Dendritic Cells, the T-cell-inflamed Tumor Microenvironment, and Immunotherapy Treatment Response

Antibodies directed against T-cell surface coinhibitory receptors such as cytotoxic T-lymphocyte associated protein 4 (CTLA-4) and the programmed cell death protein 1 (PD-1) pathway can reinvigorate antitumor T-cell responses. These treatments, collectively termed immune checkpoint blockade (ICB), have become critical pillars in cancer treatment and demonstrate activity in a broad range of cancer types. However, the majority of patients receiving ICB do not have durable therapeutic responses. As clinical use of these antibodies has grown immensely, a key area of research aims to define patients likely to respond, or not, to checkpoint blockade. Key to this understanding are the principles of T-cell activation in cancer, and in particular how immunotherapies shift the balance of tolerance towards anticancer immunity.

Although hypotheses have been advanced as to why some patient's tumors respond, or fail to respond, to checkpoint immunotherapy, most in the field have been in agreement that T cells, particularly CD8⁺ cytotoxic T cells are the drivers of therapeutic response (1). Abundance of tumor-infiltrating CD8⁺ T cells, tumor mutational burden, and IFN signatures are correlated with response to anti-PD-1 therapy (2–4). Although biomarkers have become instrumental in understanding ICB, they do not explain the totality of treatment response and resistance. T-cell presence alone in tumors may not be enough to induce antitumor immunity as numerous nontumor-specific T cells also infiltrate tumors (5). Therefore, these suggest an important need in the field to further understand and distinguish T-cell-infiltrated from T-cell-inflamed tumor microenvironments and further elucidate factors in the tumor environment driving antigen-specific T-cell recruitment and activation.

Gene expression analysis of metastatic melanoma lesions identified a strong separation of samples based on T-cell-associated gene transcripts (6). Particularly, T-cell inflammation in tumors was associated with tumor chemokine expression, particularly CXCL9, CXCL10, CCL5, CCL4, CCL3, and CCL2. T-cell attracting chemokines can be produced by a variety of cell types, including tumor cells, dendritic cells (DC), and macrophages (6–9). The importance and relative contribution of each cell type may depend upon the context and cancer type. However, some tumors grow progressively even when infiltrated with antigen specific T cells, which may be due to immune suppressive feedback and exhaustion of T-cell responses (10, 11). T-cell activation is tightly controlled and requires initiation signals provided by antigen presenting cells, predominately DCs, such as TCR stimulation, costimulatory receptor ligation, and cytokine support. Sustained CD8⁺ T-cell responses against tumors are associated with DC supportive niches within the tumor bed, and patients experiencing progressive disease exhibit breakdown of these niches (12). Moreover, DC activation phenotypes, as measured by DC gene signatures, positively correlate with the T-cell-inflamed state as well as response to inhibition of the PD-1/PD-L1 pathway (13–15), suggesting a priority need in the field to further understanding of how DCs populate tumors and how to activate DCs to facilitate anticancer immunity.

DCs arise from a bone marrow-derived DC-specific pre-cursor cell (pre-cDC) and depend upon factors such as FMS-like tyrosine kinase 3 ligand (Flt3L) and GM-CSF for their development. DC ontogeny and classification has been thoroughly reviewed by other authors (16–18) and here we will focus only on key DC functions and subtypes. Pre-cDC seed tissues and proliferate to form peripheral DC networks. DCs can be resident within lymphoid organs, or they can surveil peripheral tissues and blood. Inflammatory conditions, including cancer, enhance DC accumulation within tissues from bone marrow sourced pre-cDCs (19). DCs endocytose material from their environment, and if appropriate stimulatory triggers are present, such as pattern recognition receptor (PRR) ligation, DCs mature and transition from antigen sampling to antigen presentation functions. Once mature, DCs upregulate surface expression of the chemokine receptor CCR7 and are attracted towards the CCR7 ligand CCL21 (produced by lymphatic endothelium) to migrate from peripheral tissues to T-cell zones of local draining lymph nodes (20). DCs traffic from tissues to draining lymph nodes in the steady state, although when tissues are inflamed DCs traffic in greater numbers. Homeostatic DC migration promotes immune tolerance, in contrast DCs migrating in response to inflammation engender T-cell immunity. This is likely due to the upregulation of inflammatory cytokines and costimulatory molecules in inflammation matured DCs, signals which are not present on steady-state migrating DCs (18).

