Anti-tumor immunity: Myeloid leukocytes control the immune landscape

Abstract

The immune surveillance hypothesis proposed over 50 years ago that many precancerous lesions are eliminated without a histological trace due to immunological pressure. Since then, it has become apparent that both the tumor and the anti-cancer immune response evolve over a long period to allow the eventual escape of nascent precancerous lesions into full-blown tumors. Although primarily focusing on loss of antigenicity, the immunoediting hypothesis has gradually evolved to appreciate the role of active immunosuppression in tumor progression, where myeloid leukocytes are increasingly recognized as the major driving force. This review highlights recent studies implicating how myeloid cells with antigen-presenting capabilities are co-opted by tumors to promote malignant progression. Because at least some advanced tumors remain significantly immunogenic, these new studies add a tweak to the immunoediting hypothesis as well as a rationale to block immunosuppressive mechanisms as a first-line intervention in cancer patients.

1. INTRODUCTION

Burnet [1] and Thomas [2] originally posited that nascent tumor lesions were eliminated by the immune system without a pathological trace. Since then, overwhelming experimental evidence demonstrates that both the innate and adaptive immune systems play a non-redundant role in the prevention or promotion of tumorigenesis. Immune recognition of tumor antigens lead to the formulation of the cancer immunoediting hypothesis, which supports that immune pressure - primarily mediated by T cells - results in progressive loss of antigens (editing) by tumor cells, eventually allowing them to escape from accumulating immune pressure [3]. Loss of natural, spontaneous (relevant) antigens has been conclusively demonstrated in carcinogen-induced tumor models [4]. However, T cell infiltration is clearly associated with superior outcomes in patients with many different tumors [5–8], while clinically relevant responses have been achieved against many tumors using T cell based immunotherapies [9–11]. Most importantly, emerging clinical evidence indicates that blockade of immunosuppressive signals such as CTLA4 and, especially, PD-1/PD-L1, allows the immune system to regain control of the progression of a variety of tumors [12, 13]. These clinical data and recent experimental evidence produced by our group [14] support that advanced tumors remain sufficiently immunogenic for effective control by the immune system, adding weight to the role of immunosuppression as a major driver of malignant progression. Pathological expansion of a heterogeneous population of immature myeloid cells with immunosuppressive activity is a hallmark of virtually all solid tumor-bearing hosts, and these cells are emerging as key players of immune regulation in the tumor microenvironment (TME) [11]. Paradoxically, myeloid leukocytes with antigen-presenting capabilities are required for the orchestration of tumor-specific T cell responses. Correspondingly, we recently identified a progressive phenotypic and numerical switch in dendritic cell (DC) populations in tumor-draining lymph nodes, parallel to both malignant progression and the abrogation of T cell-mediated protection [14]. The pivotal interplay between lymphoid and myeloid cells in the TME for preventing tumorigenesis vs. dampening the anti-tumor immune responses, and how to modulate it in vivo to control established tumors, will be the focus of this review.

2. Innate and adaptive immunity during tumor initiation and malignant progression

Studies using mice deficient in immune effector molecules have emphasized the critical role of innate and adaptive immunity in tumor initiation and malignant progression. Challenge of these immune-deficient mice with chemical carcinogens such as methylcholanthrene (MCA) or 7,12-dimenthylbenz[a]-anthracen (DMBA)/12-O-tetradecanoyl phorbol-13 acetate (TPA), resulted in accelerated generation of sarcomas or skin tumors compared to control WT mice with fully functional immune effector molecules (reviewed in [15]). Innate cells such as NK, NKT cells, γδ T cells, eosinophils [15, 16] and neutrophils [17–19] mediate immune protective or tumor promoting functions in experimental models for cancers, similar to what was observed in cancer patients. NK cells in particular appear to be critical for the rejection of nascent tumors [20]. However, so far only T cells in the TME have been associated with clinically relevant immune pressure against the progression of the established tumors that are detectable in the clinic. Thus, although it is theoretically possible that NK cells “hit and run”, still being important although absent from tumor locations, current clinical evidence supports that the adaptive immune system, and in particular T cells, are the crucial effector immune cells that remain able to exert some significant (although obviously suboptimal) anti-tumor activity in advanced malignancies.

