Research ArticleCancer

Adipocytes sensitize melanoma cells to environmental TGF-β cues by repressing the expression of miR-211

See allHide authors and affiliations

Science Signaling  23 Jul 2019:
Vol. 12, Issue 591, eaav6847
DOI: 10.1126/scisignal.aav6847

Fat triggers metastasis

In melanoma patient samples, tumors appear to grow in a lateral, proliferative phase within the upper epidermal layer of the skin but switch to a vertical, invasive phase when they grow into the deeper layers where adipose (fat) exists. Golan et al. examined cocultures of melanoma cells and adipocytes and identified a direct role for fat cells in this metastatic switch. The authors found that adipocytes secreted cytokines that signaled through their receptors on melanoma cells to repress the expression of a microRNA that promotes proliferative and suppresses invasive phenotypes. The microRNA also represses the expression of a receptor for the growth factor TGF-β, which is implicated in metastatic disease and abundant in the dermal layer. Blocking TGF-β signaling prevented the invasive switch in cultured melanoma cells and, therefore, may suppress metastasis in patients.

Abstract

Transforming growth factor–β (TGF-β) superfamily members are critical signals in tissue homeostasis and pathogenesis. Melanoma grows in the epidermis and invades the dermis before metastasizing. This disease progression is accompanied by increased sensitivity to microenvironmental TGF-β. Here, we found that skin fat cells (adipocytes) promoted metastatic initiation by sensitizing melanoma cells to TGF-β. Analysis of melanoma clinical samples revealed that adipocytes, usually located in the deeper hypodermis layer, were present in the upper dermis layer within proximity to in situ melanoma cells, an observation that correlated with disease aggressiveness. In a coculture system, adipocytes secreted the cytokines IL-6 and TNF-α, which induced a proliferative-to-invasive phenotypic switch in melanoma cells by repressing the expression of the microRNA miR-211. In a xenograft model, miR-211 exhibited a dual role in melanoma progression, promoting cell proliferation while inhibiting metastatic spread. Bioinformatics and molecular analyses indicated that miR-211 directly targeted and repressed the translation of TGFBR1 mRNA, which encodes the type I TGF-β receptor. Hence, through this axis of cytokine-mediated repression of miR-211, adipocytes increased the abundance of the TGF-β receptor in melanoma cells, thereby enhancing cellular responsiveness to TGF-β ligands. The induction of TGF-β signaling, in turn, resulted in a proliferative-to-invasive phenotypic switch in cultured melanoma cells. Pharmacological inhibition of TGF-β prevented these effects. Our findings further reveal a molecular link between fat cells and metastatic progression in melanoma that might be therapeutically targeted in patients.

INTRODUCTION

Cellular plasticity, defined as the ability of cells to interconvert between phenotypic states, enables rapid adaption to a dynamically changing multicellular microenvironment (1, 2). Phenotypic plasticity has a particular relevance to cancer progression as it increases tumorigenic potential and accounts for high metastatic abilities and resistance to therapy (3, 4). As imposed by microenvironmental cues, melanoma cells are able to reversibly switch between two phenotypic states: a proliferative, weakly metastatic state and a less proliferative, highly metastatic and state (57). These switches enable circulating metastatic cells to proliferate in favorable, secondary metastatic sites (1, 2). These proliferative and metastatic states are controlled by distinct transcriptional programs. One of the major factors responsible for regulating the transcriptional switch is the microphthalmia-associated transcription factor (MITF) (5, 8, 9), which plays a central role in melanoma progression (10, 11). MITF is the master regulator of the melanocyte lineage (11, 12) and regulates the expression of several microRNAs (miRNAs) including miR-211 (13) and miR-222/221 (14). It was previously suggested by us (15) and others (16, 17) that miR-211 mediates melanoma phenotypic plasticity. Microenvironmental conditions such as hypoxia (18) and inflammation (19) also influence melanoma plasticity; however, the cell types in the melanoma microenvironment that induce the phenotypic switch have not yet been identified.

It has been well established by us (15, 20) and others (21) that interactions between melanoma cells and the surrounding stroma play major roles in melanoma initiation, progression, metastatic potential, determination of metastasis location, and clinical outcome. Among the different components of the melanoma microenvironment are adipocytes, located in the deepest layer of the skin, and the hypodermis, which is mainly composed of differentiated adipocytes and adipose progenitors (22). Adipocytes promote melanoma growth and metastases (2330). However, as melanoma progresses, the melanoma cells encounter continuously changing microenvironments (14) that vary in adipocyte composition (3133). Previous studies have not investigated how the dynamic changes in the cellular microenvironment influence melanoma progression.

Transforming growth factor–β (TGF-β) superfamily members have prominent roles in embryonic development and adult tissue homeostasis (34, 35). Canonical TGF-β signaling is induced by ligand binding to the type II transmembrane receptor serine/threonine kinase TGF-β receptor II (TGFBR2), which, in turn, recruits and phosphorylates the type I receptor (TGFBR1; also known as ALK5) (36). Activated TGFBR1 phosphorylates the cytoplasmic receptor-regulated SMADs, SMAD2 and SMAD3, after hetero-oligomerization with SMAD4 (37). The SMAD complex accumulates in the nucleus, where it binds to SMAD-binding elements and regulates the transcription of target genes in a cell-specific manner (37). TGF-β signaling can act in autocrine and in paracrine manners and has two opposing roles in human cancer progression by promoting cytostatic effects or the epithelial-mesenchymal transition (3840).

Here, we found that the dynamic plasticity of melanoma is regulated by adipocyte cells at an earlier than previously reported phase, before melanoma invasion into the dermis, through an intercellular mechanism that sensitizes melanoma cells to environmental TGF-β. Moreover, we found that the phenotypic plasticity depends on proximity of adipocytes and is reversible—and thereby therapeutically targetable.

RESULTS

Subcutaneous adipocytes in proximity to in situ melanoma correlate with advanced disease

Melanoma initiates in the epidermis, and upon disease progression, cancer cells invade into the dermis (14, 20). From the dermis, cells can metastasize through the lymph system (41). We examined melanoma pathological specimens that had been staged using Clark’s system, which classifies the depth of melanoma invasion through the dermis into five levels. The highest staging level corresponds to the deepest invasion. Staging is reflective of aggressiveness and poor prognosis (42). Hematoxylin and eosin (H&E) staining demonstrated that, at higher Clark’s levels and in cases of lymph node metastases, melanoma cells are in closer proximity to the adipose cells than in lower stages (Fig. 1A). In samples of in situ, melanoma cells with characteristics of adipocytes were present in the upper dermis layer in proximity to melanoma cells (Fig. 1B and fig. S1B). Normal skin samples had well-defined adipose tissue located at the hypodermis layer of the skin (Fig. 1B and fig. S1B) (22). In vertical melanoma samples, tumor cells had invaded the adipocyte microenvironment (Fig. 1B and fig. S1B). These results were validated by immunostaining of these specimens for a melanoma marker (HMB-45 and S100) and an adipocyte-specific marker (perilipin) (Fig. 1C and fig. S1, C and D). To exclude the possibility that the dermal perilipin1-positive cells are fibroblasts (43), we stained specimens for a fibroblast marker FSP1. Perilipin1-positive cells located in the dermis near the melanoma cells did not express FSP1, whereas the surrounding dermal fibroblast cells were FSP1 positive (Fig. 1D).

Fig. 1 Subcutaneous adipocytes are observed in proximity to in situ melanoma.

