Introduction

Although immunotherapy holds much promise for treating cancer, this therapy is not consistently effective, and many patients derive little or no benefits [1]. Understanding the nature of intra-tumor immune responses and the mechanisms that enable tumors to escape immune attack remains an urgent and daunting challenge for tumor immunologists. Current efforts to improve outcomes have mostly focused on ways to increase the number and specificity of cytotoxic T lymphocytes (CTLs) and to block molecules that are thought to impair CTL function in the tumor microenvironment. Dendritic cells (DCs) are the most potent professional antigen-presenting cells and play a pivotal role in adaptive immunity [2, 3]. In patients with cancer, defective DC function is considered a key cause of impaired immune responses to antigens expressed by tumors [4]. Recent efforts to define the DC compartment in tumors have revealed an unexpected level of diversity and complexity, which has opened new opportunities to improve immunotherapy. In this review, we discuss advances in the understanding of tumor DC ontogeny, tumor DC subset heterogeneity, factors in the tumor microenvironment (TME) that effect DC recruitment, differentiation, and function, and potential strategies to optimize their function.

DC overview

Dendritic cells comprise a heterogeneous population of cells specialized in antigen capture, processing, and presentation. They occur in trace numbers in all tissues, forming a sentinel network that links innate and adaptive immunity. DCs are subdivided currently into four major categories: the conventional or classic DC (cDC); interferon-producing plasmacytoid DC (pDC); monocyte-derived DC; and Langerhans cells. cDC is further classified based on location, surface markers, function, and, more recently, by transcription factor expression.

The hallmark function of DCs is to prime naïve T cells for adaptive immunity, which is influenced by their maturation status, subtype, and cell number [2]. Under the steady-state conditions, DCs exist in an immature state characterized by avid uptake of antigens and low expression of major histocompatibility complex (MHC) and co-stimulatory (e.g., CD80, CD86, and CD40) molecules. Antigen presentation by immature DCs promotes immune tolerance [5, 6]. DC maturation refers to a process that increases cell-surface expression of MHC/peptide complexes and co-stimulatory molecules and production of cytokines critical for T-cell activation [7]. This process can be initiated by the recognition of pathogen-associated molecular patterns and endogenous “danger” signals through pattern-recognition receptors and by inflammatory cytokines [8]. Immune defenses also depend on maintaining the optimal number of DCs in peripheral tissues: DC depletion decreases adaptive immune responses to nominal antigens and pathogens [9, 10], whereas persistently high numbers of lymphoid tissue DC elevate the risk of autoimmune disease [11, 12]. Many studies in patients with cancer and in animal tumor models have documented reduced numbers and immaturity of cDCs in tumors and blood [13–19].

With the exception of LCs, which self-renew mostly in situ, DCs originate from bone marrow progenitors. Findings from our lab and others over the past decade underlie the current view of DC and macrophage ontogeny [20–22] (Fig. 1). In this model, DCs arise from a bipotent bone marrow monocyte/macrophage and DC progenitor (MDP) that give rise to common monocyte progenitors (cMoP) and common DC precursors (CDPs). CDPs differentiate into interferon-producing plasmacytoid DC (pDC), which completes their development in BM, and pre-DCs, an immediate precursor of classical DC (cDC) that migrates rapidly via blood into peripheral lymphoid and non-lymphoid tissues [23–31]. Pre-DCs differentiate into two main cDC subpopulations: (1) lymphoid-tissue resident CD8α+ CD11b− cDC and tissue CD103+ CD11b− cDCs (cDC1), which specialize in cross presentation of exogenous antigens on MHC-I molecules to CD8+ T cells [32]; and (2) CD11b+ cDC (cDC2), which have a dominant role in presenting endogenous antigens on MHCII to CD4+ T cells [33]. Recent evidence suggests that commitment towards these subsets occurs in the bone marrow [31]. Analogous counterparts of cDC1 and cDC2 in the human are CD141+ DCs (also known as BDCA3+) and CD1c+ DCs (also known as BDCA1+), respectively [21]. Circulating monocytes participate in the generation of cDC in some tissues [34–36].

