Interactions with Muscle Cells Boost Fusion, Stemness, and Drug Resistance of Prostate Cancer Cells

This article is featured in Highlights of This Issue, p. 667

Interactions between cancer cells and nonmalignant cells within the tumor microenvironment play a critical role in cancer progression from primary tumor to metastasis, which is responsible for most of the cancer-related deaths (1, 2). These interactions direct dramatic changes in the morphology and properties of the primary tumor cells, including the transdifferentiation of epithelial cell–originated prostate cancer cells, into more invasive mesenchymal cells (the epithelial–mesenchymal transition, EMT). Cells can drastically change their properties by cell-to-cell fusion. Cells fuse in fertilization, in placentogenesis, in skeletal muscle formation, and in osteoclastogenesis (3). The molecular mechanisms underlying these fusion processes are poorly understood. Several proteins, including annexin A5 (AnxA5) and syncytin 1 (Syn1), have been implicated in diverse fusion processes. AnxA5 is involved in myoblast fusion (4), osteoclast fusion (5), and in placental trophoblast fusion (6). Syn1, a captive retroviral envelope protein that acquires a functional conformation after interactions with the ubiquitous receptor ASCT2, has been implicated in placental trophoblast fusion (7) and osteoclast fusion (5, 8). It has been suggested that cell–cell fusion, possibly involving Syn1, plays an important role in cancer (9–13). However, the very low efficiency of cancer cell fusion in many experimental models has hindered elucidation of the role of cell fusion in cancer initiation and progression. The specific role and place of cell fusion in the dynamic and complex network of interactions between tumor cells and nonmalignant cells, the nature of fusion-triggering events, and proteins involved remains to be elucidated.

In this study, we focused on prostate cancer, the most frequently diagnosed cancer in American men in 2017. The prostate gland is surrounded by smooth muscle cells of the prostate stroma and, on the apical and anterior surfaces, blends with the skeletal myofibers of the rhabdosphincter. To explore the interactions between prostate cancer cells and muscle cells, we cocultured PC3 cells (human prostate cancer cell line) and primary human prostate cancer cells with primary smooth or skeletal muscle cells. The coculturing of cancer cells with different muscle cells caused the expansion of subpopulations of cancer cells with cancer stem-like cells (CSLC) characteristics, including anchorage-independent growth, CD133 expression, and drug resistance. We found that these changes in the properties of cancer cells required coculturing-induced increases in medium concentrations of IL4 and IL13, cytokines that had been linked to both prostate cancer progression (14, 15) and fusion of muscle cells (16–18). In turn, IL4 and IL13 upregulated the expression of Syn1 and AnxA5 and induced fusion of cancer cells that was dependent on these proteins. Cancer cell fusion is required for the muscle cell–dependent changes in the properties of cancer cells. These data and our finding that levels of expression of Syn1 and AnxA5 are higher in human prostate cancer tissues than in normal and benign prostatic hyperplasia tissues suggest that in the tumor microenvironment, interactions with muscle cells increase the metastatic potential and drug resistance of the cancer cells by promoting cancer cell fusion. Thus, our work identifies a direct link between Syn1 and AnxA5 upregulation, cell fusion, and prostate cancer progression.

PC3 cells (ATCC), primary human myoblasts (hMYO), and primary prostate smooth muscle cells (pSMC), both obtained from Lonza, were maintained according to the manufacturer's instructions. Primary murine myoblasts were isolated and grown as described in (4). Primary human prostate cancer cells (pPCa) developed from a human prostate adenocarcinoma sample [Gleason 8 (4+4), T2N0M0] were obtained from Celther and cultured according to the manufacturer's instructions.

Cancer and muscle cells were cocultured in a 1:4 ratio in the same medium either separately or in a direct contact. To measure RNA or protein expression levels in cancer cells or in muscle cells, to examine clonogenic potential under anchorage-independent conditions, and to examine drug resistance, we cocultured the cells without direct physical contact using 0.4-μm cell culture inserts (Millipore, catalog no. PICM03050). To examine cell fusion, cell surface CD133 expression, and CD133 promoter–driven GFP expression in cancer cells, we cocultured cancer cells and muscle cells in direct contact. In all coculturing experiments, including the controls with cancer cells without muscle cells, the cells were grown in the medium that was a 1:1 mix of the culture media conventionally used for cancer cells and for muscle cells. We observed no obvious changes in the properties of PC3 cells after growing them in the medium that we used to culture muscle cells.

To quantify fusion by content mixing and multinucleation assays, we labeled cancer cells with either CellTracker Green CMFDA dye or CellTrace Far Red dye. Labeled cancer cells were coplated to unlabeled muscle cells that had been seeded 24 hours earlier. At the end of the 72-hour coincubation of cancer cells and muscle cells, we labeled cell nuclei with Hoechst-33342 and imaged the cells on Zeiss LSM 510 laser confocal microscope. Fusion between cells labeled with different probes mixes their cytosols and generates double-labeled cells. Images of the cells in both fluorescent channels were carefully examined and all cells that showed cell-associated fluorescence in both channels were counted as double-labeled and “content-mixed.”