(A) Hematoxylin and eosin (H&E) staining of representative vertical cross-sections of melanoma patient samples (n = 5) categorized by Clark’s staging system (stages II to V) and a melanoma metastasis in the lymph node. Scale bars, 2 mm (stages II to IV and lymph) and 5 mm (stage V). Black dashed lines indicate epidermal-dermal junctions; blue lines indicate tumor margins. T, tumor; AD, adipocytes. (B) H&E staining of typical vertical cross-sections from a healthy skin sample (left) and from two melanoma patient samples at different progression stages (right). Scale bar, 2 mm. Arrows indicate adipocytes. Graph plots the mean (±SEM) distance of the adipocytes from the epidermal basal layer from three independent experiments. *P < 0.05, t test. (C) Immunofluorescence analysis of consecutive slices from similar sections described and shown in (B) with melanoma marker HMB-45 (red) and adipocyte-specific marker Plin1 (green). DAPI-stained nuclei appear in blue. Scale bar, 2 mm. (D) Additional immunofluorescence analysis of the patient melanoma sections, represented with the in situ section shown in (C), with fibroblast marker FSP1 (pink) as well as HMB-45 (red) and Plin (green). Scale bars, 50 μm (top) and 20 μm (inset/bottom).

To further examine the clinical relevance of the adipocyte cells to melanoma progression, we analyzed nine additional in situ melanoma specimens. In most samples, adipocytes were present in the upper dermis in close proximity to melanoma cells (Figs. 1, B and C, and 2A and fig. S1, B and C); however, in a few samples, adipocytes were found in the hypodermis (Fig. 2A). Multiple nests indicate a rapid disease progression (44), and we therefore measured the numbers and sizes of melanoma nests in the melanoma in situ samples (Fig. 2, B and C, and fig. S2). In samples in which adipocytes were in close proximity to melanoma, there were substantially more melanoma nests than in specimens without adipocytes (Fig. 2C). No notable difference was observed in nest size (Fig. 2C). Therefore, our findings suggest that if lesions are not removed, they progress more rapidly toward malignancy when adipocytes are in proximity to the melanoma cells. The interactions between adipocytes and melanoma cells occur earlier than previously reported, beginning at the primary radial stage before melanoma invasion into the dermis, supporting disease progression.

Fig. 2 Subcutaneous adipocytes in proximity to in situ melanoma correlate with advanced disease.

(A) Top: H&E staining of representative vertical cross-sections from four in situ melanoma patient samples with hypodermal adipocytes (left) and five with dermal adipocytes (right). Scale bar, 2 mm. Bottom: Immunofluorescence staining for melanoma marker HMB-45 (red) and adipocyte-specific marker Plin1 (green) in sections from the same patients as atop. DAPI-stained nuclei appear in blue. White dashed line demarcates the basal layer of the epidermis; arrows indicate adipocytes. Scale bar, 200 μm. (B) Immunohistochemical staining for Melan-A in tissue sections from the same patients as in (A). Red circles depict irregular melanocytic nests. Scale bar, 2 mm. (C) Graphs plot the mean number (±SEM) of melanocytic nests (left) and mean size of the nests (μm2; right) in melanoma samples that displayed dermal adipocytes compared to melanoma samples that displayed hypodermal adipocytes. n ≥ 4. *P < 0.05, t test.

Adipocytes induce melanoma plasticity in a reversible manner by inhibiting miR-211 expression

To gain insight into the effect of adipocytes on melanoma cells in the radial phase, we used an in situ melanoma model (14, 20). We chose WM3682 melanoma cells, which are highly proliferative and have no intrinsic invasive abilities and therefore resemble melanoma cells in the in situ state (table S1) (13). WM3682 melanoma cells were cocultured with primary human subcutaneous differentiated adipocytes (fig. S3A) in a system that enables free flux of medium without direct cell-cell contact (experimental design scheme in Fig. 3A, left). Monocultured melanoma cells were used as a control. After coculture for 5 days, melanoma cell migration (Fig. 3B) and invasion (Fig. 3C) abilities were substantially enhanced compared to monocultured cells. In contrast, coculture with adipocytes led to a decrease in melanoma cell proliferation compared to monocultured cells (Fig. 3, D and E).

Fig. 3 Adipocytes drive melanoma plasticity in a reversible manner by miR-211 repression.

(A) Scheme of experimental design coculture and reverse coculture assays. (B to D) Analysis of migration (B), invasion (C), and XTT [2,3-bis-(2-methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5-carboxanilide] proliferation (D) assays using WM3682 melanoma cells in the indicated conditions, described in (A). Images are representative. Scale bars, 100 μm (B) and 50 μm (C) (DAPI-stained nuclei are blue). Graphs plot means (B, at the 24-hour time point; D, relative to sample at time 0) ± SEM from three or more independent experiments. *P < 0.05, **P < 0.01, and ***P < 0.005, t test. (E) Representative Ki67 immunostaining (green) and DAPI nuclear staining (blue) in WM3682 cells in the indicated conditions, described in (A). Scale bars, 50 μm. Graph plots mean ± SEM from three independent experiments; statistical analyses as in (B) to (D). (F) Volcano plot representing the distribution of significantly (fold >1.5, P < 0.05 by t test) differentially expressed miRNAs, noting miR-211, between cocultured and monocultured WM3682 melanoma cells in two independent gene profiling experiments. (G) qRT-PCR analysis of mature and pre-miR-211 expression in WM3682 melanoma cells grown under the indicated conditions, described in (A). Graph plots mean ± SEM from three independent experiments; statistical analyses as in (B) to (D). (H) Representative in vivo bioluminescent images of mice injected with WM1716 cells stably expressing miR-211 or a scrambled control at day 70 after injection. Graph plots mean ± SEM at the indicated time points from five mice per group; statistical analyses as in (B) to (D). (I) Representative ex vivo bioluminescence images of local xenografts, livers, and lungs isolated from mice injected with WM1716 cells stably transfected with miR-211 or scrambled control at 70 days after injection. Graph shows mean bioluminescence ± SEM from four mice per group; statistical analyses as in (B) to (D). (J) Schematic of experimental design to assess the effects of miR-211 expression in melanoma cells cultured with or without adipocyte-conditioned medium. (K to M) Migration (K), invasion (L), and proliferation (M) assays using WM3682 melanoma cells in the indicated conditions, described in (J). Scale bars, 100 μm (K) and 50 μm (L). Graphs, n, and statistical analyses are as described in (B) to (D).

Cancer plasticity is characterized by reversible alternations between two phenotypic states (57, 45). Therefore, we next examined whether the effects of adipocytes on melanoma cells are reversible. To this end, we used what we refer to as a reverse coculture system: After 5 days in coculture, the adipocytes were removed, the medium was replaced with unconditioned medium, and the remaining melanoma cells were grown as a monoculture for 5 days and then subjected to further analysis (experimental design scheme in Fig. 3A, right). After reverse coculture, adipocyte-induced migratory (Fig. 3B) and invasive (Fig. 3C) capacities were abrogated, and the cells proliferated similarly to melanoma cells in monoculture (Fig. 3, D and E). Together, these results suggest that adipocytes reversibly switch melanoma phenotype from a proliferative to an invasive state.

Because a single miRNA is able to target multiple genes and thereby regulate several phenotypic outcomes simultaneously (46), we next examined whether miRNAs are the downstream effectors of adipocytes. By miRNA profiling using a microarray, we identified 24 significantly differentially expressed miRNAs (1.5 < fold, P < 0.05) between melanoma cells cocultured with adipocytes and monocultured melanoma cells (Fig. 3F and data file S1). One of the most highly down-regulated miRNAs was miR-211, which was previously suggested to be a regulator of melanoma phenotypic plasticity (1517, 47).