Fig. 1
figure 1

Current model of DC ontogeny. This figure shows the developmental pathways for DC and monocytes/macrophages in peripheral tissues and tumors. The cell-surface molecules that are used commonly to distinguish these cell populations are shown. HSC hematopoietic stem cell, CMP common myeloid progenitor, CLP common lymphoid progenitor, MDP monocyte dendritic cell progenitor, CDP common dendritic cell progenitor, Pre-cDC immediate precursor of classical DCs, pDC plasmacytoid DCs

Full size image

During the steady-state conditions, DC homeostasis in lymphoid and non-lymphoid tissues reflects a balance between the influx of new precursors, DC emigration, and cell death. Flt3 ligand (Flt3L), which is expressed by stromal cells in BM and lymphoid tissues and by activated T cells, plays an essential role by driving proliferation and differentiation of Flt3+ bone marrow progenitors and pre-DCs [37–39]. In mice and humans, overexpression or injection of Flt3L markedly increases the number cDCs and pDCs in blood and in lymphoid and non-lymphoid tissues [39–42]. Notch 2 signaling controls differentiation of pre-DC-derived cDC2 in spleen and lymphotoxin β receptor signaling regulates cDC2 development as well as cDC proliferation [43, 44]. Retinoic acid signaling in pre-cDCs influences differentiation and homeostasis of CD11b+CD8α− cDCs in the spleen and CD11b+CD103+cDCs in the intestine [45].

cDC numbers expand during inflammatory processes through recruitment of new precursors, augmentation of cDC proliferation in situ, and differentiation of monocytes into ‘inflammatory’ DC [25, 46–49]. Increased expression of granulocyte/macrophage colony-stimulating factor (GM-CSF, also known as CSF-2) contributes to this response [50, 51]. Inflammation also causes egress of tissue DC into lymph nodes by modulating chemokine receptor expression and altering the structure of regional lymphatics and lymph nodes [52–54].

Transcription factors (TFs) guide DC development from hematopoietic progenitors [55]. Some TFs, such as Ikaros and PU.1, are critical for all DC as well as other myeloid populations. Gfi-1, a transcriptional repressor, promotes DC over macrophage differentiation at the MDP stage. E2-2 expression is absolutely essential for the development and maintenance of pDC [56, 57], whereas ID2 and the recently identified TF, Zbtb46, are expressed at high levels in all cDC [29, 30, 58]. CD8+ cDC and CD103+ CD11b− cDC express high levels of IRF8, Batf3, and NFIL3 [32, 59, 60]; CD11b+ cDC express high levels of relB and IRF4 [59, 61]. These findings in mice have been shown to be highly relevant to human DC ontogeny. For example, mutation of Gata2 and IRF8 causes DC deficiency syndromes that increase susceptibility to mycobacterial and fungal infections [62, 63].

DC and anti-tumor immune responses

Inflammation is a well-recognized component of many cancers [64]. Chemokines, cytokines, and growth factors in tumors promote the influx of infiltrating cells, including DCs, and the generation of stromal elements. Cancers express a wide range of tumor-associated and tumor-specific antigens [65, 66]. It is generally thought that DCs in tumors capture tumor antigens and migrate to draining lymph nodes, where they prime and activate tumor-specific T cells [67]. Memory and effector CTLs return to the tumor to perform immunosurveillance activities. Evidence suggests that this process can prevent tumor development and influence the rate of tumor progression [68, 69]. For most patients, however, natural anti-tumor immune defenses fail to control the cancer.

The success of cancer immunotherapy relies on augmenting numbers of functional CTL in the tumor [70]. Through intensive efforts over the past few decades, it is now possible to generate large numbers of tumor-antigen-specific T cells [65]; however, the overall efficacy of this therapy is still poor. Tumors can evade immune attack through many mechanisms, including “immunoediting”, a process that progressively selects less immunogenic tumor cells [71, 72], immune checkpoint inhibition through CTLA4, PD-1/PD-L1 and other molecules, and active suppression by myeloid-derived suppressor cells (MDSC), tumor-associated macrophages (TAMs), fibroblast stromal cells, and T regulatory cells (Tregs) [73–76]. The relative importance of each of these mechanisms likely varies with both the tumor and the host.

