TGFβ and Hippo pathways cooperate to enhance sarcomagenesis and metastasis through the hyaluronan mediated motility receptor (HMMR)

Abstract

High-grade sarcomas are metastatic and pose a serious threat to patient survival. Undifferentiated pleomorphic sarcoma (UPS) is a particularly dangerous and relatively common sarcoma subtype diagnosed in adults. UPS contain large quantities of extracellular matrix (ECM) including Hyaluronic Acid (HA), which is linked to metastatic potential. Consistent with these observations, expression of the HA receptor, hyaluronan-mediated motility receptor (HMMR/RHAMM), is tightly controlled in normal tissues and upregulated in UPS. Moreover, HMMR expression correlates with poor clinical outcome in these patients. Deregulation of the tumor suppressive Hippo pathway is also linked to poor outcome in these patients. YAP1, the transcriptional regulator and central effector of Hippo pathway, is aberrantly stabilized in UPS and was recently shown to control RHAMM expression in breast cancer cells. Interestingly, both YAP1 and RHAMM are linked to TGFβ signaling. Therefore, we investigated crosstalk between YAP1 and TGFβ resulting in enhanced RHAMM-mediated cell migration and invasion. We observed that HMMR expression is under the control of both YAP1 and TGFβ and can be effectively targeted with small molecule approaches that inhibit these pathways. Furthermore, we found that RHAMM expression promotes tumor cell proliferation and migration/invasion. To test these observations in a robust and quantifiable in vivo system we developed a zebrafish xenograft assay of metastasis, which is complimentary to our murine studies. Importantly, pharmacological inhibition of the TGFβ-YAP1-RHAMM axis prevents vascular migration of tumor cells to distant sites.

One sentence summary:

YAP1 and TGFβ cooperatively enhance proliferation and migration/invasion of UPS and fibrosarcomas.

Introduction

Soft-tissue sarcomas (STS) are mesenchymal tumors arising from muscle, fat, cartilage and connective tissue etc. Due to their karyotype complexity, subtype heterogeneity, and the lack of common driver mutations, adult sarcomas are not well understood. The paucity of molecular characterization has resulted in few therapeutic advances beyond standard resection/radiation over the last 30 years (1). Our work focuses on undifferentiated pleomorphic sarcoma (UPS) and fibrosarcomas; aggressive adult tumors found in skeletal muscle and connective tissues. These are commonly diagnosed subtypes, relative to other sarcomas, and high grade occurrences are particularly resistant to standard approaches (2). We found that the central Hippo effector, Yes associated protein 1 (YAP1), is stabilized in human UPS tumors and promotes pro-proliferation and anti-differentiation transcriptional programs (3, 4). YAP1 is unusually stable in UPS and potentially other sarcomas due to epigenetic silencing of its inhibitor, Angiomotin (AMOT) (5), and Hippo kinase copy number loss (3). These perturbations stabilize YAP1 at the protein level; enhance its nuclear localization and subsequent transcriptional activity (6). In the absence of a specific inhibitor for YAP1 we identified epigenetic modulators including suberoylanilide hyroxamic acid (SAHA; Vorinostat), and the BET bromodomain inhibitor JQ1, which can reduce YAP1 activity in combination. Importantly, SAHA/JQ1 inhibited UPS growth in murine models of UPS. Though SAHA/JQ1 treatment has widespread effects, we use these tools to interrogate and then validate YAP1-mediated signaling and phenotypes.

Having established the role of YAP1 in UPS and fibrosarcomas we are currently investigating the role of YAP1-associated pathways, including transforming growth factor β (TGFβ) signaling, in sarcomagenesis. TGFβ is a morphogen that binds to specific serine-threonine kinase receptors, which phosphorylate SMAD proteins when engaged (7). Downstream of receptor activation, phosphorylated SMADs interact with transcriptional regulators, including YAP1, to promote proliferation and metastasis (8). Whereas TGFβ and YAP1 activity are linked in developmental contexts (9, 10) and in various epithelial tumors (i.e. Breast cancer) (11), their relationship in mesenchymal tumors is entirely unknown. Furthermore, both TGFβ (12) and YAP1 (13)are known to promote metastasis in various epithelial contexts, an area that also remains unclear in sarcoma. Our recent work has focused entirely on the role of YAP1 in proliferation and tumorigenesis. Here we investigate the ability of these pathways to coordinate metastasis, as well as cellular proliferation.

Our goal was to identity critical overlapping downstream transcriptional targets of TGFβ and YAP1 in sarcomas that promote primary tumorigenesis, metastatic dissemination, or both for combinatorial therapy development. Previously, we identified NF-κB as the primary driver of proliferation and tumorigenesis in UPS cells. In the present study we report that TGFβ signaling promotes NF-κB activity as well, lending further support to our assertion that Hippo pathway deregulation and TGFβ activation cooperatively enhance sarcomagenesis.

We also sought evidence that these pathways converge to regulate metastasis. The hyaluronan mediated motility receptor (HMMR) gene (14), which encodes the hyaluronic acid (HA) surface receptor RHAMM, is a known transcriptional target of TGFβ signaling (15) and has recently been linked to YAP1 (16, 17). RHAMM is also associated with fibrosarcoma progression (18). RHAMM physically interacts with HA, a component of the extracellular matrix (ECM) and a major part of the tumor microenvironment in many cancers. HA binding engages RHAMM signaling which has both cytoplasmic effects on MAPK activity and proliferation as well as nuclear effects on the transcription of key cell motility effectors(19, 20).