To test whether miR-211 mediates adipocyte-driven melanoma phenotypic plasticity, we first validated the microarray results. Upon coculturing with adipocytes, reductions in both mature miR-211 and pre-miR-211 abundance were observed compared to monocultured cells (Fig. 3G), implying regulation at the transcriptional level. The inhibitory effect of adipocytes on miR-211 expression by melanoma cells was abolished upon adipocyte removal, indicative of reversibility (Fig. 3G). Notably, the decrease in miR-211 expression was reproduced upon coculture with differentiated NIH3T3-L1 adipocytes, a well-established murine fibroblast-origin model for adipose tissue (fig. S3, A and B) (48). This suggests that adipocyte-mediated hindrance of miR-211 expression is not limited to subcutaneous adipocytes. Moreover, incubation of melanoma cells with conditioned medium obtained from a differentiated adipocyte culture reversibly decreased expression of miR-211 produced by melanoma cells compared with cells grown in unconditioned medium (fig. S3C). This suggests that adipocytes exerted these actions through soluble factors.

To further explore how miR-211 contributes to the adipocyte-derived phenotypic plasticity of melanoma cells, we first determined whether miR-211 alters the melanoma transcriptome. To this end, we chose WM1716 melanoma cells, which express low levels of miR-211 and are highly invasive with weak proliferative potential (13). WM1716 cells were stably transfected with a miR-211 mimic or with a scrambled control (fig. S3D) and subjected to mRNA profiling. In cells stably transfected with miR-211, there was a notable increase in expression of genes associated with proliferation and down-regulation of genes associated with an invasive phenotype (fig. S3E) (6, 15). To validate these results, we generated an additional melanoma cell line. WM3314 cells, which have high invasive potential, low proliferation ability, and low levels of miR-211 (13), were engineered to stably express a miR-211 mimic or the scrambled control (fig. S3A). For consistency with the proliferative transcriptome, both WM1716 and WM3314 cells had higher proliferation rates upon miR-211 mimic expression compared to the slowly proliferating control cells (fig. S3F). Consistent with these results, antagomiR-mediated depletion of miR-211 from WM3526 cells (fig. S3G), which are highly proliferative and poorly invasive and express relatively high miR-211 levels (13), resulted in attenuated cell growth (fig. S3H). Furthermore, miR-211 expression in WM1716 and WM3314 cells resulted in enhanced progression through the cell cycle compared to control cells (fig. S3I), and the invasive potential of these cell lines was considerably reduced upon miR-211 expression (fig. S3J). These results are in agreement with our own (13) and others’ (47, 49) previous observations and support the hypothesis that miR-211 maintains melanoma cells in a proliferative state.

We next used a melanoma xenograft mouse model whereby WM1716 and WM3314 cells that constitutively expressed miR-211 or scrambled control as well as a gene encoding luciferase were subcutaneously injected into immunocompromised mice. Mice injected with miR-211–expressing cells displayed higher local tumor growth rates compared to mice injected with cells that expressed the scrambled control (Fig. 3H and fig. S3K). Further, ex vivo quantifications demonstrated that larger local tumors that resulted from cells that overexpressed the miR-211 mimic were larger than those that resulted from cells that expressed the scrambled control (Fig. 3I and fig. S3L). No metastases were observed in lungs and liver upon miR-211 expression, whereas massive metastatic formation resulted from scrambled control–expressing cells (Fig. 3I and fig. S3L). This demonstrates that miR-211 suppresses melanoma metastatic potential. Metastatic lesions were verified by H&E staining (fig. S3M). Together, these observations demonstrate that miR-211 augments cell growth and suppresses metastatic capacities of melanoma cells, as previously reported (16, 17).

We next examined whether the effect of adipocytes on melanoma phenotype is miR-211 dependent and whether miR-211 reexpression hinders these effects (experimental design scheme in Fig. 3J). To avoid collateral influence of miR-211 mimic transfection, these experiments were conducted by incubation of melanoma cells with conditioned medium from an adipocyte culture. Introduction of miR-211 mimic into WM3682 melanoma cells (fig. S3N) conditioned with adipocyte medium completely abolished the adipocyte-induced migration (Fig. 3K) and invasion (Fig. 3L) abilities observed in scrambled control–transfected cells. The suppressive effect of adipocytes on cell proliferation was reversed upon transfection of melanoma cells with a miR-211 mimic (Fig. 3M). Introduction of the miR-211 mimic into WM3682 melanoma cells not grown in adipocyte-conditioned medium had little effect on cell proliferation (Fig. 3M), likely due to the high basal miR-211 expression in these cells (13). These results demonstrate that adipocytes drive melanoma plasticity by inhibition of miR-211 expression.

Adipocytes decrease miR-211 expression through secretion of IL-6 and TNF-α

The secreted cytokines interleukin-6 (IL-6) and tumor necrosis factor–α (TNF-α) are components of the adipocyte secretome (50) and are known to mediate pro-invasive propensities of cancer cells (51). Moreover, IL-6 and TNF-α decrease the abundance and transcriptional activity of MITF (52, 53), a regulator of melanoma phenotypic switch (16), which transcriptionally promotes miR-211 expression (13). To test whether adipocytes exert their effects on melanoma cells by secretion of these two cytokines, we first examined whether subcutaneous adipocytes secrete IL-6. Medium conditioned with adipocytes had substantially higher amounts of IL-6 compared to unconditioned medium (Fig. 4A). Further, treatment of WM3682 cells with both IL-6 and TNF-α mimicked the effect of culture with adipocytes: It resulted in reduced miR-211 expression (Fig. 4B), increased invasion capacity (Fig. 4C), and decreased proliferation (Fig. 4D) in WM3682 melanoma cells.

Fig. 4 Adipocytes decrease miR-211 expression through secretion of IL-6 and TNF-α.

(A) Amount of IL-6 in medium conditioned by adipocytes compared to that in control medium. Data are means ± SEM from three independent experiments. ***P < 0.005, t test. (B to D) Analysis of miR-211 expression by qRT-PCR (B), invasion (C), and proliferation by XTT assay (D) in WM3682 melanoma cells treated for 48 hours with IL-6 (40 ng/ml) or TNF-α (15 ng/ml) relative to each in vehicle-treated cells. Images (C) are representative; DAPI-stained nuclei are blue. Scale bar, 50 μm. Relative proliferation (D) was assessed at day 3. Data are means ± SEM from three independent experiments. *P < 0.05 and ***P < 0.005, t test. (E to H) Experimental schematic (E) used to assess miR-211 expression by qRT-PCR (F), invasion (G), and relative proliferation by XTT assay (H) in WM3682 melanoma cells cultured in adipocyte-conditioned medium and treated with an IL-6 inhibitor (IL-6-i) (cucurbitacin I, 0.1 μM) or a TNF-α inhibitor (TNF-α-i) (R7050, 0.5 μM) for 48 hours. Graph plots, n, statistical analyses, images, and scale bars are as described in (B) to (D). (I) Representative images and analysis of healthy donor normal skin and in situ melanoma patient sections (examined in proximity of adipocytes or not) immunofluorescently stained for melanoma marker HMB-45 (pink), IL-6 (red), and adipocyte-specific marker Plin1 (green) and counterstained with DAPI (blue). Scale bars, 50 μm (top) and 20 μm (insets/bottom). Arrows indicate IL-6. (J) Kaplan-Meier survival plot of patients bearing melanoma with no overexpression (blue) or with overexpression of the IL-6 receptor (IL-6R) (red). Data were obtained from the Cancer Genome Atlas (n = 550; P = 0.00971).

