Circulating CD8+ mucosal‐associated invariant T cells correlate with improved treatment responses and overall survival in anti‐PD‐1‐treated melanoma patients

Abstract Objectives While much of the research concerning factors associated with responses to immune checkpoint inhibitors (ICIs) has focussed on the contributions of conventional peptide‐specific T cells, the role of unconventional T cells, such as mucosal‐associated invariant T (MAIT) cells, in human melanoma remains largely unknown. MAIT cells are an abundant population of innate‐like T cells expressing a semi‐invariant T‐cell receptor restricted to the MHC class I‐like molecule, MR1, presenting vitamin B metabolites derived from bacteria. We sought to characterise MAIT cells in melanoma patients and determined their association with treatment responses and clinical outcomes. Methods In this prospective clinical study, we analysed the frequency and functional profile of circulating and tumor‐infiltrating MAIT cells in human melanoma patients. Using flow cytometry, we compared these across metastatic sites and between ICI responders vs. non‐responders as well as healthy donors. Results We identified tumor‐infiltrating MAIT cells in melanomas across metastatic sites and found that the number of circulating MAIT cells is reduced in melanoma patients compared to healthy donors. However, circulating MAIT cell frequencies are restored by ICI treatment in responding patients, correlating with treatment responses, in which patients with high frequencies of MAIT cells exhibited significantly improved overall survival. Conclusion Our results suggest that MAIT cells may be a potential predictive marker of responses to immunotherapies and provide rationale for testing MAIT cell‐directed therapies in combination with current and next‐generation ICIs.


Abstract
Objectives. While much of the research concerning factors associated with responses to immune checkpoint inhibitors (ICIs) has focussed on the contributions of conventional peptide-specific T cells, the role of unconventional T cells, such as mucosalassociated invariant T (MAIT) cells, in human melanoma remains largely unknown. MAIT cells are an abundant population of innate-like T cells expressing a semi-invariant T-cell receptor restricted to the MHC class I-like molecule, MR1, presenting vitamin B metabolites derived from bacteria. We sought to characterise MAIT cells in melanoma patients and determined their association with treatment responses and clinical outcomes. Methods. In this prospective clinical study, we analysed the frequency and functional profile of circulating and tumorinfiltrating MAIT cells in human melanoma patients. Using flow cytometry, we compared these across metastatic sites and between ICI responders vs. non-responders as well as healthy donors. Results. We identified tumor-infiltrating MAIT cells in melanomas across metastatic sites and found that the number of circulating MAIT cells is reduced in melanoma patients compared to healthy donors. However, circulating MAIT cell frequencies are restored by ICI treatment in responding patients, correlating with treatment responses, in which patients with high frequencies of MAIT cells exhibited significantly improved overall survival. Conclusion. Our results suggest that MAIT cells may be a potential predictive marker of responses to immunotherapies and provide rationale for

