Mesenchymal-Epithelial Transition in Sarcomas Is Controlled by the Combinatorial

1 Mesenchymal-epithelial transition in sarcomas is controlled by the combinatorial

2 expression of miR-200s and GRHL2

3 Jason A. Somarelli1*, Samantha Shelter1, Mohit K. Jolly2,3, Sarah Wang4, Suzanne Bartholf Downloaded from 4 Dewitt5, , Alexander J. Hish1, Shivee Gilja1, William C. Eward5, Kathryn E. Ware1, Herbert

5 Levine2,3, Andrew J. Armstrong1,6, and Mariano A. Garcia-Blanco 4,7,8

7 1Duke Cancer Institute and the Department of Medicine, Duke University Medical Center, http://mcb.asm.org/ 8 Durham, NC, 27710, USA; 2Center for Theoretical Biological Physics, 3Department of

9 Bioengineering, Rice University, Houston, TX 77005-1827, USA; 4Department of Molecular

10 Genetics and Microbiology , Duke University Medical Center, Durham, NC, 27710; 5Department

11 of Orthopaedic Surgery, Duke University Medical Center, Durham, NC, 27710, USA; 6Solid

12 Tumor Program and the Duke Prostate Center, Duke University Medical Center, Durham, NC, on August 1, 2016 by DUKE MEDICAL LIBRARY

13 27710, USA; 7Program in Molecular Genetics and Genomics, Duke Cancer Institute, Duke

14 University Medical Center; Durham, NC, 27710, USA; 8Department of Biochemistry and

15 Molecular Biology, University of Texas Medical Branch, Galveston, TX 77555-0645

17 *Contact: correspondence and request for materials should be sent to Jason A. Somarelli at

18 [email protected]

25 Keywords: epithelial plasticity, E-cadherin, gene regulatory networks, MET, ZEB1, GRHL2

26 osteosarcoma, rhabdomyosarcoma

27 ABSTRACT

28 Phenotypic plasticity involves a series of events through which cells transiently acquire

29 phenotypic traits of another lineage. Two of the most commonly studied types of phenotypic Downloaded from 30 plasticity are epithelial-mesenchymal transition (EMT) and mesenchymal-epithelial transition

31 (MET). In carcinomas, EMT drives invasion and metastatic dissemination while MET is

32 proposed to play a role in metastatic colonization. In sarcomas, phenotypic plasticity is not well-

33 studied; however, there is evidence that sarcomas do exhibit phenotypic plasticity, with a subset http://mcb.asm.org/ 34 of sarcomas undergoing an MET-like phenomenon. While the exact mechanisms by which

35 these transitions occur remain largely unknown, it is likely that some of the same master

36 regulators that drive EMT/MET in carcinomas also act in sarcomas. Here, we combine

37 mathematical models with bench experiments to develop a predictive framework to identify a

38 core regulatory circuit that controls MET in sarcomas. This circuit comprises the miR-200 family, on August 1, 2016 by DUKE MEDICAL LIBRARY

39 ZEB1, and GRHL2. Interestingly, we find that combined expression of miR-200s and GRHL2

40 further upregulates epithelial genes to induce MET. This effect is phenocopied by

41 downregulation of either ZEB1 or the ZEB1 co-factor, BRG1. In addition, an MET gene

42 expression signature is prognostic for improved overall survival in sarcoma patients. Together,

43 our results suggest that a miR-200, ZEB1, GRHL2 gene regulatory network may drive sarcoma

44 cells to a more epithelial-like state and this likely has prognostic relevance.

54 INTRODUCTION

55 Phenotypic plasticity is defined as the reversible conversion of cellular phenotypes from one Downloaded from 56 state to another. The two most commonly-studied types of plasticity are epithelial-mesenchymal

57 transition (EMT) and the reverse, mesenchymal-epithelial transition (MET). These phenotypic

58 transitions play important roles in normal development (reviewed in (1-4)) and wound healing

59 (reviewed in (5)); however, similar pathways and gene expression programs can also be co- http://mcb.asm.org/ 60 opted by the cell during fibrosis (reviewed in (6-8)) and carcinoma progression (reviewed in (9-

61 12)). In the context of carcinoma progression, a subset of cells within the tumor are thought to

62 undergo an EMT, which enables those cells to break free from the tumor mass via loss of cell-

63 cell adhesions (13, 14) and upregulate invasive programs that facilitate dissemination (13). In

64 addition to these phenotypic changes, EMT also contributes to alterations in cancer cell on August 1, 2016 by DUKE MEDICAL LIBRARY

65 metabolism (10), drug resistance (15, 16), tumor-initiation ability (17, 18) and perhaps even host

66 immune evasion (19). EMT is often accompanied by downregulation of proliferation (20, 21),

67 and, in some cases, MET is important for re-initiation of proliferation during metastatic

68 colonization (22). It is important to note that as the field of phenotypic plasticity has matured,

69 particularly in the context of carcinoma progression, EMT/MET has become recognized as more

70 of a spectrum of phenotypes, rather than discrete states of fully differentiated epithelial and

71 mesenchymal phenotypes. These metastable, or hybrid, E/M transition states have been

72 observed in clinical specimens, including the circulation (23, 24) and the metastatic niche

73 (reviewed in (25)). Yet, despite these observations, the importance of the metastable E/M state

74 in metastasis remains unknown.

75 The relevance of phenotypic plasticity in carcinoma progression has been studied for

76 decades now as a critical pathway in the metastatic cascade; however, more recent

77 observations suggest that some of the same drivers of plasticity in carcinomas may also

78 contribute to sarcoma aggressiveness. For example, the EMT transcription factor and master

79 regulator, SNAIL (SNAI1), was shown to be associated with poorer overall survival in human

80 sarcomas, and ectopic expression of SNAIL in fibroblasts induced sarcomas in immunodeficient

81 mice (26). Similarly, ZEB1, another EMT transcription factor, was upregulated in osteosarcomas Downloaded from 82 compared to normal bone, and osteosarcoma patients with metastases had higher ZEB1

83 compared to those without metastases (27). In addition, sarcoma patients with higher levels of

84 the epithelial marker, E-cadherin, have improved survival compared to those with low or no E-

85 cadherin (28). It is perhaps not surprising that just as epithelial-derived cancers are capable of http://mcb.asm.org/ 86 awakening the embryologic pathways related to phenotypic plasticity, so too could

87 mesenchymal cancers transition to a more epithelial-like phenotype.

88 In the present work, we use both predictive mathematical models and validation

89 experiments to develop a conceptual framework for the control of cellular phenotype in

90 sarcomas. Our results predict that the combined expression of the miR-200 family and on August 1, 2016 by DUKE MEDICAL LIBRARY

91 upregulation of an epithelial gene activator, GRHL2, drive MET in sarcomas. Indeed, both

92 GRHL2 overexpression and down-regulation of ZEB1 – by either RNAi-mediated silencing or

93 miR-200 overexpression – act synergistically to control the upregulation of epithelial genes,

94 including E-cadherin, and consequently, MET. Together, our results highlight the functional

95 interplay between epithelial and mesenchymal regulators in driving MET in sarcomas.

100

101

102

103

104

105 RESULTS

106 Modeling an EMT/MET regulatory network

107 E-cadherin is a key protein involved in epithelial cell-cell adhesion, and the loss of E-cadherin is Downloaded from 108 a hallmark of EMT. We used experimentally-derived information based on published literature to

109 construct a gene regulatory network based on the known mediators of E-cadherin and EMT

110 (see Materials and Methods and Supplementary text for a detailed description of the model

111 construction). The mutually inhibitory feedback loop between microRNA family miR-200 and http://mcb.asm.org/ 112 transcription factor family, ZEB1/2 (29, 30) has been proposed to act as a three-way decision-

113 making switch, enabling the co-existence of three phenotypes – epithelial (high miR-200, low

114 ZEB), hybrid epithelial/mesenchymal (moderate miR-200, moderate ZEB), and mesenchymal

115 (low miR-200, high ZEB) (31). Our model couples this miR200/ZEB loop with another mutually

116 inhibitory loop including miR-34/SNAIL (32) by incorporating their interconnections – activation on August 1, 2016 by DUKE MEDICAL LIBRARY

117 of ZEB by SNAIL (33), inhibition of miR-200 by SNAIL (30, 34), and inhibition of miR-34 by ZEB

118 (35). We treat SNAIL as the external input signal to the miR-200/ZEB circuit (Figure 1A) as

119 these interconnections do not change the qualitative behavior of EMT decision-making (31).

