Different DNA methylome, transcriptome and histological features in uterine fibroids with and without MED12 mutations

Somatic mutations in Mediator complex subunit 12 (MED12m) have been reported as a biomarker of uterine fibroids (UFs). However, the role of MED12m is still unclear in the pathogenesis of UFs. Therefore, we investigated the differences in DNA methylome, transcriptome, and histological features between MED12m-positive and -negative UFs. DNA methylomes and transcriptomes were obtained from MED12m-positive and -negative UFs and myometrium, and hierarchically clustered. Differentially expressed genes in comparison with the myometrium and co-expressed genes detected by weighted gene co-expression network analysis were subjected to gene ontology enrichment analyses. The amounts of collagen fibers and the number of blood vessels and smooth muscle cells were histologically evaluated. Hierarchical clustering based on DNA methylation clearly separated the myometrium, MED12m-positive, and MED12m-negative UFs. MED12m-positive UFs had the increased activities of extracellular matrix formation, whereas MED12m-negative UFs had the increased angiogenic activities and smooth muscle cell proliferation. The MED12m-positive and -negative UFs had different DNA methylation, gene expression, and histological features. The MED12m-positive UFs form the tumor with a rich extracellular matrix and poor blood vessels and smooth muscle cells compared to the MED12m-negative UFs, suggesting MED12 mutations affect the tissue composition of UFs.


Introduction 46
Uterine fibroids are tumors derived from uterine smooth muscle cells and are most common in gynecologic neoplasms 1 . In the last decade, somatic mutations of Mediator complex subunit 12 (MED12) 48 have been found to be reliable biomarkers of uterine fibroids 2-4 . MED12 is located on the X chromosome 49 and encodes the RNA polymerase II mediator complex and part of the transcriptional preinitiation machinery. 50 Mutations of MED12, especially mutations in exon 2, are thought to be the underlying causes of about 70% 51 of human uterine fibroids 2,5,6 . However, the remaining 30% of uterine fibroids do not have MED12 52 mutations, which indicates that the role of MED12 mutations in the pathogenesis of uterine fibroids is unclear. 53 Uterine fibroids differ in size and the number of nodules. Uterine fibroids carrying MED12 54 mutations are reported to be smaller and often more numerous than those without MED12 mutations 4 . 55 Several reports have suggested links between MED12 mutations and different phenotypes of uterine fibroids. 56 Uterine fibroids without MED12 mutations were found to have elevated erythropoietin expression in an 57 estrogen-dependent manner, while the uterine fibroids with MED12 mutation had low erythropoietin 7, 8 . 58 expression epigenetic", "liver regeneration", and "negative regulation of canonical Wnt signaling pathway" 122 were detected in the increased genes (Fig. 2d). In the decreased genes in the MED12m-negative uterine 123 fibroids (Fig. 2d), the GO terms "cell adhesion", "extracellular matrix organization", "positive regulation of 124 cell-substrate adhesion", "integrin-mediated signaling pathway", "negative regulation of transcription by 125 RNA polymerase II", "inflammatory response", "reactive oxygen species metabolic process", "positive 126 regulation of apoptotic process", and "transforming growth factor-beta receptor signaling pathway" were 127 detected. 128 reactive oxygen species metabolic process activities compared to myometrium. Figure 3 shows the 137 expression statuses of representative genes of the commonly (Fig. 3a) and oppositely (Fig. 3b) regulated 138 processes between the MED12m-positive and MED12m-negative uterine fibroids.

Weighted gene co-expression network analysis (WGCNA) 141
The preceding DEGs (Fig. 2) is based on comparing the uterine fibroids with the myometrium, 142 and this analytic approach has been used so far 10,11 . In general, in the analytic method that compares the 143 target tissues to the control tissues, there is a possibility of missing the essential character of the target tissue 144 when the cell character of the target tissue is close to that of the control tissue. Therefore, to know the 145 intrinsic character in each of the MED12m-positive and -negative uterine fibroids, we used a WGCNA 146 analysis 20,21 . WGCNA is a system biology method to describe the correlation patterns among the genes 147 across microarray samples such as transcriptome data and to find the groups with highly correlated genes 148 that work in the same biological functions 20,21 . We defined groups consisting of highly correlated genes 149 as co-expressed gene (COG) groups and detected unique properties in the MED12m-positive and -negative 150 uterine fibroids by comparing the intrinsic functions of each tissue. 151 The transcriptome data of the MED12m-positive and -negative uterine fibroids were 152 independently subjected to the WGCNA. In the MED12m-positive and -negative uterine fibroids, 153 WGCNA identified 26 and 14 COG groups, respectively (Table 1), and these genes were subjected to the 154

