Shaping of an effective immune microenvironment to and by cancer cells

Abstract

A high density of intratumoral effector memory CD8+/Th1 T cells is associated with favorable prognosis in most cancers and may be induced or increased by immunotherapy. Efficient adaptive immune reactions are shaped in tumor adjacent tertiary lymphoid structures, which exhibit all characteristics of immunity generating lymphoid formations in reactive lymph nodes. Malignant tumor cells impact favorably or deleteriously their immune microenvironment if they bear genetic mutations that result in neo-antigens or by producing chemokines and cytokines that recruit lymphocytes and myeloid cells or increase inflammation and neo-angiogenesis. This intricate network of interactions results in control or escape of tumors, and its understanding will help define goals to monitor efficiency of immunotherapies.

Background

The era of cancer immunology has come of age. It has been particularly striking at the national American Association for Cancer Research (AACR) meeting held in San Diego, 5th–9th, 2014, where a plenary session and several major symposia devoted to immunotherapy were highly attended. All aspects were highlighted, including the use of adoptive T cell therapy with tumor-infiltrating lymphocytes or engineered T cells featuring chimeric antigen receptors (CAR), the use of immunomodulatory monoclonal antibodies specific for CTLA-4, PD1 or PD-L1, as well as the identification of target tumor-specific antigens through exome studies. When successful, immunotherapies yield impressive results, illustrated by tumor responses in chemoresistant cancers, and long-term survival of some responders, who are potentially cured.

Two features appeared striking upon analysis of data coming from immunotherapy trials. The first is that only a subset of patients responded to the treatment [1–3]—although combination of immunotherapies acting through different mechanisms (CTLA-4 and PD1) may increase this subset [4]. The second is that beneficial effects seem to correlate with long-term modifications of the host immune system as illustrated by the persistence of tumor-infiltrating leukocyte (TIL) or CAR [5]. Thus, there is a need to understand both the biological features allowing patients to respond to a given immunotherapeutic approach and which are the effector immune cells necessary to maintain long-term clinical benefits. The recent understanding of the characteristics of natural immune reactions associated with a long-term survival in human may be very useful to identify the key immune components required to promote successful immunotherapies.

During the last decade, the analysis of large libraries of annotated human tumors has allowed researchers to identify robust markers of protective anticancer immune reactions. In the vast majority of cancers, a strong CD8 T cell infiltration, in conjunction with a Th1 functional orientation, correlates with better progression-free survival (PFS) and overall survival (OS) [6]. In this regard, colorectal cancer has been extensively studied and represents a paradigm of these findings. It has been shown that not only the overall density of memory Th1/CD8 T cells was important but also the localization of the immune microenvironment [7]. Thus, both the density of these cells inside the tumor core (CT) and at its invasive margin (IM) act as complementing prognostic factors [8]. These observations have led to the definition of an immunoscore, based on the quantification of infiltrating T cell densities using two markers (CD8/CD45RO or CD3/CD8) in these two regions of the tumor microenvironment defining a score ranging from 0 to 4, which predicts with high significance PFS and OS in colorectal cancers up to stage III [9]. The immunoscore is presently tested in a large international cooperation effort on over 9000 colorectal cancers [10].

Despite the well-established prognostic value of tumor-infiltrating T cells in human cancers, major questions remain unanswered. In particular, what are the key elements involved in shaping an efficient immune microenvironment, what are the contributions of the malignant cells and of the surrounding tissues in shaping the tumor’s immune contexture, and to which extent do metastatic tumors escape immune control?

Heterogeneity of the immune microenvironment between patients and between tumor types

The rationale for defining a prognostic marker such as the immunoscore is that patients afflicted by a given malignancy differ in their intratumoral infiltration of memory Th1/cytotoxic T cells. Indeed, while the percentage of patients with a high immunoscore is smaller in locally advanced primary colorectal cancers (stage T4 versus T1) and in primary tumors of metastatic patients [7], it is a prognostic factor even after correcting for the pathological stages of the disease: Indeed, patients with a high immunoscore but large tumors and even lymph node metastases have a longer PFS and OS than patients with less advanced cancers but with a low immunoscore [11]. This observation is not fully consistent with the 3E’s hypothesis, which proposes a linear evolution of the immune control on cancer progression, from an elimination phase to an escape phase, passing through an equilibrium phase [12]. Our observations support the possibility that at each cancer stage, there are patients at different E phases. They may result from the fact that, unlike in mouse models where tumors are genetically quite homogeneous, human cancers exhibit a high diversity of carcinogenesis mechanisms and inter-patient heterogeneity [13].

