Cancer cell-expressed BTNL2 facilitates tumour immune escape via engagement with IL-17A-producing γδ

Therapeutic blockade of the immune checkpoint proteins programmed cell death protein 1 (PD-1) and cytotoxic T lymphocyte antigen 4 (CTLA4) has transformed cancer treatment. However, the overall response rate to these treatments is low, suggesting that immune checkpoint activation is not the only mechanism leading to dysfunctional anti-tumour immunity. Here we show that butyrophilin-like protein 2 (BTNL2) is a potent suppressor of the anti-tumour immune response. Antibody-mediated blockade of BTNL2 attenuates tumour progression in multiple in vivo murine tumour models, resulting in prolonged survival of tumour-bearing mice. Mechanistically, BTNL2 interacts with local γδ T cell populations to promote IL-17A production in the tumour microenvironment. Inhibition of BTNL2 reduces the number of tumour-infiltrating IL-17A-producing γδ T cells and myeloid-derived suppressor cells, while facilitating cytotoxic CD8+ T cell accumulation. Furthermore, we find high BTNL2 expression in several human tumour samples from highly prevalent cancer types, which negatively correlates with overall patient survival. Thus, our results suggest that BTNL2 is a negative regulator of anti-tumour immunity and a potential target for cancer immunotherapy.

A significant recent advance in cancer immunology has been the discovery of immune resistance mechanisms in the tumour microenvironment (TME), which facilitate tumour escape from immunosurveillance. For example, it is known that IFN-γ, produced by tumour-infiltrating T cells, enhances the expression of immune checkpoint molecules such as PD-L1 (CD274, B7-H1), which in turn engage cognate receptors on effector T cells to promote exhaustion and apoptosis1,2,3,4. Antibody-mediated blockade of PD-1/PD-L1 and/or CTLA-4 immune checkpoint molecules reverses T cell dysfunction and restores effective anti-tumour immune responses5,6. This strategy, commonly referred to as immune checkpoint blockade (ICB), has been approved for the treatment of multiple cancer types7,8,9,10. Despite the dramatic results seen in some patients in response to currently available ICB therapy, the overall response rate remains disappointingly low5,6,9,11. Studies have shown that the PD-1/PD-L1 pathway is responsible for less than 40% of the immune dysfunction observed in human solid tumours, and accumulating evidence suggests that other mechanisms contribute to dysfunctional anti-tumour immunity in the TME12,13,14,15,16. Therefore, the elucidation of additional mechanisms of cancer immune evasion will both inform our understanding of ICB treatment response, while also provide novel targets to improve immunotherapeutic approaches for cancer.

γδ T cells are a non-major histocompatibility complex (MHC)-restricted lymphocyte subset closely aligned with innate immunity. γδ T cells preferentially localize within epithelial-rich tissues, such as the intestinal tract, skin and lungs. γδ T cells recognize MHC-independent nonpeptide antigens expressed during the cell stress response. γδ T cells have also been shown to execute anti-tumour-cytotoxic responses via multiple effector mechanisms, including the production and release of IFN-γ, as well as perforin and granzyme family members17. However, in certain contexts γδ T cells have also been shown capable of promoting tumour growth via the production IL‑17 A. In fact, multiple recent reports have described tumour-promoting roles for γδT17 cells, both in murine models and in human cancer18,19,20,21,22,23. In these studies, myeloid-derived suppressor cells (MDSCs) were shown to be a key downstream target of tumour-promoting γδT17 cells. MDSCs are a heterogeneous population of suppressive innate immune cells that expand in the context of several disease states, including cancer. Several reports have found the MDSC population substantially increased in tumours of patients harbouring a broad array of cancers, including colon cancer and melanoma23,24. Furthermore, high levels of circulating MDSCs are predictive of poor response to ICB therapy, suggesting that MDSCs may play a role in immune escape25,26. Although MDSCs were known to play a clear promoting role for multiple types of cancers, there are currently no approved therapeutic agents that specifically target MDSCs.

