rgamma^null (NSG) mice were engrafted with either 5 × 10⁶ MEL 526 MEL tumor cells or 5 × 10⁶ autologous primary melanoma tumor cells. On day 12, either 1 × 10⁷ sorted CD8⁺BTLA⁺ or sorted CD8⁺BTLA⁻ were adoptively transferred into tumor-bearing mice (n = 5–8 per group). Recombinant human IL-2 (Proleukin, Prometheus) was administered intraperitoneally at a concentration of 6 × 10⁵ IU immediately after the TIL transfer and daily for three days. Tumor size was measured every other day. Mice were sacrificed when tumors exceeded 15 mm diameter. Peripheral blood was collected every other day, lysed with ACK lysis buffer, then stained for AQUA (Invitrogen), anti-human CD45 (clone HI30; BD Pharmingen), and anti-human CD8 (clone RPA-T8; BD Pharmingen).

C57BL/6J mice were intravenously administered 0.5 × 10⁶ naïve pmel-1 or OT-I T cells, and vaccinated with gp100 peptide or OVA peptide (100 μg) together with anti-CD40 (50 μg) and imiquimod (50 mg). Recombinant human IL-2 at 1.2 × 10⁶ IU was administered once, and 6 × 10⁵ IU twice daily for the next 2 days (i.p.). Peripheral blood was collected every other day to determine the frequency of circulating pmel or OT-I T cells. Mice were boosted with gp100 or OVA peptide following the contraction phase and peripheral blood was collected every other day. Mice were sacrificed on day 120, and splenocytes were isolated to determine the presence of pmel T cells.

NOD-scid IL2Rgamma^null (NSG) mice were purchased from the Jackson Laboratory. OT-1 BTLA KO C57BL/6 and Pmel BTLA KO C57BL/6 mice were kindly provided from Dr. Roza Nurieva (Department of Immunology, The University of Texas MD Anderson Cancer Center, Houston, Texas and Graduate Program in Immunology, Graduate School of Biomedical Sciences, The University of Texas Health Sciences Center at Houston, Houston, Texas). Pmel BTLA WT C57BL/6 mice were kindly provided by Dr. Willem Overwijk (Department of Melanoma Medical Oncology, The University of Texas MD Anderson Cancer Center, Houston, Texas and Graduate Program in Immunology, Graduate School of Biomedical Sciences, The University of Texas Health Sciences Center at Houston, Houston, Texas). C57BL/6 mice were purchased from Charles River Laboratories. All mice were housed in a specific pathogen-free facility at The University of Texas MD Anderson Cancer Center. Animal protocols were approved by The University of Texas MD Anderson Cancer Center [#00001229 (RN)].

Statistical analyses were performed using GraphPad Prism and the R system. For all survival curve analysis data, a log-rank test was used to compare distribution of the two groups. We applied Linear models and empirical Bayes methods to compare the differential protein expression from RPPA datasets. Differences in tumor cell death of B16-OVA and B16-F10 and the percentages of positive virally transduced mouse T cells producing IFNγ and TNFα were analyzed using a two-way ANOVA, and a two-tailed Student t test was used to determine statistical significance for all other analyses.

To investigate whether BTLA is associated with melanoma patient survival, we conducted a Kaplan–Meier survival analysis in stage III metastatic melanomas according to gene expression data in tumor tissues from TCGA (The Cancer Genome Atlas; ref. 24). We found that patients expressing high transcript levels of either CD8a or BTLA had much better survival as compared with either CD8a low or BTLA low, respectively (CD8a high vs. CD8a low, P = 0.0007, N = 42; BTLA high vs. BTLA low, P = 0.001, N = 42; Fig. 1A and B). When both markers were analyzed together, the association of CD8a high and BTLA high conferred the greatest survival benefit as compared with other combinations (CD8a high BTLA high vs. CD8a low BTLA low, P = 0.0006, N = 98) (Fig. 1C). BTLA can be expressed by immune cells other than T cells such as NK cells, B cells, or dendritic cells, with the highest levels found in B cells (5). Interestingly, we did not observe improved survival in patients with high BTLA expression when associated with either high NK cells (identified by the expression of NCR1) or B cells (CD19; Supplementary Fig. S1A and S1B). These data suggest that high BTLA expression in association with high CD8⁺ signal is associated with improved melanoma patient survival. The coexpression of CD8 and BTLA cannot be inferred by gene expression profiling from tumor tissue; however, the improved survival benefit measured in conjunction with BTLA and CD8 expression is not seen when BTLA is evaluated in conjunction with markers from other cell populations known to express BTLA. These data lend support to our hypothesis that CD8⁺BTLA⁺ TILs exert efficient antitumor control based on our previous observation that this T-cell population correlated with better response to TIL ACT (4).

