Coordinated regulation of myeloid cells by tumours

Dendritic cells

DCs are terminally differentiated myeloid cells that specialize in antigen processing and presentation. DCs differentiate in the bone marrow from various progenitors^{5, 6,7, 8}. They can also differentiate from monocytes under certain conditions, although most DCs in mouse lymphoid organs are not monocyte-derived^{5, 9}. In contrast, monocytes are the major precursors of DCs in humans¹⁰.

Two major subsets of DCs are currently recognized: conventional (cDCs) and plasmacytoid (pDCs). Although these cells share some common progenitors, they differentiate along distinct genetic programs and have different morphologies, markers, and functions¹¹ (Box 1). The centerpiece of DC biology is the concept of functional activation and maturation in response to ‘dangerous’ stimuli. Differentiated DCs reside in tissues as ‘immature’ cells that actively take up tissue antigens, but are poor antigen presenters and do not promote effector T cell differentiation. Only functionally activated DCs can effectively stimulate immune responses. DCs are activated in response to stimuli associated with bacteria, viruses or damaged tissues; such stimuli are commonly referred to as pathogen-associated molecular patterns (PAMPs) and damage-associated molecular patterns (DAMPs). Activation of DCs leads to profound changes in their gene expression, resulting in increased expression of co-stimulatory molecules and cytokines that promote T cell activation, and also in upregulation of chemokine receptors that drive DC migration to lymphoid tissues. pDCs constitute a minor population of DCs that have a morphology reminiscent of plasma cells, express TLR7 and TLR9 (this receptor is not expressed on human cDCs) and produce large amounts of IFNα in response to activation of TLRs by viruses and self-DNA ¹¹. A more detailed discussion of DC biology can be found in recent reviews^{12, 13}.

Macrophages

MΦ are a group of terminally differentiated myeloid cells closely related to DCs. MΦ are tissue-resident cells derived from monocytes circulating in peripheral blood. They include a broad spectrum of cells whose markers and functions reflect their tissue microenvironment (Box 1). Their function in healthy individuals is to eliminate infectious agents, promote wound healing, and regulate adaptive immunity (reviewed in³⁶). The terminology ‘M1’ and ‘M2’ was coined to describe the different functional states of MΦ and was originally based on studies of murine macrophages ³⁷. M1 or ‘classically activated’ MΦ are activated by IFNγ and bacterial products, express high levels of IL-12, low levels of IL-10, and are tumouricidal. In contrast, M2 or ‘alternatively activated’ MΦ are activated by IL-4, IL-13, IL-10 and glucocorticoid hormones, express high levels of IL-10, low levels of IL-12, and facilitate tumour progression. As discussed by Mantovani³⁸, the M1/M2 nomenclature is useful, but oversimplified because MΦ form a continuum of phenotypes. Although there are some differences between M2-like mouse and human MΦ, phenotypically and functionally the MΦ in these two species are quite similar (Box 1).

Granulocytes

Granulocytes are myeloid cells that are characterized by the presence of cytoplasmic granules and specific nuclear morphology. The most abundant type of granulocytes in the body are polymorphonuclear neutrophils (PMNs) named after their poly-lobed nuclei. PMN possess a complex machinery to engulf and destroy bacteria. PMN are not released from bone marrow until they reach full maturity, but during inflammation, neutrophil precursors (myelocytes and promyelocytes) can be released ⁶⁴.