We next asked whether the adipocyte-induced melanoma phenotypic plasticity is IL-6 and TNF-α dependent. To this end, we examined the influence of adipocyte-conditioned medium on miR-211 expression in WM3682 melanoma cells in the presence of cucurbitacin I, an inhibitor of IL-6 signaling (54), and R7050, a specific inhibitor of TNF-α (experimental design scheme in Fig. 4E, left) (55). Treatment of WM3682 melanoma cells, conditioned with adipocyte medium, with either cucurbitacin I or R7050 restored miR-211 expression (Fig. 4F). The inhibitors also abolished the effect of adipocytes on the melanoma phenotypic switch, as demonstrated by reductions in melanoma invasion capacity (Fig. 4G) and proliferation (Fig. 4H).

In immunofluorescence data of melanoma patient samples, IL-6 was detected in upper dermal regions only when adipocytes were present in that region; IL-6 was not detected in adipocyte-free areas (Fig. 4I). To clinically examine the relevance of IL-6, we performed Kaplan-Meier analysis on melanoma patient survival using data from the cBioPortal for Cancer Genomics (56). There was a notable increase in survival in the group of melanoma patients with no amplification of the mRNA encoding the IL-6 receptor (IL-6R) compared to those with amplification (Fig. 4J). These results suggest that TNF-α and IL-6 secreted by adipocytes induce a phenotypic switch in melanoma cells by inhibition of miR-211 expression.

Notably, treatment of melanoma with IL-6, TNF-α, or adipocyte-conditioned medium reduced the expression of MITF at the mRNA level (fig. S4, A and B) and substantially reduced the activity of a reporter gene driven by the MITF promoter (fig. S4C). Moreover, in the presence of IL-6 and TNF-α inhibitors, adipocytes failed to repress MITF expression in melanoma cells (fig. S4D). This suggests that adipocytes, through the secretion of IL-6 and/or TNF-α, cause a decrease in MITF abundance, which leads to the down-regulation of the miR-211 and disease progression.

miR-211 attenuates TGF-β signaling and reduces melanoma sensitivity to TGF-β

In a search for a cancer-related signaling pathway that might account for miR-211 actions, we identified the genes that were differentially expressed upon stable expression of the miR-211 mimic in WM1716 cells compared to cells expressing a scrambled control (Fig. 5A). Gene set enrichment analysis (GSEA) identified a single set of genes related to TGF-β signaling that are down-regulated in cells that express miR-211 (data file S2 and fig. S5A). Further, we separately calculated the mean log fold ratio of up- and down-regulated genes of 10 additional cancer-related pathways [inferred from Kyoto Encyclopedia of Genes and Genomes (KEGG) database] and found that the TGF-β–mediated signaling was the most significantly altered pathway in cells that express miR-211 mimic (Fig. 5A). Furthermore, of genes known to be involved in the TGF-β pathway, the down-regulated genes were a more prominent group than the up-regulated genes (Fig. 5A and data file S3). Melanoma cells with a proliferative gene expression signature were previously shown to be more susceptible to TGF-β–mediated growth inhibition than cells with an invasive signature (6), whereas MITF depletion was demonstrated to reduce the susceptibility of the proliferative cells to TGF-β–mediated growth inhibition (6, 57).

Fig. 5 miR-211 attenuates TGF-β signaling and reduces the sensitivity of melanoma cells to TGF-β.

(A) Cell signaling pathway using GSEA analysis based on transcriptome profiling of WM1716 melanoma cells stably expressing miR-211 or scrambled control. TGF-β pathway score, P < 0.0056; down-regulated TGF-β genes, P < 0.037; up-regulated TGF-β genes, P < 0.157. See also data files S2 and S3. (B) Luciferase activity assay of a TGF-β–responsive reporter in WM1716 and WM3314 cells stably expressing miR-211 or a scrambled control. Data are means ± SEM from three independent experiments. ***P < 0.005, t test. (C) Luciferase activity assay of TGF-β–responsive reporter in WM3682 cells cotransfected with antagomiR–miR-211 or a scrambled control and either siSMAD4 or scrambled siRNA. Data are means ± SEM from three independent experiments; statistical analyses as in (B). (D and E) Western blot analysis of phosphorylated and total SMAD2 abundance (D) and SMAD4 immunostaining (red; E) in the indicated cells. β-Actin served as loading control (D); DAPI counterstained the nuclei (blue; E). Scale bar, 20 μm. (F) qRT-PCR analysis of the expression of a panel of TGF-β signaling–related genes in WM3314 cells stably expressing miR-211 relative to those expressing a scrambled control. Data are means ± SEM from four independent experiments. *P < 0.05 and ***P < 0.005, t test. (G) Luciferase activity assay of TGF-β–responsive reporter in the indicated cells upon treatment with recombinant TGF-β (2 ng/ml) or dimethyl sulfoxide (DMSO) (ctrl). Data are means ± SEM from three independent experiments; statistical analyses as in (F). (H) XTT proliferation assay of cells described and treated as in (G). Data are mean fold change relative to day 0 ± SEM from three independent experiments; statistical analyses as in (F). (I) Venn diagram showing the overlap between the top 20% of down-regulated genes in WM1716 melanoma cells upon miR-211 introduction, the miR-211–predicted targets, and the genes identified as involved in TGF-β signaling (n = 2). (J) qRT-PCR analysis of TGFBR1 mRNA expression in the indicated cells. Data are means ± SEM from three independent experiments; statistical analyses as in (B). (K) Western blot analysis of TGFBR1 protein abundance in the indicated cells. β-Tubulin served as a loading control. Blots are representative of three independent experiments. (L) Predicted miR-211 target site identified in TGFBR1 3′ UTR (red). Wild-type (WT) and mutated (MUT) miR-211 binding site sequences. (M) Luciferase activity assay of WT or MUT TGFBR1 3′ UTR reporter constructs in WM3314 and WM1716 cells stably expressing miR-211 or scrambled control. Graphs and statistical analyses are as described in (G). (N) Luciferase activity assay of TGF-β–responsive reporter in WM3314 and WM1716 cells stably expressing miR-211 or scrambled control, which were transfected with either TGFBR1 complementary DNA (cDNA) that lacks the 3′ UTR or an empty vector (ctrl). Data are means ± SEM from three independent experiments. **P < 0.01 and ***P < 0.005, t test.

To confirm that miR-211 attenuates the canonical TGF-β pathway in melanoma, we first examined its effect on a TGF-β–responsive luciferase reporter gene. miR-211 overexpression in both WM1716 and WM3314 cells resulted in reduced TGF-β–responsive reporter activity compared to control cells (Fig. 5B). Conversely, antagomiR-mediated depletion of miR-211 from WM3526 cells led to increased TGF-β–responsive reporter activity (fig. S5B). Notably, the ability of miR-211 antagomiR to induce TGF-β activity was abolished in the presence of small interfering RNA (siRNA) targeting SMAD4 (Fig. 5C and fig. S5, C and D), indicating that TGF-β operates through SMAD4-mediated signaling. Further, miR-211–expressing cells exhibited markedly lower phosphorylation of the TGF-β signaling cytoplasmic transducer SMAD2 (58) than that observed in control cells (Fig. 5D). Moreover, SMAD4 nuclear translocation was abrogated upon miR-211 expression compared to control cells (Fig. 5E). miR-211 induced the expression of pro-proliferative genes (ID3, ID1, and c-MYC) and reduced levels of pro-invasive TGF-β–driven genes (SNAIL1, HMGA2, ZEB2, ZEB1, and TWIST) (Fig. 5F) (39). These results suggest that miR-211 represses the canonical TGF-β pathway, which, in turn, alters the cellular transcriptome in favor of a proliferative phenotype.