It has been noted that more differentiated T cells secrete more cytotoxic and cytolytic proteins compared with less-differentiated T cells (25). Thus, we conducted both in vitro and in vivo tumor control studies to compare the killing capacity of the BTLA⁺ and BTLA⁻ CD8⁺ TIL subsets. To determine CTL-mediated tumor killing and ensure an equal tumor-specific population of both BTLA subsets, MEL 526 melanoma tumor cells expressing the MART-1 antigen were cocultured with either sorted MART-1 recognizing (tetramer positive) CD8⁺BTLA⁺ (CD8⁺MART-1⁺BTLA⁺) or CD8⁺BTLA⁻ (CD8⁺MART-1⁺BTLA⁻). We found comparable MART-1 antigen-specific tumor killing ability after a short term (3 hours) TIL and tumor coincubation between the BTLA⁺TIL and BTLA⁻TIL subsets in two TIL lines (Fig. 2A, left TIL#2559; middle, TIL #2765). In addition, comparable killing ability was observed using bulk nonantigen-restricted, sorted BTLA subsets in a third TIL line (Fig. 2A, right, TIL #2549). As we did not observe a difference in in vitro killing between the two subsets, but have demonstrated a better clinical outcome in patients infused with more CD8⁺BTLA⁺ TILs, we next determined whether this subset could exhibit better in vivo tumor control. Immunodeficient NOD-scid IL2Rgamma^null mice (NSG) were engrafted with 5 × 10⁶ MEL 526 human melanoma tumor cells for 10 days prior to adoptive transfer of 1 × 10⁷ human CD8⁺MART⁺BTLA⁺TILs or CD8⁺ MART⁺BTLA⁻TILs. Tumor burden was measured every other day and transferred TILs were quantified from peripheral blood based on the expression of CD45 and CD8. We observed that the CD8⁺BTLA⁺ TIL subset exhibited significantly better tumor control on day 14 as compared with its BTLA⁻ counterpart in NSG mice treated with either MART-1 tumor antigen–restricted TIL line [2559 MART⁺; P = 0.002 (Fig. 2B, left) and 2765 MART⁺; P = 0.042 (Fig. 2B, middle)] as well as in a nontumor antigen–restricted setting (2549; P = 0.049; Fig. 2B, right). In addition, we found that the level of the CD8⁺BTLA⁺ TIL subset in the peripheral blood was significantly higher than the CD8⁺BTLA⁻ subset early after therapy in mice treated with antigen-specific TIL lines [2559 MART⁺; P = 0.039 on day 4, (Fig. 2C, left) 2765 MART⁺; P = 0.0021 and 0.02 on day 2 and 4, respectively (Fig. 2C, middle)], but not in those treated with the nonrestricted antigen-specific TIL line (2549; P = 0.06 on day 2; Fig. 2C, right). In summary, these data suggests that the CD8⁺BTLA⁺ and CD8⁺BTLA⁻ TIL subsets are equally able to directly kill tumor targets in a short-term in vitro assay. However, the BTLA-positive subset provides superior in vivo tumor control and tends to persist better following TIL transfer.

As demonstrated above, the CD8⁺BTLA⁺ subset provides significantly better tumor control than its CD8⁺BTLA⁻ counterpart in vivo; however, this difference is not observed in an in vitro setting. Because our previous in vitro tumor killing experiments were conducted using population assays and for only 3 hours, this limited our ability to track T-cell fate following tumor target cell interactions. To further understand the behavioral difference between the CD8⁺BTLA⁺ and CD8⁺BTLA⁻ subsets in mediating tumor killing, we utilized Timelapse Imaging Microscopy In Nanowell Grids (TIMING), to study the dynamic interactions between individual tumor targets and effector T cells in high-throughput (26, 27). To this end, labeled tumor cells and TILs were loaded onto the nanowell chip, consisting of 28,224 wells regularly separated in 7 × 7 blocks, and the interactions were quantified at two separate effector-to-target cell ratios of 1:1 and 1:2. This assay allowed us to dissect the kinetics of T-cell killing by calculating the following sequential parameters: (i) Time needed to establish conjugation between the T cell and tumor target (t_seek), (ii) duration of the contact between the T cell and the tumor target (t_contact), and (iii) time between the first T-cell contact with the tumor cell and tumor cell apoptosis (t_death; Fig. 3A). Interestingly, the CD8⁺BTLA⁺ subset was more efficient in tumor seeking at effector-to-target cell ratios of 1:1 (N = 3,319) as the time utilized to make first contact with tumor cells was significantly less than that of the CD8⁺BTLA⁻ subset (P < 0.0001; Fig. 3B, left). This difference was not observed when 2 targets were present in the well (1:2, P = 0.22) (Fig. 3B, left). It is possible that loading 2 targets in each nanowell rendered the distance to travel to meet each target too short to be able to measure a difference in the ability of T-cell subsets to seek the target. However, in comparison with CD8⁺BTLA⁻ TILs, individual CD8⁺BTLA⁺ TILs spent longer time in contact with a tumor cell target and induced apoptosis with slower kinetics at both E:T ratios studied (t_contact; 1:1 and 1:2, P < 0.0001, t_death; 1:1 and 1:2, P < 0.0001; Fig. 3B, middle and right, respectively). Overall, individual CD8⁺BTLA⁺ TILs found their target faster, but required longer durations to complete tumor cell killing (Supplementary Movie S1 and S2).

