Spring 2020 – Systems Biology of Reproduction Discussion Outline – Ovary Systems Biology Michael K. Skinner – Biol 475/575 CUE 418, 10:35-11:50 am, Tuesday & Thursday March 5, 2020 Week 8

Ovary Systems Biology

Primary Papers:

1. Wigglesworth, et al. (2015) Biology of Reproduction 92(1):23, 1-4 2. Sagvekar, et al. (2019) Clinical Epigenetics 11:61 3. Nilsson, et al. (2010) PLoS ONE 7:e11637

Discussion

Student 19: Reference 1 above • What cumulus and mural granulosa cells? • What categories and networks were identified? • What oocyte paracrine interactions were identified?

Student 20: Reference 2 above • What are the technology used and objectives? • What epigenetic regulation and gene network were identified? • What insights are provided into the development of polycystic ovarian disease?

Student 21: Reference 3 above • What is the experimental and systems approach? • What new insights provided on primordial follicle development? • What gene signaling networks were identified for primordial follicle development? BIOLOGY OF REPRODUCTION (2015) 92(1):23, 1–14 Published online before print 5 November 2014. DOI 10.1095/biolreprod.114.121756

Transcriptomic Diversification of Developing Cumulus and Mural Granulosa Cells in Mouse Ovarian Follicles1

Karen Wigglesworth,3 Kyung-Bon Lee,4 Chihiro Emori,5 Koji Sugiura,5 and John J. Eppig2,3

3The Jackson Laboratory, Bar Harbor, Maine 4Department of Biology Education, College of Education, Chonnam National University, Buk-gu, Gwangju, Korea 5Laboratory of Applied Genetics, Graduate School of Agriculture and Life Sciences, The University of Tokyo, Tokyo, Japan

ABSTRACT to antral follicle transition. The granulosa cells of preantral follicles sort into two more differentiated populations at the Cumulus cells and mural granulosa cells (MGCs) have time of follicular antrum formation: mural granulosa cells functionally distinct roles in antral follicles, and comparison of (MGCs), which line the follicular wall, and cumulus granulosa their transcriptomes at a global and systems level can propel cells, which are associated with the oocyte. Generally, cumulus future studies on mechanisms underlying their functional cells are thought to support oocyte development, while MGCs diversity. These cells were isolated from small and large antral carry out endocrine function(s) indicative of functional as well follicles before and after stimulation of immature mice with as cellular architectural diversity. However, these are not strict gonadotropins, respectively. Both cell types underwent dramatic Downloaded from www.biolreprod.org. transcriptomic changes, and differences between them in- divisions of labor because interactions between MGCs and creased with follicular growth. Although cumulus cells of both cumulus cells can play crucial roles in oocyte maturation and stages of follicular development are competent to undergo ovulation [1–3]. The two populations have very different fates expansion in vitro, they were otherwise remarkably dissimilar after the preovulatory surge of luteinizing hormone (LH). The with transcriptomic changes quantitatively equivalent to those MGCs remain within the ovary and participate in the formation of MGCs. analysis revealed that cumulus cells of of the corpus luteum. In contrast, the cumulus cells undergo a small follicles were enriched in transcripts generally associated dramatic morphological change, producing and becoming with catalytic components of metabolic processes, while those embedded in a mucinous matrix in a process often referred to from large follicles were involved in regulation of metabolism, as cumulus expansion; they accompany the metaphase II cell differentiation, and adhesion. Contrast of cumulus cells oocyte to the oviduct during ovulation [4–6]. Before the LH versus MGCs revealed that cumulus cells were enriched in surge, the cumulus cells communicate with the oocyte via gap transcripts associated with metabolism and cell proliferation junctions, which allow the exchange of low molecular weight while MGCs were enriched for transcripts involved in cell molecules and facilitate both metabolic cooperation between signaling and differentiation. In vitro and in vivo models were cumulus cells and oocytes and regulation of meiosis [7–11]. used to test the hypothesis that higher levels of transcripts in The cumulus cells in the first layer around the oocyte are cumulus cells versus MGCs is the result of stimulation by oocyte- referred to as the corona radiata; and both the corona cells and derived paracrine factors (ODPFs). Surprisingly ;48% of the cumulus cells in outer ranks communicate with the oocyte transcripts higher in cumulus cells than MGCs were not via membrane extensions that reach around the corona cells to stimulated by ODPFs. Those stimulated by ODPFs were mainly contact the oocyte, thus forming a pseudostratified communi- associated with cell division, mRNA processing, or the catalytic cating epithelium [12]. Oocyte-derived paracrine factors pathways of metabolism, while those not stimulated by ODPFs (ODPFs), sometimes in cooperation with other ligands, play were associated with regulatory processes such as signaling, transcription, , or the regulation of metabolism. crucial roles in the development and function of the cumulus cells [13, 14]. Oocytes are required for the formation of cumulus cells, mouse, mural granulosa cells, oocyte, oocyte- cumulus cells during the preantral to antral follicle transition derived paracrine factors, ovarian follicle, transcriptome [15]. This transition is generally correlated with the acquisition of oocyte competence to resume the first meiotic division [16]. INTRODUCTION These early cumulus cells are competent to undergo expansion The cellular architecture of the ovarian follicle of most in vitro [17] in a manner that appears similar to that of cumulus mammalian species becomes clearly diversified at the preantral cells within large Graafian follicles in situ. These specific roles reveal considerable cellular and developmental complexity, but, in fact, very little is known about the global nature of the 1Supported by grants from the Eunice Kennedy Shriver National Institute of Child Health and Human Development HD23839, HD transcriptomes that support this cellular and functional 21970 (K.W., K-B.L., K.S., and J.J.E.) and a Grant-in-Aid for Scientific diversification, which is the goal of this study. Research from the Ministry of Education, Culture, Sports, Science and The levels of some transcripts that are different in the Technology of Japan (C.E. and K.S.). cumulus cell and MGC populations in large Graafian follicles 2Correspondence: John J. Eppig, The Jackson Laboratory, 600 Main even before the LH surge and the induction of cumulus Street, Bar Harbor, ME 04609. E-mail: [email protected] expansion, provide information on how the transcriptome reflects architectural diversity. For example, transcripts encod- Received: 22 May 2014. ing enzymes participating in glycolysis and cholesterol First decision: 29 June 2014. production are expressed more highly by cumulus cells than Accepted: 3 November 2014. Ó 2015 by the Society for the Study of Reproduction, Inc. by MGCs. Other transcripts enriched in cumulus cells versus eISSN: 1529-7268 http://www.biolreprod.org MGCs of Graafian follicles include those encoding an amino ISSN: 0006-3363 acid transporter SLC38A3, the androgen receptor AR, the 1 Article 23 WIGGLESWORTH ET AL. ligand AMH, and the natriuretic receptor 2 (NPR2) [3, 13, 18]. cells to oocytes via gap junctions to maintain oocyte meiotic Expression of these transcripts is seen by in situ hybridization arrest [3, 11]. Now in the current study, we have compared the as a gradient with highest expression by cells closest to the transcriptomes of cumulus and MGCs in both early antral oocyte and decreasing with increased distance outward from follicles and in large follicles after eCG-stimulation of the oocyte [3, 13, 19]. Some periantral granulosa cells express immature mice in vivo. Although the cumulus cells isolated these transcripts at levels higher than those seen in mural cells from both stages of follicular development are competent to closer to the basal lamina. This suggests that periantral MGCs undergo expansion in vitro [17], and are morphologically receive ODPF signals at levels sufficient to affect gene indistinguishable, we found that cumulus cells undergo tran- expression but that these levels are rapidly diluted with scriptomic changes that are as dramatic as those occurring in increasing distance from the oocyte. Microsurgical removal of MGCs during the SAF to LAF transition. oocytes (oocytectomy, OOX) from isolated cumulus-oocyte The second objective of this study was to test the hypothesis complexes (COCs) results in reduced expression of these that ODPFs are responsible for the higher expression of transcripts, but these levels are maintained by ODPFs, transcripts in cumulus cells than MGCs. Two approaches were particularly BMP15, GDF9, FGF8, or combinations of these used to test this: 1) comparison of the transcripts expressed at [13, 19, 20]. This demonstrates that ODPFs promote the higher levels in freshly isolated cumulus cells with those found expression of characteristic of the cumulus cell previously to be stimulated in cumulus cells by ODPFs in vitro phenotype. On the other hand, follicle-stimulating hormone [34] and 2) comparison of the higher cumulus cell transcripts (FSH) occurs in a diffusion gradient in the follicle opposite to with those found previously to be lower in cumulus cells of that of ODPFs and promotes the expression of genes mutants deficient in the ODPFs GDF9 and BMP15 versus characteristic of the MGC phenotype, such as Lhcgr encoding wild-type (WT) cumulus cells in vivo [35]. the LH-receptor and Cyp11a1 encoding the P450 cholesterol side chain cleavage enzyme [13, 21]. Actions of FSH are MATERIALS AND METHODS augmented when MGCs contact components of the follicular COCs were isolated from the SAFs and LAFs of 22-day-old B6SJLF1 mice Downloaded from www.biolreprod.org. basal lamina [22, 23]. ODPFs often abrogate the action of FSH that were raised in the research colonies of the authors at The Jackson and promote the cumulus cell phenotype instead. For example, Laboratory. The Administrative Panel on Laboratory Animal Care at The ODPFs suppress the expression of Lhcgr mRNA by granulosa Jackson Laboratory approved animal protocols, and all the experiments were cells despite stimulation with FSH and culture on basal lamina conducted in accordance with the National Institutes of Health Guide for the [24]. Cells in intermediate zones between the gradients of FSH Care and Use of Laboratory Animals. In order to obtain sufficient and uniform populations of cumulus cells and MGCs for microarray analyses, development and ODPFs exhibit intermediate phenotypes depending upon of SAFs to LAFs was stimulated by intraperitoneal injection of 5 international their relative proximity to either the basal lamina or the oocyte. units eCG 44 h before obtaining the COCs. Thus, SAFs were from Cumulus expansion in vivo occurs just before ovulation unstimulated ovaries of 22-day-old mice, while LAFs were from 22-day-old when follicles are stimulated by LH and produce EGF-like mice stimulated 44 h previously with eCG. It is uncertain whether the growth factors (EGFLGFs), which are first generated by MGCs stimulation with exogenous hormones produces the same effects on granulosa in response to LH, and then by the cumulus cells via autocrine cell transcriptomes as would occur with natural stimulation by endogenous hormones. reinforcement [1, 2]. Cumulus expansion in response to Only COCs completely enclosed by tightly packed cumulus oophorus were stimulation of the EGF receptor requires the presence of selected for analysis. Two hundred oocyte-cumulus cell complexes were ODPFs [25]. Moreover, expansion requires the expression of at collected using micropipets with the aid of a stereo microscope and washed free least four factors (HAS2, PTGS2, PTX3, and TNFAIP6) of all other cells by serial transfer through three dishes of culture medium. This because loss of expression of the genes encoding any of these removes all possibility of contamination by individual MGCs. The cumulus factors dramatically compromises expansion [5, 26–29]. In cells were then stripped from the oocytes by drawing them into a fine glass pipet with a diameter slightly smaller than the diameter of the oocytes, which addition to these expansion-related factors, the levels of many were discarded, and the cumulus cells prepared for RNA extraction. Clumps of transcripts in cumulus cells change as a consequence of MGCs, which are easily distinguishable from oocyte-cumulus cell complexes triggering cumulus expansion by gonadotropins in vivo [30– using a stereo microscope, with a total mass approximately equal to that of the 33]. However, the transcriptomes of cumulus cells and MGCs cumulus cells, were collected with micropipets, washed, and prepared for RNA during the transition of small to large antral follicles (hereafter, extraction. Given this care, the separation of cell types was absolutely clean. As SAFs and LAFs, respectively), before the initiation of cumulus verification of this, no expression of Lhcgr, a marker of MGCs [13], was found in the cumulus cell preparations, and no expression of Slc38a3, a marker of expansion and ovulation, have not been described. cumulus cells from LAFs [13], was found in the MGC preparations as shown in Clearly, more global and systems views of the transcrip- Supplemental Figure S1 (all the supplemental data is available online at www. tional complexity underlying the architectural diversification biolreprod.org). The cumulus cells and MGCs were collected from three groups can provide rationale and impetus to future studies of follicular of five pooled mice. Thus, there were three biological replicates of each sample. cellular and functional development before the LH surge. Thus, The samples were amplified and applied to Affymetrics 430V2 arrays by the Jackson Laboratory Service as described previously [19, 34]. the first objective of this study was to obtain a more global Average signal intensities for each probeset within the arrays were perspective than provided by analyses of single transcripts or calculated by the RMA function provided within the Affymetrix package for R pathways by utilizing microarrays to characterize the tran- using a custom (Ensembl Transcript) CDF file [36]. The RMA method scriptomic diversity of cumulus cells and MGCs. Analyses of incorporates convolution background correction, sketch-quantile normalization, these data are made by performing pairwise transcriptomic and summarization based on a multiarray model fit robustly using the median comparisons, with each comparison enhancing our view of the polish algorithm. For this experiment, three pairwise comparisons were used to statistically resolve transcriptomic differences between treatment levels using transcriptome of specific cell states and types. The value of this the R/maanova analysis package [37, 38]: 1) cumulus cells of SAFs versus relatively unbiased, but global approach to the transcriptome LAFs, 2) cumulus cells versus MGCs of SAFs, and 3) cumulus cells versus was demonstrated by a previous study that capitalized on MGCs of LAFs. Specifically, differentially expressed transcripts were detected microarray data to discover a key for maintaining oocyte using F, a modified F-statistic incorporating shrinkage estimate of variance meiotic arrest. From this microarray approach, we found that components from within the R/maanova package [37, 38]. Statistical natriuretic peptide NPPC is a ligand produced by MGCs and significance levels of the pairwise comparisons were calculated by permutation analysis (1000 permutations) and adjusted for multiple testing using the false this ligand binds to its cognate receptor, NPR2, which is most discovery rate (FDR), Q-value, method [39]. Differentially expressed genes are highly expressed by cumulus cells. NPR2 is a guanylyl cyclase declared at an FDR Q-value threshold of 0.05. Therefore, the FDR was limited whose product, cGMP, is then transferred from the cumulus by this selection alone to 5%. Furthermore, an additional stringency was 2 Article 23 GRANULOSA CELL TRANSCRIPTOMIC DIVERSITY

