PD-L1–Independent Mechanisms Control the Resistance of Melanoma to CD4⁺ T Cell Adoptive Immunotherapy

Immunotherapy of cancer with tumor-specific T cells is becoming more and more a reality (1–3). However, there are multiple T cell–intrinsic and T cell–extrinsic mechanisms to dampen and suppress antitumor T cell reactivity, which can lead to primary or secondary resistance to immunotherapy (4). Most notably, intrinsic mechanisms, such as checkpoint molecules, have gained a considerable amount of attention because of their dramatic effects on antitumor immunity in cancer patients. These include the Food and Drug Administration–approved checkpoint blockade molecules: Anti–programmed cell death 1 (PD-1) (nivolumab, pembrolizumab), anti–programmed cell death 1 ligand (PD-L1) (atezolizumab), and anti–CTL-associated protein 4 (CTLA-4) (ipilimumab). In addition to the PD-1 pathway, lymphocyte-activated gene 3 (LAG-3) and Tim-3 have been targeted for their roles in chronic viral infections in preclinical models (5–11). Furthermore, interfering with inhibitory pathways PD-1 and CTLA-4 (12), PD-1 and Tim-3 (13–15), and PD-1 or PD-L1 and LAG-3 (16, 17) in mice and PD-1 (18, 19), PD-L1 (20), and CLTA-4 (21) in humans with cancer has led to improved responses and increased overall survival. Anti–LAG-3, LAG-3 fusion proteins, and anti–TIM-3 Abs are currently in human clinical trials in different combinations (www.clinicaltrials.gov).

Extrinsic mechanisms of resistance to immunotherapy include regulatory T (T_reg) cells, IDO, myeloid-derived suppressor cells, and tumor-associated macrophages(22). T_reg cells suppress T cells in multiple ways (23), and depletion of them during cancer can enhance tumor therapies in preclinical models of cancer (17, 24–26), although it may also lead to increased autoimmunity. To date, progress of depleting or reprogramming T_reg cells safely in humans has been mildly successful with daclizumab (27). Thus, new ways of overcoming immunosuppression that do not enhance autoimmunity and toxicities are warranted.

Combination immunotherapy with anti–PD-1 (nivolumab) and CTLA-4 (ipilimumab) is becoming the standard of care for melanoma, resulting in deep tumor regressions and a 56% response rate (28). Yet it is very toxic and is associated with an increased incidence in immune related adverse events (28). It is also very expensive, and finding biomarkers to predict efficacy is ongoing in research (29, 30). One such effort is the correlation of expression of PD-L1 in the tumor and response rates. It has been suggested that increases in PD-L1 expression in the tumor microenvironment can predict the responses of patients receiving anti–PD-1 therapies (18). However, the expression of PD-L1 in tumors does not always correlate with responses (19, 20). Subgroup analyses of patients benefiting from anti–PD-L1 therapy, for instance, showed improvement regardless of PD-L1 expression levels (31); others showed more benefit with PD-L1 expression in another trial (20). Also, only ∼25–30% of melanoma patients respond to anti–PD-1 checkpoint therapy (30), whereas the rest succumb to primary resistance. So, the role of PD-L1 status in melanoma is still unclear, as is which checkpoints are most beneficial to patients to overcome these resistance mechanisms.

Previously, we have shown dramatic responses in treating recurring melanoma by combining melanoma-specific CD4⁺ T cell (tyrosinase-related protein 1 [TRP-1] T cells) adoptive cellular therapy (ACT) with immune checkpoint blockade. These studies involved the administration of anti–PD-L1 Abs with T_reg cell depletion, or a combination of anti–PD-L1 and LAG-3 Abs (17). However, in this particular setting, the host immune system was absent, and we could not study the mechanism of action with host immune system influences. Therefore, we set out to determine the role of host PD-L1 on the initiation and maintenance of exhaustion on tumor-specific CD4⁺ T cells in a clinically relevant model of resistance to immunotherapy (e.g., immunocompetent mice). In addition, we were interested in determining if T_reg cell depletion could be circumvented with combination immune checkpoint blockade and irradiation. We found that anti–LAG-3 was better at controlling resistance to immunotherapy and recurrence of melanoma than anti–PD-L1 when combined with irradiation and T cell–adoptive cell transfer.