In cancer therapy, it has been noted that antigen-experienced stem cell-like memory and central and effector memory T cells are superior to terminal-differentiated CTL, because of their capacity to proliferate and acquire effector functions in the tumor [65]. One aspect of tumors that has been largely overlooked until recently is how intra-tumor DCs influence CTL behavior and function. Evidence from infectious disease models indicates that antigen-experienced T cells require cognate interactions with tissue DC to expand in situ and achieve full effector functions [77–80]. Confocal microscopy studies of tumors have revealed interactions between CTL and intra-tumor APCs [81, 82]; and recent studies from our lab and others showed that intra-tumor DCs were the only cells that could stimulate CTL proliferation, at least in vitro [83, 84]. Furthermore, contrary to the prevalent view that all tumor DCs are functionally defective, it has become clear that some DC subsets are uniquely equipped to stimulate anti-tumor immunity and influence tumor biology [85]. These findings suggest that manipulating tumor DC may be a useful approach to improve responses to cancer immunotherapy.

Tumor DC heterogeneity

Multi-dimensional flow cytometry and transcriptional profiling of transplantable tumors and spontaneous tumors in mice have revealed tremendous diversity in the myeloid cell compartment [85]. All cDCs in the mouse and human express the cell-surface integrin CD11c; however, other cells, including macrophages, monocytes, and activated lymphocytes, express this molecule, albeit at lower levels than cDCs, necessitating the use of multiple parameters to define each cell population. Similar to other tissues, the TME in mice contains three main subsets of CD11c+ MHCII+ cDCs: (1) CD11b− CD103+ DCs (BATF3/IR8-dependent cDC1); (2) CD11b+ CD103− CD64− F4/80− DC (IRF4-dependent cDC2); and (3) CD11b+ CD64+ F4/80+ DC, which maybe more closely aligned with monocytes and macrophages than DC. Collectively, they constitute a relatively minor population in most tumors, accounting for 5–10 % of all myeloid cells (macrophages and neutrophils predominate in most tumors). Numerous studies have documented the presence of cDCs in human cancers (reviewed in [86]); however, the cell-surface markers available offer limited interrogation of subsets, maturity, and function. pDCs are rare in mouse tumors, but are found in variety of human tumors.

Origin of tumor DC

Pre-cDCs exist in a variety of transplantable tumor models, including B16 melanoma, CT26 colon carcinoma, Lewis lung carcinoma, and EMT6 breast carcinoma [87]. Tumor pre-cDCs are morphologically, phenotypically, and functionally indistinguishable from those isolated from BM and spleen. Adoptive transfer studies of bone marrow pre-cDCs revealed that tumors recruit pre-cDCs through a CCL3-dependent mechanism, where they differentiate into proliferating cDCs [87]. Flt3L therapy promotes intra-tumor expansion of CD103+ DC progenitors (CD11c+MHCII+CD103− CD11b−) and immature CD103+ DCs [88]. Monocytes and more primitive bone marrow progenitors have also been detected in tumors, particularly in the setting of inflammation induced by anthracycline chemotherapeutic agents, and differentiate into inflammatory DCs [89, 90].

Tumor DC plasticity

The number, phenotype, and function of cDCs can change, as the tumor progresses [91, 92]. In a model of spontaneous ovarian cancer, Scarlett et al. detected increased densities of tumor infiltrating DCs, macrophages, MDSCs, and T cells, as well as a functional switch in DC from an immunostimulatory to an immunosuppressive phenotype as the tumors grew [91]. In this model, depletion of DCs at early time points accelerated tumor growth, whereas depletion at later time points led to tumor regression. Krempski et al. also showed in a transplantable ID8 mouse model of peritoneal ovarian cancer that the number of tumor-infiltrating cDCs correlated with tumor burden [92]. Furthermore, tumor cDCs progressively expressed PD-1, as well as PD-L1, which was associated with T-cell suppression and loss of tumor-infiltrating T cells [92]. Our research has shown that the TME drives immunostimulatory Gr-1− cDCs to generate a functionally defective Gr-1+ cDC subpopulation that induces T-cell tolerance [83]. Using a transgenic mouse model that allowed in vivo tracking of DCs in tumors, we also found that immunostimulatory cDCs derived from pre-DCs can lose their DC identity and evolve into CD11c−MHCII− regulatory macrophages [93] (Fig. 2). DC-derived-macrophages (DC-d-M) potently suppressed T-cell responses through the production of immunosuppressive molecules, including nitric oxide, arginase, and IL-10. A relative deficiency of GM-CSF appeared to provide a permissive signal for DC de-differentiation, as augmenting GM-CSF expression levels in the tumor blocked this process. Collectively, these findings highlight the plasticity of DC and suggest that maintenance of DC identity and function depends, at least partly, on cues received from the TME.