We quantified multinucleation (=generation of multinucleated cells) by normalizing the total number of nuclei in multinucleated cancer cells (defined as cells with ≥2 nuclei) to the total number of nuclei in the field. Note that in this study, we are interested only in relative efficiencies of fusion under different conditions. Culturing of the cells for 72 hours complicates unambiguous quantitative analysis of the number of fusion events based on the appearance of double-labeled cells, and reliable measurements of the fusion efficiency will likely require live imaging of the cells.

To examine the role of nuclear division without cytokinesis in generation of multinucleated cells, we suppressed DNA replication by application of DNA methylation inhibitor FdUrd (fluorodeoxyuridine) to the cocultured primary prostate cancer cells and hMYOs for the last 24 hours of the 72-hour coincubation at the time of the most robust development of multinucleation.

To examine functional role of the AnxA5 in cancer cell fusion and multinucleation, we silenced AnxA5 gene expression in only PC3 cells or in both PC3 and hMYO cells with siRNA against AnxA5. PC3 cells transfected with either AnxA5 siRNA or nontargeting siRNA and 24 hours later overlaid with hMYO cells for 72 hours. To silence AnxA5 in both PC3 and hMYO cells, the cells were first overlaid and 24 hours later transfected with either AnxA5 siRNA or negative control siRNA and cocultured for additional 72 hours.

We stained cancer cells and muscle cells with CellTracker Green CMFDA and CellTrace Far Red dyes, respectively, and cocultured the cells for 72 hours. Then, we labeled nuclei with Hoechst-33342 and took images on a Zeiss LSM 510 laser confocal microscope. To evaluate the percentage of cancer cells that were involved in fusion to muscle cells (identified as double-labeled), we normalized the number of cancer cell nuclei in the cells formed by fusion between cancer cells and muscle cells to the total number of cancer cell nuclei.

Muscle cells were grown in collagen-coated 6-well plates. Cancer cells were grown on hydrophilic polytetrafluoroethylene PTFE cell culture inserts with a pore size of 0.4 μm (Millipore, catalog no. PICM03050) placed in the same plate with muscle cells. At the end of the 72-hour coculturing, RNA was extracted and analyzed.

To analyze CD133 gene expression, we transiently transfected PC3 cells with human CD133 promoter–driven GFP reporter construct obtained from Applied Biological Materials Inc. (catalog no. MT-h68878) using Trans-IT transfection reagent (Mirus, catalog no. MIR5400); 24 hours later, PC3 cells were overlaid with either hMYO or pSMC cells for 72 hours. The percentage of GFP-positive cells was determined from confocal microscopy images. The efficiency of the transfection was consistent from experiment to experiment at 80%–85% of the cells, as estimated using the GFP empty vector.

Cells (1 × 10⁶) were plated in 100-mm dishes and cultured for 2 days before being stained with CD133 antibodies and analyzed by flow cytometry. To detect cell-surface CD133 expression in cancer cells cocultured with muscle cells, PC3 cells stably expressing lentiviral GFP were cocultured with muscle cells, and FACS analysis of the cell-surface CD133-PE was limited to the GFP-expressing PC3 cells.

To examine the effects of Syn1f expression in PC3 cells on cell surface CD133 expression, we transiently transfected PC3 cells with Syn1f; 24 hours later, we labeled PC3 cells overexpressing Syn1 with CellTrace Far Red dye and overlaid them with nontransfected PC3 cells labeled with CellTracker Green CMFDA dye. After a 72-hour coculturing, we stained the cells with CD133-PE antibody and analyzed CD133-PE expression in red-labeled, green-labeled, and double-labeled (fused = yellow) PC3 cells.

Cancer cells were grown on their own, cocultured with different muscle cells, or treated with either IL4 or IL13 cytokines for 72 hours. The cells were lifted and seeded as a single-cell suspension onto an ultra-low attachment surface (Corning) in growth factor–enriched medium (see Supplementary Materials for details). After 7–10 days of culturing, we counted the numbers of colonies (tumorspheres) operationally defined as dense, expanding, but not necessarily spherical, cell colonies that were stable after gentle media agitation, had size above 40 μm, and were present 7–10 days after plating.

Paraffin-embedded surgical resection specimens of prostate cancer, as well as normal and benign prostate tissues were obtained from the BioBank at Maine Medical Center Research Institute and were generated by the Histopathology Core at Maine Medical Center Research Institute. We have also used tissue microarrays generated by Cooperative Human Tissue Network (CHTN), Virginia University. The microarrays representing the stages of tumor progression in prostate adenocarcinoma were provided with tumor Gleason scores assigned by pathologists. Syn1 and AnxA5 were detected by immunostaining of paraffin-embedded prostate tissues with Syn1 antibody (catalog no. NB-100-93579, Novus Biologicals) and AnxA5 antibody (catalog no. ab54775, Abcam). To quantify expression level of Syn1 and AnxA5, immunofluorescence staining was performed with 12 normal tissue specimens, 11 benign prostatic hyperplasia (BPH) specimens, 10 prostatic intraepithelial neoplasia (PIN) specimens, 11 cancer tissue specimens classified by pathologists as Gleason grade 5–6, 11 tissue specimens of Gleason grade 7, and 12 tissue specimens of Gleason grade 8–10.

We analyzed the data in the GraphPad Prism 6 graphing and analysis program using the Student paired t test. A P value < 0.05 was used to define statistically significant differences (*, P < 0.05; **, P < 0.001).