Though there is extensive literature linking RHAMM activity to TGFβ signaling, the connection between RHAMM and YAP1 is significantly more limited, and not well understood. Therefore, we explored the relationship between YAP1 and RHAMM, as well as TGFβ, in sarcomas. We found that YAP1 controls HMMR/RHAMM expression in both murine and human sarcoma cell lines. Consistent with these observations, pharmacological inhibition of YAP1 and TGFβ repressed HMMR/RHAMM expression resulting in reduced tumor growth in our in vivo allograft assays. We also investigated the role of TGFβ and YAP1-mediated HMMR/RHAMM expression in sarcoma cell motility and invasion, and found significant effects on these processes. Importantly, it is extremely difficult to distinguish between effects on tumor size and metastasis in vivo, when a given target affects both processes. Reduction in primary tumor size can correlate with less metastasis. This observation is particularly accurate in sarcomas, wherein low intra-tumoral oxygen (hypoxia) and tumor size are accurate predictors of clinical outcome (21-27). Therefore we employed a novel zebrafish xenograft system that allows us to specifically interrogate the effects of TGFβ and YAP1 inhibition on metastasis in vivo while bypassing the potentially confounding influence of proliferation-associated factors (28-30). Using this system we observed that pharmacological inhibition of both pathways suppressed the metastatic cascade. Collectively, our findings suggest that TGFβ signaling, which is upregulated in UPS and fibrosarcomas, cooperates with YAP1 to control HMMR/RHAMM thus promoting tumorigenesis and metastasis. Interrogation of this mechanism and its physiological impact may help elucidate important new biomarkers and therapeutic targets for the treatment of human sarcoma.

Materials and Methods

GEMM.

All experiments were performed in accordance with NIH guidelines and were approved by the University of Pennsylvania Institutional Animal Care and Use Committee. We generated Kras^G12D+; Trp53^fl/fl; tumors by injection of a calcium phosphate precipitate of adenovirus expressing Cre recombinase (University of Iowa) into the right gastrocnemius muscle of 3- to 6-month-old mice.

Allograft mouse model.

All experiments were performed in accordance with NIH guidelines and were approved by the University of Pennsylvania Institutional Animal Care and Use Committee. For RHAMM-mediated knockdown KP230 allografts, 1×10⁶ cells were injected into the flanks of 6-wk-old nu/nu mice (Charles River Laboratories) with control tumor cells (scrambled shRNA) and experimental tumor cells (RHAMM shRNA). Tumor size was measured every other day, and animals were euthanized after 19 days post-tumor cells injection. Tumor volume was calculated by using formula (ab²)π/6 , where a is the longest measurement and b is the shortest. We also performed subcutaneous tumor experiments using a cell line derived from 3'-methylcholanthrene (MCA)-induced murine fibrosarcoma in the C57BL/6 strain, obtained from the Schrieber laboratory at Washington University of St. Louis (Koebel et al. Nature 2007). Tumor cells were propagated in vitro for two passages prior to implantation and injected cells were greater than 90% viable. We implanted 1 × 10⁶ MCA tumor cells into shaved flanks of recipient C57BL/6 mice and harvested tumors at 14 days post-implantation for analysis. Single cell suspensions were generated by collagenase and DNAse digestion, incubated for 5 minutes at room temperature with anti-mouse CD16/32 Fc Block, and subsequently stained with fluorescently labeled antibodies for identification of macrophage and tumor cell populations by FACS. Cells were sorted on a BD FACS Jazz instrument and RNA was isolated from sorted cell pellets using Sigma GenElute Mammalian Total RNA Miniprep Kit.

Cell lines.

Human HT-1080 (Fibrosarcoma), HCT-116 (Colorectal Cancer) and HEK-293T cell lines were purchased from ATCC (Manassas, VA, USA). STS-109 cells were derived from human UPS tumors by Rebecca Gladdy, M.D. (University of Toronto) RE cells were derived from human UPS tumors by Kurt Weiss, M.D. (University of Pittsburgh Medical College). TC32 cells (Ewing’s sarcoma) were obtained from Patrick Grohar M.D, Ph.D. (Children’s Hospital of Philadelphia). KP230, KP250 and KIA cell lines were derived from UPS mouse tumors as described in Eisinger-Mathason et al. (4). STR analysis was performed at the time of derivation and confirmed in April 2015. Cells were purchased, thawed, and then expanded in the laboratory. Multiple aliquots were frozen down within 10 days of initial resuscitation. For experimental use, aliquots were resuscitated and cultured for up to 20 passages (4–6 wk) before being discarded. Except STS-48, STS109 and RE cells were cultured in DMEM with 20% (vol/vol) FBS, other Cells were cultured in DMEM with 10% (vol/vol) FBS and 1% penicillin/streptomycin. All cell lines were confirmed to be negative for mycoplasma contamination.

Drug Treatments and Reagent.

Cells were treated with SB525334, SAHA (Sigma Aldrich), Trametinib (Advanced Chemblock), and JQ1 (Gift from Jun Qi, Ph.D., Harvard University). Drugs were refreshed for any cells treated for longer than 48hrs.

Lentiviral Transduction.

shRNA-mediated knockdown of Yap1 TRCN: 0000095864, 0000095866, 0000095867, 0000095868; YAP1 TRCN: 0000107266, 0000107267. Hmmr TRCN: 0000071588, 0000071589, 0000071590, 0000071591, 0000071592; HMMR TRCN: 0000061553, 0000061554, 0000061555, 0000061556, 0000061557; Rela (p65 NF-kB) TRCN:0000055343, 0000055344, 0000055347 and Scramble shRNA was obtained from Dharmacon. shRNA plasmids were packaged by using the third-generation lenti-vector system (VSV-G, p-MDLG, and pRSV-REV) and expressed in HEK-293T cells. Supernatant was collected at 24 and 48 hrs after transfection and subsequently concentrated by using 10-kDa Amicon Ultra-15 centrifugal filter units (Millipore). Lenti-pcdh-EF1 was used as the empty vector control in our lab. Mouse and Human HMMR Gene cDNA clone plasmid were obtained from Sino Biological Inc.

Immunohistochemistry.