The specific T cell subsets empowered with anti-tumor activity are the subject of intense debate, because both γδ T cells and CD4 αβ lymphocytes include cells with enough spontaneous regulatory activity to dampen protective immunity [21, 22]. Consequently, some studies have restricted the protective role of tumor-infiltrating lymphocytes to CD8 T cells [23], although both CD4 and γδ T cells are known to contribute to the orchestration and maintenance of adaptive effective immune responses through a variety of cytotoxic [24–26] and non-cytotoxic mechanisms [27–29]. Nevertheless, the prognostic value of T cell responses implies that antigen-presenting cells are able to effectively prime T lymphocytes at some point during tumor progression. Consequently, it has been shown that CD8α+ DCs are important in cross-presenting the tumor antigens to CD8+ T cells, so that in Batf3-deficient CD8α+ DCs, T cell mediated tumor rejection is impaired [30]. Furthermore, we have shown that the elimination of DCs in nascent tumor-bearing hosts dramatically accelerates malignant progression in an ovarian cancer model, which is restrained by CD8 T cells [14]. Because T cell infiltration is clinically relevant, and also because antibodies against tumor antigens are detected in a variety of cancer patients (which requires the activation of at least CD4 T cells), effective cooperation between T cells and DCs presenting tumor antigens appears to be taken place at initial stages of tumor progression. However, despite the fact that immune system can mount strong anti-tumor responses, tumors still evade the immune pressure [31]. Although the “self” nature of non-viral tumor antigens partially explains suboptimal T cell responses, the activity of tumor-specific T cells is further paralyzed in the TME through multiple complementary mechanisms.

3. Tumor immunoediting

Seminal studies by Schreiber and colleagues using chemical carcinogens such as MCA in immunocompromised mice conclusively demonstrated that tumors developed in the absence of adaptive immune system are more immunogenic in subsequent transplantation into WT mice [32]. IFNγ was found to be the principle molecule involved in tumor cell editing and both CD4 and CD8 T cells are the mediators of this strong anti-tumor response [32]. These studies are the cornerstone of the immunoediting hypothesis, which is the current framework accepted by most tumor immunologists [3, 33, 34]. The immunoediting hypothesis proposes that adaptive immune response not only regulates the quantity but also the quality of anti-tumor immunity, and has three important windows in which anti-tumor immune responses occurs. It starts with an elimination phase, during which cells of the innate and adaptive immune system eliminate cells undergoing transformation. If this elimination is complete, tumors disappear at this stage. Though experimental data supports this elimination phase, it cannot be characterized in humans because these events take place before tumors become detectable, if they ever become established. If tumor cells escape immune rejection, tumor progression goes through an equilibrium phase whereby tumors are kept under the control of effective immune responses, primarily mediated by components of the adaptive immune system. This phase culminates with three possible outcomes: First, the immune system can override the tumor cells and eliminate them. Second, this phase is continual and individuals remain free of clinically relevant tumors for their life-time. The third possibility is that adaptive immunity edits the tumors in such a way that new tumor cell variants develop, for which no T cell clones exist in the immune system. In that the case, edited tumor cells were proposed to evade the immune pressure, leading to accelerated expansion and, subsequently, development of clinical symptoms [3, 33–35].