Cell culturing, reagents, plasmid, virus production and transfection, RNA isolation, qPCR, Western blot, flow cytometry, cell viability, gene silencing, cytokine expression, and IHC are described in detail in Supplementary materials.

To study the dependence of cancer cells properties on the muscle cells in tumor microenvironment, we modeled the interphase between these cells using coculturing experiments. We used highly malignant PC3 cells that have epithelial-like morphology and are widely employed to study EMT, drug resistance, and CSLC (19, 20), and primary prostate cancer cells developed from a human prostate adenocarcinoma. In contrast to PC3 cells, primary prostate cancer cells express PTEN, p53, and IL4 (Supplementary Fig. S1). Like PC3 cells, primary prostate cancer cells are deficient in androgen receptor expression.

We incubated PC3 cells for 72 hours on transwell cell culture inserts placed into the cell culture wells with either primary human skeletal muscle cells (hMYO), primary prostate smooth muscle cells (pSMC), or primary murine myoblasts (mMYO). After the coculturing, we evaluated the clonogenic potential of the PC3 cells by plating them onto an ultra-low attachment surface in a growth factor–enriched serum-free medium. The coculturing with the muscle cells increased a number of dense, three dimensionally expanding, but not necessarily spherical, larger than 40-μm colonies (“tumorspheres”; ref. 21) of PC3 cells formed under these conditions (Fig. 1A and B). Similarly, coculturing with mMYO increased the number of primary prostate cancer–formed tumorspheres (Supplementary Fig. S2).

Non-adherent anchorage-independent growth is a hallmark of the self-renewing drug-resistant prostate cancer cells, referred to as CSLC (22). A small population of these cells is hypothesized to play a key role in heterogeneity, growth, recurrence, and metastasis of tumors (22). Analysis of the changes in CSLC population requires identification of these cells using stem cell markers. Because most of PC3 cells express CD44 (23), we identified CSLC as cells with elevated level of expression of CD133, a marker of CSLC from a variety of cancer cell lines and tumors, including prostate cancer (24). When we transiently transfected PC3 cells with a human CD133 promoter–driven GFP reporter lentiviral construct (25), approximately 25% of PC3 cells showed GFP expression and, thus, had active CD133 gene promoter (Fig. 1C). In agreement with earlier studies (26), fraction of PC3 cells expressing CD133 at their surface (CD133⁺ cells), detected by antibody staining was small (∼ 1%, Fig. 1D; Supplementary Fig. S3A). Thus, many cells showing CD133 gene promoter activity displayed no detectable CD133 at their surface. This discrepancy likely reflects the fact that cell surface expression of CD133 depends not only on transcriptional activity of the CD133 gene, but also on posttranscriptional regulation. Moreover, for the same cells, apparent protein expression varies widely depending on the antibody used (27). Both aforementioned assays and direct qPCR analysis of the endogenous CD133 transcript at the level of the cell population indicate that coculturing of PC3 cells with different muscle cells upregulates CD133 expression (Fig. 1C–E). Coculturing of primary prostate cancer cells with hMYOs or pSMCs also upregulated CD133 expression at the RNA level (Supplementary Fig. S4A). Promotion of anchorage-independent growth of cancer cells and upregulation of CD133 expression for cancer cells cocultured with muscle cells points to an expansion of the CSLC population. CSLCs have a higher resistance to chemotherapy treatments than main population of tumor cells (22). We examined whether interactions with muscle cells influence drug resistance of PC3 cells. We grew these cells on the inserts placed into the cell culture wells. hMYOs or pSMCs were grown on the bottom of the same wells. After 72-hour coculturing, we lifted PC3 cells from the inserts, and incubated them or PC3 cells grown on their own with either doxorubicin or cisplatin for 24 hours. We then analyzed the viability of PC3 cells by evaluating their metabolic activity using the MTT assay (Fig. 1F and G) and by counting live cells (Supplementary Fig. S5A–S5C) and found that PC3-muscle cell coculture moderately protected PC3 cells against the drugs. In similar experiments with primary prostate cancer cells, which, on their own, were more sensitive to doxorubicin treatment than PC3 cells, the effects of the coculturing with muscle cells were even stronger (Supplementary Fig. S5D–S5F).

Expansion of CSLC subpopulations, drug resistance, and elevated metastatic potential of cancer cells are often associated with EMT (28). We found that both PC3 and primary prostate cancer cells cocultured with pSMCs or hMYOs had moderately higher levels of mesenchymal cell-characteristic proteins, including elevated levels of AKT and its phosphorylated form (pAKT), MMP9, vimentin, and Slug, while expression of epithelial cell–characteristic E-cadherin was decreased (ref. 29; Fig. 1H and I; Supplementary Fig. S5K and S5L). These changes in protein expression suggest that muscle cells promote EMT in prostate cancer cells. Because the data on CD133 expression, drug resistance and clonogenic potential suggest that only a fraction of cancer cells change their properties in the presence of the muscle cells, the analysis of the expression of different proteins in cell lysates likely underestimates the differences in the protein expression specific for the affected cells.