Fifteen human UPS paraffin embedded tissues obtained from the Surgical Pathology group at Univ. of Pennsylvania were stained for p-Smad3. Subcutaneous tumor paraffin embedded tissues were stained for CD168 and Ki67. IHC was performed on 5 micron tissue sections according to standard protocols. Slides were digitally scanned by Pathology Core Laboratory at the Research Institute at the Children’s Hospital of Philadelphia. Aperio ImageScope (Leica Biosystem, USA) software was used for slide quantification. Modified macro (Nuclear v9 parameter) was used to distinguish intensity of staining. 15 tumor sections and 9 adjacent muscle sections were quantitated. 5 areas of tumor and 5 areas of adjacent muscle (identified by H&E staining) were captured per section. Data from the Aperio IHC Nuclear algorithm were used to calculate the “average intensity x % positivity” metric. This metric is independent of the number of nuclei within a given slide, and considers the contribution of both positive and negative nuclei to staining intensity. For each slide, the “accumulated intensity” output from Aperio (the cumulative intensity of all positively stained nuclei) was divided by the total number of cells. This calculation yielded an “average intensity” value (the average intensity of both positively and negatively stained nuclei), which was then multiplied by the “percent positive” output. Values were divided by 1000 for graphical representation purposes.The following antibodies concentrations were used: rabbit anti-p-Smad3 (phospho S423/S425) (52903; 1:50) (Abcam), rabbit anti-Ki-67 (15580; 1:100) (Abcam), rabbit anti-CD168 antibody (124729, 1:100) (Abcam).

Immunoblots.

Protein lysate was prepared in SDS/Tris (pH7.6) lysis buffer, separated by electrophoresis in 8–10% SDS/PAGE gels, transferred to nitrocellulose membrane, and probed with the following antibodies: rabbit anti-YAP1 (4912; 1:1000), rabbit anti-p-Yap (Ser397) (13619; 1:1000), rabbit anti-p-Smad2 (S465/467) (3108; 1:1000), rabbit anti-Smad2 (5339; 1:1000), rabbit anti-p-Smad3 (S423/425) (9520; 1:1000), rabbit anti-p65 (8242; 1:500), rabbit anti-GAPDH (2118; 1:1000), rabbit anti-p-44/42 Erk1/4 (Thr202/Tyr204) (4377; 1:1000), rabbit anti-44/42 Erk1/4 (4695; 1:1000) (Cell signaling), rabbit anti-CD168 antibody (124729, 1:100) (Abcam), rabbit FOXM1 (sc-502, 1:500) (Santa Cruz Biotechnoloy).

Oncomine, TCGA survival and cBioPortal analysis.

We used the publically available database Detwiller et al. and Barretina et al. through Oncomine Research Premium edition software (version 4.5, life Technologies) to query RHAMM expression and survival. RHAMM gene expression in overall survival from MFH/UPS in TCGA Adult Soft Tissue Sarcoma Dataset (31). Kaplan-Meier analyses were performed for overall survival of patients. Through cBioPortal database, we queried HMMR gene amplification in 89 (60%) of 149 total sequenced patients.

Cell proliferation assay.

Cells were plated in triplicate for each time point, and incubated overnight in tissue culture dishes. The following day, DMSO or SB525334, diluted in growth media, was added to the cells. Cells were trypsinized, resuspended in PBS and counted using a haemocytometer on the days indicated.

Wound healing assay.

To determine the effect of RHAMM on sarcoma cell migration, wound healing assays were performed with RHAMM-silenced cell lines, overexpressed RHAMM cell lines and TGFβ receptor 1 inhibitor (SB525334) (Tocris) pretreated cells for 48hrs. Wounds were made through the monolayer using a 200 μl pipette tip. Wounds were measured over a time course to calculate the migration rate according to the following formula: ((Initial wound length)-(Final wound length))/(Initial wound length) x 100. The experiments were performed more than three times.

Matrigel invasion assay.

In vitro cell invasion was performed using the BD Biocoat™ Matrigel Invasion Chamber (BD Biosciences, San Jose, CA) with endogenous RHAMM-silenced cell lines HT-1080 and KP230 according to the manufacturer’s protocol. Briefly, the Matrigel coating chambers were rehydrated in 1ml DMEM/F12 for 2h in a humidified tissue culture incubator with a 37°C, 5% CO2 atmosphere prior to the experiments. 8 ×10⁴ HT-1080 cells and 1.6×10⁴ KP 230 cells were seeded on to matrigel-coated filters. After 24hrs of incubation, the filters were stained using 0.2% crystal violet and the number of traversed cells was counted on an inverted microscope. For each membrane, five random fields were selected, and the invasion rates were determined by the percent invasion formula: percentage invasion = (mean number of cells invading through the matrigel insert membrane)/(mean number of cell migrating through the control insert membrane) x100.

qRT-PCR.

Total RNA was isolated from tissues and cells using the Trizol reagent (Life Technologies) and RNeasy Mini Kit (Qiagen). Revere transcription of mRNA was performed using the High-Capacity RNA-to-cDNA Kit (Life Technologies). qRT-PCR was performed by using a ViiA7 apparatus. All probes were obtained from TaqMan “Best coverage” (Life Technologies). HPRT was used as an endogenous control.

Methods for zebrafish microinjection:

All procedures on zebrafish (Danio rerio) were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania. Fertilized zebrafish eggs of the transgenic strain expressing enhanced green fluorescent protein (EGFP) under the fli promoter (fli:EGFP) were incubated at 28 °C in E3 solution and raised using standard methods. Embryos were transferred to E3 solution containing 0.2 mM1-phenyl-2-thio-urea (PTU, Sigma) six hours post-fertilization to prevent pigmentation. Proteases were added to PTU-E3 solution on Day 1 to dechorionate the fish embryos. At 48 h post-fertilization, zebrafish embryos were anesthetized with 0.03% tricaine (Sigma) and then transferred to an injection plate made with 1.5% agarose gel for microinjection. Approximately 200–400 KIA/mCherry cells (~5 nL) suspended in complete DMEM medium further supplemented with 0.5 mM EDTA were injected into the yolk sac of each embryo using a XenoWorks Digital Microinjector (Sutter Instrument). Pre-pulled non-filamentous borosilicate micropipettes were used for the microinjection (Tip ID 50 µm, base OD 1 mm, length 5.5 cm, Fivephoton Biochemicals). After injection, the fish embryos were immediately transferred to PTU-E3 solution. Injected embryos were kept at 33 °C and were examined every day to monitor tumor migration using a dissecting fluorescent microscope (Olympus MVX10). Confocal images on embryos showing distant metastasis were captured using a spinning disk microscope (Olympus Ix81 with an Andor iXon3 EMCCD camera) and analyzed using software FIJI.