There are multiple subclassifications of DCs—which expand further with enhanced profiling approaches (21, 22), however these can broadly be grouped into conventional DC type 1 (cDC1), conventional DC type 2 (cDC2), and plasmacytoid DCs (pDCs; Table 1). An additional subset of DCs, monocyte-derived DC, have been described to arise during inflammation and promote context-dependent differentiation of CD4⁺ T cells. The role for these cells relative to cancer immunotherapy is unclear however and will not be discussed further here. cDC1 are the DC subset most well known for their ability to cross present antigens to stimulate CD8⁺ cytotoxic T cells. These DCs express specialized antigen presentation pathways, which allow exogenous cross-presented antigens to be processed and presented on MHC-I molecules on the DC's surface (23). These cells express CD11c, MHC II, CD8α, XCR1, CLEC9a, CD24, and CD103. In humans, cDC1 also express CD141. cDC1 development depends upon the transcription factors IRF8, ID2, and BATF3, and loss of these transcription factors eliminates this DC subtype (24, 25). Mechanistically, BATF3 maintains activation of IRF8 to specify cDC1 lineage commitment, as transgenic overexpression of IRF8 can rescue cDC1 development (26). In contrast to cDC1, cDC2 potently stimulate CD4⁺ T-cell responses. These cells express CD11c and MHC II, and have surface CD11b, Sirpα, and CD301b (27). CD1c is a marker of cDC2 in humans, however mice lack CD1c genes. cDC2 function and development requires the transcription factors IRF4, RBPJ, KLF4, and RELB (18, 27). pDCs are major producers of type I IFN when activated through TLR7 or 9, although identification of the Axl⁺ Siglec6⁺ DC (asDC) subpopulation has called the T-cell stimulatory ability of pure pDCs into question (22). pDCs require the transcription factors TCF4 (E2-2), IRF8, and RUNX1 for development. These cells express CD11c and MHC II, and they can be distinguished by B220, Siglec-H, CD317, CD123, and CLEC4C in humans. pDC are present in blood and lymphoid organs, and can sometimes be found in tissues. In the setting of viral infection, pDC are known to augment CD8⁺ T-cell responses through type 1 IFN activation of cDC1(28), although this mechanism is unexplored in cancer.

Analysis of T-cell-inflamed melanomas demonstrated positive correlation with cDC1 gene expression signatures, indicating that the degree of tumor T-cell infiltration is strongly tied to cDC1 presence within tumors (13). In contrast to the hypothesis that only overall antigenic load dictates ICB response, it seems that the recruitment and activation status of tumor DCs, in particular cDC1, factor strongly into tumor T-cell inflammation. In support of this, melanoma cell intrinsic secretion of CCL4 can attract cDC1, although this can be blocked by activated β-catenin signaling (29). Studies of hepatocellular carcinoma have also connected the β-catenin pathway with cDC1 tumor infiltration and anti-PD-1 response (30). Indeed, activated β-catenin is associated with non–T-cell-inflamed tumors across a range of cancers (31, 32). Inhibition of tumor DC recruitment appears to be a dominant mechanism of tumor intrinsic β-catenin activation, although it is unclear whether β-catenin activation has diverse signaling effects across cell types. Tumor intrinsic mutations that strongly alter host immune parameters argue for a cancer cell dependent mechanism of DC recruitment to tumors.