INTRODUCTION
Melanoma characteristics, such as PD-L1 expression, levels of immune cell infiltration, alterations in antigen processing and presentation, mutational burden and anatomic location, have been correlated with clinical responses to immune checkpoint inhibitors (ICIs). 1 However, much remains unknown about the complexity of immune interactions that result in successful antitumor immunity, including analysis of mucosal-associated invariant T (MAIT) cells, which is missing in most previous studies. Unlike conventional T cells that express highly variable T-cell antigen receptors (TCRs) recognising peptide antigens presented by highly polymorphic conventional HLA class I and II molecules, MAIT cells are a population of innatelike T cells with a restricted TCR repertoire, 2-4 which recognises small molecules derived from microbial riboflavin synthesis presented by the non-polymorphic MHCI-related molecule MR1. 2,4,5 MAIT cells can comprise up to 10% of circulating T cells in healthy adults and accumulate at mucosal sites including the gut, lungs and liver, where they can account for up to 50% of CD8 + T cells. 4 This high frequency makes MAIT cells the most abundant currently known population of T cells with a single antigen specificity. 6 In humans, MAIT cells are best identified by staining with MR1-tetramers loaded with the microbial ligand 5-(2-oxopropylideneamino)-6-Dribitylaminouracil (5-OP-RU). 7 Upon activation, MAIT cells rapidly respond by producing cytokines including tumor necrosis factor-a (TNFa), interferon-c (IFNc) and interleukin-17 (IL)-17, and upregulate costimulatory molecules, including CD40 ligand (CD40L, CD154). 4,8 Furthermore, MAIT cells have direct cytotoxic functions on both virally infected and tumor cells. 9 Finally, because of their pre-primed/memory nature, MAIT cells express high levels of several cytokine receptors, including IL-18Ra, IL-12R and IL-15R, which, upon engagement, lead to MAIT cell activation in the absence of TCR triggering, resulting in rapid cytokine secretion and cytotoxic functions, including the production of perforin, granzyme B (GZMB) and granzyme K (GZMK). 4,10 To date, the role of MAIT cells in cancer remains inconclusive. [11][12][13] For example, MAIT cells were found to exert tumor-promoting functions, with improved outcomes in MAIT cell-deficient MR1 À/À mice. 11 Similarly, hepatocellular carcinoma patients with high levels of tumor-infiltrating MAIT cells showed worse overall survival (OS) compared to those with low levels, 14 and a regulatory subset of CD4 + MAIT cells was found to correlate with tumor bacterial load in human colorectal tumors. 15 However, separate studies of oesophageal cancer, multiple myeloma and colorectal cancer patients found that higher frequencies of MAIT cells were associated with improved outcomes. 12,13,16 A potential role for MAIT cells in ICI therapies is not well described. Therefore, in this prospective study, we sought to determine the frequency and functional status of MAIT cells in patients with melanoma, and whether these parameters might correlate with clinical outcomes in ICI-treated patients. We further investigated the activation of MAIT cells in the peripheral blood and tumors from a variety of metastatic sites in melanoma patients. Our results show that MAIT cells are reduced in the circulation of melanoma patients, but that their frequency is restored in patients that respond to ICI therapy, while non-responding patients remain low in MAIT cell count. Strikingly, high frequencies of circulating MAIT cells while on treatment are also associated with improved OS. Altogether, our results reveal circulating MAIT cell frequencies as a potential prognostic correlate of therapeutic responses to ICI immunotherapy and suggest that manipulating MAIT cells might be advantageous to anti-melanoma immune responses.