Butyrophilin-like protein 2 (BTNL2) is a transmembrane immunoregulatory protein that is highly expressed in the gastrointestinal tract. BTNL2 belongs to the butyrophilin-like family of proteins, and several of the butyrophilins and butyrophilin-like proteins, such as BTN2A1, BTN3A1, skint-1, BTNL1 and BTNL6, have been shown to play an essential role in regulating γδ T cell development and differentiation27,28,29,30,31,32,33,34. BTNL2 contains four extracellular immunoglobulin (Ig) domains, comprised of two IgV-IgC domain pairs35. Previous studies reported that a BTNL2-Fc fusion protein directly inhibited CD4+ T cell activation in vitro, although the cognate receptor for BTNL2 on T cells has remained unknown36,37. Intriguingly, it has also been reported that polymorphic variants of BTNL2 are associated with susceptibility to several autoimmune diseases and cancer, including lung and prostate adenocarcinoma38,39,40,41,42,43,44,45,46,47. One recent clinical study found that the expression of BTNL2 and other immune checkpoint molecules, including CTLA-4, was increased following anti-PD-1 therapy, further suggesting that BTNL2 may represent a novel mechanism of cancer immune evasion48.

In this study, we provide evidence that BTNL2 inhibits anti-tumour immunity by acting on local γδ T cell populations to promote IL-17A production, which enhances tumour immune resistance via recruitment of MDSCs. Blockage of BTNL2 with a novel monoclonal neutralizing antibody has a significant therapeutic effect for multiple mice tumours, and has a synergistic effect with anti-PD-1 blockage. Importantly, BTNL2 is expressed in multiple human solid cancers, and its expression level is negatively correlated with patient survival. In summary, we report that BTNL2 is a promising cancer immunotherapeutic target, with potential to enhance the efficacy of currently available immunotherapies, and possibly also to offer alternative stand-alone therapy in patients who are resistant to the current cancer immunotherapy.

BTNL2 blockade has a therapeutic anti-tumour effect in multiple murine models

Given the compelling human genetic and translational data in the literature, we hypothesized that antibody-mediated blockade of BTNL2 would enhance the anti-tumour immune response. To examine this, we generated a library of rat monoclonal antibodies directed against murine BTNL2, and conducted functional in vitro screening that identified a single clone (mAb-2, referred to as BTNL2 mAb) capable of rescuing the BTNL2-Fc-mediated inhibition of CD4+ T cells (Supplementary Fig. 1a). To further validate this antibody, we made BTNL2-KO mice by crispr-cas9, and deleted exon 1–3 of BTNL2 gene (Supplementary Fig. 1b). We found that BTNL2 mAb-2 antibody can be used for western blot and flow cytometry analysis (Supplementary Fig. 1c–e). As BTNL2 mAb-2 also recognized a non-specific protein of 95 kD, we performed the experiment of membrane-cytoplasm isolation, and found that this non-specific protein of 95 kD was exclusively expressed in the cytoplasm (Supplementary Fig. 1d). This data also suggest that this non-specific protein will not affect the in vivo function of mAb-2. Strikingly, both intraperitoneal (i.p.) and intravenous (i.v.) delivery of anti-BTNL2 mAb to tumour-bearing mice significantly reduced Lewis lung cancer (LLC) tumour growth (Fig. 1a). A similar anti-tumour effect was also found following i.p. injection of BTNL2 mAb into CT26 (murine colonic adenocarcinoma) or A20 (murine B cell cancer) tumour-bearing mice (Fig. 1b, c). In the A20 tumour model, five out of seven mice demonstrated a complete response with regression of all macroscopic tumour burden following anti-BTNL2 mAb treatment. We then re-implanted A20 tumour in the contralateral flank of these mice and failed to detect any evidence of tumour engraftment after re-implantation, indicating the development of long-term anti-tumour immune memory in response to treatment with anti-BTNL2 mAb (Fig. 1d). Following intravenous delivery of A20 tumour cells, a model of widely metastatic disease burden, anti-BTNL2 mAb substantially prolonged the survival of tumour-bearing mice (Fig. 1e). Next, we found that combinational treatment with anti-BTNL2 mAb and anti-PD-1 mAb had an additive anti-tumour effect compared to a single treatment, and anti-BTNL2 mAb and anti-PD-1 mAb had similar anti-tumour effects (Fig. 1f–h).

