Evidence links chronic inflammation with cancer, but cellular mechanisms involved in this process remain unclear. We have demonstrated that in humans, inflammatory conditions that predispose to development of skin and colon tumors are associated with accumulation in tissues of CD33+S100A9+ cells, the phenotype typical for myeloid-derived suppressor cells in cancer or immature myeloid cells (IMCs) in tumor-free hosts. To identify the direct role of these cells in tumor development, we used S100A9 transgenic mice to create the conditions for topical accumulation of these cells in the skin in the absence of infection or tissue damage. These mice demonstrated accumulation of granulocytic IMCs in the skin upon topical application of 12-O-tetradecanoylphorbol-13-acetate (TPA), resulting in a dramatic increase in the formation of papillomas during epidermal carcinogenesis. The effect of IMCs on tumorigenesis was not associated with immune suppression, but with CCL4 (chemokine [C-C motif] ligand 4)-mediated recruitment of IL-17–producing CD4+ T cells. This chemokine was released by activated IMCs. Elimination of CD4+ T cells or blockade of CCL4 or IL-17 abrogated the increase in tumor formation caused by myeloid cells. Thus, this study implicates accumulation of IMCs as an initial step in facilitation of tumor formation, followed by the recruitment of CD4+ T cells.
Inflammation develops in response to foreign agents or injuries and is crucial to the healing process. However, persistent unresolved inflammation can contribute to the development of cancer (Coussens et al., 2013). Despite the wealth of information implicating inflammation in tumor development, the specific role of different cells, especially myeloid cells, in this process remains largely unclear. Myeloid cells are an important component of inflammation. There is now ample evidence of abnormalities in the myeloid compartment in cancer, which manifests in inhibition of differentiation of DCs, polarization of macrophages (MΦs) toward M2 functional state, and dramatic expansion of myeloid-derived suppressor cells (MDSCs; Gabrilovich et al., 2012). MDSCs represent a heterogeneous population of pathologically activated myeloid cells that includes precursors of neutrophils (polymorphonuclear neutrophils [PMNs]), MΦs, DCs, and cells at earlier stages of myeloid cell differentiation (Gabrilovich and Nagaraj, 2009; Peranzoni et al., 2010; Youn et al., 2012). In mice, these cells are defined as Gr-1+CD11b+ cells with the recognition of polymorphonuclear (PMN-MDSC) and mononuclear (M-MDSC) subsets based on the expression of Ly6C and Ly6G markers (Fridlender et al., 2009; Peranzoni et al., 2010; Brandau et al., 2011; Youn et al., 2012). MDSCs are characterized by a potent immune-suppressive activity and the ability to promote tumor angiogenesis, tumor cell invasion, and metastases (Bierie and Moses, 2010; Gabrilovich et al., 2012; Talmadge and Gabrilovich, 2013). In tumor-free mice, cells with the same phenotype represent immature myeloid cells (IMCs) lacking immune-suppressive activity. Expansion of MDSCs is considered a consequence of tumor progression. However, in recent years, it has become clear that the cells with phenotypes and functions attributed to MDSCs are readily detectable in different conditions associated with chronic inflammation not directly linked to cancer (Cuenca et al., 2011; Nagaraj et al., 2013). We hypothesize that these cells can contribute to tumor development associated with inflammation. Understanding the role of specific components of inflammation in tumor development is difficult because of the fact that inflammation is a multicomponent complex process. To address this question, we focused on S100A9 protein. This is the member of the S100 family of Ca2+-binding proteins with diverse biological activity, including chemotaxis of myeloid cells, fatty acid transport, production of reactive oxygen species, etc. (Markowitz and Carson, 2013). Expression of S100A9 together with its dimerization partner S100A8 is found predominantly in cells of the myeloid lineage. Differentiation of myelocytes/granulocytes is associated with an increase of S100A8/A9 expression, whereas differentiation of MΦs and DCs is associated with loss of their expression (Leder et al., 1990; Ehrchen et al., 2009; Srivastava et al., 2012; Markowitz and Carson, 2013). Cells of the lymphoid lineage do not express these proteins. We and others have previously found that accumulation of MDSCs and inhibition of DC differentiation in cancer were closely associated with up-regulation of S100A8/A9 (Cheng et al., 2008; Sinha et al., 2008; Ichikawa et al., 2011). The expansion of MDSCs was significantly reduced in S100A9-deficient mice treated with complete Freund’s adjuvant or tumor-bearing mice (Cheng et al., 2008). In contrast, overexpression of S100A9 in mice resulted in accumulation of cells with MDSC phenotype (Chen et al., 2013). Furthermore, these data were consistent with a recent report that DCs derived from S100A9-deficient mice induced stronger response of allogeneic T cells (Shimizu et al., 2011). In recent years, S100A9 was identified as a marker of MDSCs in peripheral blood of cancer patients and tumor-bearing mice (Feng et al., 2012; Källberg et al., 2012; Zhao et al., 2012). We asked whether regulation of S100A9 level in myeloid cells could be used to dissect the possible role of MDSCs during early stages of tumor development. Here, we report that cells with the MDSC phenotype accumulate in tissues of patients with precancerous inflammation and that these cells play a major role in tumor development in mice. Importantly, this effect was not directly mediated by immune-suppressive mechanisms but by the recruitment of IL-17–producing CD4+ T cells.