While progressive loss of antigenicity has been experimentally supported and has provided a valuable framework for years, most data supporting the editing hypothesis derive from chemically (MCA or DMBA/TPA) induced tumors, or cell lines derived from them (reviewed in [15]). The value of artificial antigens that do not reflect the mild responses induced by tumor antigens (such as ova [36]) should be interpreted with more caution. The issue associated with chemically induced tumor models is that high degree of variability in mutated antigens between each mouse in the study groups as chemicals induce random mutations. Therefore, immune responses as well as editing will be variable in these models. Furthermore, experiments in which secondary transfer of transplanted tumor cells result in tumor escape may be due to deregulation and enhancement of the proliferative capacities of tumor cells [37], and not necessarily alterations in the antigenic repertoire. A seminal step to define these mechanisms was provided by a recent cancer exome analysis of MCA induced sarcomas, which identified spectrin-b2 as a potential tumor rejection antigen in MCA and T cells selectively exclude the cells expressing this mutations during the course of tumor evasion [4]. However, the fact that multiple established tumors become at least partially immunologically controlled simply by blocking T cell checkpoints [12], indicates that tumors cells can be still recognized by the immune system and therefore retain the expression of relevant antigens, which is further supported by our recent experimental observations [14]. The cancer immunoediting hypothesis has correspondingly evolved to integrate immunosuppression (in addition to loss of immunogenicity) as a relevant mechanism behind the escape phase of tumor progression [38]. The question is which is the predominant mechanism in the oncogene-driven tumors that take place in humans?

4. T cell unresponsiveness in the tumor microenvironment

Recently the hallmarks of cancer have been modified to incorporate additional characteristics of cancer in the context of how cancer subverts the immune system. These include tumor-promoting inflammation, reprogramming energy metabolism, and evasion of the immune system [39]. Thus, solid tumors maintain an immunosuppressive, hypoxic and hostile environment that directly affects the effector function of T cells. Sustained exposure to suboptimal antigen levels and multiple suppressive factors can result in unresponsiveness through T cell exhaustion, anergy or senescence, 3 mechanisms that use different molecular pathways [40]. Studies in chronic viral infections have unveiled that T cell exhaustion is characterized by a progressive weakening of effector activity, expression of inhibitory receptors (e.g., PD-1 TIM3, LAG-3 and CTLA-4 (reviewed in [41]) and a transcriptional state that includes the overexpression of Blimp-1 and T-bet, along with up-regulation of NFAT2 in the presence of suboptimal levels of AP1 [40]. An identical phenotype is identified in the microenvironment of many tumors, particularly in CD8 T cells, where the expression of inhibitory receptors is required for induction and maintenance of T cells in exhausted state. Ligands for these receptors are generally expressed by regulatory DCs and myeloid derived suppressor cells (MDSCs), in addition to tumor cells.

In contrast to the progressive nature of T cell exhaustion, anergy is rapidly initiated at the time of priming, and is characterized by the up-regulation of Rnf128, Egr2 and Egr3, and diminished Ras activation, along with excessive NFAT [42]. Maintenance of anergy is antigen independent while maintenance of exhaustion dependents on persistent antigen availability/TCR signaling [43]. Importantly, both exhaustion and anergy can be reversed through, respectively, the blockade of inhibitory pathways and cytokines [42][44, 45]. As commented above, emerging clinical evidence supports the promise of blocking some of these inhibitory receptors [12].

In addition to anergic and exhausted T cells, senescent lymphocytes with shortened telomeres that have reached their terminal replicative potential are also found in the TME, particularly in elderly patients. These cells are characterized by the expression of CD57 and the absence of CD28 [46], and unresponsiveness is considered to be permanent. Irreversible cell cycle arrest can also be caused by a signal transduction program induced by cellular stress [47], although these molecular pathways remain largely uninvestigated in T cells.

Besides intrinsic transcriptional programs leading to T cell unresponsiveness, many factors in the TME abrogate the activity of effector T cells. Interestingly, some of these mediators are not only produced by tumor cells, but also by DCs that, rather than promoting anti-tumor immunity, are transformed into immunosuppressive players. Those factors include Indoleamine 2,3-dioxygenase (IDO) and L-arginase, enzymes secreted by tumor cells, CD8α⁺ DCs with tolerogenic phenotypes and MDSCs [48, 49]. These enzymes deplete Amino Acids that are required for T cell functions from the TME [49, 50]. IDO catalyzes the tryptophan degradation in the kynurenine pathway [51]. Both the reduction in tryptophan concentration as well as accumulation of tryptophan metabolites is immunosuppressive. In addition, tumor-infiltrating DCs and MDSCs actively contribute to the suppression of anti-tumor CD8 T cells through the production of L-arginase [49, 52, 53]. Other potent immunosuppressive factors are secreted by both myeloid leukocytes and tumor cells, including TGFβb [41]. Therefore, tumor immune evasion is the outcome of complex immunosuppressive mechanisms paradoxically driven by myeloid leukocytes, which eventually paralyze protective T cell responses.