GO enrichment analysis. 155
In the MED12m-positive uterine fibroids, three of the 26 COG groups had significant GO terms, 156 while in the MED12m-negative uterine fibroids, five of the 14 COGs groups had significant GO terms (Table and Group5), respectively. The commonly detected GO terms between the MED12m-positive and -160 negative uterine fibroids included "RNA splicing, via transesterification reactions", "mRNA splicing, via 161 spliceosome", ncRNA processing", "mRNA processing", and "RNA splicing", which are related to 162 transcription and translation, and "ribonucleoprotein complex biogenesis" and "DNA replication", which are 163 related to cell proliferation (Fig. 4). 164 The extracellular matrix-related terms including "extracellular structure organization" and 165 "extracellular matrix organization" were also commonly found. The gene ratios were much larger in the 166 MED12m-positive uterine fibroids than those in the MED12m-negative uterine fibroids, suggesting that the 167 number of extracellular matrix-related genes was larger in the MED12m-positive uterine fibroids than that 168 in the MED12-negative uterine fibroids (Fig. 4). 169 We then focused on specific GO terms in each of the MED12m-positive or -negative uterine 170 fibroids. There were no specific GO terms to the MED12m-positive uterine fibroids (Fig. 4). On the 171 other hand, the MED12m-negative uterine fibroids had specific GO terms related to cellular protein synthesis 172 ("Golgi vesicle transport", "ER to Golgi vesicle-mediated transport", and "Rho protein signal transduction"), 173 anti-apoptosis ("regulation of apoptotic signaling pathway"), muscles ("muscle system process" and "muscle 174 contraction", and "mitochondrial ATP synthesis coupled electron transport"), and angiogenesis ("blood of extracellular matrix organization compared with MED12m-negative uterine fibroids. On the other hand, 179 MED12m-negative uterine fibroids showed increased activities of angiogenesis and smooth muscle cell and 4). We histologically examined the amount of collagen fibers in the MED12m-positive and -negative 188 uterine fibroids, and myometrium. Immunofluorescence staining showed that the amount of collagen fibers 189 was significantly larger in the MED12m-positive uterine fibroid than in the myometrium and MED12m-190 negative fibroids ( Fig.5a and 5b). There was no significant difference between the myometrium and 191 MED12m-negative uterine fibroids. We next examined the number of blood vessels in the MED12m-192 positive and -negative uterine fibroids, and myometrium. The result showed that the number of blood 193 vessels was significantly higher in the MED12m-negative uterine fibroids than in the MED12m-positive 194 uterine fibroids ( Fig. 5c and 5d). There was no significant difference between the MED12m-negative 195 uterine fibroids and myometrium. We also examined the ratio of smooth muscle cells in total cells in the was significantly higher in the MED12m-negative uterine fibroids than the MED12m-positive uterine 198 fibroids and myometrium ( Fig. 5e and 5f). There was no significant difference between the MED12m-199 positive uterine fibroids and myometrium. fibroids showed higher DNA methylation levels (more than 15% DNA methylation) than the myometrium, 207 respectively ( Fig. 6a and Supplemental Table S5 online). In NRG1, 100% (9 of 9 samples) of the 208 MED12m-positive uterine fibroids and 75% (9 of 12 samples) of the MED12m -negative uterine fibroids 209 showed higher DNA methylation levels than the myometrium, respectively ( Fig. 6b and Supplemental Table  210 S5 online). 211 The mRNA expression levels of SATB2 in all nine of the MED12m-positive uterine fibroids and 75% 212 (9 of 12 samples) of the MED12m-negative uterine fibroids were more than twice those in the myometrium 213 ( Fig. 6c and Supplemental Table S5 online), while the mRNA expression levels of NRG1 in all the methylation and mRNA expression of at least one of SATB2 and NRG1 were higher in all the MED12m-217 positive and -negative uterine fibroids than they were in the myometrium. Since DNA hypermethylation 218 and increased expression of SATB2 or NRG1 were observed regardless of MED mutations, these 219 characteristics are unlikely to depend on MED12 mutations. 220 221