We next examined whether one subset was more potent in its overall tumor killing capacity by evaluating tumor target cell survival over the 8-hour coculture period. Overall, both T-cell subsets killed 18% of the tumor targets at an E:T ratio of 1:1 (Fig. 3C, red inner circle). However, at the 1:2 ratio, the CD8⁺BTLA⁺ subset was able to kill a total of 21% of the tumor targets upon contact as opposed to only 14% for the CD8⁺BTLA⁻ subset (Fig. 3C, red and green sections, outer circle). Second, when presented with two tumor targets, the CD8⁺BTLA⁺ subset was twice as likely to kill both targets (14% for the BTLA⁺ vs. 7% for the BTLA; Fig. 3C).

Examination of the kinetics of tumor killing offered clues to understand the differences in the overall killing potential of the two subsets. The analysis of the full 500-minute assay revealed a different killing pattern early after tumor encounter (first 250 minutes) in comparison with the second half of the incubation (last 250 minutes). A close look at the first 250 minutes demonstrated that the CD8⁺BTLA⁻ subset was more effective in tumor killing than the CD8⁺BTLA⁺ subset as the survival of the contacted tumor targets dropped significantly faster when coincubated with CD8⁺BTLA⁻ TILs at effector-to-target cell ratios of 1:1 and 1:2 (left, 1:1, P < 0.03; right, 1:2, P < 0.0001; Fig. 3D). However, the overall percentage of killed targets at both ratios at the 250 minutes mark is the same between both TIL subsets. Therefore, the killing potential of the two subsets is comparable in the first 250 minutes but their kinetics of killing are different.

Over the last 250 minutes of the assay, the picture changes (Fig. 3E). The blue curve (BTLA⁺) now dips below the red curve (BTLA⁻), signifying that the BTLA⁺ subset is catching up with the BTLA⁻ subset in terms of the speed at which it kills (E:T ratio of 1:1, P = 0.54), or takes a significant lead (E:T ratio of 1:2, P < 0.0001). It is also worth noting that very limited killing occurs with the BTLA⁻ in the second half of the experiment at any ratio while the BTLA⁺ subset achieves higher killing when more targets are present in the well (1:2). We reasoned that this situation could be explained by a differential ability for serial killing, or the ability to kill a second target, which can only be detected if there is more than one target in the well therefore at a 1:2 ratio (Fig. 3E).

After activation, T cells can be eliminated through the process of activation-induced cell death (AICD). The ability of the CD8⁺BTLA⁺ subset to kill multiple targets could be linked to an enhanced ability to survive AICD after completing its effector function. To test this hypothesis, we quantified effector T-cell death after a tumor killing event. We found that in fact, CD8⁺BTLA⁻ TILs are more susceptible to undergo apoptosis as compared with their CD8⁺BTLA⁺ TIL counterpart (P < 0.01; Fig. 3F). These data demonstrate that the CD8⁺BTLA⁺ subset survives better after killing a tumor target and thus is able to sustain its cytotoxic functionality and eliminate additional tumor cells. In summary, these results argue that both subsets kill equally well in a short-term killing assay (250 minutes) but the differences emerge when the T-cell and tumor coincubation is prolonged to 8 hours wherein the CD8⁺BTLA⁺ TIL subset is found to have an increased propensity to survive a killing event and therefore function as serial killers, thus confirming that the CD8⁺BTLA⁺ TIL subset is qualitatively better consistent with the in vivo results (Supplementary Movie S3).