S100A9-positive myeloid cells in tissues of patients with inflammation or cancers
Although parameters for detection of MDSCs in peripheral blood of cancer patients by flow cytometry are well established, immunohistochemical detection of these cells in tissues has not been developed. Thus, we designed a double-staining protocol of tissues for the CD33 myeloid marker and S100A9 using confocal microscopy (Fig. 1 A). S100A9 was used for MDSC detection because this protein is directly implicated in accumulation of these cells in cancer and is shown as an MDSC marker in peripheral blood of cancer patients. The presence of CD33+S100A9+ was then assessed in skin of patients with melanoma and basal cell carcinoma (BCC), as well as in patients with several inflammatory skin diseases (Fig. 1 B). No increase in the presence of these cells was found in patients with lichen planus or psoriasis as compared with controls. In contrast, patients with spongiotic dermatitis (eczema) had significantly (P = 0.001) higher numbers of CD33+S100A9+ cells in skin. The presence of these cells was a little higher in patients with BCC. Patients with melanoma had a dramatic increase in number of these cells (Fig. 1 C). CD33+S100A9+ cells can represent not only IMCs or MDSCs but also PMNs because the expression of S100A9 is retained in these cells as opposed to MΦs or DCs (Leder et al., 1990; Ehrchen et al., 2009; Srivastava et al., 2012; Markowitz and Carson, 2013). Therefore, we evaluated the presence of PMNs in skin using a specific marker neutrophil elastase. No difference in the number of PMNs was found for any of the conditions except in patients with melanoma (Fig. 1 C). CD163+ MΦs were modestly increased in skin of patients with eczema, BCC, and melanoma (Fig. 1 C). We also evaluated colon tissue samples from a cohort of patients with ulcerative colitis with evidence of dysplasia and colon from cancer patients. Dramatic (>10-fold) increase in the presence of CD33+S100A9+ cells was found in tissues from patients with colitis and colon cancer, whereas the number of PMNs and MΦs increased less than twofold (Fig. 1 E). Lichen planus and psoriasis are not associated with the development of skin cancer, whereas history of eczema is an independent predictor of BCC (Dyer et al., 2012). Ulcerative colitis, especially in patients with dysplasia, is associated with colon cancer (Dalgleish and O’Byrne, 2006; Dyson and Rutter, 2012). These results indicate that the population of CD33+S100A9+ cells represented by IMCs or MDSCs is dramatically increased in select inflammatory conditions, which are associated with an increased risk for developing cancer.
The role of IMCs in tumor development
Next we asked what role, if any, these cells can play in tumor development. To address this question, we used mice with overexpression of S100A9 protein in hematopoietic cells (S100A9Tg mice; Cheng et al., 2008). S100A9Tg mice (FVB/N background) had no detectable abnormalities during the first months of life (Chen et al., 2013). Compared with WT mice, S100A9Tg mice expressed higher amounts of S100A9 protein in BM, spleen, LNs, and lung, but not in skin or liver (Fig. 2 A). Two models of skin carcinogenesis were used. Tg.AC mice (also FVB/N) express the mutant v-h-ras transgene under the ζ-globin promoter. When treated topically with the tumor promoter 12-O-tetradecanoylphorbol-13-acetate (TPA), Tg.AC mice develop multiple papillomas, some of which progress to a malignant stage (Humble et al., 2005). In C57BL/6 mice, skin tumor formation was induced by a single application of the carcinogen 7,12-dimethylbenz(a)anthracene (DMBA), followed by 12-wk topical treatment with TPA.
Topical treatment of WT mice with TPA for 4 wk did not result in detectable levels of S100A9 in skin. In contrast, untreated S100A9Tg mice had a low level of the protein, and TPA treatment of S100A9Tg mice caused a substantial increase in the amount of the protein (Fig. 2 B). Although treatment of WT mice with TPA resulted in only a small increase in the number of Gr-1+ cells in skin (evaluated by immunohistochemistry [IHC]), in S100A9Tg mice these cells increased approximately threefold (P < 0.01; Fig. 2, C and D). Evaluation of cells derived from skin sample digests by flow cytometry revealed that all the Gr-1+ cells also expressed CD11b and thus have phenotypic markers of MDSCs. The proportion of Ly6ChiLy6G− monocytic cells among CD11b+ cells in TPA-treated skin was the same in WT and S100A9Tg mice, whereas the proportion of Ly6CloLy6G+ granulocytic cells was significantly higher in S100A9Tg than in WT mice (Fig. 2 E). No differences in the numbers of F4/80+ MΦs between WT and S100A9Tg skin were found (Fig. 2 F), whereas the number of CD11c+ skin DCs was significantly lower in both acetone- and TPA-treated S100A9Tg than in WT mice (Fig. 2 G). We investigated whether the dramatic increase of granulocytic cells in the skin of S100A9Tg mice was the result of their expansion in BM. There was a slight increase in the presence of these cells in BM of TPA-treated S100A9Tg versus WT mice, but the ratio between granulocytic and monocytic subsets of IMCs was the same (Fig. 2, H and I). These data indicate that, consistent with previous observations (Cheng et al., 2008), the overexpression of S100A9 in myeloid progenitors did not cause expansion of myeloid cells but rather diverted their differentiation toward immature granulocytic cells. As a result, topical application of TPA in S100A9Tg mice caused substantial accumulation of IMCs with granulocytic phenotype, whereas the presence of MΦs and DCs was either unchanged or decreased.
To study skin carcinogenesis, the S100A9Tg mice were crossed with homozygous Tg.AC mice, and hemizygous monotransgenic (Tg.AC) and bitransgenic (S100A9; Tg.AC) littermates were topically treated twice weekly for 6 wk with the tumor promoter TPA. Papillomas developed much faster and at significantly (P < 0.001) higher numbers in bitransgenic mice expressing S100A9Tg than in monotransgenic mice (Fig. 3 A). S100A9 expression could be induced in keratinocytes under inflammatory conditions (Sorenson et al., 2012). To clarify the role of infiltrating BM-derived cells in increased tumorigenesis in S100A9Tg mice, lethally irradiated Tg.AC mice were reconstituted with BM from WT or S100A9Tg mice, and papilloma formation was similarly evaluated after TPA treatment. Recipients of S100ATg BM developed significantly (P < 0.05) more papillomas than recipients of BM from WT mice (Fig. 3 B), demonstrating that up-regulation of S100A9 in BM-derived hematopoietic cells was responsible for increased tumor formation. To confirm a specific role of IMCs in tumor development, Gr-1+ cells were depleted in Tg.AC mice with Gr-1 antibody during the first 3 wk of TPA treatment. This treatment significantly (P < 0.05) reduced papilloma formation (Fig. 3 C). To evaluate the direct effect of accumulation of IMCs on proliferation of keratinocytes, WT and S100A9Tg mice were treated with TPA for 5 wk, and the proportion of proliferating keratinocytes (cytokeratin 14+Ki67+ cells) was evaluated. TPA treatment caused proliferation of keratinocytes in WT mice, which was significantly (P < 0.01) enhanced in S100A9Tg mice (Fig. 3 D).
