Background
In recent years, it has become evident that the immune system can recognize and sometimes efficiently clear transformed neoplastic cells [
1]. One example of this is colorectal cancer (CRC) where lymphocyte infiltration into tumors clearly correlates to patient outcome [
2‐
4]. In this context, presence of CD8
+ memory cells and Th1 cells improve patient prognosis, while alternatively activated macrophages and Th17 cells correlate with a worse prognosis. Indeed, the immunoscore, a scoring system that quantifies tumor-infiltrating immune cells, is a more accurate method than tumor staging to predict patient outcome in CRC [
5]. Cancer immunotherapies, in particular immune checkpoint blockade, are now established targeted therapies, which shift the balance between immunostimulating and immunosuppressive forces to the benefit of the patient [
6]. Immune checkpoint blockade has had considerable effect in the treatment of several solid cancer forms, e.g. lung cancer, advanced malignant melanoma and renal cell cancer, but a favorable effect in colon cancer has only been shown for microsatellite instability high (MSI-H) tumors, while microsatellite stable (MSS) tumors show poor response [
7,
8]. MSI-H tumors have a deficient DNA repair machinery and a subsequent increase in mutational load, creating an immunogenic tumor environment and higher lymphocyte infiltration [
9].
In addition to continued efforts in improving immune checkpoint blockade there is a need for alternative strategies for immunotherapy. One target for such treatment is regulatory T cells (Treg), which are accumulating in most solid tumors and are associated with a poor patient prognosis [
10]. Initially, several studies in colorectal cancer patients showed a correlation between tumor-infiltrating Foxp3
+ putative Treg cells and a favorable prognosis [
11‐
14]. However, a more recent study by Saito et al
. showed that a high infiltration of carefully identified Treg in the tumor actually correlates to a poor prognosis in CRC [
15], and De Simone et al
. also correlate intratumoral Treg numbers with a less favorable patient outcome in CRC [
16]. It is clear from functional studies that tumor-infiltrating Treg suppress conventional T cell proliferation and cytokine production [
15‐
18]. Several studies also indicate that they may actually be more suppressive than Treg found in the unaffected colon mucosa, based on expression of immunoregulatory surface receptors or suppression of effector cell proliferation [
15,
16,
18,
19]. Thus, the contribution of tumor-infiltrating Treg to patient prognosis in colorectal cancer is still not completely defined, but it would appear that Treg infiltration into established tumors promotes tumor progression.
Considering that Treg in cancer patients would probably have to be targeted on a subpopulation basis due to extensive autoimmune side-effects when targeting all Treg [
20], the functional differences between intratumoral Treg subsets needs to be investigated. To better define intratumoral Treg, we performed mass cytometry analyses of Treg from tumors and unaffected mucosa from colon cancer patients. We identified three Treg clusters that accumulate in colon tumors, which were distinguished by their CD39 expression, and two that were reduced in the tumors. Subsequent flow cytometry analyses showed that colon tumors could be divided into two types, one with a high expression of CD39 by Treg and one with an intermediate expression. In patient outcome analyses, patients with a high CD39 expression by tumor-infiltrating Treg had a worse outcome.
Discussion
In this study, we present a detailed examination of Treg phenotypes in colon tumors and preliminary survival data for colon cancer patients, which suggests that a large intratumoral accumulation of eTreg, defined by their expression of CD39, may correlate to a poor patient prognosis.
It is well known that Treg accumulate in solid tumors, and here we provide an unbiased multiparameter assessment of the specific Treg populations that are infiltrating human colon tumors. Based on suppression assays conducted in vitro, tumor-infiltrating Treg in colorectal cancer can exert immunosuppressive functions via programmed death-ligand 1 (PD-L1), IL-10 and TGF-β [
16,
19]. In addition, tumor-infiltrating Treg will probably also act to suppress surrounding effector immune cells through the activity of CD39 and ultimately adenosine signaling, similar to what is suggested for intratumoral Treg in breast cancer patients based on in vitro experiments [
29].
A striking feature of the enriched Treg populations detected in our study is their activated phenotype with expression of the co-stimulatory molecules ICOS and OX-40, combined with the effector molecule CD39. There is also an almost uniform expression of the exhaustion marker TIGIT among the tumor-infiltrating Treg, while other immune checkpoint receptors associated with exhaustion (PD-1 and LAG3) were not expressed to any larger extent in the enriched clusters. This is somewhat surprising, but might indicate that the Treg accumulating in colon tumors represent activated eTreg that are not yet under the influence of immunomodulatory checkpoint signaling. The ongoing proliferation in two clusters and the generally high expression of activation markers in the enriched clusters would also support this conclusion. Unfortunately, the cell numbers that can be retrieved from tumor tissue only occasionally allows for cell sorting and functional assays. The Treg accumulating in tumors were also low in CD103, which has been used as a marker of tissue resident memory T cells. However, recent findings indicate that this is primarily true for CD8
+, but not CD4
+, T cells in the human intestinal mucosa [
30]. We could also identify one enriched cluster which is distinguished by a strong expression of CD56 (cluster 9). This may seem contra-intuitive, as CD56 is mainly expressed by NK cells and by unconventional T cells. However, a recent study identified high frequencies of TGF-β-producing CD56
+ Treg in active lesions from Langerhans cell histiocytosis [
31]. Suppressive CD56
+ Treg have also been detected in hepatocellular carcinoma [
32]. These authors speculated that the tumor microenvironment may convert infiltrating CD56
+ T cells into functional Treg, possibly by TGF-β, but if this is the case in CRC remains to be examined. Taken together, the common feature of all the enriched Treg clusters in the tumors is a strong expression of ICOS, OX-40, and CD39, which are not seen in the contracting clusters. Of these markers, CD39 is a well-established effector molecule of potent Treg, and correlates well to an activated phenotype [
17,
18,
24,
25].