The frequency of MAIT cells is decreased in the circulation of melanoma patients
We began our analysis by comparing the frequency of MAIT cell subsets in the peripheral blood of treatment na€ ıve (TN, n = 33) melanoma patients, regardless of stage and healthy donors (HD, n = 11) with a similar age and sex distribution (Supplementary table 1). MAIT cells were identified using flow cytometry as live, CD45 + CD3 + CD161 + 5-OP-RU-Tetramer + cells, and further stratified into subsets based on CD4 and CD8 expression (Figure 1a). The frequency of circulating total and CD8 + MAIT cells (Figure 1b-d) was reduced in TN melanoma patients compared to HD. The reduction was apparent regardless of whether it was expressed as a proportion of total CD45 + (Figure 1b) cells or total CD3 + T cells (Figure 1c). We observed no statistically significant changes in the frequency of CD4 + MAIT cells (Figure 1b-d). However, the frequency of CD4 + MAIT cells as a proportion of MAIT cells was increased in melanoma patients than HD (Figure 1e). Conventional T cells, the majority of non-MAIT (NM)-CD8 + and NM-CD4 + T cells, defined as 5-OP-RU-MR1 Tetramer À , are usually associated with the capacity to mount antitumor immune responses. 17 No significant differences in the frequency of either NM-CD4 + or NM-CD8 + T cells were observed between HD and melanoma patients (Supplementary figure 1a and b), demonstrating that the decreased frequency of MAIT cells in the blood of melanoma patients is specific to this T-cell population.
Melanoma-dependent changes in MAIT cell frequencies are associated with decreased cytokine production Human MAIT cells produce the cytokines IFNc and TNFa as well as the effector molecules GZMB and GZMK upon activation. 4,18,19 To determine whether there are functional differences in MAIT cells in melanoma patients, we measured the expression of these molecules in MAIT cells from HD and melanoma patients upon PMA/ionomycin stimulation. The percentages of polyfunctional (IFNc + TNFa + ) total MAIT cells and CD4 + MAIT cells were reduced in melanoma patients compared to HD, while no changes in the production of these cytokines by CD8 + or double-negative (CD4 À CD8 À , DN) MAIT cells were observed ( and Supplementary figure 1g). Increased PD-1 expression and low levels of CD25 and CD154 on conventional CD8 + T cells have been associated with chronic TCR stimulation in an immunosuppressive microenvironment. 23,24 Therefore, the phenotype of tumor-infiltrating MAIT cells could reflect chronic TCR stimulation, although we cannot exclude that other mechanisms, such as tumor microenvironment or currently unknown factors, may also be responsible. It is important to note that melanoma tumors analysed in this study were not responding to therapy at the time they were surgically excised, and MAIT cells may be significantly different in lesions that respond to therapy.
Non-MAIT, MR1-restricted T cells were recently shown to recognise tumor cells in an MR1dependent manner, 25 suggesting the existence of tumor-derived antigens presented by MR1. Whether such tumor-derived compounds also contain MAIT cell antigens is currently unknown. Another recent report demonstrated that the frequency of MAIT cells co-expressing CD39 and PD-1 correlates with TCR signalling driven by intratumoral bacteria. 15 Such bacteria could potentially represent a source of MAIT antigens in the tumor. In agreement with this possibility, we observed higher frequencies of CD4 + , and to a lesser extent CD8 + , CD39 + PD-1 + MAIT cells in melanoma tumors from mucosal sites (nasopharynx and jejunum). However, MAIT cells with a similar phenotype were also observed in metastases to the adrenal gland and axillary lymph node, which are not normally associated with bacterial colonisation (Figure 2h). Thus, the identification of CD39 + PD-1 + MAIT cells in our samples might be indicative of the recognition of unknown tumor-derived MAIT antigen(s). Experiments are underway to further explore this possibility. As CD39 expression has been associated with T-cell exhaustion, 26 these results may indicate an exhausted MAIT cell phenotype. 14 ICI treatment is associated with an increased frequency of circulating CD8 + MAIT cells in melanoma patients responding to therapy We next sought to determine whether the decreased frequency of MAIT cells observed in the circulation of melanoma patients could be influenced by ICI immunotherapy and whether this might correlate with responses to treatment. We first analysed the frequency of circulating MAIT cell subsets, comparing longitudinally collected pre-treatment and on-treatment [following the third anti-PD-1 (pembrolizumab) infusion] time points in responding (R, n = 14) and non-responding (NR, n = 6) melanoma patients. We found that the frequency of total MAIT cells increased in the responders but not in the non-responders (Figure 3a and b). In addition, closer examination of MAIT cell subsets revealed a statistically significant increase in the frequency of circulating CD8 + MAIT cells in responders but not in non-responders, with a reciprocal increased frequency of CD4 + MAIT cells in nonresponders but not in responders, while there were no changes in the frequency of DN MAIT cells in either patient population (Figure 3a  To further determine whether the frequency of circulating MAIT cell subsets was altered in earlystage disease (stage I/II), advanced disease (stage III and stage IV), or in patients who respond to anti-PD-1 therapies compared to those who do not, we analysed a larger cohort of patients with data collected at varying single time points during their treatment. We found that total MAIT cells, expressed as an absolute number or a proportion of total T cells in the circulation, are increased in stage IV patients who respond (n = 17) to ICI therapies compared to non-responders (n = 11) and treatment na€ ıve patients (n = 9; Figure 4a    significantly increased in responders compared to non-responders (Figure 4a and b and Supplementary  figure 2a and b). This was also apparent for stage III patients who did not progress (NP) while on therapy (n = 7) compared to those who progressed (P) while on therapy (n = 10) and TN patients (n = 8). Surprisingly, the increase in circulating MAIT cells in responding patients was not because of increased proliferation of the MAIT cells (Figure 4c). Although age and body mass index (BMI) have been shown to correlate with the frequency of MAIT cells in healthy individuals, 4 such correlations were not observed in melanoma patients regardless of whether they had received an ICI or not (Supplementary figure 3a and  b). These results suggest that the presence of the melanoma is directly affecting MAIT cells, altering their frequency and/or localisation. In addition, we observed no statistically significant differences in the frequencies of MAIT cell subsets when comparing stage III and IV patients who received either singleagent anti-PD-1 (pembrolizumab or nivolumab) vs. combination anti-CTLA-4 (ipilimumab) plus anti-PD-1 (nivolumab; Supplementary figure 4a).
Circulating CD8 + but not CD4 + MAIT cells are associated with improved overall survival The improved clinical responses that we observed in patients with high frequencies of MAIT cells were intriguing. We therefore sought to determine whether they might be associated with improved OS. To reduce the confounding factors of stage and treatment, only stage IV melanoma patients who had received ≥ 3 doses of anti-PD-1 and were on treatment at the time of the blood draw were included in the analysis (n = 27). The median follow-up time for this group of patients was 1.5 years (0.084-2.3 years). We found that patients with high levels of either total MAIT cells or CD8 + MAIT cells have significantly improved OS compared to those with low frequencies of these populations (Figure 4d). In contrast, patients with high frequencies of circulating CD4 + MAIT cells have significantly decreased OS (P < 0.0015; Figure 4d). Double-negative MAIT cells were not associated with overall survival (Figure 4d