FIG. 1. Photomicrographs of histological sections of (A) a small antral follicle (SAF) typical of those in the ovaries of 22-day-old B6SJLF1 mice not stimulated with eCG and (B) a large antral follicle (LAF) typical of those in the ovaries 44 h postinjection with 5 international units eCG showing mural granulosa cells (MGC), cumulus cells (CC), and oocyte (oo). C) The follicular remnants after puncturing an LAF with a needle to express (yellow arrow) Downloaded from www.biolreprod.org. MGC and the oocyte cumulus cell complex. Note that many MGCs most immediately adjacent to the follicular basal lamina remain unexpressed and therefore not included in the microarray analyses. Bar ¼ 100 lm. imposed by requiring a 1.25-fold difference between contrasted groups for cumulus and MGCs of both groups appear morphologically inclusion of transcripts in the cohort considered to be significantly enriched. indistinguishable by histological examination (Fig. 1, A and The effective FDR was, therefore, less than 5%. Moreover, only those B). In addition to morphological resemblance, the COCs transcripts encoded by genes annotated in Mouse Genome Database (MGD) as having known biological functions (http://www.informatics.jax.org/function. isolated from both groups were competent to undergo shtml) are presented. Transcripts levels whose FDR Q-value was .0.05 and expansion in vitro ([17] and results presented here). During fold difference was ,1.25 were considered to be not different. VLAD, a the collection process, rupturing follicles with needles expels VisuaL Annotation Display tool (v. 1.5.1) (http://proto.informatics.jax.org/ COCs and clumps of MGCs. Examination of ovarian remnants prototypes/vlad/) was used to identify biological function gene ontology (GO) after puncture of follicles showed that some MGCs were not terms defined by sets of enriched transcripts. GO terms, established by the GO extruded upon puncture, particularly those most closely Consortium (http://www.geneontology.org/), provide a standardized way to group genes or according to their biological or molecular functions. associated with the basal lamina (Fig. 1C), suggesting that Transcripts allocated to a specific GO term by electronic annotation alone were the MGCs used for analyses are probably a population biased excluded. Therefore, assignment of a transcript to a GO term was based on toward the periantral MGCs. curation of experimental data. Supplemental Figure S2 provides an illustration Our previous studies [17] showed that COCs from SAFs to aid in interpreting VLAD graphics. underwent expansion in vitro, and, moreover, these experi- For practical reasons, presentation of transcripts in contrasted groups was ments were conducted using FSH to stimulate expansion, that focused in two ways. First, only the transcripts associated with GO terms and is, prior to demonstration of the prominent role of EGFLGFs in annotated with known biological or molecular functions are presented; the entire lists of transcripts have been deposited in the Gene Expression Omnibus promoting cumulus expansion. Therefore, to provide more (http://www.ncbi.nlm.nih.gov/geo, dataset GSE55845). Additional transcrip- information on competence of the cumulus cells used for the tomic analyses showing the effects of ODPFs on cumulus cell transcripts and transcriptomic studies described here, we tested the ability of on the effect of Gdf9þ/ Bmp15/ double mutation were deposited previously these COCs to undergo expansion in response to EGF (GSE47967) [34] and (GSE7225) [19], respectively. Second, the top 50 stimulation in vitro. As shown in Supplemental Figure S3, transcripts exhibiting the greatest fold differences within the contrast are EGF dramatically stimulated the expression of the expansion- presented. Also for practical reasons, many references to general biological/ related transcripts and in COCs molecular functions of the numerous specific gene products referred to here are Has2, Ptx3, Ptgs2, Tnfaip6 not cited but can be found by searching the Mouse Genome Informatics from both the SAFs from unstimulated ovaries and the LAFs database (http://www.informatics.jax.org/) or other resources. from eCG-stimulated follicles. Nevertheless, in spite of the fact To assess the ability of cumulus cells to elevate expression of cumulus that COCs from both groups appeared maximally expanded expansion related transcripts Has2, Ptx3, Ptgs2, and Tnfaip6, in response to (not shown), levels of these transcripts in the cumulus cells of EGF, COCs were isolated from small and large follicles of ovaries of 22-day- SAFs were 50%–75% of those from LAFs. old mice and cultured for 6 h with or without 10 ng/ml EGF. Transcript levels In our previous studies, we reported differences in transcript were assessed by quantitative RT-PCR (qRT-PCR). All procedures and reagents were exactly as described previously [15]. levels between cumulus cells and MGCs using in situ hybridization and/or qRT-PCR or RNase protection assays [3, 13, 19, 20, 40–42]. It was shown that the following RESULTS transcripts are expressed more highly in cumulus cells than in General Considerations of the Samples and Analyses of the MGCs: Slc38a3, Npr2, Mvk, Fdps, Pfkp, Ldha, Nog, Smad7, Microarray Data Tpi1, and Eno1. In contrast, the following transcripts are expressed more highly in MGCs than in cumulus cells: Grem1, Cumulus and MGCs were obtained from either SAFs (200– Twsg1, Tob1, Nrip1, Cyp19a1, Cyp11a1, Lhcgr, Star, and 350 lm diameter) of unstimulated ovaries or from LAFs (450– Nppc. All of these transcripts fell within the cutoffs defining 550 lm diameter) of eCG-stimulated ovaries (Fig. 1). The the groups that were higher in cumulus or MGCs in the 3 Article 23 WIGGLESWORTH ET AL.

account for blurring of the boundary that would put transcripts inside or outside the inclusion groups. These include micro- array probe factors, statistical confounding, biological factors (such as bias in the MGC populations that were analyzed toward the periantral MGCs, which express some transcripts typical of cumulus cells at higher levels than the MGCs located nearer to the basal lamina), or other artifacts that could be created by sample collection. Nevertheless, the results here, considered in their global perspective, are probably conserva- tive, although any future studies based on the data presented here should include experimental verification of expression profiles for transcripts of interest.

Contrast: Cumulus Cells Isolated from SAFs Versus FIG. 2. Number of transcripts expressed at higher levels in SAFs of unstimulated ovaries (blue bars) versus those of LAFs 44 h post-eCG- Cumulus Cells Isolated from LAFs stimulation (red bars) in cumulus cells or MGCs. Only transcripts that In the contrast between cumulus cells from SAFs versus encode proteins with annotated biological functions are included. LAFs, 535 transcripts were expressed at higher levels in cumulus cells from SAFs while 580 were expressed more microarray analyses presented here. As indicated in Materials highly in cumulus cells from LAFs (Fig. 2). Transcripts and Methods, the cutoff to discriminate between higher or not enriched in the cumulus cells from SAFs encode proteins that higher in cumulus cells or MGCs in this study was set at fold participate in metabolic processes, including those with catalytic activities (e.g., GO:0044281, small molecule meta- difference . 1.25 and Q , 0.05. Thus, although arbitrary, the Downloaded from www.biolreprod.org. set discriminators for being different or not different produced bolic process), such as those having transferase (GO:001670) outcomes that were highly correlated with results produced by or oxidoreductase activities (GO:0016491). In contrast, those enriched in cumulus cells from LAFs encode proteins involved the various other quantitative approaches. However, Amh, Ar, with the regulation of metabolic and developmental processes, and were excluded from the higher in cumulus Sqle, Cyp51 including those with -binding functions, such as those cells versus MGCs group despite evidence using alternative, having protein- or small molecule-binding activities (e.g., and probably more reliable, methods showing that they are GO:0048522, positive regulation of cellular processes). The expressed at higher levels in cumulus cells than MGCs. The relative enrichment of transcripts in the top GO terms are microarray-based fold increase and Q values for Amh were shown in Figure 3 and/or Supplemental Table S1, which 1.25-fold and Q ¼ 0.09, for Ar they were 1.12-fold and Q ¼ presents the complete list of transcripts expressed differently in 0.08, and for Cyp51 they were 1.26-fold and Q ¼ 0.1. Thus, SAFs and LAFs for each GO term. these transcripts fell just outside the boundary that would have Transcripts with greatest fold differences. A list of the 50 included them in the higher in cumulus cell cohort. The top transcripts expressed more highly in cumulus cells from microarray data were validated for seven additional transcripts SAFs versus LAFs is shown on Table 1 and Supplemental by qRT-PCR (Supplemental Fig. S1). Several issues could Table S2. Serpinf1, encoding an antiangiogenic factor, heads

TABLE 1. Top 50 fold difference in cumulus cells of SAFs versus LAFs; gene names are presented in Supplemental Table S2.