120 The miR-200 family consists of two subgroups – one containing miR-141 and miR-200a,

121 and the other includes miR-200b, miR-200c, and miR-429. The ZEB transcription factor family

122 includes two orthologues – ZEB1 and ZEB2. The 3’ UTR region of ZEB1 has a total of eight

123 conserved binding sites for miR-200 - three for the first subgroup and five for the second

124 subgroup. The 3’ UTR region of ZEB2 has nine conserved binding sites for miR-200 (three for

125 the first subgroup and six for the second subgroup). Experiments suggest that the expression of

126 miR-200c can restore E-cadherin and induce MET (36). Therefore, in our miR-200/ZEB module,

127 we considered six binding sites (number of binding sites of the second subgroup on ZEB2) on

128 the 3’ UTR region of ZEB for the binding of miR-200. Also, the members of miR-200 family are

129 located on two different chromosomes – miR-200c and miR-141 on chromosome 12 (with three

130 conserved ZEB-binding sites); and miR-200b, miR-200a, and miR-141 on chromosome 1 (with

131 two conserved ZEB-binding sites). We consider three ZEB binding sites on the promoter region

132 of miR-200 (29, 30). ZEB can also activate its own transcription both via stabilizing SMAD

133 complexes (37) and by activating CD44s (38); hence, we assume two binding sites for ZEB self- Downloaded from 134 activation. SNAIL, a well-known EMT inducer, can transcriptionally inhibit miR-200 and activate

135 ZEB (30, 33, 34). Both of these regulatory steps have been postulated to happen through two

136 binding sites (31).

137 GRHL2, a key regulator of morphogenesis (39), forms a mutually inhibitory loop with http://mcb.asm.org/ 138 ZEB. The promoter region of GRHL2 has three binding sites for ZEB1, and the promoter of

139 ZEB1 has one binding site for GRHL2 (40, 41). GRHL2 can also activate E-cadherin (42), but E-

140 cadherin is repressed by ZEB1 via binding to E-boxes in its promoter region (Figure 1A) (43).

141 Also, E-cadherin sequesters β-catenin at the membrane which can translocate to the nucleus

142 and activate ZEB via binding to TCF/LEF (44). This cascade has been represented as an on August 1, 2016 by DUKE MEDICAL LIBRARY

143 inhibitory link from E-cadherin to ZEB. Thus, GRHL2 acts in two ways to upregulate E-cadherin:

144 1) GRHL2 represses the E-cadherin repressor, ZEB1 (45); and 2) GRHL2 directly activates E-

145 cadherin transcription (42). This model predicted that activating GRHL2 would increase E-

146 cadherin expression to levels commensurate with MET (Figure 1B).

147

148 Combined expression of GRHL2 and miR-200s increases E-cadherin and induces MET in

149 sarcoma cells

150 Based on our model, we sought to experimentally validate the importance of GRHL2 in driving

151 MET and E-cadherin expression. We selected several model cell lines to better understand the

152 potential ability for GRHL2 to influence epithelial plasticity in cells from different lineages (e.g.

153 epithelial vs. mesenchymal). The MDA-MB-231 cell line is a post-EMT breast cancer model that

154 is frequently used in studies of epithelial plasticity. RD and 143B cells represent models of soft

155 tissue sarcoma (specifically, rhabdomyosarcoma) and osteosarcoma, respectively. To test the

156 capacity of GRHL2 to activate E-cadherin and induce MET, we ectopically over-expressed

157 GRHL2 using lentiviral transduction in each of the three cell lines (Supplementary Figure 1).

158 GRHL2 was expressed to similar levels in all three cell lines, as quantified by RT-qPCR

159 (Supplementary Figure 1A) and localized to the nucleus in all three lines (Supplementary Figure Downloaded from 160 1B). Surprisingly, GRHL2 expression had no significant effect on ZEB1 in any of the cell lines

161 (Figure 1C); however, GRHL2 expression led to a relatively modest, but significant induction of

162 E-cadherin mRNA (Figure 1D) and protein (Figure 1E) in MDA-MB-231 cells. Conversely,

163 GRHL2 expression had no effect on E-cadherin in either of the sarcoma lines (Figure 1D and http://mcb.asm.org/ 164 E). We postulated that GRHL2 expression was insufficient to overcome the repression of E-

165 cadherin by other factors. RT-qPCR of five major EMT-TFs in RD cells revealed that the E-

166 cadherin transcriptional repressor and EMT master regulator, ZEB1, was expressed at least 35-

167 fold higher than the other EMT-TFs (Figure 2A). This prompted us to ask whether repression of

168 ZEB1 by the miR-200 family was able to enhance the effect of GRHL2 on E-cadherin and MET on August 1, 2016 by DUKE MEDICAL LIBRARY

169 in RD cells. Transfection of miR-200 family members miR-200a, b, and c (hereafter referred to

170 as ‘miR-200s’) had no obvious effect on the morphology of RD cells. However, addition of miR-

171 200s in the context of GRHL2 expression led to a morphological change consistent with MET;

172 cells changed from a spindle-shaped appearance with few cell-cell contacts to a rounded,

173 cobblestone-like appearance, with increased cell-cell contacts (Figure 2B). The morphological

174 change was also accompanied by upregulation of epithelial markers E-cadherin (Figure 2C),

175 EpCAM (Supplementary Figure 2A), and TJP1 (also known as zona occludens 1, or ZO-1) at

176 cell-cell contacts (Supplementary Figure 2B, white arrows). Interestingly, E-cadherin was

177 upregulated with addition of miR-200s, but the effect of miR-200s on E-cadherin was enhanced

178 in a synergistic fashion in the presence of GRHL2 over-expression (Figure 2D; note log scale).

179 Although GRHL2 had no effect on ZEB1, transfection of miR-200s reduced ZEB1 mRNA (Figure

180 2E). To test whether the synergistic effect of GRHL2 and miR-200s was due to sequence

181 elements within the E-cadherin gene, we created a luciferase-based reporter in which firefly

182 luciferase (FFLuc) is driven by the E-cadherin promoter. In addition to promoter control, the

183 FFLuc open reading frame is interrupted by a constitutively spliced intron, modified from (46),

184 harboring a region of E-cadherin intron 2 in which GRHL2 has been demonstrated to bind (42)

185 (Figure 2F). Consistent with the results from the endogenous gene, transfection of this E- Downloaded from 186 cadherin/GRHL2 reporter into RD cells in the presence of GRHL2 and miR-200s led to a

187 synergistic increase in FFLuc expression (Figure 2F). Soft agar assays revealed that MET

188 induction in RD cells significantly reduced the number of colonies capable of growing in an

189 anchorage-independent setting (Figure 2G). This MET-associated reduction in anchorage- http://mcb.asm.org/ 190 independent growth was driven exclusively by the miR-200s (Figure 2G).

191 Similar to the RD cells, the 143B osteosarcoma cell line also expressed high levels of

192 ZEB1 compared to other EMT-TFs (Figure 2H), and combined transfection of miR-200s with

193 GRHL2 expression led to a multiplicative increase in E-cadherin (Figure 2I). Also similar to the

194 results in RD cells, GRHL2 alone had no effect on ZEB1, and miR-200s inhibited ZEB1 (Figure on August 1, 2016 by DUKE MEDICAL LIBRARY

195 2J). Conversely, in MDA-MB-231 breast cancer cells, which have almost equal expression of

196 both Zeb1 and Slug (Supplementary Figure 3A), overexpression of miR-200 was sufficient to

197 induce E-cadherin, EpCAM, CGN, and downregulation of ZEB1. GRHL2, either transfected

198 alone or in combination with miR-200s, had only a modest effect on the expression of E-

199 cadherin, ZEB1, or other MET markers (Supplementary Figure 3) in MDA-MB-231 cells. We

200 also tested whether combined expression of GRHL2 and miR-200s induced MET using three

201 additional lines, including BT-549 triple negative breast cancer cells (Supplementary Figure 4),

202 U2-OS osteosarcoma cells (Supplementary Figure 5), and SK-LMS leiomyosarcoma cells

203 (Supplementary Figure 6). Overall, the results were consistent with an increased effect on MET

204 induction with combined expression of GRHL2 and miR-200s, with the miR-200s driving the

205 majority of MET-like gene expression changes (Supplementary Figures 4-6).