Tyrp1^B-wRAG^−/− TRP-1–specific CD4⁺ TCR transgenic mice (B6.Cg-Rag1^tm1Mom Tyrp1^B-w Tg[Tcra,Tcrb]9Rest/J) were described previously (32). Foxp3 diphtheria toxin (DT) receptor wild-type (WT) mice were provided by A. Rudensky (Memorial Sloan Kettering Cancer Center, New York, NY) (33). PD-L1^−/− (B7-H1) mice were a gift from Koji Tamada (University of Maryland School of Medicine). Foxp3-DTR mice were crossed with transgenic TRP-1 mice to create tyrp1^B-wRAG^−/− Foxp3-DTR TRP-1–specific CD4⁺ TCR transgenic mice. RAG1^−/− (Rag1^tm1Mom) mice were purchased from The Jackson Laboratory. PD-L1^−/− WT mice were crossed with RAG1^−/− mice to generate PD-L1^−/−RAG^−/−mice. All mice were used in accordance with guidelines from the University of Maryland Institutional Animal Care Committee. Mice received 550 rad [5.5 Gy] of total body irradiation using a Cesium irradiator as previously described (34).

B16.F10 (H-2^b), hereafter called B16, is a TRP-1⁺ spontaneous murine melanoma that was obtained from the American Type Culture Collection and maintained in culture media, as described previously (32). Tumors were injected s.c. at 3 × 10⁵ cells per mouse. Tumors were measured blindly with digital calipers. The perpendicular diameters were determined and multiplied to generate the area (in square millimeters), as described previously (32).

TRP-1 Foxp3-DTR CD4⁺ T cells were sorted from spleens of donor Tyrp1^B-wRAG-1^−/− Foxp3-DTR TRP-1–specific transgenic male mice. Spleens were harvested and made into single-cell suspensions. Cells were made devoid of RBCs by using ACK lysis buffer. Subsequently, cells were counted and enriched for CD4⁺ T cells by magnetic bead sorting using a CD4⁺ T cell enrichment kit (Miltenyi Biotec). Enriched CD4⁺ T cells were counted and resuspended in PBS and used in adoptive-transfer studies (2 × 10⁵ cells per mouse). T cells were injected i.v. via the tail vein. For early dysfunctional TRP-1 Foxp3-DTR CD4⁺ T cells, tyrp1^B-wRAG^−/− Foxp3-DTR TRP-1–specific CD4⁺ TCR transgenic mice were given B16.F10 tumor at 3 × 10⁶ cells, and T cells were harvested from the spleen, as described above, 7 d after tumor inoculation, or when tumors reached >225 mm². Fully dysfunctional T cells were isolated from tumor-bearing RAG^−/− mice that had recurring tumors after successful therapy with TRP-1 T cells.

Anti-CD4 (RM4-5) was obtained from BD Biosciences. Anti–PD-1 (RMP1-30) was obtained from eBioscience. Anti–LAG-3 (C9B7W) and anti-TIGIT (1G9) were obtained from BioLegend. All flow cytometry scales are log scales, if not otherwise specified. All samples were run on a BD Biosciences FACSCalibur (Department of Surgery, University of Maryland School of Medicine) and analyzed by FlowJo software (Tree Star).

Anti–PD-L1 (10F.9G2) and anti–LAG-3 (C9B7W) were purchased from Bio X Cell (West Lebanon, NH). Anti–PD-L1 (200 μg/injection) was given at days 7, 9, and 11 after tumor inoculation. Anti–LAG-3 (200 μg/injection) was given at days 7, 9, and 11 after tumor inoculation. DT was purchased from Sigma-Aldrich and was reconstituted according to the manufacturer’s protocol. Frozen DT stocks were thawed once, and 50 mg/kg DT was injected i.p., as described previously (33).