Fig. 2
figure 2

Model of tumor DC plasticity. Circulating pre-cDC recruited into tumors differentiate into Gr-1−cDC, which possesses the capacity to stimulate proliferation and expansion of CTLs. Under the influence of the tumor microenvironment, Gr-1−cDCs generate: (1) Gr-1+cDC, a subpopulation of maturation-resistant, IL-10-producing DCs that induce T-cell anergy, and (2) DC-derived-macrophages (DC-d-M) that potently suppress CTL proliferation by releasing IL-10, arginase-1, and nitric oxide

Full size image

DC defects in cancer

Antigen cross presentation by DCs plays a critical role in the generation of anti-tumor responses [9, 94, 95]. DCs can acquire tumor antigens from multiple sources: (1) apoptotic/necrotic tumor cells [96–98]; (2) chaperone proteins, such as heat shock proteins, that bind soluble tumor antigens [99, 100], (3) secreted vesicles (e.g., exosomes) from tumor cells [101], (4) gap junctions that transfer small antigenic protein fragments [102]; and (5) tumor plasma membrane fragments [103]. In addition, DCs can acquire preformed peptide-MHC Class I complexes through direct contact with tumor cells, a process resembling trogocytosis (also known as cross dressing) [104, 105]. Cross presentation of tumor antigens generally requires stable and high antigen expression levels and tumor cell apoptosis/necrosis to release the antigens [106]. These conditions are frequently unmet in untreated cancers, but can be induced with chemotherapy, radiation therapy, and oncolytic viral therapy.

The functional status and subset of tumor DCs affect their apparent cross-presentation capacity and immunogenicity [82, 107–109]. Stoitzner et al. reported that DCs sorted from B16 melanoma could not induce proliferation of tumor-antigen-specific CD8+ T cells or CD4+ T cells [110]. Engelhardt et al. showed by live cell imaging that tumor-antigen-bearing DCs engage tumor-antigen-specific T cells; however, these interactions were unable to fully activate T-cell effector functions, including the capacity to kill tumor cells [82]. By contrast, treatment of tumor-bearing mice with agonistic anti-CD40 monoclonal antibodies and other DC activating agents, such as TLR9 agonists, led to the influx of large numbers of T cells that were capable of eradicating established tumors [107].

Recent reports have highlighted the importance of tumor CD103+ cDCs in stimulating anti-tumor immune responses in primary cancers and metastases [84, 111, 112]. On a per cell basis, tumor CD103+cDCs stimulate naïve and primed tumor-antigen-specific T cells more effectively than tumor CD11b+cDC, which is attributed to more efficient cross-presentation machinery and higher expression levels of IL-12 [84, 113]. Tumor CD103+ also specializes in the transport of intact tumor antigens to tumor draining lymph nodes [88]. Targeted reduction/elimination of tumor CD103+DCs in BATF3 knockout mice and in Zbtb46-diphtheria toxin receptor (Zbtb46-DTR) transgenic mice attenuated responses to cancer immunotherapy. Spranger et al. reported that active oncogenic WNT/β-catenin signaling within melanoma cells in Braf V600E /Pten −/− /CAT-STA tumors inhibited T-cell priming by suppressing the CCL4-dependent recruitment of dermal CD103+ DCs. Injecting these tumors with poly:IC activated Flt3 ligand-induced bone marrow-derived DCs restored responses to anti-CTLA4 and anti-PD-L1 monoclonal antibodies [114]. Similarly, expansion and activation of CD103+ cDC in B16 melanoma with Flt3L and poly I:C treatment, respectively, enhanced responses to PD-L1 and BRAF blockade [88]. The relevance of these findings to human cancer remains to be clarified. Notably, the human equivalents of CD103+ DC and CD11b+ DC from lymphoid tissues (BDCA3+ and BDCA1+ DCs, respectively) show less striking differences in cross-presentation activity [115]. Furthermore, mouse CD11b+ cDC exhibits robust cross-presentation capacities under appropriate conditions [116], and was critical to chemotherapy-induced immune responses [89].