In summary, interactions with muscle cells expanded subpopulations of cancer cells demonstrating characteristic features of CSLC including anchorage-independent growth, elevated CD133 expression, and drug resistance. Because these changes in the properties of cancer cells did not require direct contact between muscle and cancer cells, we focused our attention on the extracellular factors.

PC3 cells grown in the conditioned medium collected from PC3 cells cocultured for 72 hours with different muscle cells demonstrated a moderate increase in resistance to doxorubicin (Fig. 2A) and in the clonogenic potential (Supplementary Fig. S6A) compared with PC3 cells cultured in a fresh medium. Culturing primary prostate cancer cells in the conditioned media collected from either primary prostate cancer/pSMCs or primary prostate cancer/hMYO cocultures also increased primary prostate cancer cells’ resistance to both doxorubicin and cisplatin (Supplementary Fig. S5G–S5J). Culturing PC3 cells in the conditioned media from different muscle cells grown on their own did not change the clonogenic potential, drug resistance, and the numbers of CD133 gene–expressing PC3 cells (Supplementary Fig. S6B–S6D). We also found no appreciable changes in the properties of PC3 cells grown on their own in the same medium for longer than 72 hours.

To identify coculture-dependent factors in the conditioned medium and to examine the mechanisms by which muscle cells influence the properties of cancer cells, we focused on two related cytokines IL4/IL13 that share receptors and have been suggested to promote progression of prostate cancer (14, 15). While we did not detect IL4 in the conditioned media of either primary prostate cancer, pSMC, or hMYO cells cultured on their own, the media collected after coincubation of primary prostate cancer cells with either hMYOs or pSMCs contained significant amounts of IL4 (Fig. 2B). In contrast, concentration of IL13 in the medium of the cocultured primary prostate cancer cells and hMYOs was as low as in the medium from primary prostate cancer cultured on their own (Fig. 2C). The medium collected from primary prostate cancer cells/pSMCs coculturing had IL13 concentration that was like that for pSMCs cultured on their own and much higher than the concentration of IL13 in the medium from primary prostate cancer cells cultured alone (Fig. 2C). Thus, coculturing with pSMCs exposed primary prostate cancer cells to higher IL13 concentrations than that characteristic for primary prostate cancer cells cultured on their own. The conditioned media from the cocultures of PC3 cells with hMYOs or pSMCs also contained higher amounts of IL13 than the medium collected from PC3 cells cultured on their own (Fig. 2C). The concentrations of IL4 in the media from PC3 cells either cultured on their own, or cocultured with hMYOs or pSMCs were too low for detection (Fig. 2B).

To individually evaluate the expression of the cytokines in cancer and muscle cells, we separately lysed cancer cells grown on cell culture inserts and muscle cells grown on the bottom of the wells and analyzed the level of IL4 and IL13 RNAs in lysates of the cells of each type. Expression of both cytokines was strongly increased in PC3 cells cocultured with hMYOs (Fig. 2D). In hMYOs, coculturing with PC3 cells boosted IL13 expression but not IL4 expression (Fig. 2E). The experiments, in which we cocultured PC3 cells with mMYOs and then analyzed the conditioned medium with a mouse ELISA test kit (Supplementary Fig. S7), showed upregulation of the expression of both IL4 and IL13 in muscle cells. Thus, elevated amounts of IL4 and IL13 in the coculture media, dependent on the cytokine and on the cells, may be generated by either cancer or muscle cells or both.

Neutralizing antibodies to IL4 and IL13 suppressed the muscle cell–induced promotion of the anchorage-independent colony formation (Fig. 2F–H) and blocked an increase in the number of cell surface CD133⁺ PC3 cells (Fig. 2I; Supplementary Fig. S3B). These data suggested that muscle cell effects on cancer cells depend on the increased concentrations of IL4 and IL13 in the coculturing medium.

We then examined whether application of recombinant IL4 or IL13 to cancer cells can substitute for the coculturing with muscle cells. Indeed, PC3 cells cultured in the medium supplemented with IL4 or IL13 demonstrated more efficient anchorage-independent growth (Fig. 3A and B), elevated expression of CD133 gene (Fig. 3C), expanded subpopulations of PC3 cells expressing the CD133 gene (Fig. 3D and E), and CD133⁺ cells identified by cell surface labeling with the CD133 antibody (Fig. 3F; Supplementary Fig. S3C). To test whether application of IL4 or IL13 increases drug resistance of cancer cells, we incubated PC3 cells with each cytokine for 72 hours and found that the cytokine-treated cells better withstood the subsequent 24-hour-long application of doxorubicin or cisplatin (Fig. 3G and H). Like PC3 cells cocultured with muscle cells, PC3 cells treated with either IL4 or IL13 demonstrated moderately elevated levels of mesenchymal cell–characteristic proteins AKT and pAKT, MMP9, Slug (for IL4-treated cells) and vimentin and somewhat lowered expression of E-cadherin (Fig. 3I and J).

In brief, our results suggest that increased concentrations of cytokines IL4 and IL13 play a key role in muscle cell–dependent promotion of the stemness and drug resistance of cancer cells. The effects of IL4 and IL13 can be modulated by other cytokines upregulated by the interactions between cancer and muscle cells.