A Student unpaired t test was used to compare the differences between cell numbers. Tumor curves were compared using a two-way ANOVA, with nonrepeated measures. Survival curves were compared using a Log-rank (Mantel–Cox) Test. A p value ≤ 0.05 was considered significant. All statistical analysis was performed on Prism 5.0f software (GraphPad).

We previously established the PD-1/PD-L1 pathway to be an important factor in the exhaustion of melanoma-specific CD4⁺ T cells (TRP-1 T cells) in lymphopenic mice. Transient lymphopenia is required before transfer of T cells for optimal function and persistence in mice and in humans with cancer (32, 35, 36). To remove the effects of the host, we previously used RAG^−/− mice, which are chronically lymphopenic (17). In this setting, by targeting the PD-1/PD-L1 pathway early in the immune response, partially dysfunctional TRP-1–specific CD4⁺ T cells mediated durable regression of established melanoma (17). In these experiments, partially dysfunctional (previously termed tumor-sensitized) TRP-1 T cells are defined as TRP-1 CD4⁺ T cells that have been isolated from B16 tumor–bearing TRP-1 CD4⁺ transgenic mice after tumors have become apparent (∼10 d). These T cells are considered partially exhausted (expressing low levels of PD-1, LAG-3, and TIM-3) and contain a small (<5%) population of Foxp3⁺ TRP-1–specific T_reg cells (32). However, targeting the PD-1/PD-L1 pathway late, (e.g., recurrence of melanoma) required combination immunotherapy with anti–PD-L1 and depletion of TRP-1 T_reg cells to work (17).

Therefore, we set out to determine a similar role for the PD-1 pathway in a clinically relevant model of immunotherapy. Instead of immune-incompetent lymphopenic RAG^−/− hosts, immunocompetent WT and PD-L1^−/− mice were used. Irradiated WT and PD-L1^−/− mice received naive TRP-1 T cells and were monitored for tumor growth. There was an initial response to immunotherapy followed by recurrence in all mice, irrespective of host PD-L1 expression (Fig. 1). Although there was a slight delay in tumor growth in PD-L1^−/− mice, this was not significant when compared with treatment in WT mice: 0 of 5 mice made it to the benchmark, day 40, in WT mice treated with TRP-1 T cells and radiation (vertical dotted line), and 2 of 5 made it to day 40 in PD-L1^−/− mice treated with the same regimen.

Because other cells, such as adoptively transferred T cells and tumor cells, can express PD-L1, we wanted to tease apart the role of host/tumor PD-L1 without the added effects of various contributors in the WT setting (i.e., extrinsic suppressive factors). To this end, we generated PD-L1^−/− mice on the RAG^−/− background. Initial experiments showed that naive TRP-1 T cells could completely treat tumors in PD-L1^−/−RAG^−/− mice (Fig. 2A). Because naive TRP-1 T cells contain very few T_reg cells and are not activated (32), we compared RAG^−/− and PD-L1^−/−RAG^−/− mice receiving early dysfunctional TRP-1–specific CD4⁺ T cells. When these CD4⁺ T cells are adoptively transferred into tumor-bearing RAG^−/−mice, tumors consistently relapsed by 21 d after adoptive transfer (Fig. 2B). Transfer of early dysfunctional CD4⁺ T cells into tumor-bearing PD-L1^−/−RAG^−/− mice, however, led to durable regression of established B16 tumors, although tumor destruction was not as deep as for naive T cells (Fig. 2A, 2B). These data illustrate an early role of PD-1 on TRP-1 T cell exhaustion mediated by host PD-L1.

To understand the role of PD-L1 on the host and tumor, we compared T cell numbers, indicative of T cell expansion in vivo, in hosts with and without PD-L1 and with and without tumor. We found that PD-L1 in the host had the greatest effect on T cell expansion, as TRP-1 T cells in PD-L1^−/−RAG^−/− mice with tumor demonstrated larger expansion than RAG^−/− hosts with tumor (Fig. 2C). However, the same results were obtained even without tumor, suggesting a minimal role for PD-L1 on tumor cells themselves (Fig. 2C).