Higher in SAF versus LAF cumulus cells Higher in LAF versus SAF cumulus cells

Fold difference Gene symbol Fold difference Gene symbol Fold difference Gene symbol Fold difference Gene symbol

9.51 Serpinf1 2.71 Mybpc3 7.78 Grem1 3.26 Scd1 7.29 Wt1 2.71 Mgp 7.53 Cbfa2t3 3.03 Aire 7.00 Nrip2 2.70 Gulo 7.40 Mrap 3.02 Fabp3 5.19 Lmo7 2.69 Cxcr7 6.02 Tubb2b 3.01 Vegfa 5.13 Ccbe1 2.66 Derl3 5.86 Cacna1a 3.00 Aicda 4.88 Gatm 2.66 Cspg4 5.57 Afap1l2 2.98 Mia1 4.45 Dsg2 2.66 Arnt2 4.97 Masp1 2.95 Dab1 4.42 Pik3ip1 2.65 Ror1 4.43 Vim 2.94 Hmgn3 4.06 Tdrd5 2.63 Perp 4.31 Plxnb1 2.91 Sdpr 3.92 Krt20 2.62 Cpeb2 4.02 Cml3 2.91 Lrp8 3.79 Plxna2 2.58 Tns3 4.01 Rbfox3 2.91 Adamts5 3.78 Slc18a2 2.58 Agtr2 3.99 Ppp1r3g 2.91 Rian 3.66 Maob 2.55 Sgms1 3.95 Cyp19a1 2.88 Gsta4 3.43 Calb1 2.54 Vnn1 3.89 Inhba 2.85 Fzd2 3.34 St3gal6 2.54 Plagl1 3.82 Cebpd 2.82 Cdkn1a 3.29 Itga6 2.54 Hnmt 3.56 Slit3 2.82 Gadd45g 3.20 Slc40a1 2.51 Sepp1 3.51 Timp2 2.81 Nmb 3.19 Kazald1 2.47 Hipk3 3.39 Meg3 2.79 Ralb 3.19 Mgst2 2.47 Abca1 3.38 Samd4 2.76 Fbxl15 3.11 Plin2 2.47 Slc38a3 3.36 Inhbb 2.74 Drd4 3.07 Runx1 2.45 Ntn4 3.31 Id2 2.70 Tmsb4x 2.98 Igfbp4 2.42 Irf6 3.31 P4ha2 2.69 Slc2a1 2.95 Eftud1 2.41 Cntn4 3.30 Gch1 2.65 Pfkp 2.91 Cd83 2.41 Hmga2 3.27 Ppfia3 2.65 Serpina5 2.76 Gadd45a 2.39 Sik1 3.26 Id1 2.65 Gsg1l

4 Article 23 GRANULOSA CELL TRANSCRIPTOMIC DIVERSITY

GO:0008150 biological_process

GO:0016043 GO:0008152 GO:0048513 GO:0044699 cellular GO:0065007 single-organism process metabolic process component organ biological (3.73e-03) (3.47e-23) organization development regulation (7.74e-09) (9.03e-07) (2.77e-03)

GO:0071704 GO:0044238 GO:0044763 GO:0044710 GO:0065009 organic GO:0065008 single-organism primary GO:0050789 regulation of single-organism substance regulation of cellular process metabolic process metabolic process regulation of molecular metabolic process biological quality biological process function (2.48e-12) (7.04e-17) (4.33e-23) (1.73e-06) (3.47e-03) (4.72e-08) (1.59e-21)

GO:0050793 GO:0051128 GO:0043436 GO:1901575 GO:0019222 regulation of oxoacid regulation of organic cellular regulation of developmental

metabolic process Downloaded from www.biolreprod.org. metabolic process substance component process (4.66e-12) catabolic process (6.90e-11) (1.66e-04) organization (3.50e-11) (9.15e-08)

GO:0031323 GO:0048522 GO:0009083 GO:0044724 GO:0048523 single-organism regulation of positive GO:0030155 GO:0042127 GO:0010941 negative branched-chain cellular regulation of regulation of regulation of regulation of carbohydrate regulation of catabolic process metabolic process cellular process cell adhesion cell proliferation cell death cellular process catabolic process (1.27e-09) (7.35e-06) (6.61e-06) (1.18e-05) (4.66e-06) (6.09e-08) (5.29e-12) (6.73e-08)

GO:0006355 GO:0045937 GO:0045647 regulation of positive regulation negative regulation transcription, of phosphate of erythrocyte DNA-templated metabolic process differentiation (5.12e-06) (7.91e-06) (6.91e-06)

biological_process SAF LAF

FIG. 3. VLAD depiction of GO terms in the contrast of cumulus cells from SAFs versus LAFs. The top 25 GO terms, selected by VLAD on the basis of local maximum P value, are shown. A local maximum term is one whose significance (log [p]) is greater than its immediate neighbors (children or parents). The size of the rectangle is proportional to the P value; the larger the rectangle the lower, and more significant, is the P value. The green portion of the rectangle shows the relative contribution of cumulus cell transcripts from SAFs to that GO term while the red portion shows the relative contribution of transcripts from LAFs. Supplemental Table S1 presents a complete list of transcripts expressed differently in SAFs and LAFs for each GO term. the list followed by the Wt1 transcription factor and Nrip2, which encodes a nuclear corepressor of hormone action. The transcription factor Runx1, encoding a transcription factor, and Agtr2, encoding an angiotensin receptor, are also much more highly expressed by cumulus cells of SAFs. Grem1, Inhba, and Inhbb, encoding ligands in the TGFb superfamily, were among the transcripts included on the list of the top 50 transcripts most highly expressed in cumulus cells of LAFs compared to SAFs (Table 1 and Supplemental Table S2). Included on this list are Id1 and Id2, which encode DNA-binding inhibitors involved in developmental processes promoted by various bone morphogenetic proteins (BMPs), Vegfa, Fzd2, and Cdkn1a (also known as p21), which encodes a cell cycle-related protein associated with FIG. 4. Number of transcripts expressed at levels higher in cumulus cells cell differentiation. Oocytes probably regulate the expression versus MGCs (blue bars) or higher in MGCs versus cumulus cells (red bars). Samples were taken from the SAFs of 22-day-old mice or the LAFs of of Id1 and Id2 in cumulus cells ([43, 44] and results 22-day-old mice 44 h poststimulation with eCG. Only transcripts that presented here). encode proteins with annotated biological functions are included. 5 Article 23 WIGGLESWORTH ET AL.

GO:0008150 biological_process

GO:0008152 GO:0016043 GO:0009987 cellular GO:0065007 GO:0032502 metabolic process cellular component biological developmental (2.09e-43) process organization regulation process (3.81e-21) (1.53e-34) (1.04e-16) (3.77e-27)

GO:0071704 GO:0044238 GO:0044237 GO:0044763 GO:0033554 organic primary cellular single-organism GO:0065008 GO:0050789 GO:0065009 cellular response regulation of regulation of substance metabolic process metabolic process cellular process to stress regulation of biological quality biological process molecular function metabolic process (8.76e-51) (1.22e-57) (1.94e-44) (8.61e-27) (4.95e-14) (1.04e-18) (7.96e-50) (1.15e-18)

GO:0048583 GO:0048518 GO:0019222 GO:0051128 GO:0050793 GO:0051239 positive regulation of regulation of regulation of regulation of GO:0050794 GO:0032879 regulation of response to multicellular regulation of regulation of regulation of cellular component developmental biological process cellular process metabolic organization process stimulus organismal process localization (7.59e-29) (5.72e-38) (6.65e-12) process (4.21e-21) (1.33e-20) (2.45e-23) (4.43e-14) (4.18e-25)

GO:0048523 GO:0031323 GO:0080090 GO:0060255 GO:0042127 negative GO:0030155 GO:0010941 regulation of regulation of regulation of regulation of regulation of regulation of regulation of cell proliferation cellular process cellular primary macromolecule cell adhesion cell death metabolic process metabolic process (4.84e-14) (5.91e-29) (8.56e-15) (1.45e-17) metabolic process (1.56e-27) (3.61e-27) (1.21e-26) Downloaded from www.biolreprod.org.

GO:0042981 GO:0019220 GO:2000112 regulation of cellular regulation of regulation of macromolecule apoptotic process phosphate metabolic process biosynthetic process (2.05e-17) (3.36e-15) (6.14e-20) biological_process Cumulus MGC

FIG. 5. VLAD depiction of GO terms in the contrast of cumulus cells versus MGCs from SAFs. The top 25 GO terms, selected by VLAD on the basis of local maximum P value, are shown. The size of the rectangle is proportional to the P value; the larger the rectangle the lower, and more significant, is the P value. The green portion of the rectangle shows the relative contribution of cumulus cell transcripts from SAFs to that GO term while the red portion shows the relative contribution of MGC transcripts from SAFs. Supplemental Table S3 presents a complete list of transcripts expressed differently in SAFs and LAFs for each GO term.

As shown in Figure 3 and/or Supplemental Table S1, and 1814 transcripts were higher in MGCs of LAFs (Fig. 4); transcripts in cumulus cells of SAFs relative to LAFs are 878 transcripts were commonly higher in MGCs in both SAFs enriched in transcripts generally associated with small and LAFs. molecule metabolic pathways (GO:0044281, GO:0019752, In SAFs, transcripts encoding proteins involved in cell GO:0009083). In contrast, cumulus cells of LAFs are enriched proliferation (e.g., GO:0022402; cell cycle process), cellular in transcripts associated with enzymatic pathways involved in response to stress (GO:0033554), and ribosome biogenesis monosaccharide catabolic processes (GO:0046365), including (GO:0042254) were enriched in cumulus cells relative to glycolysis (GO:0006096), and cholesterol biosynthetic pro- MGCs. In contrast, transcripts encoding proteins participating cesses (GO:0006695) in accordance with our previous results in the regulation of cell communication (GO:0010646), [19, 20, 40]. Moreover, the cumulus cells of LAFs were regulation of signal transduction (GO:0009966), cell morpho- enriched in transcripts associated with developmental processes genesis (GO:0000902), regulation of cell adhesion (GO:0032502), regulation of molecular function (GO:0030155), the regulation of cellular component movement (GO:0065009), positive and negative regulation of cellular (GO:0051270), and the negative regulation of cell proliferation processes (GO:0048522 and 0048523), and positive regulation (GO:0008285) were enriched in MGCs relative to the cumulus of transcription (GO:0045893). Cumulus cells, therefore, cells in SAFs (Fig. 5 and Supplemental Table S3). become enriched in transcripts encoding components of As shown in Figure 6 and/or Supplemental Table S4, metabolic pathways important for supporting cellular develop- cumulus cells of LAFs were enriched, relative to MGCs, in mental processes and oocyte metabolism during the SAF to transcripts associated with cell cycle process (GO0022402), LAF transition. noncoding RNA processing (GO:0034470), primary metabolic processes (GO:0044238), which includes cholesterol biosyn- Contrast: Cumulus Cells Versus MGCs thetic process (GO:0006695), glycolysis (GO:0006096), and Cumulus cells expressed 1385 transcripts at levels higher cellular response to stress (GO:0033554). In contrast, MGCs than in MGCs of SAFs, and 2318 were higher in cumulus cells were enriched, relative to cumulus cells, in transcripts encoding of LAFs (Fig. 4); 944 were commonly higher in cumulus cells proteins involved in the regulation of signaling (GO:0023051), versus mural cells in both SAFs and LAFs. Therefore, in SAFs phosphorylation (GO:0016310), protein catabolism and LAFs together, 2759 different transcripts were expressed at (GO:0006511 and GO:0016567), steroid biosynthetic process higher levels in cumulus cells than MGCs. MGCs expressed (GO:0006694), and the regulation of cell communication 1217 transcripts at levels higher than in cumulus cells of SAFs (GO:0010646). 6 Article 23 GRANULOSA CELL TRANSCRIPTOMIC DIVERSITY