206

207 ZEB1 repression and GRHL2 expression induce E-cadherin expression

208 Given the relatively broad effects of miR-200s on potentially hundreds of targets, we next asked

209 whether the combinatorial effect we observed on E-cadherin and MET in sarcoma cells was

210 through ZEB1. To this end, we performed siRNA-mediated knockdown of ZEB1 using two Downloaded from 211 independent siRNAs to mitigate the potential for off-target effects. Both siRNAs significantly

212 reduced ZEB1 protein (Figure 3A) and mRNA (Figure 3B) in the RD cells. Consistent with the

213 hypothesis that miR-200s repressed ZEB1 to facilitate GRHL2 activation of E-cadherin, we

214 observed a larger increase in E-cadherin with ZEB1 knockdown and GRHL2 expression than http://mcb.asm.org/ 215 with either factor alone (Figure 3C). Similarly, we also found this combinatorial upregulation of

216 another epithelial cell-cell adhesion molecule, claudin 4 (CLDN4), with ZEB1 knockdown and

217 GRHL2 expression (Figure 3D). Conversely, a third epithelial cell adhesion molecule, cingulin

218 (CGN) appeared to be solely regulated by GRHL2; knockdown of ZEB1 had no effect (Figure

219 3E). Together, our results suggest that downregulating ZEB1 via miR-200s has no effect on any on August 1, 2016 by DUKE MEDICAL LIBRARY

220 of these epithelial genes without the presence of an epithelial gene activator, GRHL2.

221

222 A revised model predicts GRHL2 exists in two possible states

223 Based on our observed experimental results, we revised our predictive model. In the revised

224 model, we parsed GRHL2 into two categories: inactive (i) and active (a) forms. Although we

225 refer to the inactive and active forms as conceptual constructions to view the potential states of

226 GRHL2, these states can potentially reflect two scenarios – GRHL2 being blocked from

227 activating E-cadherin (inactive) or GRHL2 capable of activating E-cadherin (active). We

228 constructed the model so that, rather than inhibiting GRHL2 levels directly, ZEB1 prevents

229 GRHL2 from transitioning to the active state (Figure 4A). This realization is consistent with our

230 experimental data that shows upregulation of epithelial gene targets by GRHL2 only upon ZEB1

231 knockdown (Figure 3C and D). Importantly, the revised model recapitulates the synergistic

232 effect of miR-200s and GRHL2 on E-cadherin. GRHL2 has little effect on E-cadherin (Figure 4B,

233 blue line and quantified in Figure 4C) while addition of miR-200s alone increases E-cadherin

234 (Figure 4B, red line and quantified in Figure 4C). Overexpression of both GRHL2 and miR-200s

235 is predicted to enhance E-cadherin expression in a synergistic fashion (Figure 4B, green line

236 and quantified in Figure 4C). When we plot various degrees of overexpression of GRHL2 and Downloaded from 237 miR-200s on the x- and y-axes, respectively, we observe the largest effect on E-cadherin levels

238 with the activation of miR-200s, while GRHL2 alone contributes very little to E-cadherin

239 activation (Figure 4D). Yet, the effect of GRHL2 in increasing E-cadherin is visible only once

240 miR-200s reach a threshold value and activate E-cadherin (Figure 4D). While the exact http://mcb.asm.org/ 241 numerical values of fold-change of levels of E-cadherin depend on changes in model

242 parameters, the relatively small effect of GRHL2 in activating E-cadherin in the absence of miR-

243 200s and the synergistic induction of E-cadherin levels upon combinatorial overexpression of

244 miR-200s and GRHL2 are robust features of the model (Supplementary Text). All of these

245 observations are consistent with our experimental data, suggesting that the modeled regulatory on August 1, 2016 by DUKE MEDICAL LIBRARY

246 network is capable of generating biologically-relevant outputs.

247

248 GRHL2-mediated regulation of E-cadherin and cellular phenotype depends on BRG1 and

249 GRHL2 methylation

250 The revised predictive model prompted us to hypothesize that regulation of E-cadherin by

251 GRHL2 could be controlled in one of two ways as follows: 1) GRHL2-mediated activation of E-

252 cadherin and/or 2) GRHL2 expression levels. To address the first hypothesis, we knocked down

253 the chromatin remodeling protein, BRG1, which has been shown to act as a co-factor of ZEB1

254 in repressing E-cadherin by condensing the chromatin around the E-cadherin promoter (43). We

255 knocked down BRG1 using two independent siRNAs in RD cells containing either an empty

256 vector or GRHL2 (Figure 5A). While siRNA-mediated silencing of BRG1 had no effect on E-

257 cadherin expression in empty vector cells, knock down of BRG1 in the context of GRHL2

258 expression led to a significant increase in E-cadherin mRNA (Figure 5B). The increase in E-

259 cadherin was also dose dependent; E-cadherin was higher in cells for which BRG1 was

260 knocked down to a greater degree (Figure 5B). These results suggest that GRHL2-based

261 activation of E-cadherin may be dependent on the degree to which the chromatin surrounding

262 the E-cadherin promoter is available to be bound by GRHL2. Downloaded from 263 To test the second hypothesis that GRHL2 itself could be regulated as a mechanism to

264 control E-cadherin expression and cellular phenotype, we queried data from The Cancer

265 Genome Atlas (TCGA). Interestingly, we found that methylation of the DNA upstream of GRHL2

266 was significantly higher in mesenchymal-derived sarcoma cancer specimens compared to http://mcb.asm.org/ 267 epithelial-derived breast, colorectal (CRC) and prostate cancers (Figure 5C). When we further

268 analyzed GRHL2 methylation by sarcoma subtype, we found that GRHL2 methylation was

269 lowest in synovial sarcomas (Figure 5D). Synovial sarcomas are classically characterized by a

270 biphasic histologic appearance consisting of a distinct cellular population with mesenchymal

271 morphology (and expressing mesenchymal biomarkers) and another distinct population with on August 1, 2016 by DUKE MEDICAL LIBRARY

272 epithelial morphology (and expressing epithelial biomarkers). In addition, when we used a

273 previously published signature of EMT (47) to separate sarcomas into more epithelial-like and

274 mesenchymal-like samples, we found that GRHL2 methylation was significantly higher in the

275 more mesenchymal-like samples (Supplementary Figure 7). While this was not the case for

276 breast and prostate cancers, GRHL2 methylation was also significantly higher in mesenchymal-

277 like CRC samples compared to epithelial-like samples (Supplementary Figure 7). These

278 analyses suggest that GRHL2 methylation patterns differ based on their cellular phenotype.