We next investigated the ability of late dysfunctional (fully exhausted) CD4⁺ T cells to be functionally restored in PD-L1^−/−RAG^−/− mice. These cells were obtained from tumor-bearing mice with recurring tumors measuring over 15 × 15 mm². Expression of PD-1, LAG-3, and TIGIT was elevated on these cells (Fig. 3A). In addition, >40% of the tumor-specific CD4⁺ T cell population was T_reg cells (Fig. 3B). The function of these exhausted CD4⁺ T cells is not restored, even in the absence of host PD-L1. Despite a slight growth inhibition in PD-L1^−/−RAG^−/− mice, compared with RAG^−/− mice, there was no regression of primary B16 tumors when using fully dysfunctional TRP-1 CD4⁺ T cells (Fig. 3C, 3D). Together, these results suggest nonredundant pathways of suppression need to be overcome in tumor-bearing hosts with combination therapy.

Previously we showed that depletion of T_reg cells was required to enhance immunotherapy to recurring tumors during checkpoint blockade with anti–PD-L1 (17). Therefore, we could not rule out the impact TRP-1 T_reg cells have on maintaining the unresponsive state of exhausted TRP-1 CD4⁺ T cells. Our transgenic system provides a means of eliminating TRP-1–specific T_reg cells with the administration of DT. This modality was applied to the setting of adoptively transferred exhausted CD4⁺ T cells. Without the depletion of T_reg cells in RAG^−/− hosts, B16 tumors grew relatively unimpeded, whereas tumor growth was slightly diminished when DT was given (Fig. 3C, 3E). We observed modest tumor regressions in PD-L1^−/−RAG^−/− hosts receiving the exhausted CD4⁺ T cells, with all mice relapsing by day 30–35 after ACT (Fig. 3D). The depletion of TRP-1 T_reg cells from PD-L1^−/−RAG^−/− hosts, however, demonstrated a dramatic tumor response, in which tumors regressed and did not begin to relapse until 50–55 d after adoptive transfer (Fig. 3F). Although 90% of PD-L1^−/−RAG^−/− mice treated with exhausted CD4⁺ T cells and DT were relapsing by this time, none of these mice were relapsing by the benchmark, day 40 after tumor inoculation, compared with >80% PD-L1^−/−RAG^−/− mice treated with exhausted CD4⁺ T cells only and 100% of both groups of RAG^−/− mice (Table I). These results are consistent with our published findings that combination anti–PD-L1 Ab and T_reg cell depletion enhance immunotherapy during recurrence (17) and that other pathways of resistance are present. The clinical feasibility of this approach, however, severely limits its applicability (37).

We have demonstrated previously that dual anti–PD-L1 and anti–LAG-3 Ab treatment restored the functional state of exhausted CD4⁺ T cells in lymphopenic RAG^−/− mice that had relapsing melanoma (17). However, because lymphopenic RAG^−/− mice do not contain a competent immune system, it was not possible to study the effects of extrinsic factors. Inhibitory receptor blockade therapy was therefore employed in irradiated immunocompetent WT mice. Abs were administered at the time of adoptive transfer of TRP-1 T cells and two subsequent times, for a total of three injections. When Ab treatment was given in the absence of cellular therapy, anti–PD-L1 and anti–LAG-3, singly or in combination with irradiation, were ineffective in mounting an immune response (Fig. 4A–C). Naive TRP-1 T cells provided a transient antitumor response in irradiated WT mice (Fig. 4D). Without blocking Ab intervention, all mice were progressing 40 d after tumor inoculation. The administration of anti–PD-L1, in conjunction with CD4⁺ T cell therapy, unexpectedly provided a modest survival benefit (Fig. 4E), resulting in ∼66% of mice progressing by day 40 (Table II). Combination therapy with anti–PD-L1, anti–LAG-3, and irradiation caused dramatic responses (Fig. 4F).