Statistical Analysis:

Statistical analysis was performed using Prism (Graph Pad Software). Data are shown as mean± SEM or SD. Data were reported as biological replicates, with technical replicates indicated in figure legends. Student t-tests (unpaired two tailed) were performed to determine whether a difference between two values is statistically significant different, with a P-value <0.05 considered significant. ANOVA was used for assays with multiple variables to compare.

Results

TGFβ signaling promotes sarcoma cell migration and is associated with disease progression

Our recent work demonstrated that de-regulation of the Hippo pathway, which results in stabilization of YAP1, promotes UPS initiation and growth (3, 5). In some contexts YAP1-mediated transcription is dependent on its interaction with phosphorylated SMAD3 (p-SMAD3). Therefore, we investigated the role of TGFβ, which signals through activated SMAD3, in UPS. We observed that like YAP1 nuclear p-SMAD3 staining is significantly elevated in human UPS relative to skeletal muscle, suggesting a link between TGFβ and YAP1-mediated UPS development (Fig. 1A). We also interrogated TGFβ activation in the genetically engineered mouse model of UPS, LSL-Kras^G12D/+; Trp53^fl/fl (KP). In this model, adenovirus expressing Cre recombinase is injected into the gastrocnemius muscle, activating oncogenic Kras expression and deleting p53 in muscle progenitor cells (32, 33). This model and a similar one, Kras^G12D/+; Ink4a/arf^fl/fl (KIA), generate sarcomas that are histologically, transcriptionally, and morphologically identical to UPS and are thus the standard GEMMs in UPS studies (32, 33). Though Kras mutation is rare in human sarcomas, activation of MAPK downstream of KRAS is common in UPS and is a reliable prognostic indicator (34). To fully characterize MAPK status in our experimental systems we evaluated p-ERK1/2 levels and sensitivity to MEKi in HT-1080 human fibrosarcoma cells, which express mutant oncogenic N-Ras (ATCC), and murine cells derived from KP tumors. Interestingly, we observed that HT-1080 cells maintain higher basal levels of MAPK activity than KP cells (Supplementary Fig. S1A). Consistently, we found that KP cells are less sensitive to MEKi (Trametinib) and require higher doses of Trametinib to completely inhibit proliferation (Supplementary Fig. S1B, C). Lastly, we found that treating each cell type with a dose of Trametinib that halts proliferation (HT-1080; 50 nM and KP; 150 nM) results in roughly the same cellular responses. Specifically, using Western blot analysis we determined that p-ERK1/2, RHAMM, and other proliferation targets were reduced in response to Trametinib in both cell types (Supplementary Fig. S1D, E).

Figure 1.

TGFβ stimulates sarcoma cell proliferation and migration. (A) H&E and IHC of p-SMAD3 from human UPS tumor and skeletal muscle sections. Scale bar=50 µM, tumor n=15, muscle n=9 *P <0.001. SD. (B) H&E and IHC for p-Smad3 from KP230 tumor and skeletal muscle sections. Scale bar=50 µM, *P <0.001. SD. (C) cBioPortal analysis of TGFβ pathway genetic alterations in MFS and UPS patients from the TCGA Soft tissue Sarcoma in Adults dataset. (D) Proliferation assay of HT-1080 and (E) KP230 cells treated with SB525334 1μM, 5μM and 10μM, *P <0.05. SD. (F) Scratch migration assay of HT-1080 and (G) KP230 cells pre-treated with SB535334 (1μM) for 48h. *P <0.0001, n=3. SD.

Importantly, Yap1 is stabilized in KP GEMM tumors (5), and deletion of YAP1 in this system (KPY) increased tumor latency and reduced tumor weight and volume, relative to KP (5). Our assessment of nuclear p-SMAD3 levels in KP tumors phenocopies human UPS. Consistent with our observation in human UPS, p-SMAD3 staining is significantly upregulated in KP tumors relative to murine skeletal muscles (Fig 1B). Using cBioPortal analysis of The Cancer Genome Atlas (TCGA) Soft Tissue Sarcoma Adult Dataset annotated in 2017 (31) we also observed that more than 50% of human UPS and myxofibrosarcomas, which are genetically indistinguishable from UPS (31), bear alterations in the TGFβ pathway (Fig. 1C). We observed similar alterations in liposarcomas using cBioPortal analysis of the MSKCC Broad dataset (Supplementary Fig. S2A). The majority of tumors with enhanced expression of TGFβ pathway members including, TGFβR1, TGFβ1, and SMAD3, also bear mutations in common sarcoma tumor suppressors like TP53 and ATRX, and express high levels of FOXM1, a YAP1 transcriptional target. Together, these data suggest that TGFβ activity plays a role in UPS initiation or progression. In contrast, mutations in the TGFβ pathway are uncommon in colorectal cancer (CRC) (Supplemental Fig. S2B) and CRC cell proliferation is independent of TGFβ (35-38). These data support the conclusion that TGFβ signaling promotes sarcoma cell proliferation but may not be involved in proliferation of certain carcinoma cells. To validate these observations we treated HT-1080 and KP cells with increasing doses of the TGFβ inhibitor SB525334 (SB) (Fig. 1D, E). Treatment with 1μm SB for 5 days had no effect on HT-1080 proliferation, though higher doses were significantly more impactful. Interestingly, KP cell proliferation was more sensitive to TGFβ inhibition. However, pretreatment with 1μm SB treatment had significant effects on migration in both cell types, suggesting that cell migration is more uniformly sensitive to TGFβ signaling among sarcoma subtypes (Fig. 1F, G). To demonstrate the specificity of our proliferation results to sarcoma we stained a tissue microarray of normal colon and CRC samples with p-SMAD3 and treated HCT-116 CRC cells with SB. We observed that CRC tissues stain positive for p-SMAD3, relative to normal tissue, (Supplemental Fig. S2C, D). However, TGFβ inhibition has extremely minor effects on CRC proliferation (Supplemental Fig. S2E, top). These findings are consistent with earlier reports indicating that TGFβ signaling specifically promotes migration and invasion in CRC (39). Lastly, we evaluated the effect of SB on an additional sarcoma subtype, Ewing’s sarcoma. 5μm SB dramatically reduced proliferation in the Ewing’s cell line TC32 (Supplemental Fig. S2E, bottom). We conclude that TGFβ signaling plays important and perhaps specific roles in sarcoma growth and dissemination, whereas its effects are limited to tumor progression in CRC.