Many tumor cells are genomically unstable, which results in the activation of cell intrinsic DNA sensing pathways such as the STING pathway (33). Tumor cGAS/STING is induced by DNA damage, notably that seen in the context of irradiation (34, 35). This pathway is activated upon double-stranded DNA (dsDNA) binding to cyclic guanosine monophosphate-adenosine monophosphate (cGAMP) synthase (cGAS). cGAMP then functions as a second messenger and binds STING to activate downstream signaling (36). This triggers interferon response genes and production of CCL5, which can attract cDC1s (37). Extracellular tumor DNA, or cGAMP can activate tumor DCs and initiate T-cell responses in the draining lymph node (34, 38, 39). Tumor cell loss of LKB1 is associated with poor T-cell infiltration and nonresponse to ICB therapy (40, 41), and correspondingly STING signaling pathways can be inhibited by LKB1 loss. This blunted T-cell recruitment to tumors may explain why LKB1 mutant tumors respond poorly to ICB (42). STING activity is regulated by several proteins notably including the DNA exonuclease Trex1(35), which degrades cytosolic DNA substrates that activate cGAS/STING, or via viral proteins that bind directly to STING (43). Further mechanisms of cGAS/STING pathway inhibition will likely emerge and blockade of these may represent therapeutic strategies to induce type I IFN.

Downmodulation of tumor suppressor genes and oncogene expression are now frequently associated with immune exclusion phenotypes, although at this point only β-catenin activation has been mechanistically tied to cDC1 biology (44). Absence of PTEN leads to decreased T-cell infiltration in melanoma with levels of CXCL10 transcripts decreased in PTEN mutant tumors (45). Emerging evidence demonstrates that IDH1 mutations are strongly correlated with non–T-cell-inflamed tumors and reductions in factors such as CXCL10 (46, 47). The oncometabolite R-2 hydroxyglutarate (R-2-HG), produced by mutant IDH1, has also been linked to direct metabolic suppression T-cell function (48). DC presence and activity has not yet been assessed in mutant LKB1, PTEN, or IDH1 tumors. It will be of substantial interest to learn whether lack of DCs mediates the non–T-cell-inflamed tumor microenvironment in these tumors.

Other mechanisms also likely influence cDC1 recruitment noting that NK cells are primary producers of the cDC1 attractive chemokines CCL5 and XCL1 (49), and the DC differentiation factor Flt3L (14). NK cells precede cDC1 recruitment in some models, suggesting that molecular mediators controlling NK accumulation within tumors could be important determinants of cDC1 infiltration. Tumor cell derived prostaglandin E₂ (PGE₂) can inhibit both NK and cDC1 recruitment (50), leading to a non–T-cell-inflamed tumor microenvironment. Oncogene influence on NK-cell responses is not well described, although DNA damage responses within tumors are also known to upregulate NK-activating ligands, suggesting that tumor intrinsic features could drive NK activation and recruitment (51).

Alternatively, tumor alterations have been identified that promote T-cell-inflamed tumors. Exacerbating tumor intrinsic nucleic acid sensing, via deletion of the double-stranded RNA (dsRNA) binding protein Adar1, leads to activation of the PKR and MDA5 cytosolic dsRNA sensing pathways, type I IFN induction, and enhanced response to anti-PD-1 (52). In this setting, tumor intrinsic type I IFN production is independent of RIG-I, another cytosolic dsRNA sensor, however tumor cell intrinsic RIG-I has been implicated in anti-CTLA-4 response, and combining anti-CTLA-4 with RIG-I agonist could enhance antitumor responses through cDC1 cross-presentation of tumor antigens (53). TREX1 knockdown in irradiated tumor cells led to enhanced levels of cytosolic DNA that rendered tumor cells more immunogenic through activating the cGAS/STING pathway (35). Loss of function of PBRM1 in murine melanoma sensitizes tumor cells to IFNγ and T-cell killing (54), likewise PBRM1 loss of function in clear cell renal cell carcinoma patients was associated with response to ICB therapy (55).

Collectively, these studies suggest a model where tumor intrinsic handling of innate immune pathways, in particular nucleic acid sensing pathways, plays a critical role controlling tumor immunogenicity through downstream recruitment and activation of DCs. These responses are then critical for initialization of the T-cell response and the T-cell-inflamed tumor state (Fig. 1).