Next, we assessed the expression of BTNL2 in the mouse tumour microenvironment (TME). Interestingly, we observed a protein band at ~55 kDa that was strongly induced in engrafted LLC tumours compared to LLC cells, while the 72 kDa band was unchanged following engraftment. Expression of BTNL2 mRNA was significantly induced in engrafted LLC and CT26 tumours compared to primary tumour cells, which was similar to PD-L1 mRNA induction (Supplementary Fig. 2a,b). Notably, BTNL2 mRNA induction was much greater than PD-L1 induction in LLC tumours, which may explain at least in part the significant impact of anti-BTNL2 mAb treatment on LLC tumour growth relative to anti-PD-1 mAb treatment (Supplementary Fig. 2c). Interestingly, BTNL2 protein expression was significantly increased in LLC tumours after anti-PD-1 mAb treatment (Supplementary Fig. 2d), mirroring the prior report in humans that BTNL2 expression was upregulated following anti-PD-1 treatment48. After treatment with glycosylation inhibitor PNGase F, the intensity of the 72 kDa BTNL2 band decreased while the previously observed 55 kDa BTNL2 band appeared, indicating that the 55 kDa BTNL2 band represents the native non-glycosylated form of BTNL2 (Supplementary Fig. 2e). Following site-directed mutagenesis of four predicted glycosylation sites on BTNL2 (N210S, N296S, N427S and N432S), we observed a return to the predicted molecular weight by SDS-PAGE, which indicates that BTNL2 is glycosylated at these four sites (Supplementary Fig. 2f). Flow cytometric analysis of the TME indicated that BTNL2 was primarily expressed on CD45- tumour cells; however, 48.47% of CD45+ leukocytes did also express BTNL2 (Supplementary Fig. 2g).

BTNL2 inhibition reduces tumour-infiltrating γδT17

BTNL2 belongs to the butyrophilin-like family of proteins, and many of the butyrophilins and butyrophilin-like proteins have been shown to play an essential role in the regulation of γδ T cell development and differentiation27,28,29,30,31,32,33,34. Interestingly, infiltration of CT26 and A20 tumours by γδT17 as well as serum IL-17A concentration were all greatly reduced following anti-BTNL2 mAb treatment compared to treatment with isotype control antibody (Fig. 2a, b). We also performed intravenous injection of MC38 and CT26 cells, and found that survival of tumour-bearing mice was significantly extended by anti-BTNL2 mAb treatment compared to controls (Supplementary Fig. 3a–d). It was reported that there were different types of immune cells infiltrated in the CT26 tumour, and we examined whether there was any difference in terms of other immune cell populations infiltration in the CT26 tumour after anti-BTNL2 mAb blockage49,50. Notably, there was no significant difference in the number of tumour-infiltrating IL-17A- or IFN-γ-producing CD4+ T cells, Treg cells or NK cells (Supplementary Fig. 3e–g). Next, we found that both wild-type (WT) BTNL2-Fc and N4S-BTNL2-Fc recombinant proteins were capable of inducing the production of IL-17A by γδ T cells in splenocytes (Supplementary Fig. 4a and Fig. 2c), and this effect was enhanced by IL-1β and IL-23 co-treatment (Fig. 2d)51. Notably, the glycosylation site mutant N4S-BTNL2-Fc promoted significantly greater numbers of γδT17 and production of IL-17A than did WT-BTNL2, either with or without co-stimulation with IL-1β and IL-23 (Fig. 2c, d). Consistent with this, N4S-BTNL2-Fc promoted greater gene expression of the IL-17A transcriptional factor RORC than did WT-BTNL2-Fc (Fig. 2e). This was consistent with the finding that non-glycosylated BTNL2 was mainly induced in the TME (Supplementary Fig. 2a).

To further examine the role of IL-17A in tumour regression following BTNL2 blockade, we neutralized IL-17A in tumour-bearing mice and found that this completely abolished the anti-tumour effect of BTNL2 blockade (Fig. 2f). To further confirm the important role of IL-17A in the BTNL2 blockage, we purified murine IL-17A-Fc recombinant proteins (Fig. 2g, right panel), and found that treatment with recombinant IL-17A-Fc also abolished the anti-tumour effect of BTNL2 blockade (Fig. 2g, left panel). These results indicate that BTNL2 in the TME acts on local γδ T cell populations to induce the production of IL-17A by γδ T cells, which in turn promotes tumour immune resistance.