To better ascertain the possible role of IMCs in tumor development, we also used mice with targeted deletion of S100A9 (S100A9KO mice, C57BL/6 background; Manitz et al., 2003). In contrast to FVB/N mice, untreated WT C57BL/6 mice had traces of S100A9 protein in the skin, which was increased after 4 wk of TPA treatment (Fig. 4 A). S100A9KO mice had no detectable S100A9 protein in the skin (Fig. 4 A). Treatment of WT C57BL/6 mice with TPA greatly increased the number of Gr-1+ myeloid cells in the skin, whereas in S100A9KO mice, this increase was very modest and lower than in WT mice (Fig. 4 B). In contrast to S100A9Tg mice, S100A9KO mice developed significantly (P < 0.01) fewer skin tumors than their WT C57BL/6 counterparts after treatment with both DMBA and TPA (Fig. 4 C). Furthermore, mice reconstituted with BM from S100A9KO mice had significantly (P < 0.01) fewer lesions than mice reconstituted with WT BM (Fig. 4 D), confirming the critical role of BM-derived cells in modulating skin tumorigenesis. To further verify a specific role of BM-derived cells in tumor development, WT and S100A9 KO mice were reconstituted with BM from WT mice. No differences were found in the numbers of papillomas between WT and S100A9KO mice recipients treated with DMBA followed by TPA (Fig. 4 E). Collectively, these data demonstrate that development of skin tumors directly depends on the accumulation of granulocytic IMCs in the skin. Next we investigated the mechanism or mechanisms by which IMCs mediated tumor promotion.
Accumulation of IMCs in skin did not result in immune suppression
We tested the hypothesis that IMCs accumulated in TPA-treated S100A9Tg skin had immune-suppressive activity and thus could be defined as MDSCs. S100A9Tg mice were backcrossed for nine generations to C57BL/6 mice. Mice were treated for 6 wk with topical application of TPA. Ly6G+ granulocytic IMCs were isolated from the skin and were co-cultured with splenocytes from OT-1 mice stimulated with control or specific peptides. In contrast to the cells with the same phenotype isolated from spleens of tumor-bearing mice (MDSCs), skin-infiltrating granulocytic cells from TPA-treated mice did not suppress antigen-specific T cell response (Fig. 5 A). Similarly, Gr-1+CD11b+ cells isolated from BM of WT and S100A9Tg mice treated with TPA also lacked immune-suppressive activity (Fig. 5 B). In contrast, cells with the same phenotype isolated from BM of tumor-bearing mice (MDSCs) demonstrated profound suppressive activity (Fig. 5 B). Because S100A9Tg mice had reduced presence of CD11c+ DCs in the skin as compared with WT mice, we analyzed in more detail the different populations of DCs in skin by flow cytometry. Only slight differences between WT and S100A9Tg mice were seen in the proportion of subsets of skin DCs (Fig. 5 C), with the exception of CD207+CD11b−CD103+ DCs, which were significantly elevated in S100A9Tg mice compared with WT mice (Fig. 5 C). S100A9Tg had a modestly (P < 0.01) reduced number of Langerhans cells (LCs) in epidermis (Fig. 5 D). To determine whether those changes would translate into decreased migration of DCs to draining LNs, WT and S100A9Tg mice were treated topically with acetone vehicle or TPA for 4 wk, after which the skin was painted with fluorescent dye DDAO. The next day, skin-draining LNs were collected and DDAO+ DCs were counted. We found significantly lower numbers of DDAO+CD86+IAbright migratory DCs in LNs of S100A9Tg mice than in WT mice (Fig. 5 E).
We wondered whether reduction in DC numbers and migration in S100A9Tg skin could result in impaired priming of CD8+ T cells. DDAO-labeled OVA-specific OT-1 T cells were transferred to WT or S100A9Tg C57BL/6 mice pretreated for 4 wk with TPA. OVA protein was applied to the same part of the skin as TPA, and 3 d later, proliferation of OT-1 T cells was evaluated in LNs and spleen. Robust proliferation of OT-1 cells was observed in all mice. No differences were found between WT and S100A9Tg mice (Fig. 5 F). These results indicate that despite reduced presence of DCs in the skin, antigen-specific response was unaffected in S100A9Tg mice. Together with the data indicating lack of immune-suppressive activity of IMCs, this finding suggests that immune suppression is not the primary reason for increased tumor formation in S100A9Tg mice.
IMCs recruit CD4+ T cells to the skin
We evaluated the presence of lymphocytes in the skin of vehicle- and TPA-treated mice. No significant differences between WT and S100A9Tg mice were found in the presence of B lymphocytes, NK cells, or CD8+ T cells (not depicted). In contrast, treatment with TPA resulted in the marked accumulation of CD4+ T cells in the skin that was significantly (P < 0.01) higher in S100A9Tg mice than in WT mice (Fig. 6 A). A small statistically nonsignificant increase was observed in the population of γδ T cells (Fig. 6 B). Conversely, in S100A9KO mice, TPA only caused a modest increase in skin CD4+ T cells as compared with the prominent accumulation observed in WT C57BL/6 mice (Fig. 6 C). Skin CD8+ T cells in both WT and S100A9KO mice were comparably low and unaffected by TPA treatment (not depicted).