Discussion 222
The present study showed that the DNA methylation profiles of the MED12m-positive and -negative 223 uterine fibroids differed. Since DNA methylation is cell/tissue-specific, uterine fibroids with MED12 224 mutations differ from uterine fibroids without MED12 mutations at the molecular level. This prompted us 225 to clarify the difference between the two types of uterine fibroids in this study. 226 In the MED12m-positive uterine fibroids, DEGs were enriched to the GO terms related to extracellular 227 matrix organization. The WGCNA analysis also showed the activated extracellular matrix organization in 228 the MED12m-positive uterine fibroids. On the other hand, in the MED12m-negative uterine fibroids, most 229 of the genes in the GO terms related to extracellular matrix organization were down-regulated in comparison 230 with the myometrium and MED12m-positive uterine fibroids. Furthermore, the TGF-beta signaling 231 pathway, which contributes to fibrosis 22,23 , was downregulated in the MED12m-negative uterine fibroids. 232 These results suggest that MED12 mutations activate extracellular matrix organization in uterine fibroids. 233 In fact, our histological results showed that the amount of collagen was enriched in the uterine fibroids with 234 contribute to collagen synthesis, and increased collagen deposition in uterine fibroids 11,24 . Also, many of 236 the uterine fibroids in those studies may have been MED12m-positive because more than 70% of uterine 237 The WGCNA analysis in the MED12m-negative uterine fibroids detected a COGs group that is related 239 to muscles. In our histological results, the amount of smooth muscle cells was larger in the uterine fibroids 240 without MED12 mutations compared with the uterine fibroids with MED12 mutations, which is consistent 241 with a previous report demonstrating a high ratio of smooth muscle cells to fibroblasts in the uterine fibroids 242 without MED12 mutations compared to that with MED12 mutations 9 . That study, together with the present 243 results suggest that uterine fibroids without MED12 mutations are enriched in smooth muscle cells and 244 contain a low amount of collagen fibers, and that MED12 mutations are associated with collagen-rich uterine 245

fibroids. 246
The GO enrichment analyses with DEGs and WGCNA analysis showed that both MED12m-247 positive and -negative uterine fibroids have increased cell proliferation and transcription activities. This 248 well reflects one of the characters of uterine fibroids, which is the activated cell proliferation of smooth 249 muscle cells or fibroblasts. It is interesting to note that the activity of the Wnt signaling pathway was 250 decreased in MED12m-negative uterine fibroids. The Wnt signaling pathway has been reported to play an 251 important role in the growth of uterine fibroids 22 . This may be because the major type of uterine fibroids 252 included in those reports was MED12m-positive uterine fibroids in which Wnt signaling pathway is activated. 253 We speculate that the growth of MED12m-negative uterine fibroids is regulated by signaling pathways other than the Wnt signaling pathway. As shown in Table 2, multiple signaling pathways are involved in cell 255 proliferation in both MED12m-positive and -negative uterine fibroids. 256 VEGF expression is reported to be upregulated in uterine fibroids 25 , which suggests that 257 angiogenic activity in increased in uterine fibroids. On the other hand, our GO enrichment analysis 258 suggested that angiogenesis is downregulated in MED12m-positive uterine fibroids and upregulated in 259 MED12m-negative uterine fibroids. There seems to be a discrepancy between the previous reports and our 260 results. That may be due to the difference in cellular components of the tissue samples of uterine fibroids. In addition, the response to estrogen was found to be upregulated in MED12m-positive uterine 267 fibroids. Since fibroblasts were reported to proliferate or produce collagen in response to estrogen while 268 smooth muscle cells proliferate in response to progesterone 9 , collagen-rich MED12m-positive uterine 269 fibroids seem to well respond to estrogen. 270 Our results also suggested that the immune response and reactive oxygen species metabolic 271 processes are decreased in both MED12m-positive and -negative uterine fibroids. That is not surprising 272 because tumorigenesis is well known to occur under the suppressive environment of immune responses and reactive oxygen species 25 26 27,28 . 274 High mobility group AT-hook2 (HMGA2) mutation is considered to be one of the mutations 275 driving the development of uterine fibroids 3,29,30 , and the MED12 mutations and rearrangement of HMGA2 276 have been shown to occur in a mutually exclusive manner 4,5 . However, our results showed that some 277 uterine fibroids carried both the MED12 mutation and HMGA2 overexpression, and that only half of the 278 MED12m-negative fibroids had increased HMGA2 expression (Supplemental Fig. S2). Previous reports 279 also indicated that the MED12 mutations and increased HMGA2 expression co-existed in the same uterine 280 fibroid nodule 10,31 . 281 We previously reported that SATB2 and NRG1 act as upstream regulatory factors in the 282 pathogenesis of uterine fibroids 19,22,23 . Whether or not MED12 has a mutation, SATB2 and NRG1 were 283 more strongly expressed in uterine fibroids than in the myometrium, which indicates that the upregulation 284 of NRG1 and SATB2 are independent of MED12 mutations. Our results also indicated that all the 285 MED12m-positive and -negative uterine fibroids had DNA hypermethylation and increased mRNA 286 expression in either SATB2 or NRG1, suggesting that the dysregulation of upstream regulatory factors such 287 as SATB2 and NRG1 is involved in the pathogenesis of uterine fibroids. 288 One may question the relationship between MED12 mutation and DNA methylation; whether MED 289 12 mutation changes DNA methylation status. It is unlikely that a MED 12 mutation could change the 290 DNA methylation status because the DNA methylation profile of subtype-1 MED12m-negative uterine 291 fibroids was identical to the MED12m-positive uterine fibroids. Further studies are needed to identify the differences among the three subtypes of uterine fibroids without MED12 mutation. 293 In conclusion, the present study shows that uterine fibroids with and without MED12 mutations 294 clearly differ in DNA methylation, gene expression, and histological features. The DNA methylome 295 indicated that the uterine fibroids carrying MED12 mutations differed from the uterine fibroids without 296