Next, we explored the IL-17A-producing cells in the TME, and found that the tumour-infiltrating IL-17A-producing cells were mainly γδ T cells in CT26 and A20 tumours (Fig. 3a, b). Mouse γδ T cells can be divided into Vγ1, Vγ4, Vγ5, Vγ6, Vγ7 γδ T cell subgroups according to TCR V gene utilization17, while Vγ4 and Vγ6 γδ T cells have been reported to be the major IL-17A-producing γδ T cells in multiple mouse tumour models18,19,20,21. Surprisingly, we found that the majority of tumour-infiltrating γδT17 cells were Vγ1 γδ T cells (Fig. 3c, d). Vγ1 γδ T cells were reported as an IL-17A-producing cell in certain circumstances, such as in the lethal pulmonary aspergillosis infections17,52. It was reported that IL-17A-producing γδ T cells were CD27 negative, and we indeed found that these subcutaneously infiltrated γδT17 cells did not express CD27, which was consistent with the previous report (Fig. 3c, d)21,53,54. Depletion of Vγ1 γδ T cells also had a therapeutic effect on subcutaneous tumours, and following Vγ1 γδ T cell depletion the anti-tumour effect of BTNL2 blockade was comparable to that of controls, which indicates that the anti-tumour effect of BTNL2 blockade is dependent upon Vγ1 γδ T cells (Fig. 3e). Furthermore, we found that Vγ1 γδ T cell depletion decreased IL-17A secretion by CT26 tumour-infiltrating cells, with similar IL-17A secretion seen after treatment with anti-BTNL2 and controls (Fig. 3f). Consistent with our prior findings, these data further confirmed that Vγ1 γδ T cells were the major IL-17A-producing γδ T cells in tumours, and the anti-tumour effect of BTNL2 blockade was dependent on the regulation of Vγ1 γδT17 cells. This finding is also consistent with a previous report that Vγ1 γδ T cells play a tumour-promoting role in the tumourigenesis55. As BTNL2 regulates tumourigenesis through Vγ1 γδ T cells, one of the possible mechanisms is that BTNL2 directly binds Vγ1 γδ T cells through Vγ1 γδTCR. To explore this hypothesis, we first examined whether BTNL2 can specifically bind Vγ1 γδ T cells. Interestingly, we did find that BTNL2-FC bound Vγ1 γδ T cells but not Vγ2 and Vγ7 γδ T cells (Supplementary Fig. 4b). Consistent with the finding that BTNL2 inhibited the activation of CD4+ T cells, BTNL2-FC also bound CD4+ T cells (Supplementary Fig. 4b). To examine whether BTNL2 binds Vγ1 γδTCR, we reconstituted the TCR Vγ1.1 and TCR Vδ6.3 in 293 T cells, as it was reported that Vγ1.1 paired with Vδ6.3 in the γδ T cells56,57. We did not find that BTNL2 bind TCR Vγ1.1Vδ6.3, and the data indicate that the receptor on γδ T cells is not TCR Vγ1 γδTCR (Supplementary Fig. 4c).

BTNL2 blockade decreases intra-tumoural accumulation of MDSC populations

MDSCs were previously shown to be important downstream target cell populations of tumour-promoting γδT17 cells18,21. MDSCs can be further subdivided into two subsets: so-called monocytic (CD11b+Ly6G-Ly6Chigh) and granulocytic (CD11b+Ly6G+Ly6Clow) MDSCs, which have the same surface markers of monocytes and neutrophils58,59. We detected a substantial decrease in tumour infiltration by neutrophils following anti-BTNL2 mAb treatment compared with controls (Fig. 4a). We also detected two distinct CD45+Ly6G− sub-populations that were greatly affected by anti-BTNL2 mAb treatment: a CD45 mid-expression population that was greatly reduced, and a CD45 high-expression population that was significantly increased after anti-BTNL2 mAb treatment. We designated these two populations ‘1’ and ‘2’ for convenience, and found that they were, respectively, positively and negatively associated with tumour size (Fig. 4a). We suspected that the ‘1’ population represented MDSCs, and indeed, found that over 90% of the ‘1’ population highly expressed monocytic MDSCs markers CD11b and Ly6C (Supplementary Fig. 5a). Of the remaining cells, ~1% represented dendritic cells (DCs), and 6% represented macrophages (Supplementary Fig. 5b). Interestingly, almost all of the tumour-infiltrating neutrophils were phenotypically consistent with granulocytic MDSCs, which indicates they are the same cells (Supplementary Fig. 5a). These data indicate that BTNL2 blockade decreases tumour infiltration by both monocytic and granulocytic MDSCs. Consistent with this finding, the CD45+CD11b+Gr-1+MDSC population was in fact reduced following anti-BTNL2 mAb treatment, which held true in CT26 tumour models (Fig. 4b). Given the side- and forward-scatter characteristics of the ‘2’ population, we suspected that these cells were predominantly lymphocytes. Indeed, we found that 85% of the ‘2’ population was composed of CD8+, γδ+ and CD4+ T cells (Fig. 4c). Infiltration of CD45+CD8+ T cells in the TME was greatly increased after anti-BTNL2 mAb treatment compared with anti-control Ab treatment (Fig. 4d). Importantly, the CD8+IFN-γ+ T cell population was significantly increased after BTNL2 blockade (Fig. 4e). BTNL2 inhibition led to decreased MDSC infiltration, but an increased accumulation of CD8+IFN-γ+ T cells, in the TME in the A20 tumour model (Fig. 4f, g). Following depletion of CD8+ T cells in tumour-bearing mice using anti-CD8a neutralizing antibody, the anti-tumour effect of anti-BTNL2 treatment was completely abolished, indicating that CD8+ T cells are the major tumour-cytotoxic cell following BTNL2 blockade (Fig. 4h, i). To further examine the role of MDSCs in the anti-tumour effect of BTNL2 blockade, we depleted MDSCs in tumour-bearing mice using CD11b neutralizing antibodies, but did not utilize anti-Gr-1 mAb, as we detected CD8+ lymphocytes expressing Ly6C in the TME (Supplementary Fig. 5c). Depletion of MDSCs completely abolished the previously observed increase in CD8+IFN-γ+ T cell tumour infiltration as well as the therapeutic anti-tumour effect of BTNL2 blockade (Fig. 4h–j). Consistent with previous reports, MDSCs significantly inhibited the proliferation of CD8 + T cells and OVA peptide-stimulated OT-1 CD8 + T cells in the in vitro system (Supplementary Fig. 5d, e). Consistent with this, IL-17A neutralization also led to decreased MDSC and increased CD8+IFN-γ+ T cells infiltration in the TME (Fig. 4k). Treatment with recombinant IL-17A-Fc protein abolished the BTNL2 blockade-mediated reduction in MDSC and increase in CD8+IFN-γ+ cell infiltration of the TME (Supplementary Fig. 5f, g). These results further demonstrate that the anti-tumour effect of BTNL2 blockade is mediated at least in part by a reduction in γδT17 cell infiltration in the TME, which as a consequence decreases tumour infiltration by MDSCs, which in turn leads to increased cytotoxic CD8+ T cell infiltration.