TGFβ and YAP1 cooperatively regulate HMMR/RHAMM expression

In previous studies we reported that YAP1 mediates proliferation in UPS cells and tissues by upregulating NF-κB activity. Our findings revealed that regulation of NF-κB is a critical function of YAP1 in muscle-derived sarcomas. Therefore, we hypothesized that TGFβ signaling is also required for NF-κB activity in UPS. To test this idea we treated KP230cells with 1μm SB and observed decreased phosphorylation of p65/NF-κB, which is indicative of transcriptional inactivation (Fig. 2A). We also saw loss of Foxm1 expression in response to TGFβ inhibition, which is consistent with loss of YAP1 activity. FOXM1 is a pro-growth factor that promotes UPS proliferation. Though we have done significant work on YAP1 and FOXM1 sarcoma cell proliferation and tumor formation, we have yet to determine whether YAP1 also enhances metastasis in this context or the cellular functions preceding metastatic dissemination, including cell migration and invasion. YAP1 is known to promote these processes in epithelial cancers. However, its impact on metastasis of mesenchymal tumors is unknown. In breast cancer YAP1 regulates HMMR/RHAMM, a surface receptor that responds to HA and is associated with metastasis. Here we observed that TGFβ inhibition decreases RHAMM expression in UPS cells (Fig. 2A). The specific effects of TGFβ inhibition on sarcoma cell proliferation, migration, and invasion suggest high levels of TGFβ cytokine production in the UPS tumor microenvironment. To validate Tgfβ expression in vivo, and to determine the source of Tgfβ production, we fractionated KP as well as 3’-methylcholanthrene (MCA)-induced fibrosarcomas and isolated tumor-associated macrophages (MACs), which are known to produce significant levels of cytokines including Tgfβ. We compared Tgfβ expression in isolated KP and MCA fibrosarcoma tumor cells to infiltrating macrophages. Tgfβ mRNA expression was ~6 fold higher in infiltrating MACs than KP tumor cells (Fig. 2B; left) and ~30 fold higher in MACs from MCA-induced sarcomas relative to tumor cells (Fig. 2B; right). Together, these data suggest that macrophages produce Tgfβ in the sarcoma microenvironment.

Figure 2.

TGFβ and YAP1 cooperatively regulate RHAMM expression in soft tissue sarcoma. (A) Western blot of KP230 cells treated with SB525334 1μM 48h. (B) qRT-PCR of mRNA derived from autochthonous KP and MCA-induced tumor cells (TCs) as well as associated macrophages (MACs) *P <0.0006. SD. (C) qRT-PCR of HT-1080 (D) STS109 (human UPS) and (E) KP230 cells transduced with lentivirus expressing two shRNAs targeting Yap1, *P <0.01. SD. (F) Western blot of HT-1080, STS109, KP230, and RE (human UPS) cells expressing control or YAP1 shRNAs.

Collectively, our data suggest that TGFβ signaling, which is activated by tumor infiltrating macrophage secretion of TGFβ and YAP1 cooperatively regulate HMMR/RHAMM expression. Therefore we tested the contribution of YAP1 to HMMR/RHAMM expression using multiple independent YAP1-specific shRNAs. Inhibition of YAP1 suppressed HMMR levels in HT-1080 (Fig. 2C) STS-109 human UPS (Fig. 2D), and KP230 cells (Fig. 2E). We observed similar results at the protein levels by Western blot of YAP1 shRNA expressing UPS and fibrosarcoma cells (Fig. 2F). We conclude that both YAP1 and TGFβ control RHAMM expression in UPS and fibrosarcoma cells.

HMMR expression is linked to clinical outcome

Our laboratory studies revealed a clear link between HMMR/RHAMM expression and the Hippo-TGFβ axis. Next, we investigated expression of HMMR in human sarcomas and the relationship between HMMR and patient outcome. We found that HMMR is uniformly upregulated in UPS and myxofibrosarcoma (MFS) tumors (Fig. 3A). MFS is genetically indistinguishable from UPS (31). We also observed upregulation of HMMR mRNA in the MSKCCBroad (liposarcoma) and TCGA Soft Tissue Sarcomas in Adults datasets (UPS) via cBioPortal analyses (Supplementary Fig. S2A). Importantly, high HMMR expression correlated with poor clinical outcome in MFS patients (Fig. 3B). Next, we investigated whether HMMR is co-expressed with other YAP1 targets in UPS/MFS and liposarcoma. Genes that co-expressed significantly with HMMR in the TCGA Soft Tissue Sarcomas in Adults” datasets included, FOXM1, and BUB1B as well as others that we identified as YAP1 targets in previous studies (3) (Fig. 3C). Based on cBioPortal analyses of human sarcoma datasets, these targets are upregulated in human UPS/MFS (Supplemental Fig. S3A). They are also inhibited in response to SAHA/JQ1 in KP cells (Supplemental Fig. S3B). Here we also show individual co-expression plots for HMMR vs. FOXM1 or BUB1B in UPS/MFS and liposarcoma (Fig. 3D and Supplementary Fig. S4A-C). Lastly, we determined that RHAMM and the HA binding protein HABP are highly expressed in human tissues by IHC (Supplementary Fig. S4D). These findings highlight the importance of HMMR/RHAMM expression in human sarcoma, as well as the relationship between HMMR and YAP-mediated transcription in these tumors.

Figure 3.