Our data suggest a difference in survival of BTLA⁺ antitumor CD8⁺ T cells rather than a change in killing potential. We next sought to evaluate the long-term fate of CD8⁺BTLA⁺ antigen-specific T cells after in vivo antigenic challenge in an immunocompetent animal. Our study of the expression of BTLA in human and mouse CD8⁺ T cells showed a different expression pattern (Supplementary Figs. S2 and S3). Murine naïve CD8 (CD44 low CD62L high) are negative for BTLA, but expression of BTLA is acquired as the cells differentiate to memory (CD44 high) while human naïve CD8 cells (CCR7⁺, CD45RA⁺) are positive for BTLA and expression is progressively lost as the CD8 cells differentiate to terminal effector cells (EMRA, CCR7⁻, CD45RA⁺). However, memory stages of differentiation are positive for BTLA in both mouse and human. On the basis of this, we reasoned that a vaccination model would be the most appropriate in vivo antigen trigger to measure the role of BTLA on the function of CD8. We aimed to measure the impact of BTLA expression on newly generated memory CD8⁺ T cells on their capacity to respond to a second antigen exposure (recall). In this study, we sought to determine the role of BTLA in the persistence of antigen-specific CD8⁺ T cells in vivo after antigenic stimulation. We used TCR transgenic CD8⁺ T cells that are either WT or KO for BTLA and study their response to immunization and memory-recall response using a mouse vaccination model as described previously (28). Splenocytes (5 × 10⁵) from either Pmel-1 WT or Pmel-1 BTLA KO (recognizing gp100 peptide) were adoptively transferred into C57BL/6 recipient mice. On the following day, mice were vaccinated with gp100 peptide together with anti-mouse CD40. Imiquimod cream 5% was also applied on the vaccination site to boost the innate immune response. In addition, IL-2 was also provided to support in vivo T-cell proliferation following vaccination (Fig. 4A and B). The frequency of Pmel-1 T cells was tracked in peripheral blood after vaccination (Fig. 4C). On day 20 following priming, we observed a significantly higher frequency of Pmel-1 WT in peripheral blood when compared with Pmel-1 BTLA KO. The frequency of both Pmel-1 WT and Pmel-1 BTLA KO contracted to prepriming frequency by day 30. Boost vaccination was performed on day 60. We observed that Pmel-1 WT T cells had a robust recall response that was absent in the Pmel-1 BTLA KO group (Fig. 4C). To determine the impact on the generation of a long-term memory response, we assessed the frequency of Pmel-1 T cells in the spleen on day 120. We found a significantly higher percentage of Pmel-1 WT T cells in the spleen as compared with Pmel-1 BTLA KO T cells (Fig. 4D, P = 0.038). Similar data were obtained with the transfer of BTLA WT or KO OT-1 T cells and vaccination with OVA peptide SIINFEKL (Supplementary Fig. S4). In this context, the use of BTLA KO OT-1 T cells led to a marked attenuation of the recall response, but not the complete lack observed with the Pmel-1 T cells. Nonetheless, these results suggest that BTLA is critical for the generation of a recall memory response.

Our in vivo experiments with human TIL and single-cell killing analyses have suggested that the CD8⁺BTLA⁺ subset is endowed with a superior prosurvival function leading to better antitumor activity. Conceivably, the superiority of this subset could result from either (i) properties of less-differentiated T cells marked by BTLA expression or (ii) signaling through the BTLA molecule itself. Our vaccination study suggests a direct involvement of the BTLA molecule. However, to further test the contribution of the BTLA signaling, we performed functional studies of BTLA by overexpressing wild-type BTLA (WT BTLA) or variants of BTLA's intracellular motifs in OT-1 T cells from OT-1 BTLA KO mice. To dissect the signaling contributions from the ITIM/ITSM and the Grb2 motifs independently, retroviral constructs were generated with point mutations by substitution of tyrosine for phenylalanine in either the Grb2-binding motif (ΔGrb2) or the ITIM and ITSM motifs (ΔITSM; Fig. 5A). OT-1 BTLA KO T cells were transduced with constructs containing WT BTLA, ΔGrb2, and ΔITSM as well as an empty vector control (EM). Virally transduced OT-1 T cells were cocultured with either B16-F10 (negative for OVA antigen) or B16-OVA tumor cells at an effector-to-target ratio of 1:10, 1:3, and 1:1. HVEM expression on both tumor targets was confirmed by flow cytometry (Supplementary Fig. S5). Indeed, the tumor killing capacity of the TILs in this short-term in vitro assay was comparable regardless of the presence of WT BTLA or BTLA mutants at all ratios tested (ratio 1:1, P = 0.67; ratio 1:3, P = 0.46; ratio 1:10, P = 0.29; Fig. 5B). We did not observe differences in the number of T cells producing IFNγ or TNFα among these different groups (IFNγ, P = 0.11; TNFα, P = 0.59; Fig. 5C, left). However, the amount of effector cytokines being made by the WT BTLA-transduced T cells was significantly less than that made by those transduced with the empty vector alone following restimulation with dendritic cells pulsed with OVA peptide (IFNγ MFI; EM vs. WT, P = 0.04, TNFα MFI; EM vs. WT, P = 0.0045; Fig. 5C, right). The BTLA-mediated inhibition of IFNγ and TNFα production was relieved by the disruption of the ITIM/ITSM motifs; however, it was not significantly impacted by the disruption of the Grb2 motif, suggesting that ITIM and ITSM motifs are mainly involved in regulating the amount of effector cytokines produced following TCR triggering.