Tumour expression of BTNL2 plays a major role in anti-tumour immune escape

Next, we explored the physiological role of BTNL2 by examining the phenotype of BTNL2-KO mice. When compared to control mice, BTNL2-KO mice did not show any significant abnormality in survival or weight at 6 months of age (Supplementary Fig. 6a, b). Interestingly, BTNL2 protein was exclusively expressed in the mouse gastrointestinal tract, with particularly robust expression observed in small intestinal epithelial cells (Supplementary Fig. 6c, d). Histologic examination of BTNL2-deficient mice revealed large bowel inflammation, as well as loss of colonic crypt architecture (Supplementary Fig. 6e). Consistent with these findings, the expression of pro-inflammatory cytokines, including TNFα and IL-1β, was significantly increased in the colonic tissue of BTNL2-KO mice compared with control mice (Supplementary Fig. 6f). Examination of the intestinal epithelial lymphocyte population revealed significantly decreased γδT17 cells in BTNL2-KO mice compared with littermate controls, whereas the IFN-γ-producing γδ T cell population was unchanged (Supplementary Fig. 6g). These data indicate that BTNL2 is a physiologic regulator of γδ T cell differentiation in the gut, and may be utilized by cancer cells to escape anti-tumour immunosurveillance.

We then explored the impact of genetic BTNL2 deficiency on anti-tumour immune responses. Wild-type LLC tumour showed attenuated growth in BTNL2-KO mice compared to that in control mice; however, the difference did not reach statistical significance (Fig. 5a). LLC tumour cells with genetic deletion of BTNL2 (BTNL2-KO LLC) had significantly slower growth in vivo than did wild-type LLC tumours, however there was no difference in tumour cell proliferation between control and BTNL2-KO LLC cells in vitro (Fig. 5b–d). Consistently, BTNL2-KO LLC tumours had significantly fewer numbers of infiltrating γδT17 cells and MDSCs, and conversely higher numbers of infiltrating CD8+IFN-γ+ T cells, as compared to wild-type LLC tumours (Fig. 5e). There was no detectable difference in tumour growth between BTNL2-KO LLC tumours implanted in control mice or BTNL2-KO mice (Fig. 5f). Together, these data indicate that tumour cell-autonomous expression of BTNL2 plays a major role in tumour escape from immunosurveillance in the TME, and we think that the non-tumour cell-expressed BTNL2 plays a less important role in this process. Importantly, prior studies have classified the LLC tumour model as a non-immunogenic “cold” mouse tumour due to the lack of response to anti-PD-1 therapy60,61. The significant therapeutic effect against this tumour suggests that antibody-mediated inhibition of BTNL2 may hold promise in combination with anti-PD-1/PD-L1 ICBs, and furthermore, may open the door to immunotherapy for tumours that have previously been considered “cold”.