Immaturity and co-inhibitory molecules

Multiple inhibitory (e.g., PD-L1, TIM-3, LAG-3, CD200, CTLA4) and activating (4-1BBL, ICOS-L, CD80, and CD86) immune checkpoint molecules influence the activation, differentiation, and proliferation of T cells. T-cell interactions with immature DCs can lead to T-cell tolerance through various mechanisms, including deletion, anergy, and the generation of regulatory T cells [6, 117, 118].

PD-1 and its corresponding ligands PD-L1 and PD-L2 negatively regulate T-cell priming and effector functions [119]. PD-L1 and PD-1 expressions on tumor DCs correlated with cancer progression in an ovarian cancer model [92]. In a melanoma model, CD103+ DCs, from tumor-draining lymph nodes express high amounts of PD-L1 as compared with CD103+ DCs isolated from non-draining lymph nodes [88]. Antibody blockade of PD-L1 and PD-1 mitigates DC dysfunction as evidenced by increased NF-κB activation, increased IL-12, TNF, and IL-1β production and co-stimulatory molecule expression, and enhanced T-cell stimulatory capacity [120, 121].

Tumor DCs also suppress tumor-specific T cells through the release of biomolecules, such as nitric oxide (NO), arginase I, and indoleamine 2,3-dioxygenase (IDO) [122]. Arginase I degrades arginine, an essential amino acid for CD4+ T-cell proliferation and differentiation [119]. High expression levels of arginase I promote the accumulation of reactive oxygen intermediates, such as NO, which block CD8+ T-cell responses [123]. IDO is a tryptophan-catabolizing enzyme that plays an important role in inducing and maintaining tolerance. Tumor cells, DCs, and regulatory macrophages can express IDO. IDO activates Tregs and creates a milieu deficient in tryptophan, another amino acid used by T cells during activation [124–126]. Furthermore, IDO triggers the production of tryptophan metabolites that induce T-cell apoptosis and suppress T-cell proliferation [125].

Mechanisms of tumor DC dysfunction

Many studies document the presence of immature DCs in blood and tumors in humans and tumor-bearing mice [127, 128]. The expansion of immature DCs is often considered to be a systemic process, in which tumor-derived factors skew differentiation of bone marrow progenitors [128]. This view is based partly on in vitro studies of bone marrow cells stimulated with GM-CSF and tumor-conditioned media, and studies of mice with large tumor burdens. Recent evidence indicates that GM-CSF-stimulated bone marrow DCs are poor representatives of natural DC, and as such, may not be the best tool for evaluating the effects of cancer [129, 130]. Others and we have shown that hematopoiesis and DC development proceed normally in mice with small transplantable tumors and spontaneous tumors. In addition, studies of human breast and other cancers have shown that while intra-tumor DCs are immature, DCs at the margins of the tumor are mature [86, 131]. Collectively, these findings suggest that the local TME rather than systemic factors is more relevant. Understanding how tumors promote tumor DC dysfunction is expected to help advance cancer immunotherapy. Some of the factors and their associated mechanisms of action are reviewed below.

IL-6

Cancer cells and tumor-infiltrating DCs, monocytes, and macrophages can produce IL-6. IL-6 targets genes involved in cell-cycle progression and suppression of apoptosis [132, 133], accounting for its known role in oncogenesis [64, 134, 135]. Although IL-6 is often considered a proinflammatory cytokine, IL-6 dampens immune responses in some settings [136, 137]. High serum concentration of IL-6 correlates with poor outcome and abnormal immune responses in patients with various epithelial and lymphoid cancers and hepatocellular carcinoma [138–140]. High serum IL-6 concentrations in multiple myeloma patients were associated with lower absolute numbers of circulating precursors of myeloid DCs [141]. Peripheral blood DCs from these patients also showed significantly lower expression of HLA-DR, CD40, and CD80, and impaired stimulation of allogeneic T-cell proliferation as compared with DCs isolated from healthy controls.

In vitro studies have shown that IL-6 inhibits LPS-induced DC maturation, suppresses intracellular MHC Class II expression, and lowers CCR7 expression, the chemokine receptor critical for DC migration to lymphoid tissues [136, 142–144]. IL-6 also inhibits DC differentiation from human monocytes and CD34+ myeloid progenitor cells in vitro, while promoting the development of macrophages with an alternatively activated phenotype that is associated with wound healing [145, 146]. This effect of IL-6 was linked to its ability to increase macrophage colony-stimulating factor receptor (Csf1r) expression levels [147]. Our group reported that IL-6 promotes the differentiation of immunosuppressive Gr-1+ cDCs from pre-cDCs in tumors [83]. IL-6−/− mice had a threefold reduction in the frequency of tumor Gr-1+ DCs as compared with wild-type mice, which was associated with a higher frequency and rate of proliferation of IFN-γ+ CTLs, and better responses to adoptive T cell immunotherapy [83].