DISCUSSION
Our results demonstrate that the frequencies of different MAIT cell subsets in the circulation of melanoma patients undergoing ICI immunotherapy represent a potential novel prognostic correlate of immunotherapy success. We acknowledge the small number of patients as a limitation of this study. However, we believe that these results are important and open novel lines of future research. These findings corroborate and further expand upon a recent study that found similar results in melanoma patients. 28 The role of MAIT cells in cancer remains debatable and might vary depending upon the nature of the cancer as well as the treatment administered. However, the strikingly improved OS of patients with high levels of circulating CD8 + MAIT cells provides strong rationale for further investigation into the potential role of MAIT cells in antitumor immunity. Previous studies have not always separated MAIT cells into subsets based on CD4 and CD8 expression, which may partially explain some discrepancies regarding the role of MAIT cells in cancer. 11,14 In this study, we found that CD8 + MAIT cells were positively associated with clinical responses and positively correlated with improved OS in melanoma patients. By contrast, increased frequency of CD4 + MAIT cells negatively correlated with OS, suggesting major differences in functions between these two subsets. These data highlight the need for a greater understanding of the functional differences between MAIT cell subsets in humans. Studies in our laboratory are ongoing to examine what such differences might be and how each subset may contribute to both response to therapy as well as OS. It will also be of interest to determine whether MAIT cells can directly kill melanoma tumor cells or whether they play an indirect role in antitumor immunity. Answers to these questions will be informative and might open new therapeutic avenues aimed at manipulating MAIT cell functions during immunotherapy treatments. Although we did not examine the microbiome in this patient cohort, it is tempting to speculate that the correlations between the contents of the microbiome and responses to anti-PD-1 immunotherapy might be, at least partly, mediated by microbial effects on MAIT cells. 29 Nevertheless, this study should be interpreted considering its limitations, such as the relatively low numbers of matched samples from patients pre-and on-therapy. Despite these limitations, our results highlight the potential role of MAIT cells in antitumor immunity.

Patient consent for publication
All patients provided written informed consent for sample and clinical data collection as well as data publication.

Sample preparation
PBMCs were isolated by Ficoll (Cytiva Life Sciences, Marlborough, MA, USA) density centrifugation. Tumors were minced to produce pieces < 1 mm 3 and then incubated with 5 µg mL À1 Liberase DL (Roche, Wilmington, MA, USA) at 37°C for 40 min and passed through 40 lm cell strainer to create a single cell suspension.

Statistical analysis
Statistical analyses were performed using one-way ANOVAs and two-tailed paired as well as unpaired t-tests in Prism (GraphPad, San Diego, CA, USA). Error bars represent standard error of the mean (SEM). High and low levels of cell types were identified based on the median, with statistical significance of survival based on high or low levels determined using the Log-rank test in Prism (GraphPad). To reduce selection bias, the pre-treatment blood draws were excluded from analysis unless they were being directly compared. Levels of significance are denoted as follows: ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.