a Littermate control mice or BTNL2-KO mice were implanted WT LLC cells, and tumour growth kinetics of mice was shown (n = 12 for each group). b WT LLC cells or two clones of BTNL2-KO LLC cells were implanted subcutaneously in WT mice (3 × 105/mouse), and tumour growth kinetics was shown (n = 14 for each group, P = 0.0001 for LLC-WT vs LLC-BTNL2-KO-1 and P < 0.0001 for LLC-WT vs LLC-BTNL2-KO-2). c WT or BTNL2-KO LLC cells were analyzed by anti-BTNL2 mAb-2 for flow cytometry analysis (upper panel). Tumour image from b was shown (lower panel). d WT LLC cells or BTNL2-KO LLC cells were cultured in vitro for indicated times, and cell proliferation was shown by cell counting. e Cells from WT LLC tumours or BTNL2-KO LLC tumours were isolated, followed by the flow cytometry analysis (n = 14 for each group, P = 0.0022 for LLC-WT vs LLC-BTNL2-KO-1 γδT17 cell percentage and P = 0.0006 for LLC-WT vs LLC-BTNL2-KO-2 γδT17 cell percentage, P = 0.0189 for LLC-WT vs LLC-BTNL2-KO-1 MDSCs cell percentage and P = 0.0064 for LLC-WT vs LLC-BTNL2-KO-2 MDSCs cell percentage, P = 0.0011 for LLC-WT vs LLC-BTNL2-KO-1 CD8α+IFN-γ+ cells percentage and P = 0.0001 for LLC-WT vs LLC-BTNL2-KO-2 CD8α+IFN-γ+ cells percentage). f Littermate control mice or BTNL2-KO mice were implanted with WT or BTNL2-KO LLC cells, and tumour growth kinetics of mice was shown (n = 12 for each group, P < 0.0001 for control mice were implanted with WT vs BTNL2-KO LLC cells, NS for control mice or BTNL2-KO mice were implanted with BTNL2-KO LLC cells and control mice or BTNL2-KO mice were implanted with WT LLC cells). All data are mean ± s.e.m. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 based on two-way ANOVA for a, b, d, f, and Dunn’s multiple comparisons test for e. Data are representative of three independent experiments.

Full size image

BTNL2 expression in human cancers correlates with patient’s prognosis and γδT17 infiltration

BTNL2 polymorphisms are associated with susceptibility to lung adenocarcinoma40,41, and γδT17 cells have been reported to play key roles in human colorectal cancer (CRC) progression through recruitment of MDSCs23. We therefore tested the hypothesis that a correlation might exist between BTNL2 expression level and patient outcomes in these two cancer types. Patients with lung adenocarcinoma and colon adenocarcinoma that expressed low levels of BTNL2 had significantly improved survival compared to those expressing high levels of BTNL2 (Fig. 6a, b). We expanded our investigation to include other cancer types, and found that BTNL2 was indeed widely expressed, with expression detected in all of the examined human cancer samples (Supplementary Fig. 7a, b). In unaffected tissue adjacent to cancerous lesions, expression of BTNL2 was significantly decreased compared to that observed in lung adenocarcinoma and colon adenocarcinoma lesions, which is quite similar to the expression pattern of PD-L1 (Supplementary Fig. 7c–e). Importantly, ~38% of lung adenocarcinoma samples had medium-to-high levels of BTNL2 expression, but low levels of PD-L1 expression. This finding suggests that BTNL2 may be a complementary therapeutic target in this subpopulation of patients (Fig. 6c).

In lung adenocarcinoma samples, BTNL2 was mainly expressed by cancer cells (Supplementary Fig. 7f). The abundance of γδT17 cells in freshly isolated lung adenocarcinoma samples was also significantly higher than in para-cancerous samples, and, importantly, protein levels of BTNL2 were much higher in almost all of the examined cancer samples when compared to matched para-cancerous tissue (Fig. 6d–f). Notably, in 7 out of 23 samples (no.1, 3, 7, 9, 13, 15 and 20), tumour lesions had significantly higher expression of BTNL2 than did match para-cancerous tissue, while PD-L1 expression showed no difference. This finding suggests the possibility that BTNL2, more so than PD-L1, may represent a critical mechanism of immune evasion in these tumours (Fig. 6e). Interestingly, in hepatocellular carcinoma, 9 out of 27 samples showed higher tumour expression of BTNL2 compared to matched para-cancerous tissue (no. 1, 14, 15, 18, 19, 22, 24, 25 and 26), while 10 samples (no. 3, 4, 5, 6, 8, 10, 13, 16, 17, 20) showed lower expression of BTNL2 in tumour compared to para-cancerous tissue. In 25 out of 27 hepatocellular carcinoma samples, the pattern of PD-L1 expression (relative expression in cancer vs. para-cancerous tissue) generally reflected the pattern of BTNL2 expression (Fig. 6g). This data suggests that BTNL2 may also play a role in tumour immune escape in hepatocellular carcinoma. Of note, a majority of the hepatocellular carcinoma samples were obtained from patients with chronic hepatitis B virus infection (no. 1–7, 11–27), thus it is possible that BTNL2 expression in the para-cancerous tissue from these samples may not be reflective of expression in uninfected liver tissue.