Next, we determined whether miR-211 confers TGF-β–resistance to melanoma cells. TGF-β stimulation failed to increase the TGF-β–responsive reporter activity in both WM1716 and WM3314 miR-211–expressing cells, in contrast to enhanced reporter activity observed in TGF-β–sensitive control cells (Fig. 5G). Both WM1716 and WM3314 cells lost their susceptibility to TGF-β–mediated growth inhibition upon miR-211 expression compared to control cells (Fig. 5H). As expected, the level of endogenous miR-211 inversely correlated with TGF-β susceptibility, as highly proliferative WM3526 cells, which express high levels of endogenous miR-211 (13), were less responsive to TGF-β (fig. S5E) than were WM1716 and WM3314 cells, which express low amounts of miR-211 (Fig. 5H) (13). The effect of miR-211 on melanoma sensitivity to TGF-β–mediated invasion could not be evaluated because of the high intrinsic invasive abilities of WM1716 and WM3314 cells (13).

To identify the cell type in the melanoma environment that is the source of TGF-β, we examined the amount of TGF-β in primary human skin cells that are the major components of the epidermis (differentiated and undifferentiated keratinocytes), the dermis (fibroblasts and endothelial cells), and the hypodermis (adipocytes). TGF-β was highly expressed in the dermis cells compared to cells that compose the epidermis or the hypodermis. This suggests that the dermis layer is the main source of TGF-β in the melanoma environment (fig. S5F).

Melanoma cells are resistant to the activation of TGF-β signaling (6, 59), and our findings suggest that miR-211 attenuates TGF-β–mediated signaling in melanoma cells and reduces their sensitivity to extrinsic TGF-β. To identify miR-211 downstream targets, we overlapped the top 20% of genes down-regulated upon miR-211 expression, potential miR-211 targets obtained from TargetScan prediction algorithms, and TGF-β signaling–related genes inferred from KEGG database (Fig. 5I and data file S4). TGFBR1 and SP1 were present in all three sets. The crucial role of TGFBR1 in transducing TGF-β signaling has been established (36). The transcription factor SP1 regulates multiple signaling and cellular pathways (60). TGFBR1 expression in WM3314 and WM1716 cells was reduced upon miR-211 expression at both mRNA (Fig. 5J) and protein levels (Fig. 5K). AntagomiR-211 introduction into WM3526 cells increased expression of TGFBR1 mRNA (fig. S5G) and TGFBR1 protein (fig. S5H). However, SP1 expression was not decreased by miR-211 (fig. S5I). The activity of a luciferase reporter gene fused to the TGFBR1 3′ untranslated region (3′ UTR) was decreased upon miR-211 expression in WM3314 and WM1716 cells, and specific mutations in the miR-211 binding sites resulted in derepression (Fig. 5, L and M). Last, in rescue experiments, overexpression of TGFBR1 in WM1716 and WM3314 cells stably transfected with miR-211 blocked the inhibitory effect of miR-211 on TGF-β–responsive reporter activity (Fig. 5N). Together, these findings suggest that miR-211 suppresses endogenous TGF-β signaling in melanoma cells and confers TGF-β resistance, at least, in part, by directly suppressing the expression of TGFBR1.

Adipocytes sensitize melanoma cells to TGF-β by repressing miR-211 expression

Adipocytes switch the melanoma cell phenotype from proliferative to invasive by down-regulating miR-211 expression. Given the parallels between miR-211 expression and TGF-β resistance in melanoma cells, we reasoned that adipocytes enhance melanoma cell sensitivity to environmental TGF-β. First, we investigated whether adipocytes could reversibly induce endogenous TGF-β signaling in melanoma cells. WM3682 melanoma cells had higher levels of SMAD2 phosphorylation upon coculture with adipocytes than did monocultured melanoma cells, and SMAD2 phosphorylation was diminished upon coculture reversal (Fig. 6A). Notably, culture of WM3682 cells with conditioned medium from an adipocyte culture resulted in increased expression of TGFBR1 (Fig. 6A). Moreover, cocultured WM3682 cells had higher expression of the pro-invasive TGF-β–driven genes (TWIST, ZEB2, MDM2, and HMGH2) (39) and lower expression of genes more characteristic of the proliferative phenotype (c-MYC, ID3, and ID1) (39) in comparison to monocultured cells (Fig. 6B). These effects were reversed upon adipocyte removal from the culture (Fig. 6B). The adipocyte-induced transcriptional signature of the invasive phenotype was the inverse of that of the miR-211–induced transcriptome (Fig. 5F).

Fig. 6 Adipocytes sensitize melanoma cells to TGF-β by repressing miR-211 expression.

(A) Western blot analysis of phospho-SMAD2 (pSMAD2) and SMAD2 (top) and TGFBR1 (bottom) protein expression in WM3682 melanoma cells in indicated conditions (see Fig. 2A for experimental scheme). β-Actin and β-tubulin served as loading controls. (B) Luciferase activity assay of TGF-β signaling–related genes in WM3682 cells cocultured or reverse cocultured with adipocytes relative to monocultured cells. Data are means ± SEM from three independent experiments; *P < 0.05 and ***P < 0.005. (C) Experimental schematic in (D) to (G). (D and E) Luciferase activity of a TGF-β–responsive reporter in WM3682 cells either transfected with a miR-211 mimic or scrambled control (D) or treated with TGF-β (2 ng/ml) or DMSO (ctrl) (E). (F and G) Invasion assay of WM3682 melanoma cell grown in adipocyte-conditioned medium, followed by conditioned medium removal (reverse) in the indicated conditions, upon treatment with TGF-β (2 ng/ml; F) or TGF-β inhibitor SB431542 (5 μM; G), each compared with those treated with DMSO. Images are representative; DAPI-stained nuclei are blue. Scale bars, 50 μm. Data in (D) to (G) are means ± SEM from three independent experiments. *P < 0.05, **P < 0.01, and ***P < 0.005, t test. (H) Proposed mechanism of adipocyte-mediated regulation of TGF-β signaling in melanoma cells. In proliferative melanoma (melanoma in situ), miR-211 represses the expression of both TGFBR1 (RI) and TGFBR2 (RII), suppressing endogenous TGF-β signaling in melanoma cells and conferring TGF-β resistance. As a result, cells are highly proliferative and weakly metastatic. When adipocytes are present in the melanoma microenvironment, IL-6 and TNF-α released from the adipocytes suppress miR-211 expression. The subsequent expression of TGF-β receptors enhances sensitivity to environmental TGF-β and therefore promotes the melanoma phenotypic switch from the proliferative to the highly invasive state.

Next, rescue experiments were performed to confirm that adipocytes increase TGF-β signaling in melanoma cells through miR-211 repression (experimental design scheme in Fig. 6C). Using a TGF-β–responsive reporter gene, we demonstrated that miR-211 mimic introduction into WM3682 melanoma cells treated with adipocyte-conditioned medium completely abolished the adipocyte-derived induction of TGF-β reporter activity (Fig. 6D).

Next, we aimed to determine whether adipocytes could reversibly sensitize melanoma cells to TGF-β. Melanoma cells treated with adipocyte-conditioned medium displayed higher TGF-β–responsive reporter activity upon treatment with external TGF-β compared to cells treated with vehicle control (Fig. 6E). No substantial change in TGF-β–responsive reporter activity was detected in cells incubated in unconditioned medium (Fig. 6E). The adipocyte-driven TGF-β sensitivity was abolished upon conditioned medium removal (Fig. 6E), indicative of reversibility. Further, melanoma cells grown in conditioned medium from adipocytes exhibited a reversible susceptibility to TGF-β–mediated invasion, whereas TGF-β resistance was observed in cells grown in unconditioned medium (Fig. 6F). Together, these results suggest that adipocytes reversibly evoke TGF-β sensitivity in melanoma cells.