Human tumours can be infiltrated with mature granulocytes whose numbers can also constitute independent prognostic factors for recurrence ^65–68. Recent evidence has linked granulocytes, and particularly PMNs, with tumour angiogenesis and metastasis, and has provided initial clues about the immunoregulatory role of these cells in cancer. Tumor and tumor-associated stromal cells produce PMN-attracting CXC-chemokines and the orthologue of the secreted protein Bv8, prokineticin 2^{69, 70}. Tumour-released G-CSF also mobilized granulocytes to pre-metastatic niches in the lung and supported subsequent metastasis formation, whereas prokineticin 2 aided tumour cell migration through activation of the Bv8 receptor, prokineticin receptor 1⁷¹. It is likely that granulocytes facilitate the angiogenic switch by expressing matrix metalloproteinase 9 (MMP9), which promotes tumour angiogenesis by inducing VEGF expression within neoplastic tissue⁷². MMP9 also causes the release of elastase, which enters endosomal compartments of neoplastic cells and degrades insulin receptor substrate 1 (IRS1). Degradation of IRS1 facilitates interaction between phosphatidylinositol 3-kinase (PI3K) and the mitogen platelet-derived growth factor receptor, thus promoting tumour cell proliferation⁷³. In contrast to these observations, a recent study demonstrated that in 4T1 breast tumour-bearing mice, PMN inhibited tumor metastases via direct antitumor effects mediated by reactive oxygen species⁷⁴. These new data revisited the old concept of tumour cytotoxic PMN and suggest a possible dichotomic polarization of PMNs. Similar to M1 and M2 polarization in MΦ, PMNs have been shown to shift from an anti-tumoral ‘N1 phenotype’ to a pro-tumoral ‘N2 phenotype’ in the cancer environment⁷⁵ TGFβ drives the N2 phenotype, whereas TGFβ blockade promotes an N1 phenotype with antitumor activity. In lung adenocarcinoma and mesothelioma models, TGFβ favours a pro-tumour phenotype among tumour-infiltrating PMNs, which are characterized by ARG1 expression and low levels of TNF, CCL3, and ICAM1. In tumour-bearing animals, depletion of N2 PMN led to an increase in CD8⁺ T cell activity⁷⁵. In line with these findings, serum amyloid A1 protein induced the expansion of IL-10-secreting PMN that were able to suppress antigen-specific proliferation of CD8⁺ T cells in human melanomas⁷⁶. However, IL-10 production by activated human PMNs was not confirmed in a subsequent study ⁷⁷

Myeloid cells as a single integrated system

Neoplastic cells condition distant sites, such as the bone marrow and spleen, by releasing soluble factors that drive the accumulation of myeloid cells; these myeloid cells subsequently promote neovascularization and metastasis. This creates de facto a tumour-driven ‘macroenvironment’. As discussed above this macroenvironment conditions DCs, MΦ and granulocytes to become immunosuppressive. However, the most prominent effect is accumulation of highly immunosuppressive, immature myeloid cells. These cells were named MDSCs to highlight their common myeloid origin and immunoregulatory properties⁷⁸ (Box 1). Immature myeloid cells with the same phenotype as MDSCs are continually generated in the bone marrow of healthy individuals and differentiate into mature myeloid cells without causing detectable immunosuppression. However, in cancer, normal myeloid cell differentiation is diverted from its intrinsic pathway of terminal differentiation of mature MΦ, DCs, or granulocytes and instead favours differentiation of pathological MDSCs (Fig. 3).

Common mechanisms of tumour impact on myeloid cell recruitment and function

Neoplastic and tumour-associated stromal cells release multiple tumour-derived soluble factors that perturb the myeloid compartment. Cytokines such as GM-CSF, G-CSF, M-CSF, SCF, VEGF, and IL-3 promote myelopoiesis and contribute, in part, to a blockade of myeloid cell maturation^{88, 107} (Fig. 5). Tumour-derived soluble factors that are pro-inflammatory, such as IL-1β, IL-6, and S100A8-9^134–136, as well as cytokines released by activated T cells, such as IFN-γ, IL-4, IL-13, IL-10¹²⁷, initiate the immunosuppressive pathways that commit immature myeloid cells to become MDSCs and then further promote MDSC differentiation towards immune suppressive MΦ and DCs (Fig. 5). The tumour-derived factors CCL2, prokineticin 2, CXCL5, S100A8-9, and CCL12 recruit immature myeloid cells to tumor stroma^{70, 137, 138}. Immature myeloid cells are also chemoattracted by CCL2 that is nitrated/nitrosylated within the tumour environment. In contrast, effector CD8⁺ T cells are not recruited by modified CCL2, which may explain the selective enrichment of myelomonocytic cells within mouse and human tumours¹¹⁷. LPS, in combination with IFNγ, promotes MDSC population expansion, probably by inhibiting DC differentiation¹³⁹. Tumor-derived TGFβ also regulates MDSC accumulation¹⁴⁰ and neutrophil polarization⁷⁵ (Fig. 5). Neoplastic cells and their associated stromal cells also release into the bloodstream subcellular components known as exosomes, which contain signal peptides, mRNAs, microRNAs, and lipids and promote MDSC expansion (reviewed in¹⁴¹).