Beyond the association of DCs and the T-cell-inflamed tumor microenvironment, the presence of the Batf3 transcription factor within DCs is becoming apparent as a necessary condition for anticancer immunity. This has been emphasized by Batf3 murine knockout systems, which manifest with cDC1 deficiency (25). Batf3 deficiency prevents spontaneous immunogenic tumor regression and impacts on therapeutic efficacy of many types of cancer immunotherapy, including but likely not limited to anti-PD-1/L1 response, adoptive transfer T-cell therapy, and tumor specific CD8⁺ T-cell responses (9, 25, 56–58). Specific dissection of cDC1 function has identified that cross-presentation defective mouse strains, lacking Wdfy4 or Sec22b, still retain cDC1 but these cells are incapable of inducing CD8⁺ T-cell responses to reject tumors (59, 60). cDC1 deficiency in Batf3 knockout results in similar findings as CD8⁺ T cells are not primed effectively and produce far less IFN- compared with T cells primed from wild-type DCs (58). cDC1 are also important T-cell chemoattractors to the tumor microenvironment through their production of CXCL9 and CXCL10, although it is unclear if cross-presentation is required in this setting (9, 61). CXCR3, the receptor for CXCL9 and CXCL10, is expressed by CD8⁺ T cells and is required for anti-PD-1 therapy in murine models (8). Anti-PD-1 responsive mouse tumors such as MC38 and MCA1956 had substantially higher CXCL9 and CXCL10 producing DCs compared with anti-PD-1 resistant models such as B16F10 and AT-3, and patients responding to anti-PD-1 had higher induction of these chemokines (8).

cDC1 also provide cytokine support to T cells assisting in their effector functions. IL12 is a well-known cDC1 produced factor that can drive CD8⁺ T-cell cytolytic activity and secretion of IFNγ (62, 63). Supplying exogenous IL12 intratumorally in combination with a T-cell agonizing CD137 antibody results in durable tumor rejection in MC38 murine models, however this effect is lost in Batf3 knockout mice (58). Direct administration of IL12 alone is sufficient to regress tumors in wild-type mice (64), suggesting that T-cell priming by Batf3 dependent cDCs is required for response to IL12 and that T cells are stimulated by IL12. This is further supported by the finding that tumor infiltrating T cells must receive both T-cell stimuli and IL12 to upregulate IFNγ production. Specific T cells, and the degree to which they are exhausted, responding to IL12 are unclear, although T cells isolated from mouse and human tumors can respond to IL12 (63). IL12 can drive CD8⁺ T cells to short-lived effector cell fates in models of infectious disease, and in this setting IL12 can be supplied by non-antigen-presenting bystander DCs (65).This could possibly be relevant in cancer as compartmentalized models of DC function have been proposed, but these models remain to be tested.

T-cell engagement can occur locally in the tumor microenvironment, or in the context of the lymph node after tumor antigen bearing DCs migrate. cDC1 migrating from the tumor are the major antigen carriers that initiate CD8⁺ T-cell responses (66), however some important differences exist between studies. For example, migratory cDC1 have been shown to carry antigen to the B16 melanoma draining lymph node (57), however both cDC1 and cDC2 have also been shown to traffic antigen to the lymph nodes in other systems (67). Aggregation of evidence strongly supports that migratory DCs in cancer, as the mature DC, are the most efficient at stimulating T-cell responses (57, 66–68). This is in contrast to viral infection mechanisms that demonstrate antigen handoff from migratory DC to resident DC, with resident DC as the T-cell stimulatory population (69, 70). It is possible that the lack of strong inflammatory triggers, which are present in viral infection, in cancer prevents antigen handoff mechanisms that would otherwise amplify the immune response. Likewise, augmenting tumor nucleic acid sensing pathways, pathways which also serve antiviral function can give stronger inflammatory triggers and render tumors immunogenic.

To date, evidence suggests a dominant role for cDC1 in antitumor immunity, however tumors are also frequently infiltrated by cDC2. These cDC2 are best able to initiate CD4⁺ T-cell responses upon migrating to the lymph node. T regulatory cells (T_reg) restrain cDC2 antitumor responses that activate CD4⁺ T cells (67) and T_reg can inhibit DC maturation while hampering migration to draining lymph nodes. Removing this suppression enables IFNγ, producing CD4⁺ T cells to accumulate in the tumor microenvironment and is independent of CD8⁺ T cells, suggesting that IFNγ responses may provide the dominant antitumor effect (52, 71). Tumors can also be shaped by MHC II neoantigens (72) as tumor cell expression of both MHC I and MHC II neoantigens elicited strong antitumor immunity. In this instance the CD4⁺ T cells assist in CD8⁺ T-cell priming and maturation. Abscopal antitumor effects were not seen unless the tumor had expression of both neoantigens, suggesting that local CD4⁺ T-cell activation in the tumor site, or draining lymph node, is required for a tumor immune response. These data imply that cDC2, which initiate CD4⁺ T-cell responses, can provide antitumor functions in some settings.