RHAMM expression correlates with poor outcome and YAP1-mediated sarcomagenesis (A) Oncomine gene expression analysis of RHAMM mRNA levels in normal tissues including skeletal muscle and adipose tissue as well as multiple sarcoma subtypes from the Detwiller et al. and Barretina et al. datasets. Each bar represents an individual patient sample. (B) Kaplan-Meier curve of overall survival of MFS patients (n=17; TCGA sarcoma dataset). (C) cBioportal analysis of genes co-expressed with HMMR in the Adult Soft Tissue Sarcoma TCGA dataset (MFS and UPS only). (D) Co-expression graphs of YAP1-associated targets CDKN3 and FOXM1.

RHAMM expression is sensitive to epigenetic modulation of YAP1 transcriptional activity

In previous studies, we reported that YAP1 activity is nearly abolished in sarcoma cells treated with a combination of epigenetic modulators including the histone deacetylase (HDAC) inhibitor SAHA (Vorinostat), and the BET bromodomain inhibitor, JQ1 (5). Individually, SAHA is more effective at reducing proliferation and YAP1 expression relative to JQ1. However, we observed that the combination of these drugs is by far the most effective pharmacological method for inhibiting YAP1 expression and activity. Importantly, SAHA/JQ1 treatment does not trigger cell death, but rather induces differentiation in sarcoma cells (5, 40). Though SAHA/JQ1 treatment has pleotropic effects, we use these tools to interrogate and then validate YAP1-mediated signaling and related phenotypes. The observation that YAP1 regulates HMMR/RHAMM expression led us to hypothesize that SAHA/JQ1 treatment would similarly inhibit this target. We treated KP230 cells and performed microarray analysis as well as qRT-PCR validation and found that both Yap1 and Hmmr levels were significantly inhibited by drug treatment (Fig. 4A-C). We also performed Western blot of SAHA/JQ1 treated cells for 0–72hrs and found that RHAMM expression is lost in HT-1080 (Fig. 4D), KP230 (Fig. 4E), and KIA (Fig. 4F) sarcoma cells. Consistent with our prior studies, we observed less p65/NF-κB phosphorylation and a decrease in overall YAP1 levels in treated cells. Together, our findings indicate that RHAMM expression is dependent on YAP1 and TGFβ signaling and could be pharmacologically repressed using inhibitors that modulate these pathways. The role of YAP1 in controlling NF-κB activity led us to ask whether HMMR/RHAMM is a target of this pathway or is regulated in an NF-κB-independent manner. Therefore, we inhibited NF-κB in KP230 cells using Rela specific shRNA and observed decreases in Hmmr (Fig. 4G) and Rhamm expression (Fig. 4H). We conclude that RHAMM expression is regulated by the YAP1-NF-κB axis in addition to TGFβ signaling.

Figure 4.

TGFβ and YAP1 cooperatively regulate RHAMM via NF-κB signaling pathway. (A) Broad institute Morpheus software heatmap based on gene expression from microarray analysis of KP230 cells treated with DMSO or SAHA (2 µM)/JQ1 (0.5 µM) for 48 hrs. (B) qRT-PCR of HT-1080 and (C) KP230 cells treated as in A. *P <0.01. n=3. (D) Western blot of HT-1080 (E) KP230 and (F) KIA treated as in A. (G) qRT-PCR of KP cells p65/Rela shRNA, *P <0.01. n=3. (H) Western blot of KP cells expressing Rela shRNA.

RHAMM modulates UPS growth

Our current understanding of the YAP1-NF-κB axis is restricted to its effects on proliferation. Therefore, we began our functional analysis by modulating HMMR/RHAMM expression and evaluating subsequent effects on proliferation. We used multiple independent shRNAs targeting HMMR in HT-1080 and KP230 cells (Fig. 5A-C). Loss of HMMR/RHAMM significantly reduced cell proliferation in HT-1080 (Fig. 5D) and KP230 cells (Supplementary Fig. S5A). We also performed the reverse experiment by ectopically expressing RHAMM in HT-1080 cells (Fig. 5F) and observed significantly enhanced cell proliferation (Fig. 5G). Though the effects were more modest, we observed similar effects in KP cells (Supplementary Fig. S5B). We attribute the modesty of the effects to the significantly weaker expression of pcdh-Rhamm in KP cells (Supplementary Fig. S5C, D) than pcdh-RHAMM in HT-1080 cells (Fig. 5E,F). We investigated the effects of Hmmr/Rhamm loss in vivo in an allograft KP model by subcutaneously implanting KP cells expressing control or Hmmr/Rhamm specific shRNAs and found that Rhamm expression is significantly lost based on IHC staining of tumor sections (Fig. 5H-I). Inhibition of Hmmr/Rhamm reduced both Ki67 staining and tumor volume/weight (Fig. 5J-L). We also confirmed that Rhamm protein expression was decreased in explanted tumor tissue by Western blot (Fig. 5M). Lastly, we tested the hypothesis that altered Rhamm expression resulted in a feed-back loop, which modulated SMAD3 phosphorylation and TGFβ activity. Loss of RHAMM in HT-1080 and KP230 cells had no reproducible effects on p-SMAD3 (Supplementary Fig. S5E). Collectively, these findings indicate that Rhamm expression promotes UPS proliferation and tumorigenesis in vivo, but does not impact upstream TGFβ signaling.

Figure 5.

RHAMM promotes sarcomagenesis in vivo. (A) qRT-PCR of HT-1080 and (B) KP cells expressing multiple independent shRNAs targeting HMMR/Hmmr. (C) Western blot of HT-1080 and KP230 cells expressing multiple independent shRNAs targeting HMMR/Hmmr, *P <0.01. (D) Proliferation assay of HT-1080 cells expressing control or multiple independent RHAMM shRNAs, *P <0.01. (E) Proliferation assay of HT-1080 cells ectopically expressing pcdh-RHAMM, *P <0.01. (F) qRT-PCR and (G) Western blot of HT-1080 cells ectopically expressing pcdh-RHAMM, *P <0.01. (H) IHC Images and (I) quantification of Rhamm and (J) Ki67 staining from 5 unique subcutaneously implanted KP230 cells expressing control or Hmmr/Rhamm targeting shRNAs. *P <0.01. Scale bar=50 µM. (K) Tumor volume; SEM and (L) tumor weights; SEM from subcutaneous allografts of KP230 cells treated with lentivirus expressing shRNAs; *P <0.01. (M) Western blot of 3 allograft tumors harvested 17 days after subcutaneous injection.