The characteristics of CD4+ T cells in skin of TPA-treated mice were studied using flow cytometry. Consistent with the data obtained by IHC, the proportion of CD4+ T cells among hematopoietic CD45+ cells in skin was significantly (P < 0.001) higher in S100A9Tg mice than in WT mice (Fig. 6 D). The prevalence of different subsets within the population of CD4+ T cells in skin was measured using intracellular cytokine staining. No substantial differences were found between WT and S100A9Tg mice in the proportion of most CD4+ T cell populations. However, S100A9Tg skin showed higher proportions of IL-22+ and IL-10+ cells and FoxP3+CD4+ T reg cells and a significantly lower proportion of IL-13+CD4+ cells than WT skin (Fig. 6 E). No differences were seen between WT and S100A9Tg skin treated with acetone vehicle alone (not depicted). It was possible that although the proportion of cytokine-producing CD4+ T cells was not changed, the level of cytokine production on a per cell basis could be different. However, no differences were found in the fluorescence intensity of those cytokines that showed no changes in the proportion of cells (not depicted). To investigate the ability of IMCs to directly polarize CD4+ T cells in vitro, naive CD4+CD62L+ T cells were co-cultured with IMCs isolated from BM of TPA-treated WT or S100A9Tg mice. No evidence of CD4+ T cell polarization was found (Fig. 6 F). Thus, accumulation of IMCs in skin of S100A9Tg mice was associated with substantial increase in the presence CD4+ T cells. Gr-1+ IMCs and CD4+ T cells localized in the same vicinity in dermis of TPA-treated mice. In contrast, γδ T cells resided in the different area of the skin (close to epidermis; Fig. 6 G). We proposed that although IMCs did not cause specific polarization of CD4+ T cells, they could recruit a substantial proportion of proinflammatory T cell populations. This hypothesis was tested in further experiments.
We investigated whether enhanced tumor formation in S100A9Tg mice was mediated by CD4+ T cells. S100A9Tg;Tg.AC bitransgenic mice were treated with control IgG- or CD4-specific antibody before and during the first 3 wk of TPA application. CD4 antibody significantly (P < 0.01) reduced the formation of papillomas in these mice (Fig. 7 A), indicating that the tumor-promoting effect of IMCs is mediated via recruitment of CD4+ T cells. We asked whether depletion of CD4+ T cells could affect TPA-inducible accumulation of IMCs in skin. To address this question, the presence of CD4+ T cells and Gr-1+ IMCs was evaluated in skin of mice treated with TPA and CD4 antibody for 5 wk. This treatment resulted in abrogation of CD4+ T cell accumulation but did not affect Gr-1+ cells (Fig. 7 B).
Because IMCs attracted a large number of CD4+ to the skin, even without causing preferential recruitment of T cell populations, the total number of proinflammatory CD4+ T cells in the skin would substantially increase. We asked whether the levels of cytokines produced by CD4+ T cells were different in skin of S100A9Tg and WT mice. The cytokines were also measured in spleen and BM, where no significant changes in myeloid or lymphoid cells were observed. No differences were found in the amount of IL-10 or IL-22 between TPA-treated WT and S100A9Tg mice. In contrast, the amount of IL-17A was significantly higher in the skin but not spleen and BM of S100A9Tg mice than in WT mice (Fig. 7 C). CD4+ T cells appear to play a major role in IL-17A accumulation in skin because depletion of CD4+ T cells abrogated TPA-inducible increase in IL-17A in the skin (Fig. 7 D). These data suggested that IL-17 could play an important role in IMC-mediated promotion of tumor formation. To test this hypothesis, we used IL-17 neutralizing antibody in C57BL/6 S100A9Tg and WT mice treated with DMBA and TPA. Blockade of IL-17 dramatically reduced papilloma formation in S100A9Tg mice (Fig. 7 E). Thus, accumulation of granulocytic IMCs in skin resulted in recruitment of CD4+ T cells and to a lesser extent γδ T cells that could promote tumor development via release of IL-17.
The mechanism of IMC-inducible recruitment of CD4+ T cells
How could IMCs recruit CD4+ T cells? We evaluated the ability of IMCs to attract CD4+ cells in vitro. Supernatants from normal BM-derived IMCs had no effect on CD4+ T cell migration. However, IMCs activated with LPS or TPA secreted potent chemotactic activity for CD4+ T cells (Fig. 8 A and not depicted). Importantly, chemotaxis of CD4+ T cells was observed only if T cells were also stimulated before the assay (Fig. 8 A). When IMC supernatants were placed in both chambers during the assay, CD4+ T cell migration was significantly reduced, indicating that the effect was indeed the result of chemotaxis (not depicted).
Using RNA isolated from skin, we examined expression of chemokines known to be involved in the migration of CD4+ T cells. Expression of ccl3 (chemokine [C-C motif] ligand 3) and ccl4 RNAs was significantly (P < 0.05) up-regulated in the skin of TPA-treated S100A9Tg mice compared with WT mice, whereas the expression of ccl22 was modestly increased (Fig. 8 B). However, no differences in the amount of CCL3 and CCL22 proteins were found in skin lysates of TPA-treated S100A9Tg and WT mice (Fig. 8 C). In contrast, the amount of CCL4 in the skin of TPA-treated S100A9Tg mice was significantly higher than in WT mice. No differences in the amount of CCL4 were observed in spleens, BM, lung, or liver (Fig. 8 D). Because skin was the primary site of IMC accumulation in TPA-treated S100A9Tg mice, these data are consistent with the role of IMCs as the primary source of CCL4. To verify that IMCs are indeed able to produce CCL4, this chemokine was measured in supernatants from BM IMCs isolated from TPA-treated WT and S100A9Tg mice. A large amount of CCL4 was found in supernatants from stimulated IMCs (Fig. 8 E). Next, we evaluated the nature of stimuli that could induce expression of ccl4 in IMCs isolated from BM of naive mice. IMCs were treated for 24 h with several proinflammatory cytokines. IFN-γ caused more than fourfold up-regulation of ccl4 expression. The effect of TNF was significant but less potent, whereas IL-1β at a selected concentration did not up-regulate ccl4 expression in IMCs (Fig. 8 F). We assessed the expression of ccl4 in cells isolated directly from skin of WT or S100A9Tg mice. Skin Gr-1+ IMCs from TPA-treated WT or S100A9Tg mice expressed a high level of ccl4, whereas no expression was detected in Gr-1− cells (Fig. 8 G).