279

280 An epithelial-like gene expression profile has prognostic value in sarcomas

281 Our results suggest that some of the same gene expression pathways that control

282 EMT/MET in carcinomas may also regulate these phenotypic transitions in sarcomas. To better

283 understand the clinical significance of these phenotypic plasticity pathways in sarcomas, we

284 analyzed RNA-Seq data from 250 soft tissue sarcomas from patients available through The

285 Cancer Genome Atlas. Tumors were separated into two groups based on a previously

286 published EMT signature (47). Partitioning the tumors in this way revealed a significant

287 difference in patient prognosis based on EMT signature; patients with a low EMT score (more

288 epithelial-like tumors) had a more favorable prognosis compared to patients with a high EMT Downloaded from 289 score (more mesenchymal-like tumors) (Figure 6A). Similarly, the sum of RNA-Seq scores from

290 GRHL2 and miR200 family members was prognostic for overall survival, albeit with modest

291 significance, with higher levels of GRHL2 and miR200s having a more favorable prognosis

292 (Figure 6B). As expected, the combined RNA-Seq values from GRHL2 and miR200s negatively http://mcb.asm.org/ 293 correlated with EMT score (Figure 6C). Analysis of GRHL2 and ZEB1 expression in an

294 independent osteosarcoma microarray data set (n = 84; GSE28974; (48)) showed that patients

295 with increased GRHL2 expression had significantly improved metastasis-free survival and

296 trended toward improved overall survival (Supplementary Figure 8A and B). Conversely,

297 patients with higher ZEB1 trended toward having worse metastasis-free survival (p = 0.056) and on August 1, 2016 by DUKE MEDICAL LIBRARY

298 significantly worse overall survival (p = 0.032) (Supplementary Figure 8C and D). Together,

299 these data indicate that a more epithelial-like gene expression program may be prognostic for

300 improved clinical outcomes in sarcomas. It is worth noting that although more epithelial-like soft

301 tissue sarcoma samples were evident within the group of soft tissue sarcomas (Supplementary

302 Figure 9A), the EMT biomarker expression and EMT scores were strongly mesenchymal-like

303 when normalized to all cancers (PANCAN normalized) within the TCGA dataset (Supplementary

304 Figure 9B). These observations suggest that the more “epithelial-like” sarcomas, while being

305 shifted more toward an epithelial phenotype, still maintain the signature of a mesenchymal-like

306 lineage when compared to epithelial-derived cancers (Supplementary Figure 9).

307

308 DISCUSSION

309 Maintenance of cellular phenotype is critical to the proper homeostasis of multicellular

310 organisms. For proper division of labor to take place within tissues and organs, each cell must

311 operate within the context of its lineage. The breakdown of this lineage commitment can lead to

312 dedifferentiation, trans-differentiation, and/or other events related to phenotypic plasticity, all

313 phenomena that are observed across multiple cancer types and that correlate with poor

314 prognosis. In the context of sarcoma, phenotypic plasticity appears to have varying associations Downloaded from 315 with clinical aggression. For example, dedifferentiated liposarcomas are particularly aggressive

316 and have a high rate of metastasis compared to their well differentiated and intermediate grade

317 counterparts (49). Similarly, others have argued that sarcomas arise from reprogramming of

318 mesenchymal stem-like cells that become fixed in a more undifferentiated state by oncogenes http://mcb.asm.org/ 319 (50). Conversely, in the case of MET, it appears that the transition to a more epithelial-like gene

320 expression program is associated with a favorable prognosis in a subset of soft tissue sarcomas

321 and osteosarcomas (26-28, 51, 52). Interestingly, our computational model and laboratory

322 experiments suggest that sarcoma cells may require multiple inputs to undergo a phenotypic

323 transition to a more epithelial-like state. It is possible that, under normal conditions, this multi- on August 1, 2016 by DUKE MEDICAL LIBRARY

324 level regulation limits phenotypic transitions and provides a system to buffer noise from internal

325 or external cues that feed into the transcriptional circuit. Only when the signals are strong

326 enough to activate multiple factors is the phenotypic transition able to occur. More intriguing is

327 the question of what factors would favor phenotypic transitions in sarcomas. In both carcinomas

328 and sarcomas, presumably a more mesenchymal-like state would favor the fitness of the cancer

329 cell, especially in the setting of competition for limited local resources, where such a state would

330 permit metastasis and colonization of a new, less crowded niche (10-12). What, if any, selective

331 pressures would favor MET by sarcomas have yet to be explored. It is possible that such

332 transitions in sarcomas lack the same selective pressures in carcinomas or even that such

333 transitions reflect innate biologic traits of particular tumors and hinder those tumors from lethal

334 behaviors.

335 One of the mechanisms by which a cell can be poised toward phenotypic plasticity is

336 through epigenetic modifications (53). Along these lines, differentiated mammary cells with high

337 levels of E-cadherin carry an active H3K4me3 modification on the E-cadherin promoter while

338 the more stem-like cells harbor both the active H3K4me3 mark and a repressive H3K27me3

339 modification (53). As Tam and Weinberg postulate, it is logical to conclude that the bivalent

340 modifications enable a rapid transition to a differentiated, epithelial-like state (53). Indeed, the Downloaded from 341 fact that GRHL2 methylation is significantly increased in sarcomas compared to epithelial-

342 derived cancers supports this notion. Similarly, our results showing that downregulation of the

343 chromatin remodeling protein BRG1 phenocopies the effect of ZEB1 knockdown on E-cadherin

344 expression (Figure 5B) also fits with the idea that the chromatin conformation may dictate where http://mcb.asm.org/ 345 on the spectrum of phenotypes a given cell may exist. Further study of the relationship between

346 epigenetic modification, phenotype, and capacity for phenotypic plasticity in sarcomas may

347 elucidate the context of phenotypic plasticity among mesenchymal tumors.

348 Our coupling of dynamic modeling with experimentally-based refinements and

349 validations has enabled the establishment of a framework by which to understand regulation of on August 1, 2016 by DUKE MEDICAL LIBRARY

350 MET in sarcomas, a phenomenon that appears to have clinical relevance (28, 51, 52). A deeper

351 understanding of how these transitions occur in sarcoma and what selective pressures influence

352 them could offer immediate prognostic value and, one day, lead to therapeutically shifting

353 sarcomas to a more epithelial-like state, thereby attenuating their aggressiveness and improving

354 patient outcomes.

355

356 MATERIALS AND METHODS

357 Model construction and sensitivity analysis 358 359 We generalize the modeling framework of miRNA-mRNA interactions developed for miR-

360 200/ZEB/SNAIL circuit by Lu et al. (31) to include the connections with E-cadherin and GRHL2.

361 The initial model (Figure 1A) consists of eight species – SNAIL mRNA, SNAIL protein, ZEB

362 mRNA, ZEB protein, the miR-200 family (miR-200a, b, c, 141, 429), E-cadherin mRNA, GRHL2

363 mRNA, and GRHL2 protein. In the revised model (Figure 4A), we introduce two species to

364 consider the active and inactive forms of GRHL2 protein. The model construction and parameter

365 estimation details and the sensitivity analysis are given in Supplementary text. The model was

366 simulated in MATLAB (Mathworks Inc.), and bifurcation plots were constructed using

367 MATCONT (54). The term 'relative activity' in Figures 1B, 4B, 4D represents the external Downloaded from 368 activation that was applied on respective moieties.

369

370 Cell culture and GRHL2 transduction

371 All cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM), high glucose http://mcb.asm.org/ 372 supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (pen-strep) at

5 373 37°C and 5% CO2 in a humidified incubator. To generate lentiviruses, 3 x 10 HEK293T cells

374 were co-transfected with either 2 μg of empty vector (pCMV-UBC-EGFP) or pCMV-GRHL2-

375 UBC-EGFP, 1.8 μg of pΔ8.9 and 0.2 μg of pGAG-POL. Viral media was collected two and three

376 days later, filtered, and added to target cells, along with X ug/ml X. EGFP+ cells were sorted by on August 1, 2016 by DUKE MEDICAL LIBRARY

377 flow cytometry at the Duke University Flow Cytometry Shared Resource.

378

379 SiRNA-mediated knockdowns and miR-200 transfections

380 A total of 20 nM of Allstars non-silencing (Qiagen) or target gene siRNA (Qiagen) was diluted in

381 Opti-MEM and mixed with Lipofectamine RNAiMax diluted in Opti-MEM. Lipofectamine-siRNA

382 mixtures were incubated for 20 minutes at room temperature, and cells were reverse

383 transfected overnight in DMEM/10% FBS/1% pen-strep. Media was replaced the next day. The

384 same protocol was used for reverse transfection of miR-200s.