Tumour-derived soluble factors regulate myeloid lineage cells on multiple levels involving a variety of transcription factors^{88, 107, 131} (Fig. 5), with STAT3 playing a major role. Early studies identified STAT3 as a critical regulator of DC and MΦ defects^{142, 143} and MDSC expansion^144,145,146. STAT3 not only prevents apoptosis and promotes cell proliferation via up-regulation of BCL-XL, MYC, cyclin D1, or survivin^{107, 147}, but also regulates expression of multiple proteins critical for differentiation of myeloid cells. One such pathway involves the calcium-binding pro-inflammatory proteins S100A8 and S100A9¹⁴⁸. STAT3-mediated up-regulation of these proteins in myeloid progenitors inhibits DC differentiation and promotes MDSC accumulation¹⁴⁹. S100A8 and S100A9 also enhance MDSC suppressive activity and recruit MDSCs to the tumour site¹³⁵. Myeloid cell NADPH-oxidase (NOX2) is another important target of STAT3. STAT3 up-regulation of the NOX2 components p47^phox and gp91^phox increases ROS levels, thereby making MDSCs more suppressive⁹². STAT3 also down-regulates PKCβII, which is required for DC differentiation and thus prevents the development of HPCs into mature cells ¹⁵⁰. In addition, STAT3 regulates the transcription factor CCAAT-enhancer-binding protein beta (C/EBPβ). C/EBPβ regulates myelopoiesis in healthy individuals¹⁵¹ and plays a crucial role in controlling differentiation of myeloid progenitors to functional MDSCs¹³⁴. STAT3, at least partially, induces MDSC expansion via up-regulation of C/EBPβ and plays an indirect role in myeloid cell mobilization, accumulation, and survival¹⁵².

The transcription factor STAT1 regulates subsets of myeloid cells via its effects on NOS2 and is crucial for MΦ and MDSC-mediated immune suppression^{127, 153,154}. Other characteristics of MDSCs and MΦ, including the up-regulation of ARG1^155–157, increased TGFβ production^{124, 158}, and possibly expansion of MDSCs¹⁵⁹, are controlled by STAT6. IL-4-induced polarization of TAMs activates STAT6, which binds to the promoter of the gene-encoding the demethylase Jumonji domain-containing 3 (JMJD3). Activated JMJD3 demethylates histone H3 lysine-27, which then increases expression of ARG1, YM1, and FIZZ1, resulting in M2 polarization¹⁶⁰ However, the genetic ablation of Jmjd3 gene in mice caused a defective and irf4-dependent M2 polarization in response to M-CSF, helminth infection or chitin administration, but not following IL-4 stimulation, suggesting a more complex regulatory network ¹⁶¹.

The TLR family also plays a prominent role in myeloid cell development, primarily via the activation of MYD88 and the downstream induction of NF-κB. NF-κB signaling is important for mobilization of myeloid cells to sites of infection, injury, or tumour growth^{162, 163}. TLR4 regulates inflammation-driven MDSC suppressive potency through an NF-κB-dependent mechanism¹⁶⁴. The pro-inflammatory mediators COX2 and PGE₂, which enhance MDSC accumulation and suppressive activity^{140, 165–167}, are also potential targets for NF-κB¹⁶⁸.

Two-stage model of MDSC involvement in tumour progression

Recent studies of autochthonous tumor formation in transgenic mice indicate that cells with an MDSC phenotype probably intervene in the very early stages of cancer progression. Mice with autochthonous pancreatic cancer undergo progressive waves of myeloid cell recruitment after initiation of the transforming programme driven by the Kras oncogene¹⁶⁹. Recruited myeloid cells contribute to the local production of IL-6 and IL-11 that activate STAT3. STAT3, in turn, induces anti-apoptotic and pro-proliferative genes, fuelling tumour initiation, promotion, and progression^{170, 171}. During early events of colitis-associated cancer, myeloid cells act as tumour promoters by enhancing proliferation of tumour-initiating cells and by protecting premalignant intestinal epithelial cells from apoptosis¹⁷². The oncogenic fusion protein RET/PTC3 (RP3) in thyroid carcinomas directly regulates CCL2 and GM-CSF production, which recruits CD11b⁺Gr-1⁺ cells^{173, 174}.

An important question is whether these early recruited cells are immunosuppressive MDSCs. Unfortunately, only a few studies in autochthonous tumor models have determined the immunosuppressive activity of the tumour-associated CD11b⁺Gr-1⁺ cells. In models of spontaneous breast, pancreatic, or lung cancer, accumulated myeloid cells had both the phenotype and immunosuppressive features of MDSCs^{111, 169, 175}. In a recent study, conditional deletion of p120ctn in mice caused formation of invasive squamous cell cancer and desmoplasia associated with production of GM-CSF, CCL2, M-CSF, and TNF. These events resulted in accumulation of immunosuppressive CD11b⁺Gr-1⁺CD124⁺ MDSCs that promoted tumor progression by activating stromal fibroblasts¹⁷⁶.

