Introduction
Mucosal-associated invariant T (MAIT) cells are unconventional T cells typically expressing a semi-invariant T cell receptor (TCR) with V
α7.2 joined to J
α33, 20, or 12, which in turn is paired with a limited number of
β-chains [
1,
2]. Most MAIT cells are CD8
+, but small populations of both CD4
+ and double negative (DN) MAIT cells are also present in the circulation and peripheral tissues [
2]. The MAIT TCR recognizes metabolites from the riboflavin synthesis pathway present in many, but not all, bacteria and fungi [
3]. These ligands are presented by the invariant major histocompatibility complex class I-related protein 1 (MR1), which shows a remarkable conservation between species [
4,
5]. MAIT cells can be found in most lymphoid and mucosal human tissues, but are most frequent in the liver [
6]. In mucosal tissues, MAIT cells have a phenotype reminiscent of resident memory T cells [
7‐
9], and they provide rapid responses to infection with bacteria and viruses. In particular, MAIT cells are programed for immediate cytokine production and cytotoxicity following activation, and both polyclonal and antigen-specific TCR-mediated stimulation results in production of IFN-γ, TNF, and IL-17 and up-regulation of intracellular Granzyme B (GrB). In addition, pro-inflammatory cytokines, for example the combination of IL-12 and IL-18, will also induce cytokine production in MAIT cells, independent of TCR signaling [
10]. The propensity of MAIT cells to produce different cytokines further appears to be influenced by local tissue factors [
11]. While circulating, intestinal, and hepatic MAIT cells preferentially produce TNF and IFN-γ, MAIT cells from infected lungs, the female genital tract, and breast epithelial ducts are much more prone to IL-17 and IL-22 production [
8,
12‐
15].
The cytokine production and cytotoxicity of MAIT cells indicate that they may also contribute to immunity against tumors. It was recently demonstrated that MAIT cells accumulate in colon tumors, compared to the unaffected colon mucosa of the same individuals [
14,
16,
17]. The tumor-infiltrating MAIT-cells produced IFN-γ and TNF, but only little IL-17. Their production of IFN-γ was, however, considerably lower than that of MAIT cells from the unaffected mucosa, while cytotoxic potential was similar in the two sites [
14,
18]. In addition to the killing of tumor cells by CD8
+ cytotoxic T cells, there is strong evidence that Th1 responses, and more specifically the balance between Th1 and Th17 responses, are important for patient outcome in colon cancer [
19]. Therefore, MAIT cell contribution of IFN-γ may be an important part of anti-tumor defense, and the down-regulation of IFN-γ-production in tumor-associated MAIT cells may represent a tumor immune evasion mechanism. It should be noted, though, that MAIT cell enrichment in colon and hepatocellular carcinoma (HCC) correlates to a worse patient outcome [
16,
20,
21].
Tumor-infiltrating conventional CD8
+ T cells can enter a stage of exhaustion, due to sustained TCR activation, which is associated with reduced effector functions. This is a reversible state, and it is sustained by expression of inhibitory receptors such as programmed cell death protein 1 (PD-1), T cell immunoglobulin and mucin domain-3 (Tim-3), T cell immunoreceptor with Ig and ITIM domains (TIGIT), Lymphocyte activation gene 3 (LAG-3), and B and T lymphocyte attenuator (BTLA). T cell exhaustion was first defined in chronic viral infections [
22], but has since been convincingly demonstrated also in tumors [
23,
24].
Exhaustion has also been suggested in MAIT cells, mainly in the setting of chronic bacterial and viral infections. Generally, chronic infection or inflammation results in reduced frequencies of circulating MAIT cells, and the remaining MAIT cells display a prominent expression of PD-1 and reduced cytokine production. [
7,
25‐
27]. Furthermore, the observation that blocking of Programmed death-ligand 1 (PD-L1)/PD-1 interactions in vitro partly restores IFN-γ-production in MAIT cells from tuberculosis patients further supports the notion of MAIT cell exhaustion [
25]. Here, we investigated if tumor-infiltrating MAIT cells show signs of exhaustion, and if this would explain the poor cytokine responses in MAIT cells isolated from tumors. We could show that MAIT cells in colon tumors have an exhausted phenotype compared to MAIT cells from the unaffected colon mucosa and blood, express fewer effector molecules upon stimulation, and display partly increased activation following blocking of PD-1.