385

386 RNA extraction, reverse transcription, and RT-qPCR

387 RNA was extracted using either the Promega RNA Mini-prep kit or the Zymo RNA extraction kit

388 following the manufacturers’ protocols. A minimum of 100 ng of total RNA was used in each

389 reverse transfection reaction using random hexamer primers in a 20 μl total reaction volume

390 following the ABI High Capacity cDNA Reverse Transcription Kit protocol (ThermoFisher). RT

391 reactions were incubated following the manufacturer’s protocol in a SimpliAmp thermocycler

392 (Life Technologies). RT reactions were diluted 1:5 in nuclease-free H2O, and RT-qPCR was

393 performed using 2 μl of RT, 0.06 μl of each primer (10 μM stock), and 5 μl of KAPA SYBRFAST Downloaded from 394 Universal 2X qPCR Master Mix in a 10 μl total reaction volume. Reactions were performed

395 using a Vii7 real time-PCR detection system (Applied Biosystems).

396

397 Western blotting and immunofluorescence staining http://mcb.asm.org/ 398 For western blotting, total protein was extracted from cultured cells by washing twice in

399 1X ice-cold phosphate buffered saline (PBS), and lysing cells in 1X RIPA buffer supplemented

400 with 1X protease inhibitor cocktail (Roche). Cell lysates were incubated at 4C for 15’ with

401 rocking, clarified by centrifugation at high speed in a benchtop centrifuge for 5’, and clarified

402 lysates were incubated for 5’ at 95C in 1X Laemmli sample loading buffer. Lysates were on August 1, 2016 by DUKE MEDICAL LIBRARY

403 separated in 4-12% NuPAGE Novex Bis-Tris gels (ThermoFisher). Proteins were transferred

404 onto nitrocellulose membranes (GE Healthcare Life Sciences) in 1X NuPAGE Transfer Buffer

405 (ThermoFisher) at 50V for 1-2 hours at room temperature. Membranes were blocked for one

406 hour at room temperature or overnight at 4C in Starting Block (ThermoFisher). Primary

407 antibodies were diluted in Starting Block, incubated for one hour at room temperature or

408 overnight at 4C, washed twice for five minutes in 1X PBS with 0.05% Tween-20 (PBS-T), and

409 incubated with IRDye-coupled donkey anti-rabbit or donkey anti-mouse secondary antibodies

410 diluted 1:20,000 in Starting Block. Images were captured on an Odyssey Fc imager (LI-COR).

411 For immunofluorescence imaging, cells were fixed for 15’ at room temperature in 4%

412 paraformaldehyde, permeabilized for 30’ at room temperature in PBS with 0.2% Triton X-100,

413 blocked for 30’ at room temperature in 5% bovine serum albumin (BSA) diluted in PBS, and

414 incubated overnight at 4C in primary antibodies diluted in BSA/PBS. Subsequent to washing

415 with PBS, cells were incubated with AlexaFluor secondary antibodies (1:2000 dilution in

416 BSA/PBS) along with 1 μg/ml of Hoechst dye (Sigma) for one hour at room temperature in the

417 dark. Images were captured on an inverted Olympus IX 71 epifluorescence microscope at 200X

418 total magnification unless otherwise specified in the figure legends. A complete list of primary

419 antibodies and their dilutions is provided in the Supplementary Text. Downloaded from

420

421 TCGA Analysis of Methylation Data

422 GRHL2 and E-cadherin methylation (DNA Methylation 450K) data were downloaded from the

423 UCSC Cancer Genomics Browser (https://genome-cancer.soe.ucsc.edu/). Boxplots were http://mcb.asm.org/

424 created using RStudio (version 0.98.1091). To ask whether GRHL2 or E-cadherin methylation

425 were significantly different in epithelial-like or mesenchymal-like patient samples, each patient

426 sample was assigned a value of “epithelial-like” or “mesenchymal-like” by parsing samples

427 based on their RNA-Seq data using a previously published EMT signature (24302555). The on August 1, 2016 by DUKE MEDICAL LIBRARY 428 EMT signature is as follows: (FN1 + VIM + ZEB1 + ZEB2 + TWIST1 + TWIST2 + SNAI1 +

429 SNAI2 + CDH2) - (CLDN4 + CLDN7 + TJP3 + MUC1 + CDH1). Kaplan-Meier curves were

430 created using the KM function, and the value of the upper or lower curve was used to set the E-

431 like vs. M-like state. Any sample above the value for the line on the Kaplan-Meier curve

432 corresponding to the E-like state was considered “mesenchymal-like” while any sample equal to

433 or below the value for the E-like state was considered “epithelial-like”.

434

435 Analysis of EMT gene expression profiles and clinical outcomes

436 TCGA RNA-Seq data from n = 250 soft tissue sarcomas was downloaded using the UCSC

437 Cancer Genomics Browser. Samples were designated “mesenchymal-like” or “epithelial-like”

438 using the upper and lower values on the KM curve function, as described above, and Kaplan-

439 Meier curves were generated in JMP Pro 12. Kaplan-Meier curves for osteosarcoma microarray

440 data were generated using the Kaplan-Meier plot function in R2: Genomics Analysis and

441 Visualization Platform (http://hgserver1.amc.nl/cgi-bin/r2/main.cgi) with the ‘Kaplan scan a

442 single gene’ function and the ‘cutoff_modus’ set to scan.

443 Downloaded from 444 Statistical analyses

445 All data were analyzed for statistically significant differences using the Student’s t-test (for two

446 comparisons) or analysis of variance (for multiple comparisons) in JMP Pro 12. Any p-value <

447 0.05 was considered statistically significant. http://mcb.asm.org/ 448

449

450

451

452 on August 1, 2016 by DUKE MEDICAL LIBRARY

453

454

455

456

457

458

459

460

461

462

463

464

465

466

467 REFERENCES

468 1. Nakaya Y, Sheng G. 2013. EMT in developmental morphogenesis. Cancer Lett 341:9- 469 15.

470 2. Lim J, Thiery JP. 2012. Epithelial-mesenchymal transitions: insights from development. Downloaded from 471 Development 139:3471-3486. 472 3. Thiery JP, Acloque H, Huang RY, Nieto MA. 2009. Epithelial-mesenchymal transitions 473 in development and disease. Cell 139:871-890. 474 4. Thiery JP, Sleeman JP. 2006. Complex networks orchestrate epithelial-mesenchymal 475 transitions. Nat Rev Mol Cell Biol 7:131-142. 476 5. Weber CE, Li NY, Wai PY, Kuo PC. 2012. Epithelial-mesenchymal transition, TGF- 477 beta, and osteopontin in wound healing and tissue remodeling after injury. J Burn Care 478 Res 33:311-318. 479 6. Galichon P, Finianos S, Hertig A. 2013. EMT-MET in renal disease: should we curb http://mcb.asm.org/ 480 our enthusiasm? Cancer Lett 341:24-29. 481 7. Carew RM, Wang B, Kantharidis P. 2012. The role of EMT in renal fibrosis. Cell Tissue 482 Res 347:103-116. 483 8. Willis BC, Borok Z. 2007. TGF-beta-induced EMT: mechanisms and implications for 484 fibrotic lung disease. Am J Physiol Lung Cell Mol Physiol 293:L525-534. 485 9. Ye X, Weinberg RA. 2015. Epithelial-Mesenchymal Plasticity: A Central Regulator of 486 Cancer Progression. Trends Cell Biol 25:675-686. 487 10. Li L, Li W. 2015. Epithelial-mesenchymal transition in human cancer: comprehensive on August 1, 2016 by DUKE MEDICAL LIBRARY 488 reprogramming of metabolism, epigenetics, and differentiation. Pharmacol Ther 150:33- 489 46. 490 11. Tsai JH, Yang J. 2013. Epithelial-mesenchymal plasticity in carcinoma metastasis. 491 Genes Dev 27:2192-2206. 492 12. Bitting RL, Schaeffer D, Somarelli JA, Garcia-Blanco MA, Armstrong AJ. 2014. The 493 role of epithelial plasticity in prostate cancer dissemination and treatment resistance. 494 Cancer Metastasis Rev 33:441-468. 495 13. Schaeffer D, Somarelli JA, Hanna G, Palmer GM, Garcia-Blanco MA. 2014. Cellular 496 Migration and Invasion Uncoupled: Increased Migration Is Not an Inexorable 497 Consequence of Epithelial-to-Mesenchymal Transition. Mol Cell Biol 34:3486-3499. 498 14. Mathow D, Chessa F, Rabionet M, Kaden S, Jennemann R, Sandhoff R, Grone HJ, 499 Feuerborn A. 2015. Zeb1 affects epithelial cell adhesion by diverting glycosphingolipid 500 metabolism. EMBO Rep 16:321-331. 501 15. Ware KE, Hinz TK, Kleczko E, Singleton KR, Marek LA, Helfrich BA, Cummings CT, 502 Graham DK, Astling D, Tan AC, Heasley LE. 2013. A mechanism of resistance to 503 gefitinib mediated by cellular reprogramming and the acquisition of an FGF2-FGFR1 504 autocrine growth loop. Oncogenesis 2:e39. 505 16. Yauch RL, Januario T, Eberhard DA, Cavet G, Zhu W, Fu L, Pham TQ, Soriano R, 506 Stinson J, Seshagiri S, Modrusan Z, Lin CY, O'Neill V, Amler LC. 2005. Epithelial 507 versus mesenchymal phenotype determines in vitro sensitivity and predicts clinical 508 activity of erlotinib in lung cancer patients. Clin Cancer Res 11:8686-8698. 509 17. Jolly MK, Huang B, Lu M, Mani SA, Levine H, Ben-Jacob E. 2014. Towards 510 elucidating the connection between epithelial-mesenchymal transitions and stemness. J 511 R Soc Interface 11:20140962. 512 18. Mani SA, Guo W, Liao MJ, Eaton EN, Ayyanan A, Zhou AY, Brooks M, Reinhard F, 513 Zhang CC, Shipitsin M, Campbell LL, Polyak K, Brisken C, Yang J, Weinberg RA. 514 2008. The epithelial-mesenchymal transition generates cells with properties of stem 515 cells. Cell 133:704-715.