GO:0008150 biological_process

GO:0016043 GO:0051179 cellular localization component organization (1.09e-13) (1.52e-25)

GO:0044237 GO:0065009 GO:0051649 GO:0044763 regulation of GO:0065008 GO:0008104 establishment single-organism cellular molecular regulation of protein of localization function cellular process metabolic process biological quality localization in cell (2.57e-23) (4.41e-18) (6.97e-11) (4.22e-10) (3.93e-15) (1.43e-59)

GO:0044260 GO:0044238 cellular GO:0019222 GO:0023051 GO:0033554 GO:0032879 GO:0051128 regulation of GO:0016310 primary regulation of regulation of regulation of cellular macromolecule cellular metabolic process signaling response to stress phosphorylation metabolic process localization component (6.02e-24) (1.57e-21) (5.74e-19) metabolic process (7.79e-11) organization (3.84e-16) (7.84e-51) (3.39e-10) (1.12e-55) Downloaded from www.biolreprod.org.

GO:0048522 GO:0031323 GO:0080090 GO:0006464 GO:0051726 positive GO:0043067 GO:0051493 regulation of GO:0035556 regulation of GO:0034470 cellular regulation of regulation of regulation of intracellular ncRNA protein regulation of cellular signal primary processing cell cycle programmed cytoskeleton modification process cellular process cell death organization metabolic process transduction metabolic process (5.10e-20) (5.69e-19) (4.01e-27) (7.08e-12) (1.43e-09) (7.93e-29) (2.53e-14) (8.89e-31) (1.93e-29)

GO:2000112 GO:0042326 regulation of GO:0016567 negative cellular protein regulation of macromolecule ubiquitination phosphorylation biosynthetic process (1.58e-09) (3.44e-10) (5.25e-25)

biological_process Cumulus MGC

FIG. 6. VLAD depiction of GO terms in the contrast of cumulus cells versus MGCs from LAFs. The top 25 GO terms, selected by VLAD on the basis of local maximum P value, are shown. The size of the rectangle is proportional to the P value; the larger the rectangle the lower, and more significant, is the P value. The green portion of the rectangle shows the relative contribution of cumulus cell transcripts from LAFs to that GO term while the red portion shows the relative contribution of transcripts from LAFs. Supplemental Table S4 presents a complete list of transcripts expressed differently in SAFs and LAFs for each GO term.

Transcripts Expressed at Higher Levels in Cumulus Cells signaling processes. Transcripts commonly expressed in the top Versus MGCs 50 also include Rhob, a member of the Rho GTP-binding family, and Arhgdig, which affects the functioning of RHOB Transcripts with greatest fold differences. Table 2 and and, potentially, amplifies the actions of small G protein- Supplemental Table S5 show the top 50 lists of transcripts signaling pathways. Also included in this group are Shb, which expressed at higher levels in cumulus cells versus MGCs. encodes an adapter protein that interacts with FGFR1 and is Among the top 50 transcripts more highly expressed in important for oocyte and follicular development [47]; Smad7, cumulus cells versus MGCs, 23 are in common in SAF and which encodes a member of TGFb family and antagonizes the LAF contrasts; those in the top 10 are particularly notable. actions of the TGFb type 1 receptor; Dusp1, which encodes a They include two members of the FOS proto-oncogene family, phosphatase that interacts with MAPK14, MAPK1, and Fos and Fosb, and two members of the JUN proto-oncogene MAPK8 (also known as p38, ERK2, and JNK1, respectively) family Jun and Junb. Members of these two families form that are well established to participate in a wide variety of heterodimers to produce the AP1 transcription factor complex, signaling pathways in granulosa cells [48–56]; Ank1, which which has been associated with granulosa cell differentiation encodes a protein integral to the plasma membrane involved [45]. Also in these top 10 groups are Klf4, which has been cell motility and the maintenance of specialized membrane implicated in stem cell-like capacity [46], and Cyr61, which domains; Ednrb, which encodes a G protein-coupled receptor encodes a protein that binds to various integrin receptors and to of endothelin that is thought to have important roles in heparan sulfate proteoglycans involved in cell adhesion and follicular development [57], including the induction of oocyte 7 Article 23 WIGGLESWORTH ET AL.

TABLE 2. Top 50 fold difference higher in cumulus cells versus MGCs; gene names are presented in Supplemental Table S5.

Higher in SAF cumulus cells versus MGCs Higher in LAF cumulus cells versus MGCs

Fold difference Gene symbol Fold difference Gene symbol Fold difference Gene symbol Fold difference Gene symbol

33.05 Fosb* 2.85 Dusp9* 28.25 Fosb* 4.27 Tex14 14.32 Jun* 2.84 Arhgdig* 22.94 Cyr61* 4.15 Bmp6 11.99 Klf4* 2.83 Nfkbiz 14.79 Aqp8* 4.12 Zfp36* 9.71 Ier2 2.78 Slc26a3* 14.39 Egr1 4.08 Pfkp* 8.34 Junb* 2.76 Dntt 11.42 Junb* 4.07 Mgp 7.93 Dusp1* 2.75 Cnga1* 11.18 Egr2* 4.06 Rph3al 7.62 Cyr61* 2.75 Calb1 9.15 Fos* 4.02 Hps1 6.05 Fos* 2.75 Loxl2 7.59 Arhgdig* 4.01 Pdgfrb 5.02 Zfp36* 2.75 Stat3 6.96 Jun* 4.00 Dusp9* 4.63 Btg2 2.74 Eftud1 6.53 Klf4* 3.99 Slit3 4.61 Egr2* 2.74 Plp1 6.48 Fbp1 3.98 Slc26a3* 4.06 Rhob* 2.71 Klf1* 5.87 Klf1* 3.89 Maff 3.89 Mt1 2.68 Aqp8* 5.75 Dusp1* 3.88 Cxcl1 3.66 Dnajb1 2.68 Id1 5.63 Smad7* 3.79 Ablim1 3.53 Sik1 2.67 Smad7* 5.40 Otof* 3.76 Irx3 3.36 Ednrb* 2.63 Pfkp* 5.22 Shb* 3.74 Slc38a3 3.35 Lnx1 2.60 Shb* 5.11 Afap1l2 3.71 Abcc3 3.15 Runx1 2.56 Socs3 4.92 Gadd45b 3.69 Pdia2 3.14 Ucp2 2.54 Fgfbp3 4.90 Ank1* 3.68 Cnga1* 3.05 Btg1 2.54 Gadd45g 4.49 Rhob* 3.67 Alpk3 3.04 Ank1* 2.53 Chn2 4.47 Slc18a2 3.64 Apoa1 3.02 Malat1 2.50 Gatm 4.46 Nr4a1 3.58 Heyl* Downloaded from www.biolreprod.org. 3.00 Klf2 2.49 Igfbp1 4.34 Angpt4 3.56 Rgcc 2.92 Otof* 2.49 Cd83 4.34 Ednrb* 3.54 Aicda 2.90 Elf3 2.43 Heyl* 4.27 Vegfa 3.54 Trp53 * Transcripts common to both SAFa and LAFs (N ¼ 23). maturation [58]; Pfkp, whose levels are known to be regulated and is necessary for normal cumulus expansion and ovulation by ODPFs and encodes an enzyme important for providing [64]; its expression by cumulus cells is suppressed by oocytes products of glycolysis to oocytes [20, 40]; Slc26a3, which [65]. encodes a member of the sulfate anion transporter family and Transcripts on the top 50 list of transcripts that are uniquely mediates chloride and bicarbonate ion exchange, and Cnga1, expressed higher in MGCs versus cumulus cells in SAFs which encodes a cyclic nucleotide gated ion channel; and Heyl, (Table 3 and Supplemental Table S6) included those encoding which encodes a transcription factor that mediates the action of receptors GHR and PLXNB1; ligands BMP2, INHBA, and NOTCH family receptors [59] involved in cell fate decisions NTN4; and cytoskeletal- and extracellular-related components and thought to participate in the regulation of granulosa cell ACTA2, COL3A1, FBN2, KRT20, MTAP2, SPNB2, and proliferation [60, 61]. VIM. Vegfa is among the top 50 transcripts uniquely expressed The top 50 list of transcripts uniquely expressed higher in more highly in cumulus cells versus MGCs of LAFs (Table 2 MGCs than cumulus cells in LAFs (Table 3 and Supplemental and Supplemental Table S5). This transcript encodes a Table S6) include those encoding receptors LHCGR, PRLR, member of the platelet-derived growth factor family of PTGFR, AHR, and EPHA5; ligands KITL and NPPC; cell ligands thought to play a role in folliculogenesis [62, 63]. adhesion/extracellular matrix-related molecules CD34, Suggestively, Pdgfrb is also included in the top 50 transcripts VCAM1, CCBE1, MATN2, LAMA4, and VCAN; and in this contrast, although no direct association of VEGFA and enzymes ADH1, CTSH, CYP17A1, ALDH1A1, and PDGFRB has yet been demonstrated. Bmp6 transcripts are PRKAR2B. The ephrin receptor protein EPHA5 has been also higher in cumulus cells versus MGCs of LAFs as are localized in MGCs by immunocytochemistry, regulated by Trp53 transcripts encoding the tumor suppressor TRP53 (also FSH, and thought to participate in cell adhesion processes [66]. known as p53). The transcript with the greatest differential between MGCs and cumulus cells in LAFs was Comp. It was previously suggested Transcripts Expressed at Higher Levels in MGCs Versus that Comp expression could be used as a marker of terminal Cumulus Cells follicular development because it was identified as the transcript most highly regulated by gonadotropin treatment in Transcripts with greatest fold differences. Some tran- MGCs and during follicular development in vitro [67]. scripts were expressed at higher levels in MGCs than in However, deletion of the Comp gene had no apparent effect cumulus cells in both SAFs and LAFs. Among these on the top on fertility, though quantitative data on fertility were not 50 lists, 11 were common to both size follicles (Table 3 and presented [68]. Supplemental Table S6) and encode steroidogenic hormone enzymes CYP11A1 (also known as P450 side chain cleavage Role of ODPFs in the Differential Expression of Transcripts enzyme) and CYP19A1 (also known as aromatase); the ligand in Cumulus Cells and MGCs CTGF; transcription factors and coregulators CITED2, NRIP1, and PBX1; apoptosis regulators GSN, STRADB, and TIA1; These data reveal considerable differences in transcript cell adhesion PCDH8; and receptor/ion channel GABRB2. levels in cumulus cells versus MGCs. Several factors could NRIP1, also known as RIP140, is a nuclear receptor cofactor influence this differential expression: exposure to ODPFs, expressed by MGCs that regulates expression of amphiregulin direct contact with oocytes and/or communication via gap 8 Article 23 GRANULOSA CELL TRANSCRIPTOMIC DIVERSITY

TABLE 3. Top 50 fold difference higher in MGCs versus cumulus cells; gene names are presented in Supplemental Table S6.