Modulation of DCs can be approached via a variety of methods, including but not limited to targeting of innate PRRs, supplying DC growth factors, and agonizing cell surface receptors. Modulation strategies presented here are not meant to be an exhaustive list, but rather a general overview of recent attempts to activate antitumor DC responses. Diverse methods of therapeutic DC cancer vaccination have been attempted over the past 25 years. Longer summaries of such approaches have been reviewed previously (73–75).

Agonism of toll-like receptors (TLR) drives proinflammatory gene programs that can engender antitumor immunity. The endosomal TLRs (TLRs 3, 7, 8, and 9) make particularly attractive targets as these receptors naturally sense foreign nucleic acid patterns such as dsRNA, ssRNA, and unmethylated CpG and trigger type 1 IFN responses. On the basis of current understanding of IFN response and tumor immunogenicity, these pathways are attractive therapeutically toward recapitulating microenvironmental features of T-cell-inflamed tumors. Agents such as Poly I:C (TLR3 agonist), NKTR-262 (TLR7/8 agonist), CMP-001 (TLR9 agonist), and Tilsotolimod (TLR9 agonist), among others, are in clinical development and testing. Clinical trial biomarker investigation of these agents has suggested increases in intratumoral IFN responses, T-cell chemokines and tumor-infiltrating CD8⁺ T cells (76–78).

Targeting the cGAS/STING pathway through STING agonism is another emerging therapeutic approach. Host DC uptake of tumor DNA induces a STING-dependent IFN response that is needed for antitumor immunity (38). Agonizing this pathway pharmacologically could be an approach to elicit antitumor DC responses (36). Clinical data have been disclosed for ADU-S100 and MK-1454 (79), demonstrating the induction of systemic type I IFN response, however little intratumoral biomarker data have been released. In general, the intratumoral injection approaches pursued to date may require optimization to improve clinical efficacy, although randomized phase II and phase III trials have been launched for TLR9 and STING agonists in combination with ICB.

Several other intriguing approaches are also on-going in clinical trials. Combining poly I:C activation with CDX-301 (human Flt3L) and radiotherapy in an approach termed “in situ vaccination” has been described as demonstrating abscopal antitumor responses in indolent non-Hodgkin's lymphoma with effect mediated by increased tumor T-cell infiltration and tumor cross-presenting DCs (80). Other dsRNA detection pathways like the cytosolic RIG-I sensor can be activated using the synthetic RNA oligonucleotide MK-4621. Phase I/II trials are underway with this agent and have demonstrated tolerable safety profiles and could enhance IFN gene expression in tumors (81). Targeting of CD40 via agonistic antibodies has demonstrated limited single-agent activity to date, however emerging evidence suggests that effective CD40 agonism requires antibody Fc engagement of Fcγ receptor 2b. Newer CD40 agonist antibodies, such as APX-005M and 2141-V11, with enhanced Fcγ receptor 2b affinity show improved CD40 crosslinking and activation of DCs in preclinical models, and are currently undergoing early-stage clinical trials (82).

Further elucidating the biology surrounding DC control of T-cell immunity will be critical for selecting rational combination therapies moving into the future. The highest unmet need in cancer immunotherapy may be to overcome the non–T-cell-inflamed tumor microenvironment and facilitate antigen-specific T-cell recruitment. Early clinical trials of DC agonists, such as those above, have yet to consistently show such effects and novel approaches in delivery or drug development may be necessary to accomplish this. An unmet need also continues to exist in T-cell-inflamed tumors that do not response to immunotherapy. For example, in non–small cell lung cancer, the response rate to anti-PD-1 in patients with tumors that are PDL1⁺ >50% remains only approximately 45% (83).