516 19. Chen L, Gibbons DL, Goswami S, Cortez MA, Ahn YH, Byers LA, Zhang X, Yi X, 517 Dwyer D, Lin W, Diao L, Wang J, Roybal JD, Patel M, Ungewiss C, Peng D, Antonia 518 S, Mediavilla-Varela M, Robertson G, Jones S, Suraokar M, Welsh JW, Erez B, 519 Wistuba, II, Chen L, Peng D, Wang S, Ullrich SE, Heymach JV, Kurie JM, Qin FX. 520 2014. Metastasis is regulated via microRNA-200/ZEB1 axis control of tumour cell PD-L1 521 expression and intratumoral immunosuppression. Nat Commun 5:5241. Downloaded from 522 20. Vega S, Morales AV, Ocana OH, Valdes F, Fabregat I, Nieto MA. 2004. Snail blocks 523 the cell cycle and confers resistance to cell death. Genes Dev 18:1131-1143. 524 21. Huang SS, Huang JS. 2005. TGF-beta control of cell proliferation. J Cell Biochem 525 96:447-462. 526 22. Tsai JH, Donaher JL, Murphy DA, Chau S, Yang J. 2012. Spatiotemporal regulation of 527 epithelial-mesenchymal transition is essential for squamous cell carcinoma metastasis. 528 Cancer Cell 22:725-736.

529 23. Armstrong AJ, Marengo MS, Oltean S, Kemeny G, Bitting RL, Turnbull JD, Herold http://mcb.asm.org/ 530 CI, Marcom PK, George DJ, Garcia-Blanco MA. 2011. Circulating tumor cells from 531 patients with advanced prostate and breast cancer display both epithelial and 532 mesenchymal markers. Mol Cancer Res 9:997-1007. 533 24. Yu M, Bardia A, Wittner BS, Stott SL, Smas ME, Ting DT, Isakoff SJ, Ciciliano JC, 534 Wells MN, Shah AM, Concannon KF, Donaldson MC, Sequist LV, Brachtel E, Sgroi 535 D, Baselga J, Ramaswamy S, Toner M, Haber DA, Maheswaran S. 2013. Circulating 536 breast tumor cells exhibit dynamic changes in epithelial and mesenchymal composition. 537 Science 339:580-584.

538 25. Jolly MK, Boareto M, Huang B, Jia D, Lu M, Ben-Jacob E, Onuchic JN, Levine H. on August 1, 2016 by DUKE MEDICAL LIBRARY 539 2015. Implications of the Hybrid Epithelial/Mesenchymal Phenotype in Metastasis. Front 540 Oncol 5:155. 541 26. Alba-Castellon L, Batlle R, Franci C, Fernandez-Acenero MJ, Mazzolini R, Pena R, 542 Loubat J, Alameda F, Rodriguez R, Curto J, Albanell J, Munoz A, Bonilla F, Ignacio 543 Casal J, Rojo F, Garcia de Herreros A. 2014. Snail1 expression is required for 544 sarcomagenesis. Neoplasia 16:413-421. 545 27. Shen A, Zhang Y, Yang H, Xu R, Huang G. 2012. Overexpression of ZEB1 relates to 546 metastasis and invasion in osteosarcoma. J Surg Oncol 105:830-834. 547 28. Wang N, He YL, Pang LJ, Zou H, Liu CX, Zhao J, Hu JM, Zhang WJ, Qi Y, Li F. 548 2015. Down-regulated E-cadherin expression is associated with poor five-year overall 549 survival in bone and soft tissue sarcoma: results of a meta-analysis. PLoS One 550 10:e0121448. 551 29. Bracken CP, Gregory PA, Kolesnikoff N, Bert AG, Wang J, Shannon MF, Goodall 552 GJ. 2008. A double-negative feedback loop between ZEB1-SIP1 and the microRNA-200 553 family regulates epithelial-mesenchymal transition. Cancer Res 68:7846-7854. 554 30. Burk U, Schubert J, Wellner U, Schmalhofer O, Vincan E, Spaderna S, Brabletz T. 555 2008. A reciprocal repression between ZEB1 and members of the miR-200 family 556 promotes EMT and invasion in cancer cells. EMBO Rep 9:582-589. 557 31. Lu M, Jolly MK, Levine H, Onuchic JN, Ben-Jacob E. 2013. MicroRNA-based 558 regulation of epithelial-hybrid-mesenchymal fate determination. Proc Natl Acad Sci U S 559 A 110:18144-18149. 560 32. Siemens H, Jackstadt R, Hunten S, Kaller M, Menssen A, Gotz U, Hermeking H. 561 2011. miR-34 and SNAIL form a double-negative feedback loop to regulate epithelial- 562 mesenchymal transitions. Cell Cycle 10:4256-4271. 563 33. Guaita S, Puig I, Franci C, Garrido M, Dominguez D, Batlle E, Sancho E, Dedhar S, 564 De Herreros AG, Baulida J. 2002. Snail induction of epithelial to mesenchymal 565 transition in tumor cells is accompanied by MUC1 repression and ZEB1 expression. J 566 Biol Chem 277:39209-39216.

567 34. Gill JG, Langer EM, Lindsley RC, Cai M, Murphy TL, Kyba M, Murphy KM. 2011. 568 Snail and the microRNA-200 family act in opposition to regulate epithelial-to- 569 mesenchymal transition and germ layer fate restriction in differentiating ESCs. Stem 570 Cells 29:764-776. 571 35. Ahn YH, Gibbons DL, Chakravarti D, Creighton CJ, Rizvi ZH, Adams HP, 572 Pertsemlidis A, Gregory PA, Wright JA, Goodall GJ, Flores ER, Kurie JM. 2012. Downloaded from 573 ZEB1 drives prometastatic actin cytoskeletal remodeling by downregulating miR-34a 574 expression. J Clin Invest 122:3170-3183. 575 36. Hurteau GJ, Carlson JA, Roos E, Brock GJ. 2009. Stable expression of miR-200c 576 alone is sufficient to regulate TCF8 (ZEB1) and restore E-cadherin expression. Cell 577 Cycle 8:2064-2069. 578 37. Hill L, Browne G, Tulchinsky E. 2013. ZEB/miR-200 feedback loop: at the crossroads 579 of signal transduction in cancer. Int J Cancer 132:745-754.