Unexpectedly, the control group, anti–LAG-3 without anti–PD-L1, showed a dramatic reduction in the number of mice with recurring tumors (Fig. 4G), showing that anti–PD-L1 in the combination was not noticeably contributing to the enhanced immunotherapy as predicted. In some experiments, no mice had a recurrence of tumors. In total, only 13% of WT mice receiving naive CD4⁺ T cells and anti–LAG-3 Abs had progressing tumors by the benchmark, day 40, after tumor inoculation (Table II). An analysis into the superiority of anti–LAG-3 showed an increase in TRP-1 T cell numbers at 1 wk after ACT (Fig. 4H) but no other differences in checkpoint molecule expression or T_reg cell numbers (data not shown). Finally, mice receiving LAG-3 inhibitory receptor blockade in combination with irradiation and T cells survived longer when compared with mice receiving no Ab or anti–PD-L1 (Fig. 4I, p value <0.02). Although combination with anti–PD-L1 and LAG-3 was slightly better than anti–LAG-3 alone (recurrence at day 40: 6% versus 13%; recurrence overall: 11% versus 20%, Table II), the dramatic effects seen with anti–LAG-3 clearly show that the majority of the antitumor effects to be from anti–LAG-3 itself and not anti–PD-L1. Therefore, anti–LAG-3 checkpoint therapy may be more beneficial than anti–PD-L1 in combating melanoma recurrence when combined with adoptive T cell therapy and irradiation.

As T cells are engaged by chronic stimulation of Ag, PD-L1 acts on PD-1 expressed by the T cells to decrease IL-2 and IFN-γ production, decrease proliferation, and, ultimately, lead to cell death. Interfering with this pathway at the onset of exhaustion can prevent or even reverse an exhaustive state. We have established a role for host PD-L1 as an early initiator of resistance to immunotherapy and exhaustion of TRP-1–specific CD4⁺ T cells in lymphopenic mice. However, in fully competent mice, the PD-1/PD-L1 pathway is not as dominant in inhibiting CD4⁺ T cells. This is not unexpected, as there are multiple layers of resistance to immunotherapy (4).

Previously, we reported a role for the PD-1/PD-L1 pathway in suppressing antitumor CD4⁺ T cells (17). Considering our previous results were generated in lymphopenic (e.g., RAG^−/−) mice, the effect of PD-L1 on endogenous T or B cells was not a factor, nor was the role of endogenous T_reg cells. Others have shown an immunosuppressive role for PD-L1 expressed on dendritic cells (38, 39), macrophages (40), and other T cells (41). Furthermore, Mueller et al. (42) showed, by way of bone marrow chimeras, that PD-L1 expressed on nonhematopoietic cells could also inhibit T cell immunity in a model of chronic viral infection. Additionally, the presence of PD-L1 on tumor cells can impact the antitumor response through adaptive resistance (40, 43–45). Thus, multiple cells can contribute to PD-L1 immunosuppression. We show, in this study, that it contributes minimally on tumor cells when compared with the host itself. Although the expression of PD-L1 on B16 cells [as shown in reference (39)] could be an active participant in suppressing T cell responses through adaptive resistance mechanisms (18, 46), it is not the dominant pathway. Juneja et al. (45). also demonstrated that PD-L1 expression on B16 melanoma, as opposed to MC38, was not a factor in inhibiting T cell–mediated antitumor immunity. We took this further by showing that other nonredundant immunosuppressive pathways are clearly present because anti–PD-L1 Abs failed to treat melanoma, whereas anti–LAG-3 did, when combined with irradiation and adoptive T cell therapy.