Because cells can change their properties by fusion, we tested whether interactions with muscle cells resulted in fusion of prostate cancer cells. We coplated PC3 cells prelabeled with green cell tracker with PC3 cells prelabeled with red cell tracker and with either pSMCs or hMYOs. After 72-hour coincubation, we observed some double-labeled PC3 cells (Fig. 4A) that were not observed when we cocultured differently labeled PC3 cells in the absence of muscle cells (Fig. 4B and C). Exchange of cytosolic markers between the cells (hereafter referred to as content mixing) suggested that coincubation of PC3 cells with either pSMCs or hMYOs promotes fusion between PC3 cells. The finding that coincubation of PC3 and muscle cells also increased the numbers of multinucleated PC3 cells (defined as cells with two or more nuclei) (Fig. 4B and C) and the time-lapse microscopy (Supplementary Movie S1) further substantiated this interpretation. Similar muscle cell–induced promotion of content mixing and multinucleation was observed for primary prostate cancer cells (Fig. 4D and E; Supplementary Fig. S8A).

Multinucleation can develop not only from cell fusion but also from nuclear division without cytokinesis. The latter mechanism is expected to be inhibited by blocking DNA replication with fluorodeoxyuridine (FdUrd). While, as expected, FdUrd application to primary prostate cancer/hMYO cocultures suppressed cell proliferation (Supplementary Fig. S9), this reagent did not suppress the formation of multinucleated primary prostate cancer (Fig. 4D), indicating that formation of these cells involves cell fusion. Unexpectedly, we found FdUrd to increase the numbers of double-labeled cells generated by coincubation of primary prostate cancer cells with hMYOs (Fig. 4E). We speculate that suppressing cell division lowers the numbers of contacts within cell division–generated clusters of same-colored cells and, thus, increases the frequency of green cell/red cell contacts which, upon fusion, yield double-labeled cells.

In addition to homotypic fusion between cancer cells, coculturing with human muscle cells also induced heterotypic fusion between cancer cells and muscle cells, observed as the appearance of double-labeled cells after a 72-hour coincubation of green cell tracker–labeled primary prostate cancer cells with redcell tracker–labeled pSMCs or hMYOs (Fig. 4F and G). Heterotypic fusion was also observed for PC3 cells and either hMYOs or pSMCs (Fig. 4H; Supplementary Fig. S8B). Interestingly, in contrast to hMYO, murine myoblasts (mMYO) did not fuse with PC3 cells (Fig. 4H).

To summarize, coincubation of cancer cells with skeletal or smooth muscle cells promoted fusion of cancer cells, and heterotypic fusion between cancer cells and human muscle cells.

We found that coculturing of cancer and muscle cells boosts the expression levels of Syn1 and AnxA5, proteins associated with several fusion processes (4, 6, 7, 30). The coculturing with hMYOs increased levels of Syn1 and AnxA5 gene expression (Fig. 5A) in the PC3 lysates. Western blot analysis has shown a corresponding increase in the expression of these proteins, compared with PC3 cells cultured on their own (Fig. 5B and C). Primary prostate cancer cells cocultured with either pSMCs or hMYOs also showed elevated protein expression except for AnxA5 in primary prostate cancer cocultured with hMYO (Supplementary Fig. S10).

Both AnxA5 and Syn1 are already implicated in cancer biology (30, 31), and we focused on involvement of these proteins in the muscle cell-induced fusion and multinucleation of cancer cells. Anx A5 deficiency is known to suppress myoblast fusion (4). We found the siRNA suppression of human AnxA5 in only PC3 cells or in both PC3 cells and hMYOs to inhibit content mixing and multinucleation of PC3 cells observed after PC3/hMYO coculturing (Fig. 5D–F). To further explore the effects of AnxA5, we verified that mMYOs isolated from wild-type mice, promote expression of Syn1 and AnxA5 and multinucleation in PC3 cells (Fig. 5G and H). In contrast, mMYO isolated from AnxA5-deficient mice did not promote multinucleation of PC3 cells (Fig. 5H). These data indicated that muscle cell–induced fusion between PC3 cells depends on AnxA5 produced both in cancer cells and in muscle cells.

To examine the contributions of Syn1, we used a Syn1-derived synthetic peptide that blocks fusogenic restructuring of Syn1 and Syn1-mediated cell fusion (32). We found that this peptide, but not its scrambled version, lowered the numbers of multinucleated PC3 cells observed in PC3/mMYO cocultures (Fig. 5I), suggesting that PC3 fusion depends on Syn1. Because murine cells do not have Syn1 gene, Syn1 must be expressed in PC3 cells. The importance of Syn1 in muscle cell–induced fusion between cancer cells was further substantiated by the experiments with PC3 cells stably expressing either nontargeting shRNA-GFP or Syn1 shRNA-GFP (PC3-NT and PC3-Syn1 shRNA cells, respectively). As expected, in the latter case, Syn1 expression was strongly downregulated (Fig. 5J). We cocultured PC3-Syn1 shRNA or PC3-NT shRNA cells (bars 3 and 2) with hMYO cells for 72 hours and found that suppression of Syn1 expression in PC3 cells inhibited both generation of multinucleated PC3 cells (Fig. 5J) and heterotypic fusion between PC3 cells and hMYO (Fig. 5K). Murine cells do not carry receptors for Syn1 and do not support Syn1-mediated fusion (33), thus providing additional support for the conclusion that both homotypic fusion between cancer cells and heterotypic fusion of these cells with muscle cells involve Syn1 and explaining the lack of fusion between mMYOs and PC3 cells (Fig. 4H).