580 38. Preca BT, Bajdak K, Mock K, Sundararajan V, Pfannstiel J, Maurer J, Wellner U, http://mcb.asm.org/ 581 Hopt UT, Brummer T, Brabletz S, Brabletz T, Stemmler MP. 2015. A self-enforcing 582 CD44s/ZEB1 feedback loop maintains EMT and stemness properties in cancer cells. Int 583 J Cancer 137:2566-2577. 584 39. Varma S, Cao Y, Tagne JB, Lakshminarayanan M, Li J, Friedman TB, Morell RJ, 585 Warburton D, Kotton DN, Ramirez MI. 2012. The transcription factors Grainyhead-like 586 2 and NK2-homeobox 1 form a regulatory loop that coordinates lung epithelial cell 587 morphogenesis and differentiation. J Biol Chem 287:37282-37295. 588 40. Cieply B, Riley Pt, Pifer PM, Widmeyer J, Addison JB, Ivanov AV, Denvir J, Frisch

589 SM. 2012. Suppression of the epithelial-mesenchymal transition by Grainyhead-like-2. on August 1, 2016 by DUKE MEDICAL LIBRARY 590 Cancer Res 72:2440-2453. 591 41. Cieply B, Farris J, Denvir J, Ford HL, Frisch SM. 2013. Epithelial-mesenchymal 592 transition and tumor suppression are controlled by a reciprocal feedback loop between 593 ZEB1 and Grainyhead-like-2. Cancer Res 73:6299-6309. 594 42. Werth M, Walentin K, Aue A, Schonheit J, Wuebken A, Pode-Shakked N, 595 Vilianovitch L, Erdmann B, Dekel B, Bader M, Barasch J, Rosenbauer F, Luft FC, 596 Schmidt-Ott KM. 2010. The transcription factor grainyhead-like 2 regulates the 597 molecular composition of the epithelial apical junctional complex. Development 598 137:3835-3845. 599 43. Sanchez-Tillo E, Lazaro A, Torrent R, Cuatrecasas M, Vaquero EC, Castells A, 600 Engel P, Postigo A. 2010. ZEB1 represses E-cadherin and induces an EMT by 601 recruiting the SWI/SNF chromatin-remodeling protein BRG1. Oncogene 29:3490-3500. 602 44. Schmalhofer O, Brabletz S, Brabletz T. 2009. E-cadherin, beta-catenin, and ZEB1 in 603 malignant progression of cancer. Cancer Metastasis Rev 28:151-166. 604 45. Werner S, Frey S, Riethdorf S, Schulze C, Alawi M, Kling L, Vafaizadeh V, Sauter G, 605 Terracciano L, Schumacher U, Pantel K, Assmann V. 2013. Dual roles of the 606 transcription factor grainyhead-like 2 (GRHL2) in breast cancer. J Biol Chem 288:22993- 607 23008. 608 46. Oltean S, Febbo PG, Garcia-Blanco MA. 2008. Dunning rat prostate adenocarcinomas 609 and alternative splicing reporters: powerful tools to study epithelial plasticity in prostate 610 tumors in vivo. Clinical & Experimental Metastasis 25:611-619. 611 47. Salt MB, Bandyopadhyay S, McCormick F. 2014. Epithelial-to-mesenchymal transition 612 rewires the molecular path to PI3K-dependent proliferation. Cancer Discov 4:186-199. 613 48. Kuijjer ML, Rydbeck H, Kresse SH, Buddingh EP, Lid AB, Roelofs H, Burger H, 614 Myklebost O, Hogendoorn PC, Meza-Zepeda LA, Cleton-Jansen AM. 2012. 615 Identification of osteosarcoma driver genes by integrative analysis of copy number and 616 gene expression data. Genes Chromosomes Cancer 51:696-706.

617 49. Crago AM, Singer S. 2011. Clinical and molecular approaches to well differentiated and 618 dedifferentiated liposarcoma. Curr Opin Oncol 23:373-378. 619 50. Eid JE, Garcia CB. 2015. Reprogramming of mesenchymal stem cells by oncogenes. 620 Semin Cancer Biol 32:18-31. 621 51. Yin K, Liao Q, He H, Zhong D. 2012. Prognostic value of Twist and E-cadherin in 622 patients with osteosarcoma. Med Oncol 29:3449-3455. Downloaded from 623 52. Tian W, Wang G, Yang J, Pan Y, Ma Y. 2013. Prognostic role of E-cadherin and 624 Vimentin expression in various subtypes of soft tissue leiomyosarcomas. Med Oncol 625 30:401. 626 53. Tam WL, Weinberg RA. 2013. The epigenetics of epithelial-mesenchymal plasticity in 627 cancer. Nat Med 19:1438-1449. 628 54. Dhooge A, Govaerts W, Kuznetsov YA. 2003. MATCONT: A MATLAB package for 629 numerical bifurcation analysis of ODEs. Acm Transactions on Mathematical Software

630 29:141-164. http://mcb.asm.org/

631

632

633

634 on August 1, 2016 by DUKE MEDICAL LIBRARY 635

636

637

638

639

640

641

642

643

644

645

646

647

648

649 ACKNOWLEDGMENTS

650 The results published here are in part based upon data generated by the TCGA Research

651 Network: http://cancergenome.nih.gov. JAS acknowledges support from the Duke Cancer Downloaded from 652 Institute, The Duke University Genitourinary Oncology Laboratory, and the Duke University

653 Department of Orthopaedics. HL was supported by the National Science Foundation (NSF)

654 Center for Theoretical Biological Physics (NSF PHY-1427654) and NSF DMS-1361411, and as

655 a CPRIT (Cancer Prevention and Research Institute of Texas) Scholar in Cancer Research of http://mcb.asm.org/ 656 the State of Texas at Rice University. KEW was supported by the NIH F32 CA192630 MKJ and

657 HL benefited from useful discussions with Mary C. Farach-Carson, J. N. Onuchic, Samir M.

658 Hanash, Kenneth J. Pienta, and Donald S. Coffey.

659

660 on August 1, 2016 by DUKE MEDICAL LIBRARY

661

662

663

664

665

666

667

668

669

670

671

672

673

674

675 FIGURE LEGENDS

676 Figure 1. E-cadherin induction by GRHL2 is cell line dependent. A. Construction of a

677 regulatory system using literature-based interactions suggests that a ZEB1-GRHL2 feedback Downloaded from 678 loop controls E-cadherin. B. Based on the model proposed in A, activation of GRHL2 enhances

679 E-cadherin levels to control cellular phenotype. C. Ectopic expression of GRHL2 has no effect

680 on ZEB1 mRNA levels. D. and E. Expression of GRHL2 activates E-cadherin mRNA (D) and

681 protein (E) in MDA-MB-231 cells, but not in RD or 143B sarcoma cell lines. * = p < 0.05 http://mcb.asm.org/ 682

683 Figure 2. Combined expression of miR-200s and GRHL2 induces MET. A. ZEB1 is the most

684 highly expressed of five major EMT master regulators in RD cells. B. Combined expression of

685 miR-200s and GRHL2 induces a morphological change consistent with MET. C. and D.

686 Expression of miR-200s induces E-cadherin expression, while combined expression of miR- on August 1, 2016 by DUKE MEDICAL LIBRARY

687 200s and GRHL2 has a multiplicative effect on E-cadherin protein (C) and mRNA (D) levels

688 (note log scale). E. ZEB1 mRNA levels are inhibited by miR-200s, but not by GRHL2. F.