Recently, Tumeh et al. (18) described an adaptive resistance mechanism to melanoma in cancer patients. They found that CD8⁺ T cells on the tumor margin when they expressed PD-1 and PD-L1 acted as biomarkers predicative of the efficacy of anti–PD-1 therapy in patients. In this study, we clearly showed in PD-L1^−/−RAG^−/− mice that B16 tumors, which can express PD-L1, are easily treated with CD4⁺ T cells, which can also express PD-1 and PD-L1. However, in irradiated hosts, either using PD-L1^−/− mice or anti–PD-L1 Ab therapy, T cell therapy had a minimal effect on tumor growth. This latter effect was only recapitulated in PD-L1^−/−RAG^−/− hosts with late dysfunctional T cells, which contained high levels of PD-1, LAG-3, and TIGIT, and was accompanied by an increase in T_reg cells. Depleting T_reg cells mildly improved therapy further in PD-L1^−/−RAG^−/− hosts. These results show that PD-L1 is only one facet of many other pathways of immunosuppression that are present within tumors and that targeting other pathways is crucial to overcome resistance mechanisms.

As chronic stimulation persists and functional exhaustion is established, disrupting the interaction of PD-1 and PD-L1 becomes less beneficial as T cells acquire more and more exhaustion molecules on their surface (47, 48). At this point, additional intervention is required. In our hands and those of others, this involved removing T_reg cells or blocking other checkpoint molecules (17, 48, 49). We have depleted TRP-1–specific T_reg cells using targeted therapy against DTR-bearing Foxp3⁺ TRP-1 T cells. Inhibitory receptor blockade combined with T_reg cell depletion led to a substantial improvement in the treatment effect of chronically exhausted TRP-1 T cells. This therapy, however, does not always lead to durable regression of tumors. This may be because of many factors, including the induced expression of PD-L1 on the tumor through adaptive resistance (46). However, it is more likely that there is an induction of new inhibitory receptors, like LAG-3, TIM-3, or TIGIT (48). Furthermore, induction of new T_reg cells, loss of tumor Ag, loss of MHC class I or II, loss of IFN-γ responsiveness, or loss of persistence of tumor-specific T cells could all be confounding factors (28, 30, 41, 50–53). Therefore, if recurrence does occur, retreatment with irradiation, another ACT, and other types of checkpoint therapy may be beneficial.

Removing T_reg cells in patients in the manner used in this study is currently not a viable option for treating cancer, unless it could be directed specifically against tumor-antigen reactive T_reg cells. Instead, we sought to investigate the more readily practiced approach of combination therapy. Transferring this to a more clinically relevant setting, defined by immunocompetent hosts receiving some form of lymphodepletion (given as irradiation in this study), we observed a similar treatment benefit with both Abs when administered at the time of adoptive cellular transfer but not at recurrence (data not shown), which is the opposite of what we previously shown in lymphopenic RAG^−/− (17); this shows that lymphopenia, depending how it is accomplished, can have drastic effects on results. Surprisingly, anti–PD-L1 Ab treatment by itself or combined with anti–LAG-3 (Fig. 4A, 4C) was not successful, even in combination with irradiation, contradicting other reports showing that irradiation and PD-L1 can promote antitumor immunity (54). Because irradiation therapy is transient (unlike RAG^−/− mice), this may allow extrinsic host regulatory functions to reestablish and impede antitumor T cell efficacy (55), especially through increased T_reg cells (56). Surprisingly, when combined with T cell therapy and irradiation, anti–LAG-3 checkpoint therapy was remarkably efficient at maintaining the cytotoxic function of TRP-1 CD4⁺ T cells, as evident in significantly fewer relapses and increased T cell numbers.