Recent studies have indicated that neoadjuvant checkpoint inhibitor treatment can produce enhanced antitumor T-cell responses compared with adjuvant therapy, with more clonal and diverse tumor T-cell infiltrates in the neoadjuvant setting (84, 85). T-cell responses to checkpoint inhibition can be rapid, and response can be seen in blood and normal adjacent tissue (86, 87). T-cell clonotyping analysis also suggests that novel T-cell clones not previously seen in the tumor can be observed following checkpoint inhibition (85, 88). It is likely these peripherally stimulated T cells are activated via DCs—possibly within tumor draining lymph nodes. Blockade of PD-1/PD-L1 can also enhance priming of naïve T cells (15), suggesting that these therapeutics do not solely act on preexisting T-cell clonotypes. Several of these studies also associated tumor T-cell-inflamed gene signatures with therapeutic benefit (84–86). DCs can have multiple roles in this setting, from initiating T-cell expansion, to guiding T-cell entry and effector function within tumors. It will be important to examine antitumor T-cell clonalities in response to DC directed therapeutics, and the location of tumor antigens to understand where T-cell re-invigoration occurs. Antigen location should also be taken into account when using DC agonist therapy. For example, peritumoral dosing of CD40 agonist antibodies drives stronger anti-tumor responses compared with systemically delivered agonist, and CD40 agonist delivered to an irrelevant tissue site failed to generate antitumor responses (89). Determining whether the application of DC agonists is optimally to the tumor or whether cross-presentation in the draining lymph nodes is operative may impact on combination partner selection. If this is indeed the case, a reasonable hypothesis may be entertained that optimal ICB combination therapy may include anti-CTLA-4. Further, combinations of multiple DC agonists may be necessary to optimally prime tumors for ICB response (53).

Effective cancer immunotherapy has been strongly associated with the presence of the T-cell-inflamed tumor microenvironment and DC activation appears to be a primary driver of this phenotype. Conversely, a major mechanism of resistance appears to be the lack of T-cell inflammation, raising agonism of DCs to generate type I IFN, in combination with ICB, as a priority approach. Multiple molecular targets are being actively explored however novel drug delivery approaches may be necessary. Many in the field have proposed to convert cold tumors to hot or make hot tumors hotter, we will observe with interest whether this simple paradigm can expand the numbers of patients benefiting from cancer immunotherapy over the next several years.

J.J. Luke holds stock in Actym, Alphamab Oncology, Kanaph, Mavu (now part of AbbVie), Onc.AI, Pyxis, and Tempest, is a paid consultant for TTC Oncology, 7 Hills, Alphamab Oncology, Kanaph, Mavu (now part of AbbVie), Onc.AI, Pyxis, Springbank, Tempest, Abbvie, Akrevia, Algios, Array, Astellas, Bayer, Bristol-Myers Squibb, Eisai, EMD Serono, Ideaya, Incyte, Janssen, Merck, Mersana, Novartis, PTx, RefleXion, Regeneron, Silicon, Tesaro, and Vividion, reports receiving commercial research support from (all to institution for clinical trials unless noted) AbbVie, Agios (IIT), Array (IIT), Astellas, Bristol-Myers Squibb, CheckMate (SRA), Compugen, Corvus, EMD Serono, Evelo (SRA), Five Prime, FLX Bio, Genentech, Immatics, Immunocore, Incyte, Leap, MedImmune, Macrogenics, Necktar, Novartis, Palleon (SRA), Merck, Springbank, Tesaro, Tizona, and Xencor, travel expenses from Akrevia, Bayer, Bristol-Myers Squibb, EMD Serono, Incyte, Janssen, Merck, Mersana, Novartis, Pyxis, and RefleXion, and is listed as an inventor on the following patents (both provisional): Serial #15/612,657 (Cancer Immunotherapy), PCT/US18/36052 (Microbiome Biomarkers for Anti-PD-1/PD-L1 Responsiveness: Diagnostic, Prognostic, and Therapeutic Uses Thereof). C. S. Garris is listed as an inventor on the following provisional patents: serial #16/495,725 (Cancer Immunotherapy), PCT/US2018/023064 (Mitigating Fc-Fc Receptor Interactions in Cancer Immunotherapy). No other potential conflicts of interest were disclosed.

C.S. Garris was supported by NIH F32 grant no. CA250147. J.J. Luke was supported by DOD W81XWH-17-1-0265.