To verify the role of CCL4 in CD4+ T cell recruitment by Gr-1+ myeloid cells, we added neutralizing CCL4 antibody to IMC supernatants in chemotaxis assay. CCL4-specific antibody abrogated CD4+ T cell migration caused by IMCs in a dose-dependent manner (Fig. 9 A). To test the role of CCL4 in vivo, WT and S100A9Tg mice were treated for 4 wk with control or CCL4 antibody concurrent with TPA. Treatment with CCL4 antibody did not affect the presence of Gr-1+ myeloid cells or CD8+ T cells in the skin (not depicted). In contrast, neutralization of CCL4 significantly (P < 0.01) reduced the presence of CD4+ T cells in WT and S100A9Tg skin (Fig. 9 B). CCL4 antibody significantly reduced the presence of IL-17A in the skin of S100A9Tg mice, strongly suggesting the link between CCL4-mediated CD4+ T cell migration and IL-17A presence in the skin (Fig. 9 C). Weekly treatment of Tg.AC mice with CCL4 antibody during the first 3 wk of TPA promotion significantly reduced the formation of tumors (P < 0.01; Fig. 9 D). Similar experiments were performed in S100A9Tg;Tg.AC bitransgenic mice who developed higher numbers of papillomas than Tg.AC mice, but the CCL4 antibody still drastically (P < 0.01) reduced the number of tumors (Fig. 9 E).
In this study, we report, for the first time, that granulocytic IMCs directly contribute to skin tumor development via recruitment of IL-17–producing CD4+ T cells (Th17 cells). Although the link between inflammation and cancer is well established (Arthur et al., 2012; Coussens et al., 2013), the contribution of specific cell populations to the initial phase of tumor development remains unclear. Previously, MΦs were shown to be involved in early stages of tumor development and progression (Hagemann and Balkwill, 2005; Sica and Bronte, 2007; Schmid and Varner, 2012). The role of MDSCs in tumor immune escape is now well established. In recent years, evidence pointed to expansion of MDSCs in chronic infection and inflammation (Cuenca et al., 2011; Nagaraj et al., 2013). This increase in MDSCs raised the question as to whether these cells could contribute to tumorigenesis. Our data demonstrated accumulation of a large number of CD33+S100A9+ cells (the phenotype typical for MDSCs) in tissues of patients with inflammatory conditions, which predisposes to the development of cancers. Accumulation of these cells was more prominent than that of mature neutrophils and MΦs.
Because inflammation is a complex process, the specific role of individual cell populations in tumor promotion is difficult to ascertain. S100A9Tg mice provide a novel experimental model in which tissue accumulation of cells with the IMC phenotype is not caused by infection, trauma, or cancer. The vast majority of myeloid cells accumulated in skin of S100A9Tg were granulocytic IMCs, which is consistent with the concept that overexpression of S100A9 in myeloid precursors inhibits differentiation of myeloid precursors to DCs, and to some extent MΦs, and promotes appearance of immature granulocytes. Activated keratinocytes can attract granulocytes via release of CXCL8 chemokine (Huang et al., 2011). This chemokine may provide a signal for initial influx of IMCs to TPA-treated skin. Our results with BM transfers and Gr-1–specific antibody depletions in S100A9Tg mice, as well as decreased skin carcinogenesis in S100A9KO mice, demonstrate that IMCs were the primary facilitator of tumor development in this experimental system. Our data in S100A9KO mice are not consistent with a recent study indicating that S100A9 had a protective role in inflammation-induced skin carcinogenesis (McNeill and Hogg, 2014). Whether differences were caused by differences in the mouse models or other factors needs to be elucidated. IMCs accumulated in skin of mice lack immune-suppressive activity and thus cannot be defined as MDSCs. These results were consistent with recent findings of accumulation of myeloid cells with a phenotype similar to that of MDSCs but lacking immune-suppressive activity in mice exposed to cigarette smoke (Ortiz et al., 2014). These cells acquired immune-suppressive activity only after the development of lung tumors (Ortiz et al., 2014). Our data demonstrated that although the number of DCs migrating from skin to LN in S100A9Tg mice was decreased, the priming of CD8+ T cells to antigen applied over the skin was not affected. Apparently, the modest decrease in DCs in skin was not sufficient to inhibit CD8+ T cell priming. Alternatively, the significantly higher proportion of CD11b−CD207+CD103+ DCs in S100A9Tg mice may have compensated for decrease of other DC subsets because these cells play a major role in cross-presentation of antigens in the skin (Henri et al., 2010). Collectively, these data indicate that enhanced papilloma formation in S100A9Tg mice is not caused by myeloid cell–mediated immune suppression as typically observed with MDSCs in cancer.