Discussion
In this study, we show that the MAIT cells accumulating in human colon tumors differ phenotypically from those in the unaffected mucosa and blood in several ways. Tumor-infiltrating MAIT cells have a different distribution of CD8+ and DN MAIT cell subsets and display surface markers indicating an exhausted state. Furthermore, the PD-1highTim-3+ phenotypically exhausted MAIT cells have reduced effector functions, but the presence of blocking antibodies to PD-1 increase their activation.
MAIT cells have generally been described as CD8
+ or DN, while CD4
+ MAIT cells usually make up a minute population, at least in the blood and liver [
2,
20,
29‐
31]. In the current study, the DN MAIT cells were enriched in the tumors compared to both unaffected tissues and blood
. It has been suggested that DN MAIT cells develop from CD8
+ MAIT cells and that they represent a more mature MAIT cell population [
29]. TCR-mediated activation in vitro results in a profound loss of CD8 and subsequent apoptosis in MAIT cells [
29], and this process may be accelerated in the tumor microenvironment, where an impaired epithelial barrier allows influx of bacterial metabolites [
35,
36]. DN MAIT cells in the circulation express less cytotoxic markers and Th1-associated cytokines than CD8
+ MAIT cells, and are more prone to apoptosis. There is also an increased production of IL-17 from DN MAIT cells, albeit from a low starting level [
29,
31]. Also in the rectal mucosa, CD8
+ MAIT cells appear to have more pro-inflammatory functions than DN cells [
37]. If these features of DN MAIT cells are consistent also in the tumor microenvironment, DN MAIT cells may be less functional with regard to anti-tumor immunity, and their increased presence may partly explain the reduced cytokine responses recorded in tumor-infiltrating MAIT cells [
14,
20,
38].
To further evaluate the effect of the tumor microenvironment on MAIT cell effector functions, we investigated the presence of exhausted MAIT cells in the tumors. Exhaustion was originally described as a sequential, but reversible, loss of effector functions due to consistent antigen stimulation and inflammation in mice chronically infected with lymphocytic choriomeningitis virus [
22]. Exhaustion has since become evident in many infections, and also in tumors, and is sustained by expression of inhibitory receptors, and also involves changes in transcription factor expression and proliferative capacity [
22,
32]. Our analyses showed a markedly increased frequency of PD-1
highTim-3
+ exhausted MAIT cells in the tumors compared to both unaffected colon tissue and blood. Further analyses showed that PD-1
highTim-3
+ cells also co-express CD39, and that they have an increased proliferation. Tumor cells can exhibit a prominent expression of the PD-1 ligands PD-L1 and PD-L2, and also produce the Tim-3 ligand Galectin-9, and may thus directly contribute to maintain the exhausted state of infiltrating T cells. In colon tumors, however, infiltrating immune cells appears to be a major source of PD-L1 and PD-L2, as well as Galectin-9 [
39,
40], and may thus modify MAIT cell function in human colorectal cancer. It is also possible that additional factors from cancer cells contribute to reduced effector functions in MAIT cells [
14,
20]. A recent study indicates that bacteria present in colorectal tumors induce CD39 expression following TCR-mediated stimulation [
36], and chronic bacterial stimulation may thus contribute to MAIT cell exhaustion in the tumor microenvironment. CD39 is an ectoenzyme converting pro-inflammatory ATP to immunosuppressive adenosine which is expressed on Treg and exhausted conventional CD8
+ T cells [
41], and this way exhausted MAIT cells may contribute directly to reducing anti-tumor immunity. Whether MAIT cells themselves are sensitive to adenosine-mediated immunosuppression is not yet known, but presents an intriguing possibility of negative autocrine regulation of exhausted MAIT cells.
Another important observation is that colonic MAIT cells were distinctly different from circulating cells, with regard to expression of most of the exhaustion markers analyzed. These inhibitory receptors were generally higher in the tissue than in the circulation, with the exception of BTLA, which had a reversed expression. Similar results were recently also shown for PD-1, TIGIT, and LAG-3 by Schmaler et al., who compared MAIT cells from normal colon mucosa and blood [
42]. These observations underline the importance of using MAIT cells from the organ of interest, and not extrapolating from circulating cells, even though they may be isolated from a relevant patient population.