Together, our data suggest that Syn1 and AnxA5 upregulated in cancer and muscle cell cocultures are required for cancer cell fusion.

Application of recombinant IL4 and/or IL13 promoted fusion of PC3 cells (Fig. 6A and B) and primary prostate cancer cells (Fig. 6C and D; Supplementary Fig. S11) and expression of Syn1 and AnxA5 in PC3 cells (Fig. 6E and F). Moreover, neutralizing antibodies to IL4 and IL13 inhibited fusion of cancer cells (Fig. 6G).

We hypothesized that muscle cell–induced increases in stemness and drug resistance of cancer cells depend on fusion between these cells. PC3 cells that were neither incubated with muscle cells nor exposed to IL4/IL13 were transfected to express exogenous Syn1 with a fusion-enhancing deletion in the cytoplasmic domain (Syn1f; ref. 34). Syn1 overexpression in the transfected PC3 cells was confirmed by Western blots and qPCR (Fig. 7A). The transfected PC3 cells labeled with red cell tracker were coplated, 24 hours after transfection, with nontransfected PC3 cells labeled with green cell tracker. As expected, after a 48-hour incubation of the cells, we observed many double-labeled (“yellow”) and multinucleated PC3 cells (Fig. 7B). We found that PC3 cell fusion induced by exogenous expression of Syn1f increased the clonogenic potential of cancer cells under anchorage-independent conditions (Fig. 7C). Syn1f-expressing PC3 cells demonstrated dramatically increased fractions of cells expressing the CD133 gene (Fig. 7D) and CD133 protein at the cell surface (Fig. 7E). Using content mixing assay, we separated our PC3 cells into three subpopulations: yellow cells that contain only fused cells, red cells that are a mixture of fused and unfused Syn1f-transfected cells, and green cells that are neither transfected nor fused. Fraction of CD133⁺ cells positively correlated with the presence of fused cells (Fig. 7E) increasing from 1% for green cells to 20% for yellow cells. In addition, a higher percentage of CD133⁺ cells observed for yellow cells than for red cells indicated that the promotion in the CD133 expression depends on Syn1f-mediated fusion rather than just on expression of Syn1f. These findings suggested that a major increase in the stemness of PC3 cells depends on Syn1-mediated fusion. We also found that PC3 cell fusion induced by exogenous expression of Syn1 raised the AKT, pAKT, MMP-9, and vimentin content and lowered E-cadherin content, both characteristic of the EMT (Fig. 7F).

To test involvement of Syn1-mediated fusion in muscle cell–induced increase in the drug resistance of cancer cells, we blocked fusion with the Syn1 peptide. Addition of the Syn1 peptide, but not its scrambled version, inhibited increase in the doxorubicin resistance observed after GFP-expressing PC3 cells were cocultured with hMYO (Fig. 7G and H). Interestingly, Syn1 peptide inhibited drug resistance not only for the PC3 cells cocultured with hMYOs but also for the PC3 cells that were cultured on their own, suggesting a possible role of Syn1-mediated fusion in generation of the drug-resistant cells that are normally present in the PC3 cell population.

In summary, our findings support the importance of cell fusion in the muscle cell–induced increases in the numbers of CD133⁺ and drug-resistant prostate cancer cells.

Because our findings implicated Syn1 and AnxA5 in prostate cancer progression, we examined the expression of these proteins in the human prostate tissue microarrays by staining with antibodies to Syn1 and AnxA5 (Fig. 7I and J). We analyzed microarrays with normal prostate, benign prostatic hyperplasia (BPH), and malignant prostate tissue samples characterized by different Gleason score, a well-established histopathology grading system used to evaluate the aggressiveness of prostate cancer: the higher the score, the more abnormal is the tissue. We found that while in normal tissue and in BPH the expression levels of either of these proteins were indistinguishable, in moderately differentiated carcinoma (Gleason score 5–6), Syn1 expression and AnxA5 expression were 60% and 180%, respectively, higher than in the normal tissues. Syn1 expression in prostatic intraepithelial neoplasia (PIN), a prostate abnormality preceding the development of prostate adenocarcinoma, was also higher than in noncancerous and benign tissues.

The observation that Syn1 and AnxA5 expression is increased in cancer tissues compared with nonmalignant tissues is consistent with our in vitro analysis supporting the important role of these proteins in cancer progression.