We have previously demonstrated that the CD8⁺BTLA⁺TIL subset had an improved proliferative capacity in response to IL-2 (14). Thus, we sought to determine whether BTLA signaling motifs could play a role in T-cell proliferation. OT-1 BTLA KO T cells overexpressing BTLA WT or mutants were labeled with the cell proliferation dye eFluor670 and restimulated with dendritic cells pulsed with OVA peptide for two days. T-cell proliferation was comparable between control and WT BTLA groups. (MFI: WT vs. EM; P = 0.30; Fig. 5D, histogram). On the other hand, subtle differences are appreciable between T cells expressing the different BTLA constructs. We observed that OT-1 BTLA KO T cells expressing ΔITSM had a lower mean fluorescence intensity (MFI) of the eFluor670 dye as compared with the empty vector control which is indicative of a more robust proliferation (MFI: ΔITSM vs. EM; P = 0.012, ΔITSM vs. WT; P = 0.0057; Fig. 5D). On the contrary, attenuation of T-cell proliferation was observed in T cells expressing a disrupted Grb2 motif (ΔGrb2), (MFI: ΔGrb2 vs. ΔITSM; P = 0.0003, ΔGrb2 vs. EM; P = 0.0008, ΔGrb2 vs. WT; P = 0.0009; Fig. 5D). These data suggest that the T-cell proliferation post TCR triggering is regulated positively by the Grb2 motif and negatively by ITIM/ITSM motifs.

Our previous report demonstrated that CD8⁺BTLA⁺ human TILs produce more IL-2 upon activation (14). Thus, we further investigated whether BTLA signaling could be responsible for IL-2 secretion. OT-1 BTLA KO T cells overexpressing WT BTLA or its mutants were restimulated with anti-CD3 in the presence of HVEM-Fc fusion protein to engage the BTLA molecule, and IL-2 secretion was assessed. Indeed, we found that IL-2 production was significantly increased in activated T cells transduced with the BTLA WT and ΔITSM constructs, but not in ΔGrb2 and the empty vector. This suggests that the BTLA-dependent IL-2 production was mediated through Grb2 motif, independently of the ITIM/ITSM motifs (WT vs. EM; P = 0.02, ΔITSM vs. EM; P = 0.005, WT vs. ΔGrb2; P = 0.02, ΔITSM vs. ΔGrb2; P = 0.01, EM vs. Grb2; P = 0.79, WT vs. ΔITSM; P = 0.32; Fig. 5E). These data highlight that Grb2 signaling contributes to IL-2 production following BTLA ligation during T-cell activation. A summary of the phenotypes seen with the expression of BTLA or its variants is presented in Fig. 5A (right).

To identify proteins activated downstream of BTLA, we employed RPPA, a high-throughput method developed for functional proteomic studies. OT-1 BTLA KO mouse T cells overexpressing WT BTLA or BTLA mutants were stimulated with anti-CD3 with or without HVEM-FC for 8 hours. The full list of tested proteins can be found in Supplementary Table S3. The protein expression levels found in activated T cells transduced with BTLA WT was used as a reference to compare the differential protein changes in ΔGrb2 and ΔITSM. The protein expression profile of the T cells expressing the BTLA variants were also compared with that obtained with the empty vector (Supplementary Table S4). We observed that the phosphorylation of Akt at T308 as well as the phosphorylation of one of its substrates, pPRAS40 at T246, were significantly attenuated in ΔGrb2 in comparison with BTLA WT (pAkt T308; P = 0.045, pPRAS40 T246; P = 0.007; Fig. 6A, left; Supplementary Table S4) (29). As expected, the disruption of ITIM/ITSM motif resulted in a remarkable enhancement of the phosphorylation of several proteins such as Src, Chk1, Chk2, as well as members of the Wnt pathway GSK-3b, and of β-Catenin at T41 and S45 (pSrc at S527; P = 0.005, pA-Raf at S299; P = 0.01, and pC-Raf at S338; P = 0.02, pChk1 at S286; P = 0.01, pChk2 at T68; P = 0.02, GSK-3b at S9; P = 0.04, β-Catenin at T41 and S45; P = 0.04; Fig. 6A, right; Supplementary Table S4).