5. Myeloid leukocytes and tumor-induced immunosuppression

The presence of exhausted tumor-specific T cells and (CD4-dependent) tumor antigen- specific antibodies in most cancer patients indicates that at least a fraction of tumor-reactive lymphocytes are effectively primed at early stages of tumor progression. So, how are myeloid leukocytes responsible for the orchestration of adaptive immune responses turned into immunosuppressive cells in tumor-bearing hosts? The answer is that a hallmark of virtually all advanced solid tumors is excessive mobilization of bone marrow precursors of myeloid leukocytes (including macrophages, dendritic cells and granulocytes), in response to multiple inflammatory cytokines [54–57]. This heterogeneous population, globally termed MDSCs, home to tumor locations in response to multiple chemokines, but they also exert immunosuppressive activity beyond the TME (reviewed in [58]). Among the multiple tolerogenic mechanisms that they promote, nitration of tyrosines in TCR-CD8 complex appears to be particularly relevant [56, 59]. Once inside the TME, maturation of these myeloid cells into immunocompetent antigen-presenting cells is derailed, resulting in diminished adaptive immunity and eventual tumor escape. Thus, under hypoxic conditions, Ly6C⁺ MDSCs differentiate into immunosuppressive macrophages and DCs in solid tumors [60]. Correspondingly, the categorization of the highly heterogeneous myeloid populations that massively accumulate in solid tumors is complicated by a high degree of phenotypic overlap, different stages of differentiation, predominant inflammatory signals produced by every specific tumor, and the location and histological type of the tumor itself, among other factors. Applying the markers and functional attributes of leukocytes categorically defined under steady-state conditions to immune cells in the TME is therefore very challenging. Nevertheless, immunosuppressive, pro-angiogenic CD11b⁺CD68⁺MHC-II⁺ macrophages are represented in virtually all solid tumors. In addition, we have repeatedly demonstrated that the predominant leukocyte subset found in solid ovarian tumors (but not in human tumor ascites) co-expresses determinants of bona fide DCs, including CD11c, DEC205, CD86 and MHC-II, and in at least a third of clinical specimens lacks the macrophage markers CD11b and CD14 [61–67]. From their perivascular location, these myeloid cells abrogate the activity of anti-tumor T cells extravassating from blood vessels into the tumor microenvironment [66]. The expression of PD-L1 by ovarian cancer-associated DCs appears to be particularly relevant immunosuppressive mechanism, based on multiple converging lines of evidence [23, 63, 68]. Additionally, pDCs isolated from tumors in the prostate expressed high amounts of IDO and TGFβ to promote immune suppression and VEGF-A, and IL-6 to promote angiogenesis and metastasis [69]. Therefore, although the role of other immunosuppressive leukocyte subsets such as Treg is also relevant for tumor progression [21], the abundance and per cell immunosuppressive activity of myeloid cells in the TME indicates that this heterogeneous population is the major driving force for the abrogation of anti-tumor immunity in the TME. In addition, how and to what extent myeloid leukocytes control the conversion of inducible Treg remains largely uninvestigated.

6. Tumor mediated escape: Dendritic cell conversion

DCs and macrophages are sentinels in immunity, and are required to respond rapidly to infection or to be able to quickly modulate robust inflammatory responses. Because of this plasticity in function and phenotype, myeloid-derived cells are vulnerable to the polarizing signals elicited by the tumor and tumor microenvironment. For instance, mobilization of monocytes from the periphery due to recognition of bacterial ligands [70] or during inflammation in the intestine [71], can give rise to inflammatory dendritic cells and possibly conventional CD103⁺ DCs capable of inducing potent T cell responses. Conversely, tumor associated fibroblasts, through depletion of GMCSF in the tumor microenvironment, are capable of converting CD11c⁺ dendritic cells into macrophages with potent immunosuppressive capabilities [72]. To investigate the dynamics of plastic antigen-presenting cells from tumor initiation to terminal malignant progression, we recently generated an inducible model of ovarian carcinoma driven by mutations in oncogenes and suppressor genes, as it happens in humans [14]. As expected, we found that measurable tumor-specific T cell responses are orchestrated shortly after tumor initiation by immunocompetent DCs. These responses were enough to keep tumors as microscopic lesions for relatively long periods. Correspondingly, depletion of DCs 7 days after tumor challenge resulted in a dramatic acceleration of tumor growth.