RHAMM promotes UPS migration and invasion

Though our investigation into YAP1 transcriptional targets has focused primarily on proliferation modulators, YAP1 is also known to have effects on migration/invasion in vitro and metastasis in vivo (41-44). RHAMM is also known to mediate cellular changes in response to ECM and plays a role in tumor cell dissemination. Therefore, we explored the effects of RHAMM inhibition on sarcoma cell migration and invasion using scratch migration and matrigel- coated transwell invasion assays. In HT-1080 cells HMMR/RHAMM loss inhibited both migration and invasion, but more effectively reduced migration (Fig. 6A; top, Fig. 6B) relative to invasion. Invasive cells are labeled with crystal violet (Fig. 6A; bottom, Fig. 6C). In KP230 cells Rhamm loss inhibits both migration and invasion to a similar extent (Fig. 6D-F). KIA cell migration was also sensitive to Rhamm loss (Supplementary Fig. S6A, B). To determine the specificity of Rhamm as a Yap1 target in this system we performed a rescue migration assay using Yap1 targeting shRNA and pcdh-Rhamm. Ectopic expression of Rhamm increased migration in a scratch assay, whereas loss of Yap1 reduced migration. In combination, we observed that ectopic expression of Rhamm in Yap1-depleted cells returned migration to control levels (Fig. 6G). Therefore we conclude that Rhamm expression promotes migration and invasion in response to Yap1 upregulation, collectively enhancing sarcomagenesis.

Figure 6.

RHAMM promotes migration and invasion of sarcoma cells. (A) (top) Scratch migration and (bottom) and (bottom) invasion assay of HT-1080 cells expressing control or HMMR/RHAMM shRNAs. (B) Quantification of migration and (C) invasion assays from A. *P <0.01. n=3. SD (D) (top) Scratch migration and (bottom) and (bottom) crystal violet invasion assay of KP230 cells expressing control or Hmmr/Rhamm shRNAs. (E) Quantification of migration and (F) invasion assays from D. *P <0.01. n=3. SD (G) Rescue scratch migration assay of HT-1080 cells expressing shYAP1, pcdh-HMMR, or pcdh-HMMR+shYap1. *P <0.01. n=3. SD.

TGFβ and YAP1 inhibition prevent metastasis in vivo

Many proteins, including YAP1 and RHAMM, promote proliferation as well as aspects of metastatic progression. Until recently, it has been difficult to investigate molecules with these dual functions in vivo. Reduction in distant metastases cannot be specifically and conclusively attributed to changes in migratory or invasive characteristics when primary tumor size is also decreased in response to target depletion. Therefore, we have incorporated a zebrafish xenograft approach into our studies. In recent years zebrafish embryo xenograft models have gained attention for their utility in rapid assessment of cell migration in vivo, high throughput drug screening, and lack of adaptive immunity, which supports engraftment of cell lines and PDX systems (28-30). As a result of the short time frame of these experiments (3–48 hrs) and the imaging-based output, we were able to adapt this system to specifically evaluate the effects of the TGFβ-YAP1 axis on sarcoma metastasis. We injected fluorescently tagged (mCherry+) KIA cells, which are highly metastatic, into the yolk sac of a 2-day old Fli-EGFP embryo. These zebrafish express EGFP in the endothelial population. 3 hrs post-injection we observed tumor cell dissemination towards the tail and head regions of the embryo, indicating that sarcoma cells will migrate in zebrafish (Fig. 7A). Using spinning disk microscopy to generate high-resolution images of live zebrafish embryos we observed substantial migration and colonization at both the head and tail of the embryo 24 hrs post tumor cell injection (Fig. 7B). We also noticed that the caudal artery (CA), which contains the tumor cells, is significantly disrupted as they cluster and proliferate. Interestingly, at 3 hrs post-injection many tumor cells can be visualized individually and they generally remain inside the caudal artery, though the extravasation process has begun. However, by 24 hrs post-injection some cells appear clustered or fused together and many have fully extravasated (Fig. 7C). We and others have performed similar studies using MDA-MB-231 breast cancer cells and other tumor cell types and have observed predominantly single cell migration in MDA-MB-231(45), whereas other tumor cell types cluster prior to extravasation (46). This clustering may be an important contributor to extravasation and metastatic outgrowth in sarcoma and will be investigated in future studies. For more detailed analyses of sarcoma cell dissemination we performed 3D surface interaction analysis on merged fluorescent images as shown in C (Supplementary Fig. S7A). Yellow peaks indicate transendothelial migration, which is mostly complete by 24 hrs. 3D modeling of the tail region of injected embryos confirms this finding (Supplementary Fig. S7B). To address the role of Hmmr/Rhamm in tumor dissemination in vivo we used shRNA to inhibit Rhamm expression in GFP expressing KIA cells and injected them into flk-mCherry zebrafish embryos. In these fish the vasculature is labeled with mCherry. We measured metastatic capacity of control and Rhamm shRNA expressing cells and found a dramatic decrease in dissemination when Rhamm is lost (Fig. 7D, E). Metastatic capacity is defined as # of fish showing migrated cells/ # of fish injected. We also used a pharmacological approach to inhibit TGFβ and YAP1 signaling and found that both treatments suppressed metastasis by 75% (Fig. 7F). Using this powerful xenograft tool we were able to specifically assess the role of these pathways in tumor cell dissemination and observed that the TGFβ-YAP1-Rhamm axis promotes metastasis and can be therapeutically targeted in vivo.

Figure 7.