Next, we analyzed in detail the pattern of BTNL2 expression in lung adenocarcinoma. We found that BTNL2 was robustly and exclusively expressed by tumour cells, and was almost not expressed in the para-cancerous tissue (Fig. 7a, b). Additionally, we found that BTNL2 expression reliably mirrored the expression of Napsin A, which is a well-established biomarker for lung adenocarcinoma (Fig. 7a, b). This data suggest that BTNL2 is not only an attractive target candidate for cancer immunotherapy but also may have utility as a biomarker for lung adenocarcinoma.

In addition to the well-described role of CD8+ T cells in anti-tumour immunity, other cell types have also been shown to play important roles in tumour immunosurveillance, including γδ T cells, and natural killer cells17,62. One illustrative study reported that in a cohort of breast cancer patients, tumour-infiltrating γδ T cells were a strong independent predictor of disease recurrence and overall survival. Interestingly, mechanistic studies have shown that human γδ T cells are capable of both tumour-promoting and anti-tumour functions. The anti-tumour role of γδ T cells is thought to be executed by type 1 cytotoxic γδ T cells, which are defined by the expression of IFN-γ. In fact, several clinical trials have employed the adoptive cellular transfer of ex vivo expanded type 1 γδ T cells for the treatment of solid tumours17. However, a clear tumour-promoting function of γδT17 cells has also been reported in multiple human solid tumour types, including colorectal cancer, pancreatic cancer and breast cancer17,19,23,63. Here, we show that human lung adenocarcinoma samples contain significantly increased numbers of tumour-infiltrating γδT17 cells, when compared to adjacent para-cancerous tissue. This finding suggests that γδT17 cells may also play a tumour-promoting role in lung adenocarcinoma. This also suggests that inhibition of BTNL2-dependent γδT17 cell accumulation holds therapeutic potential in human lung adenocarcinoma, as well as in other γδT17-enriched cancer types (Fig. 7c).

As reported by previous studies, mouse γδ T cells can be divided into Vγ1, Vγ4, Vγ5, Vγ6, Vγ7 γδ T cell subsets, which are defined according to TCR V gene utilization17. Several previous reports have suggested that Vγ4 and Vγ6 γδ T cells are the major IL-17A-producing γδ T cell subsets in mouse tumour models18,19,20,21. However, several groups have reported that Vγ1 γδ T cells are also capable of producing IL-17A, and in one such study, Vγ1 γδ T cells were actually shown to be the dominant source of IL-17A in the lungs of mice during pulmonary aspergillosis infection17,52. Vγ1 γδ T cells have also been reported to play critical roles in multiple processes, including tumourigenesis, anti-bacterial host defence and airway hyperresponsiveness55,64,65,66. Interestingly, a previous study utilizing a syngeneic mouse tumour model reported that depletion of Vγ1 γδ T cells had a therapeutic anti-tumour effect, which was mediated by suppression of Vγ4 γδ T cell acitivity55. This study is consistent with our finding that the Vγ1 γδ T cell subset has a tumour-promoting effect. In the current study, we found that Vγ1 γδ T cells are the major IL-17A-producing cells in the TME, which differs from prior reports that Vγ4- and Vγ6-restricted cells constitute the dominant γδT17 cell population in tumours. One possible explanation for this discrepancy may be due to differences in mouse tumour models, as we mainly employed heterotopic implantation and intravenous delivery of syngeneic tumour, while the other groups primarily utilized spontaneous mouse tumour models18,19,21,67. Nevertheless, our data from multiple mouse syngeneic tumour models consistently demonstrated a critical role for Vγ1 γδ T cells in tumour progression. Although the precise mechanism by which Vγ1 γδ T cells exert this pro-tumour effect was outside the scope of the present study, we believe that this question merits additional investigation.