To exclude the bias of overexpression of WT BTLA and genetically modified BTLA in BTLA knockout mouse, we further investigated BTLA signaling pathway in human TILs, which is more physiologically and clinically relevant to our observation in our adoptive T-cell therapy clinical trial. Sorted CD8⁺BTLA⁺ human TILs were stimulated with plate-bound anti-human CD3 at concentrations of 10, 30, 100, 300, and 1,000 ng/mL with or without HVEM-Fc for 8 hours and proteins extracted from cell lysates were used to perform RPPA in 5 separate patient TIL samples. We observed a general decrease in differentially phosphorylated protein changes in several signaling pathways when T-cell activation happened in the presence of HVEM. These included the MAPK kinase pathway (pP38 at T180; P = 2.08 × 10⁻⁹, pP90RSK at T573; P = 3.26 × 10⁻⁴, pS6 at S235; P = 1.68 × 10⁻²), NFκB pathway (pNFκB p65 at S536; P = 3.15 × 10⁻²), the Akt pathway (pAKT at S473; P = 9.6 × 10⁻³), and the β-catenin pathway (GSK-3a-b pS21; P = 9.6 × 10⁻³; Fig. 6B and C; Supplementary Table S5). Inhibition of these positive signaling pathways clearly supports the role of BTLA as a coinhibitory molecule. Interestingly, we found a significant elevation of Src phosphorylation at S416 (pSrc at S416; P = 3.36 × 10⁻⁶) when the T-cell activation happened in the presence of HVEM. Our results suggest that the TCR signaling pathway is not completely suppressed by BTLA, but is specifically attenuated in certain pathways as indicated above and specifically potentiated in very select pathway(s). Unexpectedly, we found that phosphorylation of HER2 at tyrosine 1248 was prominently increased in an anti-CD3 dose-dependent manner regardless of the presence of HVEM. This strengthens the notion that the BTLA/HVEM axis acts on specific targets. To further comprehensively understand the downstream signaling pathway of human BTLA, we generated the signaling network of proteins significantly changed during HVEM ligation using Ingenuity Pathway Analysis (IPA). We observed that Src signaling node was exclusively clustered and separated from Akt, NFκB, and β-catenin signaling nodes. This suggests that the Src signaling pathway is likely not or minimally interfered by SHP1/2 (Fig. 6D).

Our recent report demonstrated that the CD8⁺BTLA⁺ TILs are less differentiated, respond better to IL-2, and persist longer in the patient postinfusion (14). In the current study, we shed light on the dichotomy of the signaling downstream of BTLA and highlight a role for BTLA in the development of CD8⁺ T-cell memory-recall and the survival of effector memory CD8⁺ T cells sustaining their antitumor function.

We observed that the CD8⁺ BTLA⁺ and CD8⁺ BTLA⁻ subsets had comparable in vitro killing, which did not explain the superiority of the CD8⁺ BTLA⁺ subset in controlling disease in patients. The use of a single-cell nanowell-based cytolytic assay helped elucidate fundamental differences between the two subsets. Importantly, the results demonstrated that both subsets efficiently kill a first target, but the CD8⁺BTLA⁺ subset is more likely to survive after this killing event and kill another target, thus providing a rationale for the improved in vivo efficacy of the CD8⁺BTLA⁺ TILs.

The defective recall response seen in BTLA KO antigen-specific T-cell in vivo models used contrasts with previous literature. A publication by Krieg and colleagues evaluates the response of the CD8⁺ BTLA-KO OT-1 T cells in a vaccination model and claims that there is more efficient memory formation in the BTLA-KO OT-1 T cells than their WT counterpart (30). This claim is based on the fact that there are twice as many CD8⁺ OT-1 BTLA-KO T cells than WT counterpart in the spleen 30 days after the primary stimulation even though the BTLA-KO OT-1 subset peaked at lower frequency in the early days of the primary response. The BTLA-KO OT-1 also had a slightly lower fold expansion early (day 4) during the secondary expansion (7-fold vs. 10-fold) following boosting with peptide-pulsed DC. This is in contrast with the data obtained when boosting with an infectious agent expressing OVA, LM-OVA, where the BTLA-KO OT-1 expand more than the BTLA WT OT-1 at day 5 (91-fold vs. 191 fold). Unfortunately the experiment was not carried out longer. The authors conclude that BTLA KO OT-1 can form memory, and that the amplitude of the recall response may depend on the type of antigenic challenge used. Interestingly, another study utilizing BTLA KO animals and LM-OVA infection concluded that HVEM expression on T cells transduces a survival signal through interaction with BTLA expressed on other cells in the host and that interaction was required for CD8⁺ T-cell memory generation (31).

There are a few key differences between our vaccination model and the reported studies. First of all, the adoptive cell transfer in the article by Krieg and colleagues is performed to allow competition between the cell subsets in the host. The authors actually transfer both BTLA WT OT-1 and BTLA KO OT-1 T cells at a 1:1 ratio to BTLA WT hosts and vaccinate the animals one day later with peptide-pulsed dendritic cells. While these experiments were designed to test BTLA's function, the known expression of BTLA's receptor, HVEM, on activated T cells, and its ability to transduce a survival signal by interactions in trans may become a confounding factor. In this system, it is possible that BTLA expressed on BTLA WT OT-1 may have interacted with HVEM on both transferred subsets which could have delivered a survival signal potent enough to limit the contraction phase and enhance memory formation for both subsets. The amplitude of the memory generation in this system could be controlled by the strength of TCR signal which clearly will be stronger in BTLA KO OT-1 as all studies are in agreement that BTLA suppresses TCR signaling.