Previous studies have suggested MAIT cell exhaustion in chronic bacterial and viral infections [
7,
25,
26], and more recently, also in HCC and colorectal cancer [
20,
36]. In HCC, MAIT cells express more PD-1 and Tim-3 than in the surrounding liver tissue, even though expression levels were generally much lower than in colon mucosa and tumors, respectively [
20]. Interestingly, co-culture with HCC cell lines resulted in increased PD-1 and Tim-3 expression on MAIT cells. MAIT cell exhaustion in the tumor microenvironment may thus be induced in parallel to, or shortly after, strong stimulation, as previously demonstrated for bacterial infections [
43]. In the current study, we could also correlate the exhausted MAIT cell phenotype with reduced functional capacity, as PD-1
highTim-3
+ tumor-infiltrating MAIT cells have reduced effector polyfunctionality. Our studies show no selective loss of one certain cytokine from exhausted MAIT cells, but rather a decline in polyfunctionality. However, GrB appears to be kept in the phenotypically exhausted MAIT cells to a larger extent than IL-2, TNF, and IFN-γ, and did not contribute to the reduced polyfunctionality. Likewise, it has previously been shown that exhausted conventional CD8
+ T cells still express high levels of GrB, and retain their cytotoxic ability [
22,
32]. This finding is compatible with our recent observation that MAIT cells from tumors and unaffected tissues have comparable cytotoxic potential [
18].
Recently, it has become clear that the pool of exhausted CD8
+ cells present in chronic viral infection and in tumors can be further divided into progenitor exhausted cells and terminally exhausted cells. Progenitor exhausted cells are distinguished by an intermediate expression of PD-1, CXCR5, and the transcription factor Tcf1, and upon TCR-mediated stimulation they can differentiate into terminally exhausted cell co-expressing high levels of PD-1 together with Tim-3 and CD39 [
41,
44,
45]. The terminally exhausted cells are still highly cytotoxic and proliferate more in the tumor microenvironment, but have reduced long-term survival and reduced polyfunctionality with regard to cytokine production [
44,
46]. The PD-1
highTim-3
+CD39
+ phenotype, increased proliferation, and reduced polyfunctionality of tumor-infiltrating MAIT cells all suggest that they are terminally exhausted. It has been shown in murine studies that only progenitor exhausted T cells can respond to anti-PD-1 immunotherapy [
44,
45], and in human tumors, the presence of progenitor exhausted cells correlates to a better response to anti-PD-1 therapy [
44,
47]. However, presence of PD-1
high cells with a phenotype reminiscent of terminally exhausted cells in non-small cell lung cancer also strongly correlated to treatment response with anti-PD-1 [
48]. Nevertheless, it is not clear if these observations apply to unconventional T cells like MAIT cells. In this study, we show that the presence of PD-1 blocking antibodies during in vitro stimulation increases activation of tumor-infiltrating MAIT cells in a subset of colon cancer patients, indicating that checkpoint blockade may act on unconventional T cells from cancer patients. A recent study showed that both PD-1 and Tim-3 can bind to Galectin-9 in vitro, and that simultaneous PD-1 signaling actually prevents apoptosis induced by the Galectin-9 Tim-3 axis [
49]. Therefore, combining blocking of PD-1 and galectin-9 might give a further mechanistic insight into the regulation of MAIT cell responsiveness. We cannot say if the reinvigorated cells responding to checkpoint therapy are only the terminally exhausted cells retrieved from the tumors, as the stimulation protocol leads to up-regulation of PD-1 in the majority of MAIT cells. Furthermore, the effect of PD-1 blocking may be underestimated in our experimental system, as there are probably substantial amounts of potentially MAIT-stimulating cytokines induced by the polyclonal stimulation or presence of bacteria. The effect of checkpoint blockade is probably weaker if the MAIT cell is stimulated by mixed cytokine and TCR signals compared to TCR-mediated signaling alone.
In the few studies available to date, MAIT cell infiltration into tumors is associated with reduced effector functions and impaired patient outcome [
14,
16,
20,
21]. This may partly be caused by an accumulation of exhausted MAIT cells in tumors, but our results suggest that immune checkpoint blockade may improve MAIT cell function, and may thus contribute to improved anti-tumor immunity in the tumor microenvironment also from resident MAIT cells. It is also encouraging to note that terminally exhausted conventional CD8
+ T cells retain efficient cytolytic activity and can kill tumor cells in vitro and reduce tumor growth in vivo [
44]. Taken together, our results indicate that the tumor microenvironment inflict a state of exhaustion in tumor-infiltrating MAIT cells, and that checkpoint blockade might reinvigorate MAIT cells to increase anti-tumor effector functions.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.