Higher in MGCs versus cumulus cells in SAFs Higher in MGCs versus cumulus cells in LAFs

Fold difference Gene symbol Fold difference Gene symbol Fold difference Gene symbol Fold difference Gene symbol

7.23 Vim 3.70 Gabrb2* 64.45 Comp 5.54 Slc26a7 7.21 Plxnb1 3.68 Fabp3 51.51 Lhcgr 5.43 Ahr 6.32 Acta2 3.68 Krt20 33.36 Cyp11a1* 5.39 Ccbe1 6.25 Gsn* 3.65 Bmp2 26.54 Gabrb2* 5.22 Fmnl2 5.99 Cald1 3.65 Ctgf* 19.74 Cd34 5.19 Nrip1* 5.97 Col11a1 3.62 Ctnna1 16.45 Cited2* 5.19 Cyp19a1* 5.28 Timp2 3.61 Gsta4 15.14 Gsn* 5.04 Satb1 4.90 Tia1* 3.59 Spnb2 13.99 Vcam1 5.00 Stradb* 4.89 Tmsb4x 3.56 Etl4 13.36 Lrp11 4.94 Pappa 4.76 Col3a1 3.54 Rbm5 13.06 Prlr 4.94 Kcne2 4.73 Ntn4 3.54 Stk38 10.06 Adh1 4.92 Prkar2b 4.73 Nrip1* 3.47 Cited2* 9.54 Cyp17a1 4.91 Vcan 4.71 Stradb* 3.47 Fbn2 9.38 Scara5 4.86 Nppc 4.71 Foxp1 3.44 F11r 9.36 Ctsh 4.85 Neb 4.37 Ghr 3.44 Mysm1 9.34 Ctgf* 4.82 Hao2 4.27 Cyp11a1* 3.42 Rgmb 8.55 Ptgfr 4.80 Matn2 4.22 Ctsf 3.40 Fbxl7 7.75 Olfm1 4.68 Epha5 4.22 Fhl1 3.39 Rala 7.62 Tnfsf11 4.68 Tia1* 3.99 Cntf 3.38 Mtap2 6.87 Tulp2 4.48 Fosl2 3.97 Cyp19a1* 3.35 Rab5b 6.44 Acsbg1 4.36 Pcdh8* 3.89 Inhba 3.35 Pbx1* 6.36 Kcnh2 4.33 Zeb2 3.86 Bcl9 3.35 Dzip1 6.29 Aldh1a1 4.32 Lama4 Downloaded from www.biolreprod.org. 3.85 Col5a2 3.34 Itch 6.11 Socs2 4.27 Malat1 3.80 Hsd3b1 3.32 Fzd2 5.79 Kitl 4.22 Dpyd 3.74 Atxn10 3.29 Pcdh8* 5.64 Pbx1* 4.21 Hunk * Transcripts common to both SAFs and LAFs (N ¼ 11). junctions, interactions with basal lamina, different hormone or both SAFs and LAFs, were compared with those increased by growth factor concentrations, or other factors influencing oocytes in OOX cumulus cells. In total, 1099 transcripts were cellular microenvironments or physical conditions. Because of higher in freshly isolated cumulus cells and were also our prior findings on the importance of ODPFs, we used two stimulated by ODPFs in OOX-cumulus cells. Unexpectedly, approaches to test the role of ODPFs in promoting higher 1660 transcripts that were higher in the freshly isolated levels of transcripts in cumulus cells versus MGCs. The first cumulus cells compared with MGCs were not significantly experimental paradigm, an in vitro approach, compared the stimulated by ODPFs in vitro. transcripts expressed at higher levels in cumulus cells freshly To further test the surprising result that many cumulus isolated from SAFs and LAFs with those elevated in cumulus transcripts expressed at higher levels than MGCs are not cells by ODPFs in vitro as reported previously [34]. The stimulated by ODPFs, we used an in vivo model. In a previous second approach, an in vivo one, compared the higher cumulus study [19], we assessed the transcriptomes of cumulus cells in cell transcripts with those expressed at lower levels in cumulus Gdf9þ/ Bmp15/ DM ovaries, wherein the oocytes are cells of Gdf9þ/ Bmp15/ double mutant (hereafter, DM) deficient in the production of ODPFs GDF9 and BMP15 cumulus cells compared to WT cumulus cells in vivo as compared to WT cumulus cells. Cumulus expansion is reported previously [19]. defective in the COCs of these mutant mice, and the defective For the in vitro analysis of the role of ODPFs on differential expansion was not remedied by coculture with WT oocytes, transcript expression, oocytes were microsurgically removed indicating that differentiation of the cumulus cells during from COCs isolated from eCG-stimulated ovaries and cultured follicular development requires these ODPFs [35]. Here we for 20 h without or with coculture with two oocytes/ll of compared the transcripts with significantly lower expression by medium. After removal of the oocytes, the cumulus cells are the DM cumulus cells versus WT cumulus cells with the cohort referred to as OOX cumulus cells. We reported effects of of transcripts found in the present study to be expressed at ODPFs on the transcriptome of OOX cumulus cells previously levels significantly higher in cumulus cells versus MGCs. [34], and the Jackson Laboratory Gene Expression Service Lower expression is the consequence of deficiencies in ODPFs carried out all the sample preparation and microarray protocols. GDF9 and BMP15 throughout follicular development. Thus, The same boundaries for significance established for the this experimental paradigm tests the effects of chronic ODPF present study, Q , 0.05 and fold change . 1.25, were applied deficiency on the cumulus cell transcriptome in vivo in contrast to all data sets. Transcripts whose steady state levels were to the model testing the acute effects of ODPFs on normal elevated by ODPFs in OOX cumulus cells were compared to OOX cumulus cells in vitro. As in the in vitro experiment, the the transcripts differentially expressed by freshly isolated cumulus cell transcripts expressed higher than in MGCs from cumulus cells and MGCs. Those transcripts expressed both SAFs and LAFs were pooled for comparison to the significantly more highly in freshly isolated cumulus cells transcripts reduced in the DM cumulus cells. A total of 843 versus MGCs and those whose expression was stimulated in transcripts expressed lower in DM cumulus cells were also OOX cumulus cells by oocytes in vitro are probably expressed expressed higher in normal cumulus cells versus MGCs. at higher levels in cumulus cells because of exposure to ODPFs Moreover, 1916 were expressed at higher levels in the normal in situ. cumulus cells versus MGCs but were unchanged in the DM All the transcripts with annotation of biological function cumulus cells versus WT. Of the 843 transcripts that were expressed at higher levels in cumulus cells than MGCs, from lower in DM cumulus cells, 487 were the same as those 9 Article 23 WIGGLESWORTH ET AL. Downloaded from www.biolreprod.org.

FIG. 7. VLAD depiction of GO terms in the contrast of consensus transcripts higher in cumulus cells than MGCs stimulated by ODPFs or not. The top 25 GO terms, selected by VLAD on the basis of local maximum P value, are shown. The size of the rectangle is proportional to the P value; the larger the rectangle the lower, and more significant, is the P value. The green portion of the rectangle shows the relative contribution of cumulus cell transcripts not stimulated by ODPFs to that GO term while the red portion shows the relative contribution of transcripts stimulated by ODPFs. Supplemental Table S8 presents complete list of transcripts expressed differently in SAFs and LAFs for each GO term. stimulated by ODPFs in OOX cumulus cells in vitro. Thus, and consensus ODPF-stimulated cumulus cells included those these 487 transcripts can be considered consensus transcripts enriched in biological function GO terms associated with whose expression in cumulus cells is stimulated by ODPFs catalytic processes in small metabolic processes, including (Supplemental Table S7). Importantly, of the 1916 transcripts glycolysis (GO:0006096, e.g., Pfkp) and cholesterol biosyn- that were lower in the DM relative to WT cumulus cells, 1301 thesis (GO:0008208, e.g., Mvk) as described previously [19, were in common with the cohort of transcripts not stimulated 20, 40]. Transcripts encoding molecules involved in other by ODPFs in vitro (Supplemental Table S7). Thus, these can be metabolic pathways such as GO:0009123 (nucleoside mono- considered consensus transcripts that are expressed more phosphate metabolic process) were also promoted by consen- highly in normal cumulus cells, but the higher levels of sus ODFPs. Previous studies have demonstrated that oocytes expression is not due to stimulation by ODPFs. promote the proliferation of granulosa cells [69–71]. Tran- The GO terms associated with consensus ODPF-stimulated versus not stimulated are compared in Figure 7 and scripts associated with cell proliferation GO terms such as Supplemental Table S8. By definition, these transcripts are GO:0022402 (cell cycle process) and GO:0006259 (DNA all more highly expressed in fresh cumulus cells than in MGCs. metabolic process) are therefore not surprisingly found in the Analysis of the GO terms associated with the consensus consensus ODPF-regulated group. The transcripts encoding transcripts for ODPF-stimulated or ODPF not stimulated receptors EDNRA, EDNRB, EGFR, and ESR1 as well as processes was similar to that found for transcripts defined by ligands IGF1, VEGFA, and VEGFB are also among those that either the in vitro or in vivo protocols, which are not shown. are more highly expressed by both cumulus cells in situ and Transcripts that were higher in freshly isolated cumulus cells stimulated by ODPFs in OOX cumulus cells. 10 Article 23 GRANULOSA CELL TRANSCRIPTOMIC DIVERSITY