Paradoxically, the initiation of malignant macroscopic expansion was dependent upon the accumulation of CD11c⁺DEC205⁺MHC-II⁺ DCs within the TME. However, these cells were not only unable to effectively present tumor antigens, but also abrogated the robust priming of T cells elicited by different immunocompetent DCs. Consistently, depletion of DCs at advanced stages of tumor progression significantly delayed tumor growth, allowing the immune system to regain control of tumors, again in the absence of any direct intervention on tumor cells [73]. These results demonstrate that myeloid leukocytes, and in particular DCs in ovarian tumors, govern malignant progression, as tumor growth can be modulated in opposite directions simply by eliminating this microenvironmental cell type at different stages. Our data also support that advanced tumors remain immunogenic, because cells from advanced tumors were able to induce significant T cell responses, particularly in lymphocytes derived from early tumors. Most importantly, our results provide a framework to understand the progression of aggressive epithelial tumors, whereby transition from microscopic lesions to exponentially growing masses could be occurring without a pre-malignant or dormant detectable lesion.

So, what factors induce the conversion of immunosuppressive DCs? Our data indicate that tumor cell derived PGE2 and TGFβ are sufficient to initiate the switch in DC phenotype from an immunostimulatory to immunosuppressive phenotype [73]. Ongoing studies should clarify which one is the predominant mechanism of tolerization in vivo. Other pathways potentially involved in the immunosuppressive activity of advanced tumor DCs include hypoxia within the TME, which induces DCs that are capable of presenting peptides but have impaired antigen processing capabilities and express significantly higher levels of VEGF-A, CXCL1, and CXCL8; chemokines all implicated in promoting angiogenesis in multiple forms of cancers [74]. Although we do not find high levels of IL-10 in human or mouse ovarian cancers, secretion of IL10 can also lead to DC-inhibition of maturation in different tumors [75], in addition to inducing the expression of immunosuppressive OX40 ligand via production of thymic stromal lymphopoietin (TSLP) [76]. Finally, increased lipid accumulation in tumor associated DCs in both humans and mice resulted in a diminished ability to process and load antigen onto MHC, resulting in ineffective antigen presentation to T cells within the TME [77].

Other tumor-derived factors that could be significant contributors to DC conversion include IL6, which induce Socs3 upregulation in tumor-associated DCs leading to inhibition of pyruvate kinase M2 (M2-PK) [78], an enzyme involved in aerobic glycolysis [79]. S100A8/A9 [80], which induce the massive recruitment of immune cells and prevent their differentiation within the TME [81], could also participate in this process.

7. Back to normal: In vivo re-programming of tumor DCs

Due to their massive accumulation and suppressive power, macrophages and DCs in the TME emerge as major therapeutic targets. Importantly, we have demonstrated that when these leukocytes receive certain activating signals, at least in mouse models, they can process full-length OVA in vitro[61] and in vivo [63, 67], and effectively present processed SIINFEKL to T cells. Therefore, interventions that achieve effective re-programing of immunosuppressive myeloid leukocytes in vivo into immunocompetent antigen-presenting cells, could be much more effective than their mere depletion, by simultaneously eliminating a major immunosuppressive driving force and boosting anti-tumor immunity in situ at tumor locations.