Interactions between tumor cells and between tumor cells and nonmalignant cells in the tumor microenvironment play important roles in cancer progression (1). For example, signaling from prostatic smooth muscle cells to the prostatic epithelium influences dedifferentiation of epithelial cells and the progression of human prostatic adenocarcinomas (35). In this study, to model the heterotypic cell–cell interactions at the interphase between the prostate gland and adjacent smooth muscle and striated muscle cells, we devised novel coculture systems for cancer cells to grow in the presence of skeletal and smooth muscle cells. We found coculturing to promote the cancer cell stemness, defined here as increases in the fractions of CD133-expressing cells, anchorage-independent growth, drug resistance, and EMT-characteristic changes in the protein expression, all known hallmarks of cancer progression. The mechanisms by which muscle cells induce these changes depend on the coculture-induced robust increases in IL4 and IL13 concentrations in the medium. In turn, elevated levels of IL4 and IL13 upregulated the expression of Syn1 and AnxA5 and promoted fusion of cancer cells. Finally, cell fusion increased stemness of cancer cells. Therefore, our findings suggest that progression of prostate cancer depends on IL4/IL13-mediated interactions between tumor cells and adjacent muscle cells. Finding that coincubation with muscle cells or recombinant IL4/IL13 similarly changes the properties of two distinct types cancer cells (PC3 and primary prostate cancer cells) in the presence of different muscle cells suggests that these changes are not unique to a specific combination of cancer cells and muscle cells. Both of our cell lines are AR-negative and although most primary prostate tumors contain AR-negative cells (36), it will be important to extend our work to AR-positive prostate cancer cells.

Progression of prostate cancer is influenced by cytokines in the tumor microenvironment and, specifically, by IL4 and IL13, (14, 15, 37). IL4 promotes antiapoptotic pathways that contribute to drug resistance of cancer cells and cancer stem cells (38) and that activate the pAKT pathway in cancer cells (39). IL4 and IL13 have been also implicated in myoblast fusion and macrophage fusion (16–18). Here we found that interactions between cancer cells and skeletal or smooth muscle cells exposed cancer cells to increased concentrations of IL4 and IL13. Several lines of evidence suggest that IL4 and IL13 play a pivotal role in muscle cell–induced fusion and elevated stemness of the cancer cells. First, we found an increased concentration of these cytokines in the medium after the coculturing. Second, muscle cell–induced fusion of cancer cells was suppressed by neutralizing antibodies to IL4 and IL13. Third, application of recombinant IL4 and IL13 to cancer cells promoted fusion and stemness, similarly to the effects of the coculturing. The concentrations of recombinant IL4 and IL13 that induce these effects were approximately 1,000-fold higher than their concentrations in the medium of the cocultured cells. This large difference may reflect the contributions of other factors present in the medium, a reduced biological activity of the recombinant cytokine, and a higher local concentration of cytokines at the surface of the contacting cells.

The mechanisms by which the coculturing of cancer cells with muscle cells upregulates expression and secretion of IL4 and IL13 in these cells are yet to be clarified. Our finding that blocking cancer cell fusion suppressed IL-4/IL-13 dependent increase in stemness, indicates that this increase depends on the cytokine-induced cell fusion. More research is needed to elucidate the roles of these cytokines and other secreted factors in the effects of heterotypic cell interactions in the microenvironment of prostate cancer on its progression.

The slow development of cancer cell fusion (2–3 days) is consistent with characteristic times of cell–cell fusion in the generation of myotubes, osteoclasts, and syncytiotrophoblasts (3). Development of the fusion competency of cancer cells is accompanied by upregulation of the expression of Syn1 and Anx5 implicated in some fusion processes (4, 30). We found both proteins to be involved in fusion of cancer cells. Heterotypic cancer–muscle cell fusion also depended on Syn1.

Multinucleated cells labeled with only one of our probes can be also generated by endoreplication, that is, nuclear divisions without accompanying cytokinesis. We present several lines of evidence (content mixing, videomicroscopy, and multinucleation in the presence of DNA replication inhibitor) that coculturing-induced multinucleated cells are formed by fusion.

Fusion between cancer cells and nonmalignant cells such as macrophages, endothelial cells, and epithelial cells, and very rare events of fusion between nonmalignant cells, have been suggested to promote the formation and progression of tumors (12, 13, 40). Fusion of cancer cells with leukocytes generates cells exhibiting increased metastatic behavior and the number of these fusion hybrids in peripheral blood of human patients with cancer correlates with disease stage (41). On the other hand, it has been suggested that fusion between cancer cells could suppress their invasive phenotype and that Syn1 expression could be used as a positive prognostic marker (42). Our work documents fusion of cancer cells in the context of their interactions with smooth and skeletal muscle cells, implicates Syn1 and AnxA5, and establishes the central place of cancer cell fusion in the pathway by which these cells acquire CSLC characteristics and become drug-resistant in vitro. Increased expression of Syn1 and AnxA5 in prostate cancer tissues compared with nonmalignant tissues substantiates the hypothesis that Syn1 and AnxA5 play important role in cancer progression in vivo. These observations also suggest that changes in the expression of these proteins could be used to evaluate a progression potential of the prostate cancer cells.

We propose that interactions with smooth and skeletal muscle cells bring about the hallmarks of prostate cancer progression by the following mechanism (Fig. 7K). First, interactions between muscle and cancer cells upregulate the expression and secretion of IL4 and IL13. Then, IL4 and IL13 upregulate expression of the fusion-related proteins including AnxA5 and Syn1. While AnxA5 is upregulated in many tumors and cultured cancer cells and has been suggested to be involved in the tumor progression (31), the specific contributions of this protein remain unclear. Our data are the first evidence that AnxA5 is upregulated in prostate cancer in vivo and is involved at the cell fusion stage of the progression. Alternatively, AnxA5 expression levels may influence prefusion processes, including cytokine expression. Similarly, while Syn1 was reported to be expressed in some cancer cell lines (9), and RNA encoding envelope protein of another endogenous retrovirus HERV-E was found in prostate carcinoma tissues and cell lines (43), our study is the first report showing upregulation of Syn1 expression in prostate cancer tissue. Moreover, we report that Syn1 contributes to cancer progression by fusing cancer cells, as evidenced by the finding that a peptide that specifically inhibits Syn1-mediated fusion also inhibits the cancer cell progression. An earlier finding that IL4 promotes ASCT2 expression in some breast cancer cells (44) suggests that, in addition to Syn1 upregulation, IL4/IL13 can further promote cancer cell fusion by upregulating expression of Syn1 receptor.