TGFβ inhibitor and SAHA/JQ1 treatment prevent metastasis in a zebrafish xenograft model. (A) KIA/mCherry sarcoma tumor cells were injected into the yolk sac of Fli-EGFP transgenic zebrafish embryos at 48 h post fertilization. Tumor cell migration and metastasis were detected under fluorescent microscopy at 3 h post injection. White arrow indicates primary injection site. Blue arrowhead indicates distant metastasis towards tail. (B) Representative images of zebrafish embryos displaying overt extravasation at 24hr post injection. Upper: Representative images showing metastasis in zebrafish head, trunk, and tail respectively. “h” marks heart. White arrow indicates primary injection site. Star indicates tumor migration into the pericardiac space. Blue arrowheads indicate metastasis. Lower: Magnified image on tail metastasis. White arrowheads indicate extravasated tumor foci. Cartoons indicate areas being imaged. (C). Representative high-resolution 3D confocal images of zebrafish tail regions showing initial and overt extravasation at 3hr and 24hr post injection respectively (Scale bar, 100 μm.). Yellow signals indicate transendothelial migration of tumor cells. Dashed boxes encircle extravasation detailed in zoom-in-views on the right. White arrowheads indicate extravasated tumor foci. Cartoons indicate areas being imaged. Annotations: A: anterior; P: posterior; D: dorsal; V: ventral. (D) KIA/Cop-GFP sarcoma tumor cells expressing Scr or Rhamm shRNA #1 were injected into the yolk sac of Flk-mCherry transgenic zebrafish embryos at 48 h post fertilization. Tumor cell migration and metastasis were detected under fluorescent microscopy at 3 h post injection. White arrow indicates metastatic cells. Cartoons indicate areas being imaged. (E) Quantification of normalized metastatic capacity. Metastatic capacity is defined as # of fish showing migrated cells/ # of fish injected (head, trunk, tail metastasis all combined), normalized to 100% for Cop-GFP expressing KIA cells as in D or (F) SB-525334- and SAHA/JQ1-treated tumor cells (n ≥ 2 injections; each injection included at least 20 zebrafish embryos). Statistical analysis was performed using one-way ANOVA followed by Dunnett's multiple comparisons test. *P<0.05. SD.

Discussion

The goal of our recent work has been to delineate mechanisms by which Hippo pathway deregulation contributes to sarcomagenesis. Our observation that constitutive NF-κB signaling is the main molecular driver of muscle-derived UPS (5) led us to explore additional pathways that promote NF-κB activity independent of or in cooperation with YAP1, the transcriptional effector of the Hippo pathway. Here we report that TGFβ signaling increases transcription of YAP1-associated pro-proliferation targets including FOXM1, enhances NF-κB activity, and promotes metastasis. TGFβ signaling has been shown to promote YAP1 activity through phosphorylated SMAD3 (47, 48). In the nucleus p-SMAD3 physically interacts with the YAP1/TEAD1–4 complex and initiates transcription. This mechanism, which has been characterized in epithelial cancer cells, now appears likely to extend to sarcomas (Fig. 8).

Figure 8.

Model summarizing the role of TGFβ and Hippo pathways in through the HMMR/RHAMM – modulated sarcomagenesis and metastasis.

Very little is known about the role and importance of TGFβ signaling in muscle-derived sarcomas like UPS. Other groups have linked this pathway to osteosarcoma, fibrosarcoma, and chondrosarcoma (49-53), but only one limited study has suggested that TGFβ is relevant in UPS (54) (formerly known as malignant fibrous histiocytoma (MFH)). Our work has revealed a clear link between TGFβ activity and UPS. We observed that p-SMAD3 expression is significantly upregulated in human and murine UPS relative to control skeletal muscle tissue. Importantly, we also show that pharmacological inhibition of TGFβ significantly reduces UPS cell proliferation, migration, and invasion. Interestingly, migration/invasion are more sensitive to TGFβ inhibition, suggesting that the critical function of the TGFβ pathway in UPS and fibrosarcoma is in control of tumor cell dissemination and metastatic processes.

Therefore, we sought to identify potential transcriptional targets of TGFβ, associated with migration/invasion, that are products of co-regulation by YAP1 in sarcoma. HMMR/RHAMM is one of the most well established TGFβ targets in various cancer contexts (15, 55, 56). RHAMM signaling specifically promotes metastasis and associated cellular processes, particularly in fibrosarcoma (14, 20, 57, 58). Recently, Wang et al. reported that RHAMM is regulated by YAP1 in breast cancer motility (16). Together, these findings suggested that RHAMM might be a critical target downstream of cooperative TGFβ and YAP1-mediated transcription in sarcoma. Moreover, RHAMM is highly expressed in human UPS and fibrosarcomas, relative to normal mesenchymal tissues, and its expression is linked to poor clinical outcome. Our findings suggest that dual inhibition of YAP1 and TGFβ may significantly reduce primary sarcomagenesis and metastasis. Furthermore, our results suggest the HMMR/RHAMM expression mediates the effects of these pathways on proliferation and dissemination.

To test the effects of pharmacological inhibition of the YAP1- TGFβ-RHAMM axis on sarcoma cell dissemination we initiated a zebrafish xenograft model that allows specific investigation of cell motility in vivo (29, 59, 60). This system has significant advantages over standard in vitro migration/invasion assays. Injected tumor cells are subjected to the fluidic properties native to circulating cells including shear stress and interaction with multiple additional cell types. This system facilitates powerful imaging approaches due to the transparent nature of 2- day old embryos, which also lack fully developed immune systems. Pre-treatment of murine UPS cells with a TGFβ inhibitor or SAHA/JQ1 significantly reduced tumor cell dissemination from the site of injection to the tail region of the fish.

Collectively, our findings suggest that TGFβ signaling cooperates with YAP1 to modulate critical genes like HMMR to promote primary sarcomagenesis and metastatic progression in UPS and fibrosarcomas. Furthermore, we’ve established the potential utility of pharmacological inhibition of the TGFβ/YAP1 axis in this context while investigating RHAMM as an important biomarker of metastasis and therapeutic intervention in this context.

Implications: These studies reveal key metastatic signaling mechanisms and highlight potential approaches to prevent metastatic dissemination in UPS.