In another model of multistep squamous carcinogenesis driven by the HPV16 early-region genes (including the E6/E7 oncogenes) under the control of the human keratin-14 promoter/enhancer, CD11b⁺Gr1⁺F4/80⁻CD11c⁻ cells constituted the most abundant leukocyte subtype in premalignant skin, and accumulated progressively in the spleen. However, these cells failed to inhibit polyclonal activation of either CD4⁺ or CD8⁺ T cells and did not produce ROS⁶².

These results, together with the data discussed above from transplantable tumour models, support a two-stage model of MDSC involvement in cancer. The almost universal feature of tumour progression is activation of abnormal myelopoiesis and recruitment of immature myeloid cells into tissues. This process is governed by diverse soluble factors and is dependent upon up-regulation of STAT3 and other key transcription factors (Fig. 5). Myelopoiesis during acute infections, stress, or trauma results in rapid terminal differentiation of myeloid cells. In contrast, cancer myelopoiesis is associated with defective myeloid cell differentiation, which results in accumulation and persistence of immature myeloid cells. Although necessary, these events are not sufficient to generate immunosuppressive MDSCs: activation of cells via a network of regulatory mechanisms is also required (Fig. 5). Activation of these mechanisms in mice with most transplantable tumours and many, but not all, spontaneous tumours, results in the accumulation of immunosuppressive MDSCs. In tumor sites these cells further differentiate into TAMs and possibly into suppressive DCs. In patients with cancer, cells with an MDSC phenotype are almost universally immune suppressive, which may reflect their isolation from patients with advanced disease. If immunosuppressive activity is not a property of the first wave of immature myeloid cells recruited to tumors, continuous stimulation of myelopoiesis and activation of immature myeloid cells by tumour-derived soluble factors may drive the subsequent accumulation of immunosuppressive MDSCs that support tumor promotion and form the metastatic niche. Accordingly, the oncogenic programme may influence the functional immunosuppressive activity more than the accumulation of CD11b⁺Gr-1⁺ cells. Thus, the transition from immature myeloid cells to MDSCs might be defective in some experimental tumour models, and different oncogenic programmes may differentially affect the kinetics of immature myeloid cell to MDSC conversion. Combinations of GM-CSF, G-CSF, IL-6 and IL-13 induce the rapid differentiation of cells similar to MDSCs from human and mouse bone marrow precursors in vitro ^{91, 134, 177,140}. These studies may provide the framework for identifying key molecules governing the stages of MDSC maturation.

Therapeutic targeting of myeloid cells

Knowledge of the molecular mechanisms responsible for accumulation of MDSCs, immune suppressive macrophages and DCs in cancer has allowed for therapeutic targeting of these cells as it is increasingly clear that successful cancer immunotherapy will require limiting the immunosuppressive effects of myeloid cells. This targeting is focused on six main goals: first, inhibiting the molecular mechanisms used by myeloid cells to block lymphocyte reactivity and proliferation; second, inhibiting the expansion of MDSCs from bone marrow progenitors or inducing apoptosis of circulating MDSC; third, forcing MDSCs to mature into proficient APCs; fourth, preventing trafficking of myeloid cells from bone marrow to peripheral lymphoid organs and to tumors; fifth, repolarizing or eliminating TAMs and replacing them with M1 macrophages; sixth, restoring the antigen-presenting capabilities of DCs and macrophages (Table 1).

The proposed two-stage model for MDSC involvement might have implications for the further development of therapies. Some immunosuppressive mechanisms are common to all myeloid cells but others are unique to individual populations. Therefore, targeting common effector molecules is likely to be more effective than targeting individual suppressive pathways.

Conclusions and perspective

It is not clear whether abnormal myelopoiesis and pathological activation of myeloid cells are temporarily regulated. Do cancer cells first condition mature leukocytes, MDSCs, DCs, and macrophages that are then recruited in response to suppressor factors produced by the progressing tumor? Or, do the two processes occur concurrently and MDSCs are recruited as a pre-requisite to tumor progression? More sophisticated tumor models and techniques will be required to address this key question. It is clear that the myeloid lineage is globally altered in cancer as a single, closely integrated system involving all terminally differentiated myeloid cells and their pathologically activated immature progenitors. Although there are a multitude of phenotypic and functional changes in different myeloid cell subpopulations, these changes are governed by common tumour-derived suppressor factors and transcriptional programmes. These commonalities provide an opportunity for therapeutic interventions that may concomitantly normalize multiple myeloid cell abnormalities.