Notable among the consensus transcripts expressed at levels In fact, the transcriptome changes in cumulus cells during the higher in cumulus cells than MGCs but not promoted by SAF to LAF development is qualitatively and quantitatively as ODPFs were the proto-oncogenes Fosb, Jun, and Junb. In significant as the transcriptomic transformation of the MGC addition to these components of the AP1 transcription population during the same follicular developmental span. complex, Ap1b1 and Ap1g1, encoding AP1 complex adaptor Previous studies demonstrated that cumulus cells provide small proteins, are in the group of consensus transcripts expressed nutritional and regulatory molecules to the oocyte via the gap higher in cumulus cells than MGCs but not regulated by junctions that couple the metabolism of these cells. In fact, ODPFs. Transcripts also apparently not regulated by ODPFs oocytes, via ODPFs, promote pathways in cumulus cells, such were apoptosis-regulating factors Bad, the transcription factor as the glycolytic and cholesterol-generating pathways, in which involved in follicular development Foxo3, all of the early the oocytes themselves are deficient [19, 40]. Moreover, growth response genes (Egr1/2/3), the ligands Efna1, Gh, Ghrl, previous studies have shown that ODPFs promote cumulus cell Shh, Tgfa, Pdgfa, and Wnt4, and transcripts encoding receptors division [69–71]. A general emerging theme is that the Acvr2b, Adra2c, Fgfr1, Igf1r, Ntrk3, and Pdgfra. In general, cumulus cell transcriptome is enriched in the transcripts of GO terms enriched with transcripts not controlled by ODPFs workhorse molecules that drive both metabolism and cell include those that encode regulators of biological processes division. As shown here, as the cumulus cells develop, they while those controlled by ODPFs encode proteins involved in also become enriched in transcripts that regulate metabolic and cell division and catalytic metabolic pathways (Fig. 7 and developmental pathways. Curiously, however, most of the Supplemental Table S8). regulatory transcripts appear controlled by factors other than ODPFs, although the catalytic pathways themselves are often DISCUSSION driven by ODPFs. Moreover, it is not yet clear how, or if, these transcriptomic changes in the cumulus cells relate to the This study has delineated, at a global level, major qualitative changes in oocytes during the SAF to LAF transcriptomic dynamics that underlie the structural and

developmental transition, when they increase their embryonic Downloaded from www.biolreprod.org. functional architecture of the ovarian follicle. The analysis developmental competence [72]. has revealed not only transcriptomic complexity, but also Follicular microenvironments are influential in defining the unanticipated regulatory differences that control cumulus cell functional diversity of the granulosa cell compartments. As diversification, development, and function. Scrutiny of gene shown both here and in other studies that focused on the lists and GO terms has revealed functions in cumulus cells and expression patterns of individual transcripts or pathways [73], MGCs not previously fully appreciated. For example, although cumulus cells and MGCs differ in their patterns of gene cumulus cells of both stages of follicular development are expression and these patterns change during the SAF to LAF competent to undergo expansion in vitro, they were otherwise transition. What determines these lineage diversifications? remarkably dissimilar with transcriptomic changes quantita- ODPFs diffusing from the oocyte are a major influence in tively equivalent to those of MGCs. GO analysis revealed that establishing the cumulus cell lineage [13]. It has been cumulus cells of small follicles were enriched in transcripts suggested that factors in apposition to the influence of ODPFs, generally associated with catalytic components of metabolic such as FSH, diffuse from outside of the follicle to establish the processes, while those from large follicles were involved in reversed gradients in levels of several transcripts, for example, regulation of metabolism, cell differentiation, and adhesion. for Lhcgr versus Pfkp. Lhcgr is expressed at highest levels near Contrast of cumulus cells versus MGCs revealed that cumulus the follicular-basal lamina [22, 74] while Pfkp is expressed at cells were enriched in transcripts associated with metabolism highest levels near the oocyte [20]. ODPFs suppressed and cell proliferation while MGCs were enriched for transcripts expression of Lhcgr by granulosa cells that was stimulated involved in cell signaling and differentiation. Together, these by both FSH and contact with basal lamina [24], suggesting findings validate that global systems approaches, such as that that ODPFs are the most influential factors in lineage taken here, can provide a richness of information that determination. In fact, evidence suggests that ODPFs orches- complements other more focused (biased) approaches. trate the rate of follicular development [75], probably by Oocytes are required for the progression of preantral promoting essential metabolic pathways and cell division. (secondary) follicles to the early antral (tertiary) follicle stage We hypothesized that the higher levels of expression of [15]. This pivotal transition in folliculogenesis propels the most transcripts expressed higher in cumulus cells are structural and functional divergence of the cumulus cell and promoted by ODPFs. There were 2759 total transcripts MGC lineages. Although it is well established that gonadotro- expressed more highly by cumulus cells than MGCs of SAFs pins drive the development of SAFs to LAFs and promote the and LAFs together. Using data from both the in vitro and in expression of genes key to steroidogenesis and the ability to vivo models used to test this hypothesis, a total of 52% (1455) respond to LH by MGCs, an analysis of the developmental of the transcripts depend upon ODPFs; 39% (843) were acutely changes in gene expression in the cumulus and MGC somatic stimulated using the in vitro model. Thus, there is a remarkable compartments before the LH-surge (or hCG-treatment) has reliance on ODPFs to promote in cumulus cells higher levels of been lacking. Such information is essential to define the transcripts that are involved in cell division or catalytic transcriptomic foundation of the cellular and functional metabolic pathways. Nevertheless, 48% of the transcripts that architecture of the follicle and to determine how differential were higher in cumulus cells were apparently not stimulated by gene expression relates to the fates of both the somatic and ODPFs. This unanticipated result indicates that factors other oocyte compartments. than ODPFs are also major determinants of transcriptomic The cumulus cells of SAFs are competent to undergo diversity of cumulus cells and MGCs. expansion in vitro in a manner that appears similar to that of Discovering the mechanisms for apparent ODPF-indepen- cumulus cells retrieved from LAFs isolated from follicles dent elevated expression of some transcripts in cumulus cells before the preovulatory LH surge ([17] and results presented versus MGCs is a future challenge. It is possible that oocytes here [Supplemental Fig. S3]). Despite this similarity, this study stimulate expression by the cumulus cells via mechanism has shown that transcripts of cumulus cells from SAFs are requiring direct contact rather than ODPFs. However, no actually quite different than those in cumulus cells from LAFs. differences were detected in the transcriptomes of cumulus 11 Article 23 WIGGLESWORTH ET AL.

decrease the levels of expression of some transcripts in MGCs relative to cumulus cells, which would account for the transcriptomic diversity observed here. Other factors present in the granulosa cell microenvironments—estrogens and growth factors, for example—probably also participate in the architecture of transcriptomic diversity of cumulus cells and MGCs fundamental to cellular function and coordination during follicular development.

ACKNOWLEDGMENT We are grateful to Tim Stearns of the Jackson Laboratory Computa- tional Sciences Service for statistical analyses of microarray data and for spearheading the Gene Expression Omnibus submission of the data, to Chrystal Snow of The Jackson Laboratory Gene Expression Service for FIG. 8. Working model depicting possible mechanisms regulating the sample preparation and running the microarrays, and to Joel Richardson of expression of two groups of transcripts enriched in cumulus cells relative the Mouse Genome Informatics Program for heroic assistance in the use of to MGCs. These groups of transcripts affect the diversity in functions of VLAD and for allowing us to have a small part in its evolution. We also these cell types. Both groups are expressed at higher levels in cumulus thank Dr. Mary Ann Handel for her comments during the preparation of cells than in MGCs. One group of transcripts is elevated in cumulus by this manuscript oocyte-derived paracrine factors (ODPFs) and the other is not. The group whose expression is stimulated by ODPFs includes, among others, transcripts enriched in the GO terms associated with cell division and REFERENCES catalytic enzymes of metabolic pathways. The transcripts not affected by ODPFs include, among others, transcripts enriched in the GO terms 1. Park JY, Su YQ, Ariga M, Law E, Jin SL, Conti M. EGF-like growth associated with the regulation of these pathways and various aspects of factors as mediators of LH action in the ovulatory follicle. Science 2004; Downloaded from www.biolreprod.org. gene expression (see Fig. 7 and Supplemental Table S8). The transcripts 303:682–684. unaffected by ODPFs may be expressed at higher levels in cumulus cells 2. Shimada M, Hernandez-Gonzalez I, Gonzalez-Robayna I, Richards JS. than MGCs because expression is suppressed in the MGCs. Paracrine and autocrine regulation of epidermal growth factor-like factors in cumulus oocyte complexes and granulosa cells: key roles for prostaglandin synthase 2 and progesterone receptor. Mol Endocrinol cells cultured either as intact oocyte-cumulus cell complexes or 2006; 20:1352–1365. 3. Zhang M, Su YQ, Sugiura K, Xia G, Eppig JJ. Granulosa cell ligand OOX cumulus cells cocultured with denuded oocytes. This NPPC and its receptor NPR2 maintain meiotic arrest in mouse oocytes. suggests that oocytes do not affect the cumulus cell tran- Science 2010; 330:366–369. scriptome via juxtacrine or gap junction-mediated communi- 4. Dekel N, Hillensjo T, Kraicer PF. Maturational effects of gonadotropins cation, at least under the culture conditions used in that study on the cumulus-oocyte complex of the rat. Biol Reprod 1979; 20:191–197. [34]. In the in vitro experimental approach, the action of 5. Chen L, Russell PT, Larsen WJ. Functional significance of cumulus ODPFs was determined by coculturing OOX cumulus cells expansion in the mouse: roles for the preovulatory synthesis of hyaluronic with denuded fully grown oocytes. Thus, the ODPFs contained acid within the cumulus mass. Mol Reprod Dev 1993; 34:87–93. 6. Salustri A, Yanagishita M, Hascall VC. Synthesis and accumulation of the natural cocktail of factors released by oocytes in vitro and hyaluronic acid and proteoglycans in the mouse cumulus cell-oocyte not arbitrarily selected recombinant ODPFs used individually complex during follicle-stimulating hormone-induced mucification. J Biol or in combination. This protocol has been used in several Chem 1989; 264:13840–13847. previous studies to demonstrate the ability of ODPFs to 7. Anderson E, Albertini DF. Gap junctions between the oocyte and promote or suppress several transcripts in cumulus cells [13, companion follicle cells in the mammalian ovary. J Cell Biol 1976; 71: 19, 24, 34]. It is possible that expression of some cumulus cell 680–686. transcripts require that ODPFs act in concert with paracrine 8. Gilula NB, Epstein ML, Beers WH. Cell-to-cell communication and ovulation. A study of the cumulus-oocyte complex. J Cell Biol 1978; 78: factors produced within the follicle or endocrine factors, such 58–75. as FSH or estrogens. 9. Schultz RM. Roles of cell-to-cell communication in development. Biol What, if not ODPFs, results in the higher levels of these Reprod 1985; 32:27–42. transcripts in cumulus cells? Clearly, this should be an 10. Albertini DF, Barrett SL. Oocyte-somatic cell communication. Reproduc- objective of future studies, but our working hypothesis is tion 2002; 49–54. presented in Figure 8. Although intrinsic developmental 11. Norris RP, Ratzan WJ, Freudzon M, Mehlmann LM, Krall J, Movsesian MA, Wang H, Ke H, Nikolaev VO, Jaffe LA. Cyclic GMP from the programs cannot be excluded, the relative level of transcript surrounding somatic cells regulates cyclic AMP and meiosis in the mouse expression in cumulus cells versus MGCs is probably the net oocyte. Development 2009; 136:1869–1878. result of stimulation by ODPFs in cumulus cells, suppression in 12. Anderson E, Wilkinson RF, Lee G, Meller S. A correlative microscopical MGCs, or both. FSH suppresses levels of Ar and Slc38a3 analysis of differentiating ovarian follicles of mammals. J Morph 1978; mRNAs in isolated COCs wherein ODPFs stimulated expres- 156:339–366. sion, thus imparting some MGCs-like phenotypic characteris- 13. Diaz FJ, Wigglesworth K, Eppig JJ. Oocytes determine cumulus cell tics to cumulus cells [13]. Therefore, FSH could be a key factor lineage in mouse ovarian follicles. J Cell Sci 2007; 120:1330–1340. 14. Zhang M, Su YQ, Sugiura K, Wigglesworth K, Xia G, Eppig JJ. Estradiol in establishing or maintaining the diversity of granulosa cell promotes and maintains cumulus cell expression of natriuretic peptide lineages and is certainly needed, as well as ODPFs, for driving receptor 2 (NPR2) and meiotic arrest in mouse oocytes in vitro. follicular development. Because FSH enters by diffusion from Endocrinology 2011; 152:4377–4385. outside the follicle, and much of it is probably bound to FSH 15. Diaz FJ, Wigglesworth K, Eppig JJ. Oocytes are required for the preantral receptors abundant on MGCs [76], the concentration of FSH granulosa cell to cumulus cell transition in mice. Dev Biol 2007; 305: probably decreases with increasing depth into the follicle, 300–311. 16. Eppig JJ, Wigglesworth K, Chesnel F. Secretion of cumulus expansion minimizing the exposure of cumulus cells to FSH in vivo. enabling factor by mouse oocytes: relationship to oocyte growth and Moreover, the follicular basal lamina, contacted by most competence to resume meiosis. Dev Biol 1993; 158:400–409. MGCs [77], augments some actions of FSH [22, 24]. Thus, the 17. Vanderhyden BC, Caron PJ, Buccione R, Eppig JJ. Developmental pattern combination of FSH and basal lamina could function to of the secretion of cumulus-expansion enabling factor by mouse oocytes 12 Article 23 GRANULOSA CELL TRANSCRIPTOMIC DIVERSITY