Importantly, the progression of cancer cells was triggered by fusion between cancer cells rather than by fusion between cancer and muscle cells. Indeed, mMYO, which did not fuse with cancer cells, promoted muscle cell–induced progression of prostate cancer cells toward their more dedifferentiated and malignant state. Whereas here we focused on the effects of muscle cells on cancer cells, interactions with cancer cells clearly influenced the muscle cells. For instance, we observed hMYOs and mMYOs cocultured with cancer cells to form some multinucleated myotubes that we have not observed for either of these skeletal muscle cells grown in the same medium on their own. Three days of coculturing with cancer cells also resulted in a notable loss of muscle cells that may resemble the loss of smooth muscle and their replacement with fibroblasts and myofibroblasts in cancerous stroma surrounding developing tumor (45). The specific effects of cancer cells on the smooth and skeletal muscle cells in their environment still wait to be explored. It will be also interesting to compare the effects of mononucleated satellite cells, precursors to skeletal muscle cells used in our study, with the effects of fully differentiated multinucleated myotubes.

In earlier studies, large multinucleated cells have been noted in several cancer cell lines and in different human cancer tissues (46, 47). Moreover, the presence of multinucleated cells in human cancer tissues has been reported to correlate with aggressiveness of the tumor and poor prognosis (46, 47). Chemotherapeutic treatments increase the numbers of large multinucleated cells in cancer tissues (48). Because multinucleated cells undergo asymmetric division (producing mononuclear cells) and self-renewal in vitro and in vivo, these cells can function as cancer stem cells (49). Muscle cell–induced and IL4/IL13–mediated fusion of cancer cells increases cell content of mesenchymal cell–characteristic proteins including AKT, and, specifically, pAKT, MMP9, vimentin, Slug, and downregulates epithelial cell–characteristic E-cadherin. These changes in protein expression in cancer cells may directly contribute to their increased aggressiveness. For instance, the EMT marker Slug plays a major role in prostate cancer progression through regulation of proliferation and promotes pAKT activity via suppression of PTEN (50). Increased Slug expression could initiate EMT in the prostate cancer–muscle microenvironment.

Importantly, we found that, within 3 days of the coculturing, the interactions with muscle cells notably changed the properties of only a fraction of cancer cells: most of the cells in the population did not express CD133 and remained mononucleated and sensitive to the drugs. In the physiologic context, the muscle cell–induced changes in cancer cells can be expected to be slower and less frequent than in our cocultures to be consistent with, fortunately, relatively inefficient progression of the primary prostate tumors.

The prostate gland is surrounded by smooth muscle cells and contacts the skeletal myofibers of the rhabdosphincter. Our study identifies a novel pathway in which interactions between prostate cancer cells and muscle cells increase local concentrations of IL4 and IL13. These cytokines upregulate expression of Syn1 and AnxA5, and promote fusion between cancer cells. In turn, this fusion expands populations of CD133-expressing and drug-resistant cells, an expansion that may represent significant progression of the cancer. This novel pathway of microenvironment-driven progression of prostate cancer identified in in vitro experiments and substantiated by the in vivo findings showing upregulated expression of Syn1 and AnxA5 in cancerous tissues may present new therapeutic targets and may have analogies in other tumors.

No potential conflicts of interest were disclosed.

Conception and design: B. Uygur, E. Leikina, L.V. Chernomordik

Development of methodology: B. Uygur

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Uygur, E. Leikina, K. Melikov, R. Villasmil

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B. Uygur, K. Melikov, R. Villasmil, S.K. Verma, C.P.H. Vary, L.V. Chernomordik

Writing, review, and/or revision of the manuscript: B. Uygur, E. Leikina, K. Melikov, S.K. Verma, C.P.H. Vary, L.V. Chernomordik

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.P.H. Vary, L.V. Chernomordik

Study supervision: L.V. Chernomordik

The research in the LVC laboratory was supported by the Intramural Research Program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development, NIH (Bethesda, MD). C.P.H. Vary is supported by NIH grants P30GM103392 (phase III: COBRE in Vascular Biology); P20GM121301 (phase I: Mesenchymal and Neural Regulation of Metabolic Networks); and U54GM115516 (Northern New England Clinical and Translational Research Center). We thank Armie Mangoba and Dr. Volkhard Lindner (Maine Medical Center Research Institute histology core) for creating prostate cancer tissue microarray. We thank Robert Ackroyd and Dr. Anne Breggia (the Maine Medical Center BioBank) for coordinating studies on clinical specimens, and Dr. Lucy Liaw for sharing these specimens with us.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.