IL-10

Many cells in the TME express IL-10, including Tregs, TAMs, MDSCs, DCs, and cancer cells [148]. Autocrine and paracrine IL-10 can increase the proliferation and survival of cancer cells (e.g., B16 melanoma, human stomach adenocarcinoma, and human glioblastoma multiforme) [135, 149]. IL-10 is a classic anti-inflammatory cytokine that inhibits multiple aspects of DC biology, including maturation, production of proinflammatory cytokines (e.g., IL-12 and IL-1β), and stimulation of T cells [150, 151]. Under some conditions, IL-10 stimulates the generation of ‘tolerogenic’ DCs, which express low levels of MHC and co-stimulatory molecules and produce high amounts of IL-10 [152]. Recent studies in a breast cancer model have shown that inhibition of IL-12 expression in tumor DCs is the dominant effect of macrophage-derived IL-10; blocking the IL-10 receptor (IL-10R) with neutralizing antibodies restored IL-12 expression in tumor DCs and anti-tumor T-cell responses [112]. Similarly, Vicari et al. reported that tumor DC function could be restored with a combination of anti-IL-10R antibodies and activation with CpG oligonucleotides [153].

The effect of IL-10 on cancer is complex, however, as anti-tumor effects have also been observed [154, 155]. Exogenous IL-10 inhibited growth and metastasis of mammary and ovarian carcinoma xenografts, partly through the downregulation of MHC Class I expression, which enhanced NK-cell-mediated tumor killing [154]. Mumm et al. reported that IL-10 promotes expansion of tumor-infiltrating IFN-γ+ CTLs, which enhanced tumor immunosurveillance and control of tumor growth. The reason for these discrepant findings remains unclear, but likely reflects differences in the models, TME, cellular source and bioavailability of IL-10, and the relative sensitivity of tumor infiltrating cells to IL-10, IL-12, and other molecules [112, 156].

Vascular endothelial growth factor (VEGF)

Serum VEGF levels correlate with poor prognosis in various human cancers [157, 158]. VEGF stimulates tumor endothelial cell proliferation and angiogenesis, which are essential for tumor growth and development [157]. VEGF binds to high-affinity membrane tyrosine kinase receptors—VEGFR-1, 2, 3—that are expressed mostly on endothelial cells and a few populations of hematopoietic cells, including DCs [157]. Earlier studies showed that VEGF inhibited DC differentiation and maturation in vitro through an NF-κB-dependent pathway, which could be blocked with neutralizing anti-VEGF antibodies [159, 160]. Administration of VEGF to tumor-free mice resulted in impaired DC development and accumulation of immature Gr-1+ myeloid cells (defined as MDSCs). Treatment of tumor-bearing mice with neutralizing anti-VEGF antibodies increased the number of spleen and lymph node DCs, improved DC function, and improved anti-tumor CTL responses [161–163].

STAT3

Constitutive activation of STAT3 occurs in many cancers [164]. STAT3 serves as an oncogenic driver that enhances tumor cell proliferation, survival, and invasion [165]. STAT3 activation up-regulates the expression of anti-apoptotic proteins, such as BCL-XL, MCL1, cyclin D1 and MYC, and pro-angiogenic factors, such as HIF-α, VEGF, MMP-2, and MMP-9 [166]. STAT3 activation also contributes to tumor progression by enhancing tumor inflammation and hampering anti-tumor immunity. STAT3 activation in DC has been linked to defects in differentiation, maturation, and function [167–170]. Both the intensity and duration of STAT3 signaling determine whether DC dysfunction develops [171, 172]. Pharmacologic inhibitors and genetic ablation of STAT3 signaling restore maturation responses to inflammatory stimuli in DCs incubated in tumor-conditioned medium. In a study using the Cre-loxP system to delete STAT3 in hematopoietic cells, tumor-infiltrating DC from STAT3−/− mice expressed higher levels of surface MHC Class II, CD80, and CD86, and produced more IL-12 [173], which was associated with better anti-tumor immune responses. Elevated STAT3 activity in tumor DCs is attributed mostly to paracrine stimulation by cytokines in the TME, such as IL-6, IL-10, and VEGF [169, 174]. Our studies show that tumor-derived factors stimulate autocrine IL-6 and IL-10 production in cDCs, which worked synergistically to promote DC dysfunction [175].