Our experiments were designed to study the response of the two subsets separately, in different animals. Our data rather shows BTLA-mediated survival of T cells following effector function, probably due to dampening of TCR signaling and prevention of AICD. We believe that BTLA acts as a rheostat reducing TCR signaling during strong or chronic antigenic stimulation while also promoting cell survival due to the presence of both activating and inhibitory signaling motifs in its intracellular signaling domain. Our vaccination was administered by injection of the cognate peptide and anti-CD40 in saline subcutaneously, with topical application of the TLR 7 agonist imiquimod on the injection site and IP IL-2. This vaccination strategy, optimized by our collaborator Drs. Hailemichael and Overwijk, results in a powerful activation of the antigen-specific T-cell response (28). This stimulation produces a robust recall response in WT OT-1 and Pmel T cells; however, in our hands the BTLA KO OT-1 cells have a reduced ability to generate a recall response and Pmel BTLA KO cannot recall at all. We reasoned that BTLA expression may be essential to suppress overstimulation conducive to AICD and allow the survival of cells to form a memory pool. The method employed to vaccinate in the study by Krieg and colleagues, the use of peptide-pulsed DCs, may also not cause as much AICD as the stimulation used in our system, thus not requiring dampening of TCR stimulation by BTLA for survival.

We have demonstrated previously that resting CD8⁺BTLA⁺TIL manifested enhanced mitochondrial function and spare respiratory capacity (SRC) as compared with their CD8⁺BTLA⁻ TIL counterpart (32). High SRC has been reported to be a feature of memory CD8⁺ T cells distinguishing them from naïve and effector subsets, and was found to play a role in survival of the memory CD8⁺ subset conducive to the establishment of a long-lived memory pool (33). Our findings of the better survival of the CD8⁺BTLA⁺ TIL subset after a killing event are consistent with a prosurvival advantage. A recent study indicated that increased SRC was strongly associated with a superior bioenergetic capacity, which helped improve T-cell survival and motility under hypoxic conditions (34). The intrinsic properties of this subset provide an intriguing possibility for why infusion of large numbers of CD8⁺BTLA⁺TILs positively correlates with clinical response to TIL ACT in metastatic melanoma patients.

We have found that a BTLA-dependent potentiation of IL-2 secretion entirely depends on Grb2 motif. Disruption of the ITIM and ITSM motifs did not alter the levels of IL-2 induced by BTLA, suggesting a signaling pathway independent of the proinhibitory influence of ITIM and ITSM. Our findings are consistent with previous reports that have shown that Grb2-linked SLP-76 and Vav interaction are involved in IL-2 production, and that recruitment of Grb2 is essential for CD28-induced IL-2 production (19, 35, 36). Our current study reveals that BTLA engagement in the context of TCR stimulation results in IL-2 production in a manner similar to CD28 which positions BTLA as both a coinhibitory and costimulatory molecule. The in vivo relevance of BTLA-dependent IL-2 production in the context of T-cell activation warrants further study. Autocrine IL-2 production by CD8⁺ T cells has been shown to be essential for secondary expansion of CD8⁺ memory T cells which could potentially explain the lack of secondary expansion seen in BTLA KO cells after boosting with the vaccinating antigen (37). The prosurvival properties conferred by BTLA on CD8⁺ T cells may come from the interaction of BTLA with HVEM on activated T cells as it was demonstrated for CD4⁺ murine T cells (38, 39). However, the HVEM interaction with its alternative ligand LIGHT rather than BTLA has been implicated in furthering CD4⁺ T-cell activation and survival (39). In our experiments, the absence of BTLA may render HVEM available to bind to its alternative ligand LIGHT to provide costimulatory signaling to T cells; however, this mechanism was not sufficient to rescue the establishment of memory cells able to respond to a recall antigen stimulation.

Conversely, we did not observe any major differences in in vitro killing capacity and cytotoxic cytokine production between BTLA KO T cells versus BTLA KO T cells overexpressing BTLA WT or its mutants. This suggests that BTLA signaling does not affect tumor killing capability. Our findings are consistent with the previous report showing that the BTLA blockade in γδT cells had no effect on the tumor lysis (40).

We did not observe a significant impact of WT BTLA when the molecule was reintroduced in BTLA KO T cells. However, malfunction of ITIM and ITSM motifs significantly enhanced T cell proliferation while Grb2 alteration reduced the proliferation. A lack of impact of BTLA on CD4⁺ T-cell proliferative response has been reported previously (12). Previous studies in both human and mouse settings indicated that CD8⁺ T cells were intrinsically less susceptible to BTLA-mediated inhibition as compared with CD4⁺ T cells (9, 41). This might explain the minimal inhibition of WT as compared with EM in CD8⁺ T cells in terms of proliferation.