Granulocytic cells may contribute to neoplastic transformation by increasing the rate of mutation and proliferation of epithelial cells via release of reactive oxygen species (Marnett, 2000; Le’Negrate et al., 2003; Balkwill, 2009), activation of cyclooxygenase-2 (Kim et al., 2005), and inhibition of nucleotide excision repair by myeloperoxidase (Gasche et al., 2001; Hofseth et al., 2003; Güngör et al., 2007). In this study, we did not evaluate the malignant conversion of tumor lesions but focused on the earlier premalignant events promoting papilloma formation. We found that granulocytic IMCs recruited CD4+ T cells to the site of their accumulation. Importantly, only activated IMCs were able to recruit CD4+ T cells. Our data indicated that IMCs did not specifically recruit Th17 CD4+ T cells or cause their conversion from precursors, but rather recruited these cells as part of large pool of CD4+ T cells. It was associated with dramatic decrease in IL-13–producing CD4+ T cells. IL-13 was previously shown to be a negative regulator of Th17 cells (Newcomb et al., 2009). This resulted in significant increase of IL-17 but not IL-22 in the skin of S100A9Tg mice as compared with WT mice. Depletion of CD4+ T cells or IL-17 in S100A9Tg mice abrogated induction of papilloma formation. These data strongly indicate that CD4+ T cells and IL-17 mediated the effect of IMCs on tumor development.
The role of CD4+ T cells in tumor progression remains enigmatic and thus requires further investigation. The CD4+ T cell deficiency on one hand enhanced cancer progression in methylcholanthrene (MCA)-initiated sarcoma development (Koebel et al., 2007; DeNardo et al., 2010), whereas on the other hand, it inhibited tumor development in the two-stage squamous carcinogenesis model (Girardi et al., 2003). Similarly, in the K14HPV16 skin cancer model, with E6/E7 oncogenes expressed in epidermis from keratin 14 promoter, CD4+ T cell deficiency modestly attenuated neoplastic progression, whereas cervical carcinoma development was significantly enhanced (Daniel et al., 2003, 2005). However, there are studies suggesting that IL-17 and IL-22 produced by Th17 cells contribute to tumor development by inducing angiogenesis or directly affecting keratinocyte proliferation and survival (Heidenreich et al., 2009; Cho et al., 2012; Fujita, 2013). IL-17 also promotes accumulation of cells with MDSC phenotype (He et al., 2010), suggesting a feedback mechanism. IL-17 KO mice in a two-stage carcinogenesis model showed delayed papilloma formation compared with WT mice (Xiao et al., 2009; Wang et al., 2010).
We identified CCL4 chemokine as the main factor responsible for IMC-mediated recruitment of CD4+ T cells. CCL4 or MIP-1β is a 69-aa member of the β (CC) family of chemokines. It is produced by various cells, including myeloid cells, and binds to CCR1, CCR5, and US28 receptors. Melanoma-associated MDSCs produce high levels of CCL4 among other chemokines, and in CCR5-deficient mice growth of transplantable tumors was delayed (Schlecker et al., 2012). Activated CD4+ T cells up-regulate CCR5 receptor, allowing these cells to be recruited to the site of CCL4 production (Castellino et al., 2006), which is consistent with our finding that IMCs attracted only activated CD4+ T cells. Although it was reported that different CD4+ T cell subsets may express unique repertoires of chemokine receptors, current evidence suggests that CCR5 can be expressed on recently activated cells of any subset (Bromley et al., 2008), consistent with lack of preferential accumulation of specific populations of CD4+ T cells in the skin of S100A9Tg mice.
Our data suggest a sequence of events that promote tumor formation in skin and possibly other tissues. Granulocytic IMCs accumulate at the site of cutaneous irritation or inflammation. Activated IMCs release CCL4 that recruits CD4+ T cells, including proinflammatory populations. These T cells release IL-17 to stimulate proliferation of keratinocytes, which contribute to tumor formation. The direct effect of IMCs on keratinocytes or epithelial cells in other tissues merits further investigation, as well as the possibility that targeting IMCs may benefit individuals with high risk of cancer development.
MATERIALS AND METHODS
De-identified colon tissues slides were obtained from St. Mark’s Hospital (Harrow, England, UK). Samples were taken from patients after obtaining informed consent and with the approval of the Outer West London Research Ethics Committee (UK). It included six samples obtained from normal colonic biopsies (control), five samples from patients with ulcerative colitis showing low- and high-grade dysplasia, and six samples from patients with colorectal cancer (moderately differentiated adenocarcinoma). Paraffin-embedded tissue blocks were retrieved using an approved Institutional Review Board protocol for de-identified archived skin biopsies through the Department of Dermatology, National Institutes of Health Skin Disease Research Center Tissue Acquisition Core (P30-AR057217), Perelman School of Medicine, University of Pennsylvania (Philadelphia, PA). It included five samples from normal skin (control), five samples from patients with lichen planus, five patients with psoriasis, five patients with spongiotic dermatitis (eczema), five patients with BCC, and five patients with melanoma.
After deparaffinization, rehydration, and blocking endogenous peroxides with fixation with 5% hydrogen peroxide in methanol, heat-induced antigen retrieval was performed using Tris-EDTA buffer. For evaluation of S100A9+CD33+ cells, tissues were stained with S100A9 antibody (diluted 1:500 in 5% BSA; Novus Biologicals) and CD33 antibody (dilution 1:100 in 5% BSA; Novocastra) followed by Alexa Fluor 594 antibody (dilution 1:400 in 5% BSA; Life Technologies) and Alexa Fluor 647 antibody (dilution 1:400 in 5% BSA; Life Technologies). For neutrophils and MΦs, a single staining protocol was used. Neutrophils were stained with neutrophil elastase antibody (dilution 1:2,000 in 5% BSA; Abcam) and MΦs with CD163 antibody (dilution 1:200 in 5% BSA; Abcam). Alexa Fluor 594 (Life Technologies) anti–rabbit antibody was used as secondary. Cell nuclei were stained with DAPI (dilution 1:5,000 in PBS; Life Technologies). Cells were analyzed using a TCS SP5 confocal microscope (Leica) and E600 upright microscope (Nikon). The number of cells per square millimeter of tissue was calculated in at least five fields using Image pro plus software (Media Cybernetics). Positive stained cells were normalized to total nuclear cell count using DAPI.