In addition to the heterogeneity of the immune reactions between patients, the analysis of different cancer types has revealed interesting patterns. Indeed, T cell infiltration, and particularly by CD8 T cells, correlates with favorable prognosis in most human cancers—it is the case for colorectal cancer, non-small cell lung cancer, head and neck, breast, bladder, ovarian, esophageal and pancreatic cancers, prostatic and hepatocellular adenocarcinomas, Merkel cell and urothelial carcinomas, as well as melanoma [6]. However, an increasing number of exceptions are being reported. They include notably clear cell–renal cell cancer [14], Hodgkin lymphoma [15] and ocular melanoma [16]. In these cancers, a high tumor infiltration by CD3 and/or CD8 T cells correlates with shorter patient survival. These apparently contradictory results have to be interpreted in view of the large complexity of the tumor microenvironment. In addition to the adaptive immune reaction illustrated by the infiltration of Th1/cytotoxic T cells, various cells can inhibit protective T cell responses, such as myeloid-derived suppressor cells (MDSC) or regulatory T cells (T_reg) cells, or generate a local inflammation. It is well established that the latter supports tumor outgrowth and spread by producing growth and angiogenic factors, such as VEGF, that increase tumor vascularization, resistance to apoptosis, and, together with suppressor cells, inhibit T cell responses [17, 18]. Thus, depending on their association to overall vascular, inflammatory and suppressive microenvironment, CD8 T cells may exhibit different functional states, and extensive infiltration may be associated with different clinical outcomes [19]. Finally, several studies described the presence of high numbers of B lymphocytes in tumors. Their deleterious impact has been reported in mouse models [20], but their density has been associated with favorable prognosis in recent reports on human colorectal [7], non-small cell lung cancers [21], melanoma [22] and breast carcinoma [23]. There remains much to understand about their function in antigen presentation, cytokine production as well as the identification of the anti-tumor antibodies they may produce.

Shaping an efficient immune microenvironment: the role of tertiary lymphoid structures (TLS)

It is generally accepted that adaptive immune reactions are generated in the draining lymph nodes of an infected or inflamed site [24]. In the case of an immune response to a solid tumor, this view is complicated by the fact that the draining lymph nodes are usually the first sites of implantation of metastatic spread—the sentinel lymph node being indeed used as a prognostic marker. Nevertheless, since the composition of the immune microenvironment of primary tumors in patients with lymph node metastases remains a prognostic factor [25], it strongly suggests that there are extranodal sites for generating the tumor-targeted adaptive immune response.

Studies in lymphotoxin alpha deficient mice which lack lymph nodes have revealed that the clearance of a lung infection in these mice was associated with the formation of TLS in the infected lungs, where T and B cell responses to the infective agent qualitatively resembling those initiated in lymph nodes were generated [26]. Indeed, TLS present all the characteristics of active immunological sites, with T cell and B cell zones, dendritic cells (DCs) making contacts with lymphocytes and a germinal center [27], and are thus by all means similar to secondary lymphoid organs (SLO), where canonical immune responses are generated. In man, with the exceptions of mucosal-associated lymphoid tissue (MALT) in the gut, such structures are not present in tissues from adults under physiological conditions. In most cases, the immune response to an infectious agent is generated in the draining lymph nodes by mature DCs (mDCs) that have internalized antigens at the infected and inflamed site and have travelled to the lymph node [28]. In pathological conditions, however, when there is a strong and persistent antigen release, TLS can form in the inflamed tissue. Such structures have notably been reported in sites of graft rejection [29], inflammatory diseases [30] and persistent infections [31].

Tumors present all the characteristics necessary to induce TLS: antigen persistence, inflammatory context and chronic immune reaction. We have searched for and successfully identified TLS in tumor sites, identified the chemokines locally expressed, characterized their structure and cellular composition and analyzed their clinical impact. The seminal study was conducted in non-small cell lung cancer where TLS, initially called Ti-BALT for “tumor-induced bronchus-associated lymphoid tissue,” were fully characterized [32]. They are indeed composed of a T cell zone and a B cell zone, with mDCs infiltrating the T cell zone and physically contacting T cells [33, 34]. B cells associate with follicular DCs (fDCs) to form a germinal center with all the characteristics of active antibody production sites, such as the expression of activation-induced cytidine deaminase (AID), necessary for somatic hypermutation and class switch recombination, or that of BCL6 [21]. They are surrounded by high endothelial venules [35]. An extensive analysis of the chemokines expressed within the TLS showed that they are reminiscent of the ones involved in SLO formation in lymph nodes, with the addition of IL-16. Thus, CCL19, CXCL13, CCL21, CCL22, CCL17 and IL-16, produced by tumor-infiltrating cells, but also, for some of them, by malignant cells, are preferentially found in TLS where lymphocytes express the corresponding receptors. Integrins such as ITGαL, ITGαD, ITGα4 and adhesion molecules such as ICAM-2, ICAM-3, VCAM-1 and MaDCAM-1 are also involved in TLS formation [33].