We noticed that the Akt signaling pathway was remarkably attenuated in BTLA ΔGrb2, while the phosphorylation of Src was enhanced in BTLA ΔITSM. This finding is consistent with a previous report demonstrating that ITIM and ITSM motifs of PD-1 inhibited signals through Akt and MAPK (10). However, in the presence of costimulation mediated through the BTLA Grb2 motif, as in the case of ΔITIM/ITSM, the phosphorylation of Akt was unaltered, suggesting that Grb2 is not directly involved in maintaining or augmenting Akt phosphorylation. Taken together, these data suggests that the Akt pathway in BTLA was likely targeted by SHP1/2, similar to what was found in PD-1.

Overexpression of Grb2 in osteoclasts was shown to promote phosphorylation of Src at Y416 and the opposite result was obtained when Grb2 expression was disrupted (42). From our data, it is clear that the phosphorylation of Src increased when Grb2 was providing BTLA signaling in the absence of functional ITIM and ITSM motifs. This could result from either the direct effect of Grb2 signal transduction or the lack of dephosphorylation normally caused by SHP1/2 via ITIM and ITSM motifs. As the phosphorylation of Src is not significantly influenced by the ITIM/ITSM cosignaling (ΔGrb2), we conclude that it is likely to be Grb2-mediated Src activation.

Our data of BTLA signaling in mouse T cells was found to be consistent with changes observed when activated human CD8⁺BTLA⁺TILs were costimulated by HVEM. Indeed, we observed an attenuation of MAPK, NFκB, and Akt signaling proteins, but phosphorylation of Src was enhanced when human TILs were stimulated with anti-CD3 and HVEM as compared with CD3 alone. This suggested that the downstream signaling pathway of BTLA in mice and humans possibly share similar downstream signaling targets.

Overall, our study sheds light on a dual role of BTLA as both a costimulatory and coinhibitory molecule. This observation is supported by a study reporting that the gene expression profile of CD4⁺ T cells activated with CD3 and BTLA engagement mimics the profile induced by costimulatory molecules (12). In our hands, the integration of the positive and negative signals transduced by BTLA promotes IL-2 secretion while reducing effector cytokine production and proliferation in certain contexts. In addition, the inherent properties of the less-differentiated T cells expressing BTLA also display enhanced resistance to apoptosis and an efficient bioenergetic profile providing a survival advantage following tumor killing. These findings support the concept that the intrinsic attribute of the less-differentiated CD8⁺BTLA⁺ human TIL subset together with balanced signals transduced by the engagement of HVEM in the tumor microenvironment on melanoma cells could provide a costimulatory signal to CD8⁺BTLA⁺ TILs promoting IL-2 secretion, T-cell survival, and antitumor function.

No potential conflicts of interest were disclosed.

Conception and design: K. Ritthipichai, C. Haymaker, L.G. Radvanyi, P. Hwu, C. Bernatchez

Development of methodology: K. Ritthipichai, L.G. Radvanyi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Ritthipichai, W.W. Overwijk, L. Vence, N. Varadarajan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Ritthipichai, C. Haymaker, M. Martinez-Paniagua, A. Aschenbrenner, W.W. Overwijk, L. Vence, J. Roszik, N. Varadarajan, L.G. Radvanyi, P. Hwu, C. Bernatchez

Writing, review, and/or revision of the manuscript: K. Ritthipichai, C. Haymaker, M. Martinez-Paniagua, A. Aschenbrenner, W.W. Overwijk, J. Roszik, N. Varadarajan, P. Hwu, C. Bernatchez

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Yi, M. Zhang, C. Kale, Y. Hailemichael

Study supervision: R. Nurieva, L.G. Radvanyi, C. Bernatchez

The authors thank the melanoma surgical staff at MD Anderson Cancer Center. We appreciate help from TIL laboratory for processing tumor samples. We are thankful for the staff at the MD Anderson Cancer Center Flow Cytometry Core Facility for technical assistance with cell sorting. We are grateful to Drs. Weiyi Peng and Ondrej Havranek for guidance on experimental design.

This work was supported by the National Cancer Institute (NCI) grants R01-CA111999 (to P. Hwu), R01CA174385 (to N. Varadarajan) and 1R21CA178580-01 (to C. Bernatchez). STR DNA fingerprinting was done by the CCSG-funded Characterized Cell Line Core at MD Anderson, NCI grant P30CA016672. This grant also supported the the Flow Cytometry Core Facility that was used for cell sorting and the Functional Proteomics Reverse Phase Protein Array (RPPA) Core. This work was further supported by the UT Health Innovation for Cancer Prevention Research Pre-doctoral Fellow, The UT School of Public Health-Cancer Prevention and Research Institute of Texas grant RP160015 (to K. Ritthipichai and A. Aschenbrenner), RP130570 to (N. Varadarajan), RP140522 (to W.W. Overwijk), NIH grant A1R03AI120027 and 1R21AI20012 (to R. Nurieva) and the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation (AMRF), Melanoma Research Alliance Young Investigator Award (509800; to N. Varadarajan), and the Mulva Foundation.

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.