Ethics Statement 306
This study was reviewed and approved by the Institutional Review Board of Yamaguchi University 307 Graduate School of Medicine. Written informed consent was obtained from the participants before 308 collecting any samples, and the specimens were irreversibly de-identified. All experiments involving the 309 handling of human tissues were performed following the Tenets of the Declaration of Helsinki.

Tissue preparation 312
Tissues of uterine fibroid and myometrium were obtained from 42 Japanese women, respectively. 313 Uterine fibroids were obtained from patients aged 33-45 who underwent hysterectomy for uterine fibroids. Three of the nine MED12m-negative uterine fibroids had expressions more than two-fold of the mean 336 expression in the myometrium (Supplemental Fig. S2 online). Moreover, one of the six MED12m-positive 337 uterine fibroids had expressions more than two-fold of the mean expression in the myometrium 338 (Supplemental Fig. S2 online). These results suggest that 1) a number of the MED12m-negative uterine 339 fibroids lack HMGA2 rearrangements, and 2) MED12 mutations and HMGA2 rearrangements can co-exist. 340 These facts led us to compare the uterine fibroids with and without MED12 mutations rather than to compare 341 uterine fibroids with MED12 mutations and HMGA2 rearrangements. The transcriptomes of myometrium, MED12m-positive, and -negative uterine fibroids were 359 analyzed as previously reported 33,34 . Total RNAs were isolated from cells by using an RNeasy mini kit 360 (Qiagen). Target cDNA for a microarray was prepared from 250 ng of total RNA with the Ambion WT 361 Genomics Suite 6.5 software program (Partech, Munster, Germany). All expression data were converted 368 to log2 values. Differentially expressed genes (DEGs) were extracted when the expressions in the 369 MED12m-positive or -negative uterine fibroids were higher than 2.0-fold or less than 0.5-fold of that in the 370 myometrium, and p < 0.05 (t-test), and the average expression levels in the tissues with higher expression 371 were more than 100. The amounts of collagen fibers, the number of blood vessels, and the percentage of smooth muscle 416 cells between the MED12m-positive and -negative uterine fibroids and myometrium were compared with 417 pairwise Wilcoxon rank-sum tests using R (function "pairwise.wilcox.test"; version 3.6.0.). p < 0.05 was 418 considered significant. 419 Encyclopedia of Genes and Genomes; KEGG), respectively 37 . In GO analysis, GO terms with adjusted p 425 (BH method) < 0.01 were considered significant enrichment. In KEGG analysis, pathways with p < 0.05 426 were considered significant enrichment. In WGCNA analysis, adjusted p < 0.1 was regarded as substantial 427 enrichment in the GO enrichment analysis. Hierarchical clustering was performed in R using the Ward 428

Competing interests 563
The authors declare no competing interests. 564 565