TIM-3

TIM-3, a receptor for galectin-9, was identified initially as a negative regulator of Th1 immunity [176], but is also expressed by myeloid cells, including DCs [177]. Chiba et al. reported that tumor-derived factors upregulate TIM-3 expression in tumor DCs [178]. TIM-3 plays dual roles in anti-tumor immunity. Under some conditions, galectin-9-TIM-3 interactions promote DC maturation and cross priming of tumor-antigen-specific T cells [178, 179]. By contrast, the interaction of TIM-3 with high-mobility group box 1 protein (HMGB1), a classic danger associated molecular pattern molecule (DAMP) in the TME, prevents tumor DCs from sensing nucleic acid danger signals released from dying tumor cells, which suppresses type 1 interferon- and IL-12-mediated anti-tumor responses [178]. Others have shown that ligating TIM-3 on bone marrow-derived and spleen DCs with cross-linking antibodies activates Bruton’s tyrosine kinase and c-SRC, which inhibits DC activation and maturation by blocking the NF-κB pathway [180].

Lipids

Intracellular fat and glycogen accumulate in DCs with exposure to maturation stimuli in vitro, and occur during normal development in lymphoid tissues [181]. Several reports suggest that accumulation of oxidized lipids, especially triacylglycerols (TAG), causes dysfunction and shortens the lifespan of DCs [19, 182–185]. DCs from mouse EL-4 lymphoma, CT-26, and B16-F10 tumors and in some cancer patients exhibit elevated TAG levels [19]. Tumor-conditioned media can stimulate the uptake of TAG in bone marrow-derived DCs and human monocyte-derived DC by regulating the expression levels of scavenger receptor A (SRA1, CD204, and MSR1), lipoprotein lipase (LPL), and fatty acid-binding protein 4 (FABP4) [19, 182, 184]. As compared with normal DCs from tumor-free mice, lipid-laden DCs stimulated T cells poorly because of defects in antigen cross presentation and increased production of IL-10 [19, 185].

Endoplasmic reticulum stress response

Tumors adapt to hypoxia, nutrient deficiency, and oxidative stress by triggering an endoplasmic reticulum (ER) stress response, also known as the unfolded protein response (UPR) [186–189]. Cubillos-Ruiz et al. recently demonstrated that the TME also induces an ER stress response in DC that is mediated by the spliced transcription factor, XBP1 [183]. Although the ER stress response contributes to normal DC development and survival [190, 191], constitutive XBP1 activation in tumor DCs induced abnormal accumulation of oxidized lipids by targeting multiple triglyceride biosynthetic genes [183]. Reactive oxygen species generated reactive lipid peroxidation byproducts (e.g., aldehyde 4-hydroxy-trans-2-nonenal (4-HNE)) that sustained XBP1 activation. Notably, lipid accumulation induced by this pathway operates independently of scavenger receptors involved in the uptake of extracellular lipids. Inactivation of XBP1 in tumor DCs improved anti-tumor T-cell immune responses and the control of tumor growth in an ovarian cancer model.

Gangliosides

Gangliosides are sialic-acid-containing glycosphingolipids located in the plasma membrane of all vertebrate cells. Gangliosides which shed from various tumors have been shown to suppress anti-tumor immune responses by affecting T cells, NK cells, and DCs [192, 193]. Gangliosides derived from neuroblastoma and melanoma inhibit DC differentiation from human monocytes and mouse bone marrow cells [194, 195]. Exposure of bone marrow-derived DCs to GM1 ganglioside inhibited maturation and up-regulation of co-stimulatory molecules, reducing their capacity to prime naïve T cells. T cells primed with pre-treated ganglioside DCs also produced significantly less IFN-γ and IL-2 upon re-stimulation [192, 195].