We next asked whether adipocyte-induced melanoma phenotypic plasticity is TGF-β signaling dependent. To this end, we examined the influence of SB431542, a specific inhibitor of TGFBR1 (33), on adipocyte-mediated invasion of WM3682 melanoma cells. SB431542 treatment resulted in complete loss of adipocyte-driven invasion (Fig. 6G). These results indicate that adipocytes elicit their effects on melanoma cells in a TGF-β signaling–dependent manner.

DISCUSSION

Cancer research has traditionally focused on cancer cell mutagenesis per se or on its effect on tumor stroma (1). Recently, accumulating evidence, however, highlights the role of the reciprocal cross-talk between the microenvironment and the tumor in establishing the tumor phenotype (1, 14, 61). The cross-talk between melanocytes and their neighboring cells is a classic example of the continuous communication that occurs during normal and disease conditions. Melanoma is a neoplasm of melanocyte origin. Melanocytes produce the pigment melanin, which is stored in melanosomes (20). In normal skin, melanosomes are transferred to the epidermis in response to ultraviolet (UV) exposure providing protection against UV-induced DNA damage (62). Melanoma maintains melanosome production and trafficking functions for reasons that are mostly unknown. We found that melanoma cells communicate with the microenvironment by miRNA trafficking through melanosomes, prompting dermal metastatic niche generation (20). It will be interesting to learn whether there are additional cases of cancer cells that keep their seemingly unnecessary cell of origin functions and whether this might predispose characterization of tumor initiation.

Over the course of progression, the cancer encounters different microenvironments. In the case of cutaneous melanoma during the radial growth phase, proliferation begins from the basal epidermis and advances toward the upper epidermis, which is populated by differentiated keratinocytes (14). Subsequently, in the vertical growth phase, the melanoma invades the dermis, which consists of blood vessels and mesenchymal factors such as fibroblasts (63). It then spreads to the subcutaneous tissue, which is mainly composed of adipocytes (29). The intercommunication between melanoma cells and the neighboring cells is mostly based on direct cell contact, growth factor secretion, circulating extracellular miRNAs, and extracellular vesicle exchange, all leading to the activation of specific pro-cancer signaling pathways (14, 20). Here, we report that, in the presence of in situ melanoma, subcutaneous adipocytes translocate from their usual location in the hypodermis layer of the skin to the upper dermal layer. We also found that the translocated adipocytes’ proximity to the melanoma cells correlates with disease aggressiveness. It will be interesting to further explore de novo melanoma microenvironments that are characterized on the basis of not only molecular changes within the cells but also changes in their location. For example, is there an immune cell that appears specifically in close proximity to in situ melanoma? Because the identification of de novo cell location depends on immunostaining with specific antibodies as we did here, this presented a technical challenge for us, which we hope future studies will overcome.

The proliferative-to-invasive phenotypic switch model for melanoma progression proposes that tumor plasticity is driven by microenvironmental changes, rather than by the accumulation of genetic events (5, 6). Although several microenvironmental conditions induce this phenotypic switch, the cell type(s) in the melanoma microenvironment that induces this crucial phenomenon was hitherto not described. Here, we show that skin adipocytes serve as the cellular inducer of melanoma plasticity. They do so by secreting IL-6 and TNF-α, which decrease miR-211 expression in melanoma, thereby leading to the proliferative-to-invasive phenotypic switch. We further demonstrate that miR-211 inhibition enhances endogenous TGF-β signaling and sensitizes melanoma cells to environmental TGF-β. According to this phenotypic switch model (5, 6), tumor cells are at the invasive stage while in the blood (45). To establish metastases, tumor cells travel via the bloodstream to the destination organ and subsequently revert into the proliferative stage again (5, 6). It will be interesting to learn what the environmental cues in the distant metastatic niche that trigger this phenotypic switch are and whether it can be used to block the formation of metastases.

Our findings are also of relevance to epidemiological and experimental studies, suggesting that obesity increases the risk of developing subcutaneous melanoma (6468). However, it is still not fully understood whether local changes in the subcutaneous adipose layer contribute to these phenotypes or whether the phenotypes result from a systemic effect. It is possible that obesity alters the subcutaneous adipose tissue and changes the skin microenvironment (67). Obesity is known to be associated with a wide spectrum of dermatologic diseases, such as psoriasis, ulcerations, infections, poor wound healing, and insulin resistance syndrome (67, 69). Furthermore, obesity increases the expression of IL-6 and the IL-6R in subcutaneous adipocytes (70), and obese patients have high serum concentrations of IL-6 compared to patients who are not overweight (71). Increased serum IL-6 is also observed in patients with metastatic melanoma (72, 73). Although most of the studies that link melanoma development to obesity have focused on the systemic effects of visceral adiposity and insulin resistance mechanisms (74), there are substantial changes in the skin microenvironment due to obesity that may contribute to melanoma development. Therefore, obesity might increase the risk of melanoma development due to the alterations in the skin microenvironment.

MATERIALS AND METHODS

Cell culture

Primary human white subcutaneous pre-adipocytes (HWP; PromoCell) were cultured in pre-adipocyte growth medium (C-27410; PromoCell). At 80 to 90% confluency, differentiation was induced by culturing the cells in differentiation medium (C-27436; PromoCell) for 3 days, followed by culturing in nutrition medium (C-27438; PromoCell) that was renewed every 2 days for 6 to 8 days. NIH3T3-L1 cells were provided by A. Munitz (Department of Microbiology and Clinical Immunology, Sackler School of Medicine, Tel Aviv University). Cells were cultured in complete Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (Sigma-Aldrich) and 1% penicillin/streptomycin/glutamine (Invitrogen). Differentiation was induced at 80% confluency using 3T3-L1 differentiation kit (DIF001; Sigma-Aldrich) containing insulin (1.5 μg/ml), 500 μM 3-isobutyl-1-methylxanthine (IBMX), and 1 μM dexamethasone for 3 days followed by culturing in complete DMEM supplemented with insulin (1.5 μg/ml) (I0516; Sigma-Aldrich) for an additional 8 days. Medium was renewed every 2 days. Differentiation was validated at day 10 after induction by oil red O staining.

Primary normal human epidermal keratinocytes (NHEK; PromoCell) were cultured in growth medium (C-20011; PromoCell) and were induced to differentiate with high calcium (1.2 mM) for 5 days. Primary normal human dermal fibroblasts (NHDF; PromoCell) were cultured in DMEM. Primary human endothelial cells (HUVEC; PromoCell) were grown in endothelial cell growth medium (C-22011; PromoCell). WM3314, WM3526, WM3682, and WM1716 melanoma cell lines were provided by L. A. Garraway (Department of Medical Oncology and Center for Cancer Genome Discovery, Dana-Farber Cancer Institute, Boston, MA). Cells were cultured in complete DMEM. To establish stable cell lines, cells were selected with puromycin (1 μg/ml) (Sigma-Aldrich). Polymerase chain reaction (PCR) detection kit (Sigma-Aldrich) tested cells for mycoplasma. Characteristics of the melanoma lines are summarized in table S1.