All animal experiments were approved by the University of South Florida Institutional Animal Care and Use Committee. Mice were housed in pathogen-free facilities. Tg.AC mice described previously (Tepper et al., 1990) were obtained from Taconic. S100A9Tg mice on FVB/N background (Cheng et al., 2008) and S100A9KO mice on C57BL/6 background (Manitz et al., 2003) were described previously. BALB/c, C57BL/6, and FVB/N mice were obtained from the National Cancer Institute, and OT-1 mice were obtained from the Jackson Laboratory. In S100A9Tg;Tg.AC bitransgenic mice, the presence of S100A9-IRES-GFP transgene was verified by the expression of GFP in leukocytes from peripheral blood, and the v-h-ras transgene was detected by genomic PCR for the SV40 sequences. In some experiments, S100A9Tg mice on C57BL/6 background were used after backcrossing S100A9Tg FVB/N mice with C57BL/6 mice for nine generations.
Female, aged-matched (7–10 wk old) littermate mice were used in experiments with Tg.AC and S100A9Tg;Tg.AC mice. Dorsal skin was shaved, and 3 nmol TPA (Sigma-Aldrich) in 200 µl acetone vehicle was applied twice a week for 4 or 6 wk. In the carcinogenesis model in C57BL/6 WT or S100A9KO mice, 100-nmol single dose of DMBA was topically applied, followed by 10 nmol TPA every 24 h for 12 wk, as previously described (Gebhardt et al., 2008). Papillomas were assessed weekly and counted when they reached 1 mm for at least 2 wk. In BM transfer experiments, lethally irradiated (950 rads) mice were injected i.v. with 106 BM cells from donor mice. Treatment with TPA or DMBA plus TPA as described above was started 3 wk after the BM transfer.
Tissue preparation and histology.
Skin pieces were snap-frozen, and slides were fixed with acetone and blocked overnight with 10% goat serum, 1% BSA, and 2.5% mouse serum in PBS at room temperature. The primary antibodies from BD were used at 1:100 dilutions: Gr1 (RB6-8C5), CD4 (H129.19), and CD8 (53-6.7). The antibodies from eBioscience were used at 1:50 dilution: CD11c (N418) and F4/80 (BM8). Biotinylated anti–rat IgG (Vector Laboratories) or anti–hamster IgG (Vector Laboratories) was used as a secondary antibody. Alkaline phosphatase kit and Vector Red substrate (Vector Laboratories) were used for visualization of the results. The tissues were counterstained with hematoxylin. Images were taken by the digital slide scanner Scanscope (Aperio) and analyzed by Aperio software. Cell number was calculated per 1 mm2. The antibodies specific for γδ T cells (GL-3), cytokeratin-14 (LL002), and Ki67 (SP6) were purchased from Abcam, and the staining was evaluated on an E600 upright microscope.
For the analysis of gene expression, total RNA was extracted with TRIzol reagent (Invitrogen), and cDNA was synthesized using the High Capacity cDNA Reverse transcription kit (Applied Biosystems). To detect chemokines, PCR was performed with 2 µl cDNA and 12.5 µl SYBR Master Mixture (Applied Biosystems) using specific primers: Ccl2, 5′-CCCAATGAGTAGGCTGGAGA-3′ and 5′-AAAATGGATCCACACCTTGC-3′; Ccl3, 5′-CCAAGTCTTCTCAGCGCCATA-3′ and 5′-GATGAATTGGCGTGGAATCTTC-3′; Ccl4, 5′-TGCTCGTGGCTGCCTTCT-3′ and 5′-CTGCCGGGAGGTGTAAGAGA-3′; Ccl5, 5′-TGCCCACGTCAAGGAGTATT-3′ and 5′-CAGGACCGGAGTGGGAGTA-3′; Ccl9, 5′-GATGAAGCCCTTTCATACTGC-3′ and 5′-GTGGTTGTGAGTTTTGCTCCAATC-3′; and Ccl22, 5′-GTGGCTCTCGTCCTTCTTGC-3′ and 5′-GGACAGTTTATGGAGTAGCTT-3′. Amplification of endogenous β-actin was used as an internal control.
For evaluation of the proteins, tissues were homogenized in RIPA buffer, and BM cells were flushed out for protein extraction. Chemokines were measured by ELISA using the kits from PeproTech: CCL3, CCL4, or CCL22. CCL4 was also measured using an R&D Systems kit. Likewise, the cytokines (IL10, IL17A, and IL22) in skin, spleen, and BM were measured by ELISA using the kits from eBioscience.
Samples (30 µg protein per lane) were subjected to electrophoresis in 12% SDS-polyacrylamide gels and then blotted onto PVDF membranes. Membranes were blocked for 1 h at room temperature with 5% dry skim milk in TBS (20 mM Tris-HCl, pH 7.6, and 137 mM NaCl plus 0.1% [vol/vol] Tween 20) and then probed with S100A9 or β-actin–specific antibodies (Santa Cruz Biotechnology, Inc.), followed by secondary antibody conjugated with horseradish peroxidase. Results were visualized by chemiluminescence detection (GE Healthcare).
Depletion of cells and neutralization of chemokines in vivo.
Mice were injected i.p. with 250 µg CD4 (clone: GK1.5; Bio X Cell) and Gr-1 (clone: RB6-8C5; Bio X Cell). Injections were performed once a week for 4 wk starting 1 wk before the beginning of TPA treatment. For neutralization of CCL4 and IL17A, 100 µg CCL4 (R&D Systems)- or 200 µg IL-17A (clone: 17F3; Bio X Cell)–specific monoclonal antibodies were used. Control groups in CD4 and Gr-1 depletion and neutralization experiments received rat IgG2b isotype (clone: LTF2; Bio X Cell); controls for IL17A neutralization experiments received mouse IgG (Sigma-Aldrich), whereas controls in CCL4 neutralization experiments received goat IgG (R&D Systems).
CD4+ T cell polarization experiments.