The key biomarker for predicting response to anti-PD-1/PD-L1 ICB therapy is the expression of PD-L1 in the TME68,69. A recent phase 3 clinical trial enrolled 1274 patients with non-small-cell lung cancer (NSLC) and found that a PD-L1 TPS (tumour proportion score) of 50% or greater predicted a superior response to anti-PD-1 (Pembrolizumab) therapy compared to systemic chemotherapy, while a PD-L1 TPS of 1–49% predicted therapeutic equivalence70. Reports have suggested that fewer than half of all NSCLC patients have a TPS of 50% or greater70, implying that a majority of NSCLC patients fail to realize enhanced clinical benefit from PD-1/PD-L1 blockade over traditional chemotherapy. Our immunohistochemical analysis of lung adenocarcinoma samples from patients revealed that ~60% of tumours had low levels of PD-L1 expression. Of these PD-L1 low-expressors, approximately 68% had medium to high levels of BTNL2 expression, and represented 38.6% of all patients (Fig. 6c). These data suggest that BTNL2 blockade may effectively complement PD-1/PD-L1 inhibition, particularly in the subpopulation of patients with low tumour expression of PD-L1 but high expression of BTNL2. Consistent with this, implanted LLC tumours, which are poorly immunogenic “cold” tumours resistant to anti-PD-1 therapy, showed higher expression of BTNL2 compared to PD-L1, and antibody-mediated BTNL2 inhibition yielded a significant therapeutic response (Fig. 1a, Fig. 5; Supplementary Fig. 2c)71. This data again supports the notion that BTNL2 is an attractive target with significant potential as an adjuvant to enhance clinical responses to anti-PD-1/PD-L1 ICBs.

Multiple cellular and molecular mechanisms contribute to tumour resistance to ICB therapy, including MDSCs, Treg cells and the expression of immune checkpoint molecules in the TME3,72,73. Prior studies have found that patients with cancer who respond poorly to ICB therapy have a high level of circulating MDSCs25,26. In patients with prostate cancer and melanoma, MDSCs have also been reported as negative predictors of response to treatment with the anti-CTLA-4 mAb ipilimumab74,75. Mirroring these findings in human patients, MDSC numbers have been found to be markedly increased in the TME in multiple mouse tumour models, where they mediate T cell immunosuppression as well as angiogenesis76. Therefore, it has been speculated that therapeutic targeting of MDSCs might offer significant additive or synergistic benefits when combined with ICB therapy, which is supported by our finding that dual-inhibition of PD-1 and BTNL2 bolsters anti-tumour immune responses (Fig. 1f–h). In the current study, we found that BTNL2 blockade reduced MDSC infiltration of the TME, which we found was primarily mediated by IL-17A production by tumour-infiltrating γδ T cells. However, we cannot exclude the possibility that other cellular populations, for example, M2 macrophages or tumour-associated macrophages (TAM), also act downstream of BTNL2 to modulate anti-tumour immunity, therefore further investigation is needed to clarify this issue. The presence of other immune checkpoint molecules in the TME also undoubtedly contributes to resistance to ICB therapy. As has been reported previously, treatment with anti-PD-1 agents may induce the expression of other checkpoint molecules, such as TIM-3 and LAG3, and dual-blockade of PD-1/TIM-3 or PD-1/LAG3 has been shown to enhance the anti-tumour response77,78. Interestingly, we found that anti-PD-1 treatment resulted in upregulation of BTNL2 in the mouse TME (Supplementary Fig. 2d), which mirrors a prior clinical report that the expression of BTNL2 was upregulated after patients received anti-PD-1 treatment48.

Here, we have demonstrated that BTNL2 is a potent suppressor of anti-tumour immunity, and acts via a previously unreported role in the regulation of γδT17 cells. However, several questions regarding this promising therapeutic target remain and should be explored in future studies. For example, useful insights will be gained by elucidating the mechanism by which BTNL2 expression is selectively upregulated in the TME, as well as by defining its cognate receptor on γδ T cells. The human and mouse BTNL2 show high protein conservative (64% in terms of amino acid sequence), and both of them contain two extracellular IgC and two IgV domains36. The high conservative between human and mouse BTNL2 suggests that human and mouse BTNL2 may function through a similar mechanism, while the function of human BTNL2 needs to be further investigated in the future study. Our findings make clear that BTNL2 is a promising therapeutic target in the ongoing effort to broaden and enhance cancer immunotherapy.