and the role of oocytes in promoting granulosa cell differentiation. Dev Analysis of Gene Expression Data. An Overview of Methods and Biol 1990; 140:307–317. Software: Methods and Software. New York: Springer; 2003. 18. Eppig JJ, Pendola FL, Wigglesworth K, Pendola JK. Mouse oocytes 38. Cui X, Hwang JT, Qiu J, Blades NJ, Churchill GA. Improved statistical regulate metabolic cooperativity between granulosa cells and oocytes: tests for differential gene expression by shrinking variance components amino acid transport. Biol Reprod 2005; 73:351–357. estimates. Biostatistics 2005; 6:59–75. 19. Su YQ, Sugiura K, Wigglesworth K, O’Brien MJ, Affourtit JP, Pangas 39. Storey JD, Tibshirana R. Statistical significance for genomewide studies. SA, Matzuk MM, Eppig JJ. Oocyte regulation of metabolic cooperativity Proc Natl Acad Sci USA 2003; 100:9440–9445. between mouse cumulus cells and oocytes: BMP15 and GDF9 control 40. Sugiura K, Su YQ, Diaz FJ, Pangas SA, Sharma S, Wigglesworth K, cholesterol biosynthesis in cumulus cells. Development 2008; 135: O’Brien MJ, Matzuk MM, Shimasaki S, Eppig JJ. Oocyte-derived BMP15 111–121. and FGFs cooperate to promote glycolysis in cumulus cells. Development 20. Sugiura K, Pendola FL, Eppig JJ. Oocyte control of metabolic 2007; 134:2593–2603. cooperativity between oocytes and companion granulosa cells: energy 41. Sugiura K, Su YQ, Li Q, Wigglesworth K, Matzuk MM, Eppig JJ. metabolism. Dev Biol 2005; 279:20–30. Fibroblast growth factors and epidermal growth factor cooperate with 21. Richards JS. Perspective: the ovarian follicle—a perspective in 2001. oocyte-derived members of the TGFbeta superfamily to regulate Spry2 Endocrinology 2001; 142:2184–2193. mRNA levels in mouse cumulus cells. Biol Reprod 2009; 81:833–841. 22. Furman A, Rotmensch S, Kohen F, Mashiach S, Amsterdam A. 42. Sugiura K, Su YQ, Eppig JJ. Does bone morphogenetic protein 6 (BMP6) Regulation of rat granulosa cell differentiation by extracellular matrix affect female fertility in the mouse? Biol Reprod 2010; 83:997–1004. produced by bovine corneal endothelial cells. Endocrinology 1986; 118: 43. Hogg K, Etherington SL, Young JM, McNeilly AS, Duncan WC. Inhibitor 1878–1885. of differentiation (Id) genes are expressed in the steroidogenic cells of the 23. Huet C, Pisselet C, Mandon-Pepin B, Monget P, Monniaux D. ovine ovary and are differentially regulated by members of the Extracellular matrix regulates ovine granulosa cell survival, proliferation transforming growth factor-beta family. Endocrinology 2010; 151: and steroidogenesis: relationships between cell shape and function. J 1247–1256. Endocrinol 2001; 169:347–360. 44. Verbraak EJ, van ‘t Veld EM, Groot Koerkamp M, Roelen BA, van 24. Eppig JJ, Wigglesworth K, Pendola F, Hirao Y. Murine oocytes suppress Haeften T, Stoorvogel W, Zijlstra C. Identification of genes targeted by expression of luteinizing hormone receptor messenger ribonucleic acid by FSH and oocytes in porcine granulosa cells. Theriogenology 2011; 75: granulosa cells. Biol Reprod 1997; 56:976–984. 362–376.

25. Peng J, Li Q, Wigglesworth K, Rangarajan A, Kattamuri C, Peterson RT, 45. Sharma SC, Richards JS. Regulation of AP1 (Jun/Fos) factor expression Downloaded from www.biolreprod.org. Eppig JJ, Thompson TB, Matzuk MM. Growth differentiation factor 9: and activation in ovarian granulosa cells. Relation of JunD and Fra2 to bone morphogenetic protein 15 heterodimers are potent regulators of terminal differentiation. J Biol Chem 2000; 275:33718–33728. ovarian functions. Proc Natl Acad Sci U S A 2013; 110:E776–E785. 46. Meissner A, Wernig M, Jaenisch R. Direct reprogramming of genetically 26. Fulop C, Szanto S, Mukhopadhyay D, Bardos T, Kamath RV, Rugg MS, unmodified fibroblasts into pluripotent stem cells. Nat Biotechnol 2007; Day AJ, Salustri A, Hascall VC, Glant TT, Mikecz K. Impaired cumulus 25:1177–1181. mucification and female sterility in tumor necrosis factor-induced protein- 47. Calounova G, Livera G, Zhang XQ, Liu K, Gosden RG, Welsh M. The Src 6 deficient mice. Development 2003; 130:2253–2261. homology 2 domain-containing adapter protein B (SHB) regulates mouse 27. Ochsner SA, Day AJ, Rugg MS, Breyer RM, Gomer RH, Richards JS. oocyte maturation. PLoS One 2010; 5:e11155. Disrupted function of tumor necrosis factor-alpha-stimulated gene 6 48. Maizels ET, Cottom J, Jones JC, Hunzicker-Dunn M. Follicle stimulating blocks cumulus cell-oocyte complex expansion. Endocrinology 2003; 144: hormone (FSH) activates the p38 mitogen-activated protein kinase 4376–4384. pathway, inducing small heat shock protein phosphorylation and cell 28. Ochsner SA, Russell DL, Day AJ, Breyer RM, Richards JS. Decreased rounding in immature rat ovarian granulosa cells. Endocrinology 1998; expression of tumor necrosis factor-alpha-stimulated gene 6 in cumulus 139:3353–3356. cells of the cyclooxygenase-2 and EP2 null mice. Endocrinology 2003; 49. Su YQ, Wigglesworth K, Pendola FL, O’Brien MJ, Eppig JJ. Mitogen- 144:1008–1019. activated protein kinase activity in cumulus cells is essential for 29. Varani S, Elvin JA, Yan C, DeMayo J, DeMayo FJ, Horton HF, Byrne gonadotropin-induced oocyte meiotic resumption and cumulus expansion MC, Matzuk MM. Knockout of pentraxin 3, a downstream target of in the mouse. Endocrinology 2002; 143:2221–2232. growth differentiation factor-9, causes female subfertility. Mol Endocrinol 50. Su YQ, Denegre JM, Wigglesworth K, Pendola FL, O’Brien MJ, Eppig JJ. 2002; 16:1154–1167. Oocyte-dependent activation of mitogen-activated protein kinase (ERK1/ 30. Hernandez-Gonzalez I, Gonzalez-Robayna I, Shimada M, Wayne CM, 2) in cumulus cells is required for the maturation of the mouse oocyte- Ochsner SA, White L, Richards JS. Gene expression profiles of cumulus cumulus cell complex. Dev Biol 2003; 263:126–138. cell oocyte complexes during ovulation reveal cumulus cells express 51. Sela-Abramovich S, Chorev E, Galiani D, Dekel N. Mitogen-activated neuronal and immune-related genes: does this expand their role in the protein kinase mediates luteinizing hormone-induced breakdown of ovulation process? Mol Endocrinol 2006; 20:1300–1321. communication and oocyte maturation in rat ovarian follicles. Endocri- 31. Shimada M, Hernandez-Gonzalez I, Gonzalez-Robanya I, Richards JS. nology 2005; 146:1236–1244. Induced expression of pattern recognition receptors in cumulus oocyte 52. Hunzicker-Dunn M, Maizels ET. FSH signaling pathways in immature complexes: novel evidence for innate immune-like functions during granulosa cells that regulate target gene expression: branching out from ovulation. Mol Endocrinol 2006; 20:3228–3239. protein kinase A. Cell Signal 2006; 18:1351–1359. 32. Xu F, Stouffer RL, Muller J, Hennebold JD, Wright JW, Bahar A, Leder 53. Diaz FJ, O’Brien MJ, Wigglesworth K, Eppig JJ. The preantral granulosa G, Peters M, Thorne M, Sims M, Wintermantel T, Lindenthal B. cell to cumulus cell transition in the mouse ovary: development of Dynamics of the transcriptome in the primate ovulatory follicle. Mol Hum competence to undergo expansion. Dev Biol 2006; 299:91–104. Reprod 2011; 17:152–165. 54. Miyoshi T, Otsuka F, Inagaki K, Otani H, Takeda M, Suzuki J, Goto J, 33. Ouandaogo ZG, Frydman N, Hesters L, Assou S, Haouzi D, Dechaud H, Ogura T, Makino H. Differential regulation of steroidogenesis by bone Frydman R, Hamamah S. Differences in transcriptomic profiles of human morphogenetic proteins in granulosa cells: involvement of extracellularly cumulus cells isolated from oocytes at GV, MI and MII stages after in vivo regulated kinase signaling and oocyte actions in follicle-stimulating and in vitro oocyte maturation. Hum Reprod 2012; 27:2438–2447. hormone-induced estrogen production. Endocrinology 2007; 148: 34. Emori C, Wigglesworth K, Fujii W, Naito K, Eppig JJ, Sugiura K. 337–345. Cooperative effects of 17beta-estradiol and oocyte-derived paracrine 55. Sriraman V, Modi SR, Bodenburg Y, Denner LA, Urban RJ. Identification factors on the transcriptome of mouse cumulus cells. Endocrinology 2013; of ERK and JNK as signaling mediators on protein kinase C activation in 154:4859–4871. cultured granulosa cells. Mol Cell Endocrinol 2008; 294:52–60. 35. Su YQ, Wu X, O’Brien MJ, Pendola FL, Denegre JN, Matzuk MM, Eppig 56. Liu Z, Fan HY, Wang Y, Richards JS. Targeted disruption of Mapk14 JJ. Synergistic roles of BMP15 and GDF9 in the development and (p38MAPKalpha) in granulosa cells and cumulus cells causes cell-specific function of the oocyte-cumulus cell complex in mice: genetic evidence for changes in gene expression profiles that rescue COC expansion and an oocyte-granulosa cell regulatory loop. Dev Biol 2004; 276:64–73. maintain fertility. Mol Endocrinol 2010; 24:1794–1804. 36. Dai M, Wang P, Boyd AD, Kostov G, Athey B, Jones EG, Bunney WE, 57. Cho J, Kim H, Kang DW, Yanagisawa M, Ko C. Endothelin B receptor is Myers RM, Speed TP, Akil H, Watson SJ, Meng F. Evolving gene/ not required but necessary for finite regulation of ovulation. Life Sci 2012; transcript definitions significantly alter the interpretation of GeneChip 91:613–617. data. Nucleic Acids Res 2005; 33:e175. 58. Kawamura K, Ye Y, Liang CG, Kawamura N, Gelpke MS, Rauch R, 37. Wu H, Kerr MK, Cui XQ, Churchill GA. MAANOVA; A Software Tanaka T, Hsueh AJ. Paracrine regulation of the resumption of oocyte Package for the Analysis of Spotted cDNA Microarray Experiments in the meiosis by endothelin-1. Dev Biol 2009; 327:62–70. 13 Article 23 WIGGLESWORTH ET AL.