Irradiation with dual-checkpoint blockade has been successful with anti–PD-1 and CTLA-4 (57), but, to our knowledge, we are the first to show that a single blockade of a checkpoint with irradiation combined with ACT can treat large established tumors. Why LAG-3 is a better choice for immunotherapy was not readily answered in our findings, except for the fact that there were more TRP-1 CD4⁺ T cells. Huang et al. demonstrated a high level of LAG-3 expression on induced, as well as natural T_reg cells, which is required for suppressing effector T cell proliferation and function. Over the 2d assay period, anti–LAG-3 Abs did not affect proliferative responses of T cells stimulated in the absence of T_reg cells, confirming that the effect of anti–LAG-3 was indeed on the T_reg cells and not the effector cells (58). The ability of anti–LAG-3 to block in vitro suppression by T_reg cells demonstrates that LAG-3 is not simply a T_reg cell–selective marker but is also a molecule that is required for maximal T_reg cell activity. Durham et al. (55) showed that LAG-3 blockade led to a skewing of naive CD4⁺ T cells to a Th1 phenotype, which we showed to be critical for antitumor immunity in our model (32). However, they also showed that T_reg cells were not less suppressive without LAG-3. LAG-3 on T_reg cells appeared to be important in their induction and expansion only. LAG-3 is expressed on TRP-1 T effector cells (Fig. 3A) and TRP-1 T_reg cells (data not shown). Because our model involves cytotoxic CD4⁺ T cells, anti–LAG-3 may possibly target these cells more effectively, owing to their MHC class II restriction. The mechanism of action of anti–LAG-3 may also favor CD4⁺ T cells because it engages MHC class II directly, blocking CD4⁺ T_reg cell expansion and preventing suppression of CD4⁺ T cell effector function, as shown by others (55). Although not studied in this article, this could explain why anti–LAG-3 in combination with radiation and ACT generated maximal antitumor effects.

Two additional observations must be discussed. First is the ubiquitous nature of the TRP-1 self-antigen. TRP-1 self-antigen could be driving T cell responses in lymphopenic mice because it is expressed in melanocytes and other tissues. However, treatment of B16 tumor is better in TRP-1 Ag knockout mice on a RAG^−/− background (tyrp1^B-wRAG^−/−) when compared with Ag-expressing mice on the same background (data not shown). Therefore, in this particular model, Ag appears to be suppressing T cell responses, not enhancing them. Second, we did not transfer exhausted TRP-1 CD4⁺ T cells into WT mice undergoing combination irradiation and checkpoint therapy. In our previous publication (17), we showed that during melanoma recurrence in RAG^−/− lymphopenic mice that had been solely repopulated with exhausted TRP-1–specific CD4⁺ T cells, only combination with anti–PD-L1 and anti–LAG-3 Abs or anti–PD-L1 and T_reg cell depletion were beneficial in reestablishing antitumor immunity. In the current study, naive TRP-1–specific CD4⁺ T cells in combination with irradiation or naive TRP-1–specific CD4⁺ T cells in combination with anti–PD-L1 and irradiation did not work well in preventing recurrence in immunocompetent mice, but anti–LAG-3 in combination with naive TRP-1–specific T cells and radiation did. Thus, we concluded that using exhausted CD4⁺ T cells would not be any more beneficial than using naive CD4⁺ T cells (32). As this may be the Achilles’ heel of the work presented in this study, we suspect that exhausted CD4⁺ T cells would need a more rigorous treatment regimen, which may include both anti-PD1/PD-L1 and anti–LAG-3 Abs or some other combination (29). Because CAR T cells and engineered TCR T cells are produced from PBLs (1–3), which can be cell sorted into subsets that are not exhausted, our concept could be a reality.

Our system is unique in that it uses effector CD4⁺ T cells. Therefore, because LAG-3 is activated by MHC class II molecules, anti–LAG-3 may be most beneficial when using tumor-specific CD4⁺ T cells to overcome primary resistance to melanoma (55, 58–60). This, however, does not preclude tumor-specific adoptive cell therapy with CD8⁺ T cells from also benefiting from anti–LAG-3 therapy and irradiation, as CD8⁺ T cells also express LAG-3 within the tumor microenvironment during acquired resistance (16). Because adoptive immunotherapy with either engineered T cells, TIL, or CAR T cells is a reality (1, 29, 61), the possible benefits to enhancing their use are dramatically increased with the findings presented in this article.

The authors have no financial conflicts of interest.