Once formed in the inflammatory milieu of a growing tumor, TLS will be sites of T and B lymphocyte activation, which will contribute to tumor growth control. Naive and central memory lymphocytes directly penetrate into TLS through high endothelial venules, without passing through the immunosuppressed milieu of the tumor [34], and T and B cells are activated and educated, respectively, by mDCs in the T cell zone and by fDCs in the B cell zone. T cells in tumors with a high density of TLS present a Th1/CD8 memory orientation [34], and the plasma cells generated were shown to produce antibodies that recognize tumor-associated antigens [21]. Additionally, educated T and B cells may leave TLS through lymphatics, and one can postulate that they migrate within the primary tumor’s microenvironment but also travel to the lymph nodes, where they are responsible for both local and systemic long-term protection. The fact that antibodies with specificity for the same tumor antigens are found produced by intratumoral plasma cells and detected in the serum of the patient [21] supports this model.

Extensively studied in non-small cell lung cancer, TLS have been observed in many other cancer types: colorectal, breast, ovary, renal cell [36] and pancreatic cancers. In all cases, their density correlates with favorable prognosis, and so do the presence of associated high endothelial venules, mDCs or follicular B cells and T follicular helper cells (T_fh) [7, 21, 32, 37]. Moreover, they modulate the clinical impact of CD8 T cells. Thus, the survival of patients with non-small cell lung cancer exhibiting both high densities of intratumoral CD8 T cells and TLS-associated mDCs has a longer survival, even at advanced stages (stage I–III), than patients with high intratumoral CD8 T cell density but low TLS-associated mDC density, underlying the fact that the site of generation of intratumoral T cells is a major modulator of their clinical impact [34]. It is likely that tumor-associated TLS represent such sites. The fact that the induction of TLS is associated with an efficient response to immunotherapies, such as in vaccination against human papilloma virus (HPV) in cervical intraepithelial neoplasia [38], support this hypothesis.

Shaping an efficient anti-tumor immune microenvironment: the role of the malignant cells

The involvement of the host and of the cancer in shaping the tumor immune microenvironment is still an open question, which can crucially help identifying patients who could benefit from cancer therapies. Although the respective impacts of two very complex systems such as the host immune system and the cancer cells are difficult to decipher, some clues are beginning to emerge. It is striking that, with a few exceptions such as immunosuppressed virally infected [39] or chemically immunosuppressed patients [40], cancers arise and develop in immunocompetent individuals. There must therefore be tumor-produced factors that impair cancer-controlling immune responses without inducing systemic immunosuppression. These factors are likely to be present locally, in the tumor microenvironment.

One characteristic of tumor cells that may favor the development of a beneficial immune microenvironment is the expression of potential neo-antigens at the surface of tumor cells, able to induce and sustain adaptive anti-tumor immune reactions. In mouse models, tumors grown in immunodeficient mice were shown to harbor neo-antigens, which elicit an immune response [41]. In human, it is illustrated by the subtype of colorectal cancer with microsatellite instability (MSI) where mutations affecting the DNA-repair machinery provoke the membrane exposure of neo-antigens, which elicit Th1/cytotoxic T cell responses, associated with favorable prognosis (Fig. 1a). A recent analysis of human cancers indeed revealed that tumors considered as immunogenic such as melanoma, colorectal cancer or lung cancers have a high rate of somatic mutations, which could be responsible for the generation of immune-recruiting neo-antigens [13].

Fig. 1.