Reagents

TGF-β1 (T7039; Sigma-Aldrich) was added to culture medium at a final concentration of 2 ng/ml 48 hours before analyses unless indicated differently. A TGF-β inhibitor, SB431542 (Sigma-Aldrich), was added at a final concentration of 5 μM at 24 hours before analysis. IL-6 (final concentration, 40 ng/ml; Abcam), TNF-α (final concentration, 15 ng/ml; Abcam), IL-6 inhibitor cucurbitacin I (final concentration, 0.1 μΜ; Tocris), and TNF-α inhibitor R7050 (final concentration, 0.5 μΜ; Tocris) were added 48 hours before analysis.

RNA purification and quantitative reverse transcription PCR

Total RNA was extracted using TRIzol (Invitrogen) according to the manufacturer’s instructions. RNA was quantified by measuring OD260/280. For quantitative reverse transcription PCR (qRT-PCR) analysis of mature miR-211, 20 ng of total RNA was subjected to a TaqMan mRNA assay (Applied Biosystems). Mature miR-211 levels were normalized to levels of RNU48. For mRNA analysis, cDNA prepared using the qScript cDNA Synthesis Kit (Quantabio) was subjected to qRT-PCR using the PerfeCTa SYBR Green FastMix (Quantabio). Data are presented as fold changes relative to control. Primer sequences and manufacturers are listed in table S2.

Oligonucleotide transfection

miR-211 mimic, antagomiR-211, scrambled control, and antagomiR control were transfected into melanoma cells using HiPerFect (Qiagen) according to the manufacturer’s instructions. Cells were transfected twice, at 24-hour intervals, with 100 pmol of oligonucleotide per 0.5 × 106 cells. Transfected cells were analyzed 48 hours after the first transfection. For data shown in Fig. 3 (K to M), transfections were conducted 3 days after incubation with conditioned medium. Sequences and manufacturers are listed in table S2.

Melanoma coculture with adipocytes

Coculture analysis was performed in 6- or 12-well Millicell Hanging cell culture inserts (1.0 μm of polyethylene terephthalate; MCRP06, MCRP12; Millipore). Pre-adipocytes were either seeded (coculture) or not seeded (monoculture; control) into hanging inserts and induced to differentiation for 9 to 11 days. At differentiation, WM3682 cells were seeded into the lower compartment of the transwell, and the cells were cocultured for 5 days in nutrition medium. For reverse coculture analysis, the hanging inserts, containing adipocytes, were removed after 5 days of coculture, medium was replaced to unconditioned medium, and cells were allowed to grow for an additional 5 days. Cells were trypsinized to maintain 70% confluency. For assays with adipocyte-conditioned medium, medium from culture of differentiated adipocytes was replaced 8 days after differentiation induction, and after an additional 2 to 3 days, supernatants were collected on ice. Cells and cellular debris were excluded from the medium by centrifugation at 12,000 rpm for 5 min.

Invasion assays

Melanoma cells (5 × 104) were seeded in duplicate in serum-free DMEM into an 8-μm pore Transwell membrane (Corning) coated with Matrigel (BD Biosciences). Invasion analysis was performed as previously described (13). Images of fixed cells from the invasion chambers were taken using an Olympus IX81 microscope and the cellSens Dimension software. In all assays, three fields per insert were photographed. The number of invaded cells was normalized to the number of total seeded cells.

Quantification of IL-6

IL-6 was quantified in adipocyte-conditioned medium using the Quantikine ELISA Kit (R&D Systems) according to the manufacturer’s instructions.

Immunofluorescence

Cells were seeded onto glass slides. After 24 hours, the cells were fixed with 4% paraformaldehyde for 10 min at room temperature, permeabilized with 0.1% Triton X-100, and incubated with 5% bovine serum albumin. The cells were stained with SMAD4 (1:50, D3R4N; Cell Signaling Technology) for 2 hours at room temperature followed by incubation with Alexa Fluor 488–conjugated secondary antibody (1:1000, Invitrogen) or with Ki67 Alexa Fluor 488–conjugated antibody (1:100, D3B5; Cell Signaling Technology) for 2 hours at room temperature. Nuclei were labeled with 4′,6-diamidino-2-phenylindole (DAPI; Vector Laboratories). Staining was analyzed using an Olympus IX81 fluorescence microscope and the cellSens Dimension software.

Plasmids and site-directed mutagenesis

For construction of miR-211 and scrambled control, expression vectors, miR-211, and scrambled sequences were cloned into modified PLKO.1 vector containing a cassette coding for luciferase, mCherry, and puromycin as previously described (13). The TGF-β–responsive luciferase construct and TGFBR1 expression plasmid were provided by Y. Henis (Neurobiology Department, The George S. Wise Faculty of Life Sciences, Tel Aviv University). pcDNA3-MITF was described in previous publications (75). The plasmid containing luciferase fused to the TGFBR1 3′ UTR was a gift from E. Galun (The Goldyne Savad Institute of Gene Therapy, Hadassah University Hospital, Jerusalem). Site-directed mutagenesis was performed on the miR-211 binding sites in the TGFBR1 3′ UTR using the QuikChange method (Stratagene) according to the supplier’s protocols. Primer sequences for site-directed mutagenesis are listed in table S2.

Transfection and luciferase reporter assay

For establishment of stable cell lines, WM3314 and WM1716 cells were transfected with PLKO.1–miR-211 or PLKO.1–scrambled control vectors using jetPEI kit (Polyplus-transfection) according to the manufacturer’s instructions. For luciferase assays, 1 μg of reporter plasmid (TGF-β–responsive reporter, TGFBR1 3′ UTR reporter, mutated TGFBR1 3′ UTR reporter, or MITF reporter) was cotransfected into cells with 40 ng of Renilla control plasmid using jetPEI (Polyplus-transfection). At 48 hours after transfection, luciferase assays were performed using the Dual Luciferase Kit (Promega) according to the manufacturer’s recommendations. For data shown in Fig. 5 (M and N), cells were transfected first with oligonucleotides, and then at 24 hours after transfection, the cells were transfected again with oligonucleotides and luciferase reporter plasmids. Luciferase assays were performed 24 hours after reporter plasmid transfection.

Western blot analyses

Equal amounts of extracted proteins (20 to 40 μg) were resolved on 10% SDS–polyacrylamide gel electrophoresis gels, and the proteins were transferred to nitrocellulose membranes (Whatman). The membranes were incubated with TGFBR1 (1:1000, 3712; Cell Signaling Technology), SMAD2 (1:1000, 3102; Cell Signaling Technology), pSMAD2 Ser465/467 (1:1000, 138D4; Cell Signaling Technology), SMAD4 (1:1000, 46535; Cell Signaling Technology), and β-actin (1:5000, 3700; Cell Signaling Technology) or β-tubulin (1:5000, 2146; Cell Signaling Technology) followed by incubation with appropriate horseradish peroxidase–conjugated antibodies at room temperature for 1 hour. Proteins were detected by an enhanced chemiluminescence solution (Thermo Fisher Scientific).

Wound scratch assay/migration assay

WM3682 melanoma cells were cocultured or treated with adipocyte-conditioned medium in six-well plates for 3 days. At confluency, the monolayer was wounded by scratching with a plastic pipette tip to create an about 450-μm-diameter strip. Images of wound closure were taken at 0, 19, 24, and 42 hours using an Olympus IX81 microscope and cellSens Dimension software and were analyzed by ImageJ software. Percentage of migration area was calculated by dividing the healed area by the wounded area. Five scratch areas per sample were analyzed.

Oil red O staining

Lipid droplets were stained as previously described (76). Cells were fixed with 4% paraformaldehyde at room temperature for 20 min, incubated with 60% isopropanol at room temperature for 5 min, and stained with 0.3% oil red O (O1391, Sigma-Aldrich) at room temperature for 20 min. After staining, cells were counterstained with Harris hematoxylin solution (HHS16; Sigma-Aldrich) for 45 s. Lipid droplets were photographed using an Olympus BX61 microscope.