TLR2

We recently reported that TLR2 activation is a critical proximal signal that contributes to tumor DC dysfunction [175]. Tumor-conditioned medium (TCM) generated from both murine and human cancer cell lines stimulated DC through TLR2 to produce autocrine IL-10 and IL-6. The ability of IL-6 and IL-10 to stimulate DC dysfunction, however, relied on TLR2-induced up-regulation of their cell-surface receptors, which markedly decreased the cytokine concentration threshold required to activate STAT3. This effect of TLR2 helps reconcile the relative tumor specificity of DC dysfunction, and suggests that cytokine receptor expression levels, rather than the source of the cytokines, may be the ultimate arbiter of DC fate and function in tumors. TLR2 deficiency improved the immunogenicity of intra-tumor DCs, enhanced CTL expansion, and improved anti-tumor responses to immunotherapy. Consistent with the earlier reports [134], we found that versican, an extracellular matrix glycoprotein that is overexpressed in many cancers, is a key TLR2 ligand in TCM. Interestingly, the level of versican expression in human cancers correlates inversely with prognosis and the density of CD8+ T-cell infiltration [196–198]. Additional TLR2 ligands released by cancer cells include laminin-β1, procollagen III-α1, Hsp60, and Hsp72 [134, 199, 200]. The microbiome also generates TLR2 ligands that could potentially influence the TME and tumor progression [201].

Strategies to optimize tumor DC function

Immunogenic cell death (ICD) is a recognized benefit of the conventional cancer therapies. Chemotherapy and radiation therapy induce ICD by stimulating the release of DAMPs [202–205]. Endoplasmic reticulum stress and autophagy expose calreticulin (CRT) on the outer leaflet of the plasma membrane; apoptosis releases adenosine triphosphate (ATP); and permeabilization of the nuclear membrane during necrosis releases HMGB1 [206]. DAMPs cause ICD partly through their interactions with DCs: CRT, ATP, and HMGB1 ligate, respectively, CD91, P2PX7 (a purigenic receptor), and TLR4, which facilitates DC recruitment, engulfment of tumor antigens, and antigen presentation to T cells. Vacchelli et al. reported that the efficacy of anthracycline-based chemotherapy requires DCs to express formyl peptide receptor 1, which promotes stable interactions with dying cancer cells expressing annexin-1 [207].

Studies of the conventional cancer therapy on ICD reveal the central importance of tumor DCs, and suggest that cancer immunotherapy should incorporate strategies to promote their activity. Administrating GM-CSF and Flt3 ligand systemically to increase tumor DC numbers fails to improve anti-tumor CTL responses significantly [208, 209], likely because neither growth factor supersedes the local mechanism(s) that cause tumor DC dysfunction. Successful approaches to improve DC function have generally targeted inhibitory molecules in the TME and their downstream signaling pathways or have included DC promoting agents that overwhelm negative regulators. Improving the function of tumor DCs presents several challenges, however, including the delivery of therapeutic agents into the TME at effective concentrations, potential off-target effects on other cells, and the rapid turnover of tumor DCs, which may necessitate on-going therapy for nascent DCs and their progenitors.

Concluding remarks

Anti-tumor immune responses elicited by adoptive T-cell therapy and active vaccination approaches rely on interactions between CTLs and cancer cells. The development of in vitro culture techniques to generate large quantifies of DCs in the early 1990s led to extensive exploration of tumor antigen-loaded DC as vaccination vehicles. Despite their ability to stimulate tumor-antigen-specific immune responses, DC-based vaccines have met with a little clinical success thus far [210]. These failures require reassessment of how CTL responses are regulated within TME. Recent evidence strongly suggests that secondary antigen presentation in the tumor modulates the number, duration, and quality of CTLs. As the principal antigen-presenting cells in tumors, DCs have a pivotal position in this process. The inherent plasticity of tumor DCs, however, renders them susceptible to the vagaries of the TME, which often leads to dysfunction. Progress in our understanding of tumor DC biology in both untreated tumors and in tumors treated with drugs, radiation, and checkpoint inhibitor antibodies have revealed potential targets to optimize their function. Although the identification of individual molecules that promote immune evasion will be important to improve T-cell responses, alternative approaches might focus on improving the intrinsic functional qualities of DCs. Our recent work shows that tumor DCs experience a phenotypical and functional de-differentiation process involving fundamental changes in TF expression [93], suggesting that targeting TF regulatory pathways in tumor DCs may improve immunotherapy.