CD39 is also known as ectonucleoside triphosphate diphosphohydrolase-1 (ENTPD1), an ectoenzyme catalyzing the first and rate-limiting step in the conversion of extracellular ATP to immunosuppressive adenosine, which is one of the main effector functions in Treg [
24,
33,
34]. A further consequence of this enzymatic activity is the removal of pro-inflammatory extracellular ATP, which activates myeloid cells through the NLRP3 inflammasome. Adenosine may act directly on tumor cells and promote oncogenic processes, but also creates an immunosuppressive environment and reduce both T cell and antigen-presenting cell functions [
23,
35,
36]. In particular, CD39
+ Treg may be most prone to suppress Th17 cells [
37,
38]. CD39
+ Treg from cancer patients also suppress transendothelial migration of conventional effector T cells [
39]. Indeed, CD39
+ Treg have a superior suppressive ability compared to CD39
− Treg both in vivo and in vitro [
24‐
26]. This effect is not necessarily dependent on only CD39 activity and generation of adenosine, as previous studies suggest that CD39
+ Treg in colon tumors display several additional markers conferring suppressive ability [
17,
18,
39‐
41]. We thus used CD39 to mark the enriched subsets of putative eTreg in our subsequent analyses of tumor-infiltrating Treg. After dividing the colon cancer patients into two groups based on the frequency of CD39
+ Treg among intratumoral Treg, we could also show that CD39
+ Treg in the tumors have higher Foxp3 expression on a per cell basis if they are isolated from tumors with a high infiltration of CD39
+ Treg compared to tumors with fewer CD39
+ Treg. This suggests that the former patients may exhibit a more tightly regulated, and less favorable, anti-tumor immune response compared to patients with fewer CD39
+ Treg and that CD39 is a good marker for activated eTreg.
The accumulation of CD39
+ Treg in colon tumors may result from specific recruitment, intratumoral proliferation, improved retention and upregulation of CD39 in the tumor microenvironment. Genetic polymorphisms in the
ENTPD1 gene encoding CD39 may also be an important determinant of this process. Three different single nucleotide polymorphisms (SNPs) in this gene have been reported to correlate with different levels of CD39 expression among Treg in peripheral blood of healthy donors [
26] and cancer patients [
17]. We observed a strong correlation between CD39
+ Treg frequencies in the different tissues examined, indicating that
ENTPD1 SNPs may affect CD39 expression in all tissues. In particular, the correlation between CD39
+ Treg frequencies in peripheral blood and tumor is of potential interest and indicates that CD39
+ Treg frequencies above approximately 60% of total Treg in peripheral blood will likely correspond to patients with a large accumulation of intratumoral CD39
+ Treg. Taken together, this suggests that frequencies of CD39
+ Treg in peripheral blood, as well as the
ENTPD1 SNPs of each patient may serve as biomarkers to assess the corresponding CD39
+ Treg frequencies in colon tumors.
A correlation between Treg accumulation in colorectal tumors and a poor patient survival has been demonstrated in recent studies [
15,
16]. However, this correlation may be weaker for colorectal cancer compared to other solid cancer forms, based on several earlier studies showing a favorable effect of Treg on patient outcome [
10,
12‐
14]. As most of these studies relied only on Foxp3 expression to define Treg, it is possible that activated conventional effector T cells were also included in the analysis [
15]. As flow cytometry allows a much more detailed delineation of cell subsets, it is a more adequate method to identify Treg. Here, we did a further subdivision, and analysed if the accumulation of eTreg, defined by CD39, would correlate to established clinical parameters.
It should be noted that we analysed stage I to III patients with no metastases at surgery and complete removal of the tumor, a patient group that has a relatively good prognosis. This is a limitation of our study, as is the relatively low number of individuals included. Still, the evaluation of patient outcome shows a tendency towards lower cancer-specific survival in the patients with a larger proportion of intratumoral eCD39+ Treg. It is worth considering that there were no differences in total Treg frequencies in the tumors between patients with high and low frequencies of CD39-expressing Treg, and the potential effect seen here is not due to the presence of Treg as such, but rather the presence of an activated effector subset. The relatively low number of events registered in terms of relapse and cancer-specific death does not allow the detection of any significant differences between groups, and multivariate analysis can not be performed at this stage. Generally, patients with MSI-H tumors have a more favorable outcome than patients with MSS tumors do. In our study, patients with high and low CD39+ Treg frequencies had an equal distribution of MSS- and MSI-tumors. Furthermore, the other clinical parameters with a direct correlation to patient outcome that were examined did not seem to influence intratumoral CD39+ Treg frequencies. Thus, CD39+ Treg frequency in colon tumors is probably independent of the most common clinical parameters associated with patient prognosis in colon cancer. As these parameters do not co-vary with frequencies of CD39+ Treg, we believe that CD39 expression on Treg may be an independent variable contributing to patient outcome.
Patients with MSI-H colon tumors respond better to anti-PD-1/PD-L1 treatment compared to patients with MSS tumors [
7]. PD-1 targeted treatment primarily improves effector T cell function, which is regulated in part by PD-L1 and PD-L2 expression on several different cell types, and the role of Treg suppression in this respect is not well known at present. If targeting of specific Treg subsets in colorectal cancer patients may help patients with MSS tumors, for example due to the inhibitory effect of Treg on the migration of effector T cells into tumors [
39], remains to be elucidated. However, this may be an important alternative to already established cancer immunotherapies.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.