Cell cycle analysis by flow cytometry

Cells were synchronized by serum starvation for 18 hours and then released for 10 hours. Cells (1 × 105 cells per sample) were fixed in cold 70% ethanol at 4°C overnight and permeabilized with 0.1% Triton X-100. To analyze DNA content, samples were treated with propidium iodide (0.05 mg/ml) (P4170; Sigma-Aldrich) and RNase A (ribonuclease A) solution (0.1 mg/ml) (R4642; Sigma-Aldrich) at room temperature for 30 min and were analyzed immediately by flow cytometry using a BD FACSCalibur flow cytometer. Data were gated by using Kaluza analysis software.

XTT cell proliferation assay

Aliquots of 2 × 103 WM3682 cells were seeded into 96-well plate in duplicate. Cell proliferation was determined every 24 hours for a period of 4 days, unless indicated differently, using the XTT Cell Proliferation Kit (Biological Industries) according to the manufacturer’s instructions. Absorbance intensity was normalized to samples at time 0.

Mice and histopathology analysis

Animal experiments were approved by the University of Tel Aviv Institutional Animal Care and Use Committee (M-11-053). WM3314 or WM1716 cells stably expressing the firefly luciferase reporter and miR-211 or scrambled control were mixed 1:1 with Matrigel (356231; BD Biosciences) and subcutaneously injected into 10-week-old NOD-SCID-IL2γ null mice (The Jackson Laboratory). Bioluminescence intensities of luciferase-expressing cells in mice were quantified at 29, 35, 41, 49, 56, 63, and 70 days after injection using an IVIS Spectrum system (Caliper Life Sciences, PerkinElmer). Mice were injected with 150 μl of d-luciferin (Promega) and then gas-anesthetized with isoflurane. A total of 20 mice were used. For ex vivo experiments, mice were sacrificed at 6 to 10 weeks after xenografting. Local xenografts, lungs, and livers were surgically removed and individually imaged. Regions of interest from displayed images were quantified as photons per second (p/s). Internal organs were fixed with 10% formalin and were paraffin-embedded followed by hematoxylin (HHS16, Sigma-Aldrich) and eosin (HT110232, Sigma-Aldrich) staining according to the manufacturer’s instruction.

Human samples and histopathology analysis

Human melanoma specimens were from the Sheba Medical Center (Helsinki ethical approval Smc-8333-10) and the Wolfson Medical Center (Helsinki ethical approval 0015-16-WOMC). Melanoma margins were determined by pathologists. The distance between the epidermis (stratum corneum) and the dermal adipocytes was measured using the ruler tool in the Aperio ImageScope software (unit: μm) at five random places throughout the tissue section, and these distances were averaged. Melanocytic nest numbers and sizes (in μm2) were calculated by analysis of Melan-A–stained patient samples within a 6-mm range. H&E staining was conducted as described in the previous section. Immunostaining was performed as previously described (14). In brief, slides were incubated, according to the manufacturer’s instructions, with FSP1 (ab27957, Abcam), perilipin1 (ab61682, Abcam), IL-6 (ab6672, Abcam), HMB45 (ab732, Abcam or Dako), MART-1/Melan-A (A103, BioSB), and S100 (ab52642, Abcam) antibodies, followed by incubating with the associated fluorophore-conjugated secondary antibodies: Alexa Fluor 488 (A11055, Invitrogen), Alexa Fluor 594 (8889, Cell Signaling), Alexa Fluor 594 (A21203, Invitrogen), or Alexa Fluor 647 (A31571, Invitrogen). Images were obtained at ×4, ×10, and ×20 magnification using fluorescence microscopy (Nikon).

mRNA profiling analysis

Total RNA from WM3682 melanoma cells that were monocultured or cocultured with adipocytes was extracted using TRIzol (Invitrogen). RNA quality was analyzed on Bioanalyzer, and RNA concentrations were determined with NanoDrop. Single-stranded cDNA was generated from the amplified complementary RNA (cRNA) with the WT cDNA Synthesis Kit (Affymetrix) and then fragmented and labeled with the WT Terminal Labeling Kit (Affymetrix). Samples were hybridized with Clarioum S Human Arrays (Affymetrix) and scanned at the Hebrew University Microarray Core Facility. Array scanning was performed according to the manufacturer’s instructions. Raw data were processed using Transcriptome Analysis Console (Thermo Fisher Scientific).

Proliferation and invasion gene signature analysis

Log fold ratio of the two genes sets (7) was used after excluding all genes with log fold change of less than 25%. P values based on two-sided Wilcoxon rank sum test that the two distributions are different were calculated using R version 3.5.1.

Signaling pathway analysis for WM1716 P value calculations

The probability that a given pathway with certain mean up- and down-regulated genes would be identified by chance was calculated using simulations as follows: The up- or down-regulation was represented as a mean log fold ratio and was calculated separately for each pathway after noise filtering (genes with fold change of less than 0.25 were not considered; other thresholds resulted in the same trend). In cases where multiple probes existed for a gene, their average were taken. P values were determined after 5000 repeated calculations of the same measurements of mean log fold ratios for each pathway after shuffling all genes in the pathways in the analysis while keeping pathway size and gene expression distributions intact. The fraction of times the mean log fold change of the down-regulated genes exceeded that of the down-regulated genes in a pathway was considered the P value of down-regulation. P values for up-regulated genes were calculated in the same manner, and the P values for both pathways were the joint probability. KEGG pathway genes were downloaded using the package “EnrichmentBrowser” version 2.10.6 (77).

Gene expression analysis

CEL files (annotation: pd.hugene.1.1.st.v1) for data from WM1716 cells were processed in R version 3.5.1 using the package “oligo” version 1.44.0 (78) with default parameters. Differential expression was calculated by the log (base 2) fold ratio of control to transfected samples after the exclusion of genes with expression of less than 6 (log 2 base).

Gene set enrichment analysis

GSEA for WM1716 cell data was done using GSEA v3.0 (79) from the Broad Institute using c2.cp.kegg.v6.2.symbols gene sets, permutation type “gene set,” and all the other parameters as default. False discovery rate of 25% was considered for significance.

Kaplan-Meier analysis

The information analyzed was taken from the Cancer Genome Atlas database, which contains data on 550 patients with melanoma. These data were generated by the TCGA Research Network (http://cancergenome.nih.gov/). The Kaplan-Meier graph was generated by using the “survival” tab of cBioPortal for Cancer Genomics (56). Patients were divided into two groups: with and without amplifications of IL6R mRNA.

SUPPLEMENTARY MATERIALS

stke.sciencemag.org/cgi/content/full/12/591/eaav6847/DC1

Fig. S1. Subcutaneous adipocytes are observed in proximity to in situ melanoma.

Fig. S2. Subcutaneous adipocytes approximate to in situ melanoma correlate with advanced disease.

Fig. S3. Adipocytes drive melanoma plasticity in a reversible manner by miR-211 repression.

Fig. S4. Adipocytes decrease miR-211 expression through secretion of IL-6 and TNF-α.

Fig. S5. miR-211 attenuates TGF-β signaling and reduces melanoma sensitivity to TGF-β.

Table S1. Characteristics of melanoma cell lines.

Table S2. Sequence data for oligonucleotides.

Data file S1. Differentially expressed miRNAs in melanoma upon coculture with adipocytes.

Data file S2. Gene set enrichment upon miR-211 expression in melanoma.

Data file S3. Pathway enrichment upon miR-211 expression in melanoma.

Data file S4. Venn diagram data.

REFERENCES AND NOTES