689 Schematic of the E-cadherin/GRHL2 reporter. The firefly luciferase transcript is driven by the E-

690 cadherin promoter. The transcript is interrupted by a portion of E-cadherin intron 2 that contains

691 the GRHL2 binding site. The E-cadherin/GRHL2 reporter recapitulates the synergistic effect of

692 GRHL2 and miR-200s on the endogenous E-cadherin gene. G. Anchorage-independent growth

693 of RD cells is inhibited by miR-200s and increased by GRHL2 in soft agar assays. H. ZEB1 is a

694 highly expressed EMT master regulator in 143B osteosarcoma cells. I. GRHL2 and miR-200s

695 have a multiplicative effect on E-cadherin mRNA in 143B cells. J. ZEB1 is downregulated by

696 miR-200s, but not by GRHL2 in 143B cells. * = p < 0.05

697

698 Figure 3. GRHL2 competes with ZEB1 to activate E-cadherin. A. SiRNA-mediated

699 knockdown of ZEB1 protein (A) and mRNA (B) in RD cells. NS = nonsilencing control siRNA,

700 si_2 and si_5 = two independent siRNAs targeting ZEB1. C. E-cadherin is upregulated upon

701 ZEB1 knockdown only in the presence of GRHL2 expression. D. The cell-cell adhesion

702 molecule CLDN4 is upregulated upon ZEB1 knockdown only in the presence of GRHL2

703 expression. E. Unlike E-cadherin and claudin 4, ZEB1 knockdown has no additional effect on Downloaded from 704 the GRHL2-induced upregulation of the cell-cell adhesion gene, cingulin (CGN). * = p < 0.05.

705

706 Figure 4. A proposed regulatory network of E-cadherin control. A. ZEB1 represses

707 GRHL2-induced activation of E-cadherin. Dashed lines indicate the transition of GRHL2 from http://mcb.asm.org/ 708 unbound to bound on the E-cadherin promoter and miR-200 translational repression of ZEB1;

709 solid arrows/bars are transcriptional activation/repression of a given gene, respectively. B.

710 According to the proposed model, activation of GRHL2 (blue line) results in almost no change in

711 E-cadherin. Conversely, miR-200 upregulation induces E-cadherin expression while the

712 combined expression of miR-200s and GRHL2 has a synergistic effect on E-cadherin. C. on August 1, 2016 by DUKE MEDICAL LIBRARY

713 Relative levels of E-cadherin upon combined miR-200 and GRHL2 upregulation are synergistic.

714 D. Heat map of miR-200- and GRHL2-induced E-cadherin expression model predicts that, at a

715 threshold of miR-200 expression, E-cadherin is upregulated. This effect is enhanced by GRHL2

716 expression (upper right, in dark red).

717

718 Figure 5. Regulatory steps in the GRHL2-mediated activation of E-cadherin. A. BRG1

719 mRNA expression upon siRNA-mediated knockdown of BRG1. B. E-cadherin is upregulated in

720 a GRHL2-dependent manner upon BRG1 knockdown. C. Analysis of GRHL2 methylation status

721 from The Cancer Genome Atlas indicates that GRHL2 methylation is significantly increased in

722 sarcomas compared to epithelial-derived breast, colorectal (CRC) and prostate cancers. D.

723 GRHL2 methylation is lowest in synovial sarcoma compared to other sarcoma subtypes. * = p <

724 0.05

725

726 Figure 6. Epithelial biomarkers are prognostic for improved survival in soft tissue

727 sarcomas. A. RNA-Seq data from The Cancer Genome Atlas was used to construct Kaplan-

728 Meier curves from patients with soft tissue sarcomas. Patients were segregated into two groups Downloaded from 729 based on a previously-published EMT signature (Salt et al. 2013; Cancer Discovery). Patients

730 with a more epithelial-like gene expression profile (red line) have a better prognosis compared

731 to patients with a more mesenchymal-like gene expression profile (blue line). B. Soft tissue

732 sarcoma patients with higher GRHL2 and miR200 expression (red line) have improved survival http://mcb.asm.org/ 733 compared to patients with lower GRHL2 and miR200s (blue line). C. Expression of GRHL2 and

734 miR200s negatively correlates with a mesenchymal-like gene expression profile. Individual

735 patients are color coded by their soft tissue sarcoma subtype. on August 1, 2016 by DUKE MEDICAL LIBRARY

Downloaded from

A B C

1.4 1.2 1 0.8 0.6 0.4 http://mcb.asm.org/ 0.2

ZEB1 ZEB1 foldchange 0 - + - + - + GRHL2 MB-231 RD 143B

D E

8 * on August 1, 2016 by DUKE MEDICAL LIBRARY MB-231 RD 143B 7

6 GRHL2 - + - + - + change 5 4 E-cadherin 3

2 β-actin

cadherin cadherin fold -

E 1 0 - + - + - + GRHL2 MB-231 RD 143B Downloaded from

Neg miRs miR200s 1.4 10000 A * B C EV GRHL2 D 1.2 EV miR200s *

- + - + 1000

1.0 E-cadherin change 0.8 * β-actin 100 0.6 GRHL2 0.4 http://mcb.asm.org/ Relative levels 10

0.2

cadherin cadherin fold -

0.0 E 1 Zeb1 Zeb2 Snail Slug Twist1 miR200s - + - + GRHL2 GRHL2 - - + +

E-cad 1.8 FF Luc G E F 120

1.6 *

1.4 * 100 on August 1, 2016 by DUKE MEDICAL LIBRARY 1.2 4 80 1 Renilla 3 0.8 60

0.6 Luc) 2 40

0.4 * * * * Foldchange ZEB1 fold change fold ZEB1 20 0.2 1 Number ofcolonies

0 (normalized to 0 miR200s - + - + 0 miR200s - + - + miR200s - + - + - - + + GRHL2 - - + + GRHL2 - - + + GRHL2 25 * 2

I J H 1.4 143B 143B 143B * 20

1.2 1.5

change

1 15 change 0.8 1 0.6 10

0.4 0.5 Relative levels

5 ZEB1 fold cadherin cadherin fold 0.2 - * E * 0 0 0 Zeb1 Zeb2 Snail Slug Twist miR200s - + - + miR200s - + - + GRHL2 - - + + GRHL2 - - + + Downloaded from

1.2 A NS si_2 si_5 B Zeb1 Hoechst 1

0.8

0.6 http://mcb.asm.org/ change

ZEB1 Fold 0.4

0.2 * * * * 0 NS si_2 si_5 NS si_2 si_5 GRHL2 - - - + + + on August 1, 2016 by DUKE MEDICAL LIBRARY C D E 20 * 16 8 18 E-cadherin CLDN4 CGN 14 * 7 * 16 *

12 6

12 10 * 5 10 * change 8 4 * 8

6 3

Fold Foldchange Foldchange * 6 4 2 4 * 1 2 2 0 0 0 NS si_2 si_5 NS si_2 si_5 NS si_2 si_5 NS si_2 si_5 NS si_2 si_5 NS si_2 si_5 GRHL2 - - - + + + GRHL2 - - - + + + GRHL2 - - - + + + Downloaded from

A SNAIL B

GRHL2i μ200 ZEB http://mcb.asm.org/

E-cad GRHL2a

C D on August 1, 2016 by DUKE MEDICAL LIBRARY Fold change miR200s GRHL2 E-cadherin expression + - 25.72 - + 1.44 + + 37.25 Downloaded from

* 5 40 A * B E-cadherin 4.5 BRG1 35 * 4 30

3.5

3 25 http://mcb.asm.org/ 2.5 20

2 * 15

Foldchange Foldchange 1.5 * 10 1 5 0.5 0 0 NS si_5 si_6 NS si_5 si_6 NS si_5 si_6 NS si_5 si_6 GRHL2 - - - + + + GRHL2 - - - + + + on August 1, 2016 by DUKE MEDICAL LIBRARY

C D

0.5 0.5 0.4 * 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0.0 0.0

-0.1 -0.1 Relative Methylation Relative

-0.2 Methylation Relative -0.2

Breast CRC Prostate Sarcoma SS

LPS

LMS UPS

MXFS

PMFH MPNST Downloaded from

A EMT score B GRHL2+miR200s

Low EMT score High GRHL2+miR200s

High EMT score Low GRHL2+miR200s

http://mcb.asm.org/ Fraction surviving Fraction Log rank = 0.0099 surviving Fraction Log rank = 0.0432 Wilcoxon = 0.0016 Wilcoxon = 0.0308

Overall survival (days) Overall survival (days) on August 1, 2016 by DUKE MEDICAL LIBRARY C Pearson correlation = -0.3510 20 p<0.0001

-10

EMT score EMT -20

-30

-40 -10 0 10 20 30 40 50 GRHL2+miR200s