Ovarian cancer represents an ideal disease for these interventions because the TME is both compartmentalized and accessible. Supporting the feasibility of this approach in preclinical models, we have demonstrated that agonistic (and clinically available) CD40 and TLR agonists synergize to transform ovarian cancer-associated myeloid cells from an immunosuppressive to an immunostimulatory cell type [67]. Building on the insight of these studies, and by taking advantage of the enhanced endocytic pathways of tumor-associated DCs [63], we have more recently combined the synergy between the intrinsic TLR agonistic activity of double-stranded RNA and CD40 activation with the immunostimulatory activity of miR-155. Thus, we demonstrated that Dicer substrates mimicking the sequence and structure of endogenous miR-155 are selectively taken-up by tumor-associated CD11c⁺MHC-II⁺ DCs in mice growing aggressive orthotopic ovarian tumors when combined with biocompatible polymers, which synergizes with CD40 agonists [82]. Two important observations can be drawn from these studies; one, that DCs are major orchestrators of the immunosuppressive microenvironment, and two, in situ delivery of a potent antigenic stimulus is sufficient to reverse the tolerogenic phenotype, which provides a rationale for subsequent clinical testing.

8. Conclusions and future perspectives

Accelerated malignant growth coincides with the massive accumulation of immature myeloid leukocytes into the TME, which eventually breaks the dynamic equilibrium between protective T cell responses and proliferating tumor cells. Although tumors also lose recognizable antigens during their progression, they appear to remain significantly immunogenic to be controlled by existing anti-tumor T cells when inhibitory checkpoints are neutralized, as supported by experimental and clinical evidence. Because of their plasticity, myeloid leukocytes are highly susceptible to endogenous and exogenous signals within the tumor milieu. The presence of these cells within the tumor microenvironment is sufficient to tip the balance in favor of exponential tumor progression and escape from the immune pressure. However, myeloid cells (DCs) initially orchestrate measurable adaptive immune responses that can keep tumors in check for relatively long periods. Consequently, emerging evidence indicates that myeloid leukocytes govern cancer progression. Most importantly, partial reversal of the immunosuppressive genetic program of tumor DCs can be achieved in vivo and in situ by combining immunostimulatory agonists and delivering immune-activating miRNA mimetics. Understanding the genetic pathways and secretory factors that influence the mobilization of myeloid precursors and how to transform them from an immature to immunosuppressive phenotype should open new avenues for effective control of established tumors, besides iterations of chemotherapeutic drugs directly targeting tumor cells.

Figure 1. Dendritic cell plasticity influences tumor progression.

During aggressive malignant expansion of tumor cells, immature myeloid cells are recruited into the tumor microenvironment (TME) by CCL20 and CXCL12 produced by tumor cells or CXCL4 and S100A9 (upregulated in hypoxic environments). In the TME immature myeloid cells are converted into suppressive regulatory DCs by TGFβ and PGE-2 produced by the tumor cells. Suppressive DCs cooperate with the developing tumor mass to promote escape by secreting VegF and Il-6 (supporting angiogenesis), producing IDO and TDO2 (establishing a more hypoxic and immunosuppressive microenvironment) and secretion of immunosuppressive factors such as TGFβ and arginase (directly impeding T cell function). T cells become exhausted characterized by upregulation of inhibitory receptors such as PD-1 and CTLA4. This immunosuppressive tumor microenvironment impairs CD8 T cell anti-tumor responses, resulting in tumor escape.

Figure 2. In situ reversal of the immature phenotype of suppressive DCs with potent antigenic stimulus.

Reversal of DC phenotype can be achieved by delivery of potent antigenic stimulation using agonistic CD40 and poly(I:C) or by delivery of immunostimulatory nanoparticles complexed with a mimetic for miR155, resulting in the conversion of immunosuppressive DCs into immunostimulatory DCs. Following stimulation, costimulatory molecules such as B7-2 are upregulated and resulting in enhanced effector T cell function and inhibition of tumor progression.

HIGHLIGHTS.

• Measurable T cell responses control tumor progression from very early stages

• Accelerated malignant expansion is driven by immunosuppressive myeloid leukocytes

• Tumor cells remain immunogenic during the escape phase of tumor progression

• Neutralizing inhibitory checkpoints in T cells restores immune control of tumors

• Immunosuppressive leukocytes can be re-programmed in vivo and in situ