Naive CD62L+CD4+ T cells were isolated from spleen of FVB/N mice using magnetic beads (Miltenyi Biotec) and co-culture with Gr-1+ cells isolated from BM of WT or S100A9Tg mice treated for 4 wk with TPA. Cells were activated for 4 d at 37°C with CD3/CD28 beads. On day 4, cells were stimulated with 50 ng/ml TPA and 700 ng/ml ionomycin for 4 h in the presence of protein transport inhibitor GolgiStop (BD) to prevent cytokine secretion during the last 3 h. The cells were harvested, followed by intracellular staining with CD4, IL-17, IL-4, FoxP3, or IFN-γ antibodies (all from BD) and analysis by flow cytometry.
Skin cell isolation and analysis.
Pieces of shaved skin were incubated in PBS containing 0.2% trypsin for 30 min at 37°C, cut into small pieces, and incubated with 0.8 mg/ml collagenase in RPMI for 40 min at 37°C. Single cell suspension was prepared by pressing tissues through a 100-µm cell strainer. For analysis of the cell phenotype, cells were used directly for flow cytometry. Nonspecific staining was blocked by Fc-block (BD), and staining was performed in 1% FBS. The cells were analyzed by flow cytometry using an LSRII cytometer (BD) and FlowJo software (8.6). Dead cells were excluded by DAPI staining, and CD45+ cells were analyzed. For the analysis of intracellular cytokines, mononuclear cells were isolated by density centrifugation on Ficoll-Paque, washed, and plated at 2 × 106/ml concentration in round-bottom 96-well plates. Cells were then stimulated with 30 ng/ml TPA and 750 ng/ml ionomycin for 4 h, the last 3 h in the presence of GolgiStop. Cells were stained by Live-Dead Violet (Life Technologies), followed by CD45 and CD4 antibodies. After fixation and permeation, cells were stained for intracellular cytokines or FoxP3 and analyzed by flow cytometry. For the isolation of myeloid cells from skin, the shaved skin was cut and washed with PBS buffer. The subcutaneous fat was scrapped off using a scalpel, and the skin tissue was divided into three to four pieces. The epidermis and dermal layer were enzymatically digested and subsequently dissociated into single cell suspension using a gentleMACS dissociator (Miltenyi Biotec). IMCs were purified from single cell suspension using biotinylated anti–Gr-1 antibody (Miltenyi Biotec), followed by streptavidin microbeads. Ly6G+ granulocytic IMCs were isolated from the skin using the Epidermis Dissociation kit (Miltenyi Biotec).
Gr-1+ cells were isolated from BM using biotinylated anti–Gr-1 antibody and streptavidin microbeads (Miltenyi Biotec). The purity of Gr-1+CD11b+ cells was >95%. Cells were cultured overnight in serum-free medium (Cellgenix) at 106/ml, in the presence or absence of 0.5 µg/ml LPS. Supernatants were collected after 24 h and added at 30% vol/vol to the bottom chamber of Transwell plates (5-µm pores) in a total volume of 500 µl serum-free medium. To activate T cells, splenocytes were cultured with 0.5 µg/ml anti-CD3 antibody for 3 d in serum-free medium. One million stimulated or nonstimulated splenocytes were resuspended in 100 µl serum-free medium and added to the top chamber of the Transwell. After 4-h incubation at 37°C, CD4+ T cells that passed through the membrane to the lower chamber were measured by flow cytometry.
Epidermal sheets were prepared from dorsal skin as previously described (Gabrilovich et al., 1994). The epidermal sheets were fixed in acetone for 1 min, blocked with 1% BSA and Fc-block, and stained overnight at 4°C with IA/IE-PE antibody (clone 2G9; BD) at 1:100 dilution. The next day, sheets were washed, stained with secondary antibody, and mounted on slides. The staining was analyzed in an SP5 confocal microscope.
DC migration and T cell proliferation.
Shaved dorsal skin was painted with 0.4 ml of a 1:1 mixture of acetone/DBP containing 250 µM DDAO (Invitrogen). 24 h later, draining LNs were collected and treated with 1 mg/ml collagenase in PBS/BSA for 30 min, and single cell suspensions were stained and analyzed by flow cytometry.
For the analysis of in vitro cell proliferation, 105 splenocytes from BALB/c mice (responders) were mixed with 105 T cell–depleted and irradiated (20Gy) splenocytes from FVB/N mice (stimulators) in U-bottom 96-well plates. Gr-1+ cells from TPA-treated S100A9Tg or WT mice were added into wells at different ratios. On day 4, cells were pulsed with [3H]thymidine (1 µCi/well; GE Healthcare) for 18 h. [3H]thymidine uptake was counted using a liquid scintillation counter and expressed as counts per minute.
For the assessment of in vivo T cell proliferation in response to antigens, splenocytes and LN cells collected from OT-1 CD45.1+ mice were labeled with 5 µM DDAO. Eight million DDAO-labeled OT-1 cells were injected into the tail vein of C57BL/6 (CD45.2+) WT and S100A9Tg mice that were pretreated for 4 wk with TPA. The day after transfer, 10 mg OVA protein in 500 µl Ultrasicc gel was applied onto the shaved dorsal skin of the mice. 3 d later, skin-draining LNs and spleen were collected. Cells were stained and analyzed by flow cytometry.
Statistical analysis was performed using an unpaired two-tailed Student’s t test with significance determined at P < 0.05. For the analysis of papilloma formation, statistical significance of repeated measurements was assessed using a two-way ANOVA test.
Support was provided by Wistar Institute imaging and flow cytometry cores.
This paper was supported by National Institutes of Health (NIH) grant CA 100062 to D. Gabrilovich and, in part, by P50 CA168536. E. Celis was supported by NIH grant R01CA157303.
The authors declare no competing financial interests.
M.L. Ortiz, V. Kumar, and A. Martner contributed equally to this paper.