Two models of tumor immunotypes. In a, we represent the classically described adaptive anti-tumor immune response, characterized by a strong infiltration of CD8+ memory T cells and the presence of TLS. Immune escape is achieved through selection of weakly immunogenic tumor cells. In b, we represent the tumor’s immune contexture in case of tumor-induced chronic inflammation. In this case, high infiltration by CD8+ T cell will lead to the expression of PD-L1 on tumor cells, while regulatory T cells and myeloid cells are recruited in the microenvironment. In this case, CD8+ T cells’ suppression and inflammatory signals foster tumor growth

There are other ways by which malignant cells may contribute to build a beneficial, or deleterious, microenvironment. The multiple genetic and epigenetic modifications in tumor cells may result in the production of angiogenic factors, chemokines and cytokines by the malignant cells that ultimately modulate the tumor microenvironment. For instance, high production of VEGF by renal cell cancer cells remodels the tumor microenvironment not only by increasing angiogenesis but also by favoring the activation of T_reg cells and MDSC [18] (Fig. 1b). We have recently shown that, in lung metastases of renal cell cancer, high expression of VEGFA mRNA is associated with high expression of Th1-related genes but also with inflammatory and immunosuppressive cytokine transcripts, ultimately correlating with low patient OS, despite a high intratumoral density of CD8 T cells [42]. We have recently compared the expression of immune-related genes in large series of cancer cell lines from renal cell and colorectal cancers and found different patterns of immune genes expression (Becht et al. unpublished). That may explain the overall contrasted clinical impact of the immune microenvironment in the two diseases [19].

In addition to the differences between cancer types, the malignant cell’s genotype and phenotype may also explain the heterogeneity of the immune landscape within a given tumor type. To address this question, we analyzed the immune-related gene expression in a cohort of colorectal cancers, which had been classified in six molecular subgroups, identified using a consensus clustering approach. These groups are associated with significantly different PFS and OS. We found high immune-related gene expression in essentially two groups, the good prognosis-MSI group and the poor prognosis-stem cell-like group [43]. However, immune-related markers’ expression qualitatively and quantitatively differed between the two groups: Whereas a strong expression of genes associated with activated DCs and T cell cytotoxicity was associated with the MSI group, a strong expression of genes associated with immature DCs, inflammatory cells and mast cells was detected in the stem cell group (Becht et al. unpublished data). In addition, the amplification or deletion of chemokine and cytokine genes in tumor cells revealed striking differences between colorectal tumors. Among those, tumors with a deletion in the CXCL13 gene had a low infiltration of both T_fh and B lymphocytes correlating with low PFS [7], whereas tumors with a deletion of the IL-15 gene had a low number of proliferating T and B cells correlating with low PFS [44]. As a whole, these observations support the hypothesis that tumor cells may locally favor host immune responses by the generation of neo-antigens through frameshift or other somatic mutations, the production of CXCL13 favoring the formation of TLS, and of IL-15 helping the proliferation of potential tumor-specific T and B lymphocytes.

Finally, in a cohort of metastatic clear cell–renal cell cancers where four molecular groups have been defined [45], high expression of genes related to inflammation, Th1/cytotoxic immune response and immunosuppression was associated with tumors displaying a CpG islands methylator phenotype, low differentiation and with low PFS and OS as well as a poor response to treatment by sunitinib, suggesting that tumor-induced inflammation fosters CD8 T cells suppression and impedes response to anti-angiogenic treatments.

Overall, these findings explore how immune-related gene expression is associated with different clinical outcomes in colorectal and renal cell cancers [19], but, more importantly match patients with different cancer phenotypes to their corresponding immune contextures, which may guide patient selection and drug discovery for immunotherapies.

Conclusion

Immune markers are strongly related to patient’s outcome. Recent advances in tumor genomics shed light on the high molecular heterogeneity within most malignancies. In the meantime, tumor immunologists identified markers of an efficient anti-tumor immune response and more recently of deleterious immune responses, leading to the concept of immune contexture. This concept encompasses both the extent of the anti-tumor response, as measured by the quantity of cytotoxic effector cells, and the quality of the response, which depends on the sites where the anti-tumor immune response is mounted and the type of inflammation that reigns in the tumor microenvironment. It appears that the molecular phenotypes of tumor cells play a major role in shaping their immune contextures, both in the primary and metastatic locations, which can help identify patients able to most benefit from immunotherapies (Fig. 1). Further understanding of the cross talk between tumor and immune cells in the microenvironment may also help develop new treatments.

Abbreviations

AID

Activation-induced cytidine deaminase

CAR

Chimeric antigen receptor

CIN

Cervical intraepithelial neoplasia

CT

Core tumor

DC

Dendritic cell

HPV

Human papilloma virus

IM

Invasive margin

MDSC

Myeloid-derived suppressor cell

MSI

Microsatellite instability

NSCLC

Non-small cell lung cancer

OS

Overall survival

PFS

Progression-free survival

SLO

Secondary lymphoid organ

Ti-BALT

Tumor-induced bronchus-associated lymphoid tissue

TIL

Tumor-infiltrating leukocyte

TLS

Tertiary lymphoid structure

T_reg cell

Regulatory T cell