59. Lavery DN, Villaronga MA, Walker MM, Patel A, Belandia B, Bevan CL. regulated transcript throughout in vitro follicle growth. Mol Reprod Dev Repression of androgen receptor activity by HEYL, a third member of the 2013; 80:132–144. Hairy/Enhancer-of-split-related family of Notch effectors. J Biol Chem 68. Svensson L, Aszodi A, Heinegard D, Hunziker EB, Reinholt FP, Fassler 2011; 286:17796–17808. R, Oldberg A. Cartilage oligomeric matrix protein-deficient mice have 60. Johnson J, Espinoza T, McGaughey RW, Rawls A, Wilson-Rawls J. normal skeletal development. Mol Cell Biol 2002; 22:4366–4371. Notch pathway genes are expressed in mammalian ovarian follicles. Mech 69. Vanderhyden BC, Telfer EE, Eppig JJ. Mouse oocytes promote Dev 2001; 109:355–361. proliferation of granulosa cells from preantral and antral follicles in vitro. 61. Zhang CP, Yang JL, Zhang J, Li L, Huang L, Ji SY, Hu ZY, Gao F, Liu Biol Reprod 1992; 46:1196–1204. YX. Notch signaling is involved in ovarian follicle development by 70. Gilchrist RB, Ritter LJ, Armstrong DG. Mouse oocyte mitogenic activity regulating granulosa cell proliferation. Endocrinology 2011; 152: is developmentally coordinated throughout folliculogenesis and meiotic 2437–2447. maturation. Dev Biol 2001; 240:289–298. 62. Artac RA, McFee RM, Smith RA, Baltes-Breitwisch MM, Clopton DT, 71. Hickey TE, Marrocco DL, Amato F, Ritter LJ, Norman RJ, Gilchrist RB, Cupp AS. Neutralization of vascular endothelial growth factor antiangio- Armstrong DT. Androgens augment the mitogenic effects of oocyte- genic isoforms is more effective than treatment with proangiogenic secreted factors and growth differentiation factor 9 on porcine granulosa isoforms in stimulating vascular development and follicle progression in cells. Biol Reprod 2005; 73:825–832. the perinatal rat ovary. Biol Reprod 2009; 81:978–988. 72. Pan H, O’Brien MJ, Wigglesworth K, Eppig JJ, Schultz RM. Transcript 63. Qiu Y, Seager M, Osman A, Castle-Miller J, Bevan H, Tortonese DJ, profiling during mouse oocyte development and the effect of gonadotropin Murphy D, Harper SJ, Fraser HM, Donaldson LF, Bates DO. Ovarian priming and development in vitro. Dev Biol 2005; 286:493–506. VEGF(165)b expression regulates follicular development, corpus luteum function and fertility. Reproduction 2012; 143:501–511. 73. Su YQ, Sugiura K, Eppig JJ. Mouse oocyte control of granulosa cell 64. Nautiyal J, Steel JH, Rosell MM, Nikolopoulou E, Lee K, Demayo FJ, development and function: paracrine regulation of cumulus cell metabo- White R, Richards JS, Parker MG. The nuclear receptor cofactor receptor- lism. Semin Reprod Med 2009; 27:32–42. interacting protein 140 is a positive regulator of amphiregulin expression 74. Amsterdam A, Koch Y, Lieberman ME, Lindner HR. Distribution of and cumulus cell-oocyte complex expansion in the mouse ovary. binding sites for human chorionic gonadotropin in the preovulatory follicle Endocrinology 2010; 151:2923–2932. of the rat. J Cell Biol 1975; 67:894–900. 65. Sugiura K, Su YQ, Li Q, Wigglesworth K, Matzuk MM, Eppig JJ. 75. Eppig JJ, Wigglesworth K, Pendola FL. The mammalian oocyte Estrogen promotes the development of mouse cumulus cells in orchestrates the rate of ovarian follicular development. Proc Natl Acad Downloaded from www.biolreprod.org. coordination with oocyte-derived GDF9 and BMP15. Mol Endocrinol Sci U S A 2002; 99:2890–2894. 2010; 24:2303–2314. 76. Richards JS, Ireland JJ, Rao MC, Bernath GA, Midgley AR Jr, Reichert 66. Buensuceso AV, Deroo BJ. The ephrin signaling pathway regulates LE Jr. Ovarian follicular development in the rat: hormone receptor morphology and adhesion of mouse granulosa cells in vitro. Biol Reprod regulation by estradiol, follicle stimulating hormone and luteinizing 2013; 88:25. hormone. Endocrinology 1976; 99:1562–1570. 67. Skory RM, Bernabe BP, Galdones E, Broadbelt LJ, Shea LD, Woodruff 77. Lipner H, Cross NL. Morphology of the membrana granulosa of the TK. Microarray analysis identifies COMP as the most differentially ovarian follicle. Endocrinology 1968; 82:638–641.

14 Article 23 Sagvekar et al. Clinical Epigenetics (2019) 11:61 https://doi.org/10.1186/s13148-019-0657-6

RESEARCH Open Access DNA methylome profiling of granulosa cells reveals altered methylation in genes regulating vital ovarian functions in polycystic ovary syndrome Pooja Sagvekar1, Pankaj Kumar2, Vijay Mangoli3, Sadhana Desai3 and Srabani Mukherjee1*

Abstract Background: Women with polycystic ovary syndrome (PCOS) manifest a host of ovarian defects like impaired folliculogenesis, anovulation, and poor oocyte quality, which grossly affect their reproductive health. Addressing the putative epigenetic anomalies that tightly regulate these events is of foremost importance in this disorder. We therefore aimed to carry out DNA methylome profiling of cumulus granulosa cells and assess the methylation and transcript expression profiles of a few differentially methylated genes contributing to ovarian defects in PCOS. A total of 20 controls and 20 women with PCOS were selected from a larger cohort of women undergoing IVF, after carefully screening their sera and follicular fluids for hormonal and biochemical parameters. DNA extracted from cumulus granulosa cells of three women each, from control and PCOS groups was subjected to high-throughput, next generation bisulfite sequencing, using the Illumina HiSeq 2500® platform. Remaining samples were used for the validation of methylation status of some identified genes by pyrosequencing, and the transcript expression profiles of these genes were assessed by quantitative real-time PCR. Results: In all, 6486 CpG sites representing 3840 genes associated with Wnt signaling, G protein receptor, endothelin/integrin signaling, angiogenesis, chemokine/cytokine-mediated inflammation, etc., showed differential methylation in PCOS. Hypomethylation was noted in 2977 CpGs representing 2063 genes while 2509 CpGs within 1777 genes showed hypermethylation. Methylation differences were also noted in noncoding RNAs regulating several ovarian functions that are dysregulated in PCOS. Few differentially methylated genes such as aldo-keto reductase family 1 member C3, calcium-sensing receptor, resistin, mastermind-like domain 1, growth hormone- releasing hormone receptor and tumor necrosis factor, which predominantly contribute to hyperandrogenism, premature luteolysis, and oocyte development defects, were explored as novel epigenetic candidates in mediating ovarian dysfunction. Methylation profiles of these genes matched with our NGS findings, and their transcript expression patterns correlated with the gene hypo- or hypermethylation status. Conclusion: Our findings suggest that the epigenetic dysregulation of genes involved in important processes associated with follicular development may contribute to ovarian defects observed in women with PCOS. Keywords: PCOS, Methylome profiling, DNA methylation, IVF, Granulosa cells, NGS

* Correspondence: [email protected]; [email protected] 1Department of Molecular Endocrinology, ICMR-National Institute for Research in Reproductive Health, J.M. Street, Parel, Mumbai, Maharashtra 400012, India Full list of author information is available at the end of the article

© The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 2 of 16

Background follicular growth and dominance, ovulation, oocyte devel- Polycystic ovary syndrome (PCOS), one of the leading opmental competence, cumulus-oophorus complex causes of anovulatory infertility, is characterized by ovar- (COC) expansion, luteal maintenance, etc., which are ian, neuroendocrine, and metabolic perturbations in under stringent control of gonadotropins and other hor- women of reproductive age group. With a global preva- mones, have been well documented so far. Therefore, lence of 6–15% [1], it generally features irregularity or identification of locus/gene-specific epigenetic alterations absence of menses and presence of multiple ovarian in the ovaries of these women is of prime significance to cysts on ultrasonography, in addition to hyperandrogen- understand the pathophysiology of this disorder. emia and systemic insulin excess. Increased pulsatility of So far, few studies conducted to identify global DNA GnRH neurons at the hypothalamo-hypophyseal inter- methylation differences in PBLs and mural as well as cu- face elevates gonadotropic secretion of LH over FSH, mulus granulosa cells (CGCs) of women with PCOS have which alongside co-gonadotropic actions of insulin, pro- yielded ambiguous findings [6, 8, 9]. Also, promoter motes increased production of androgens from ovarian methylation profiles of a few established candidate genes folliculo-thecal cells [2, 3]. Adding to the pool of free an- of PCOS including yes-associated protein (YAP1), follista- drogens in circulation is the insulin-mediated suppres- tin (FST), aromatase (CYP19A1) and luteinizing hormone sion of hepatic sex hormone binding globulin (SHBG), chorionic gonadotropin receptor (LHCGR) have been in- which increases the bioavailability of testosterone, vestigated in these cells, and ovarian tissues by some thereby affecting its clearance [4]. The cumulative ac- groups till date [10–14]. Among these genes, LHCGR has tions of all these events trigger a series of physiological been consistently reported to be hypomethylated in defects including ovarian cyst formation, amenorrhoea, women with PCOS and in animal models of PCO [14, 15]. anovulation, infertility, hyperandrogenism, insulin resist- Subsequently, a few high-throughput attempts were made ance, hyperinsulinemia, obesity, glucose intolerance, to identify some differentially methylated genes (DMGs) lipid abnormalities, type 2 diabetes mellitus (T2DM), in women with PCOS. These included the use of diverse hypertension, and cardiovascular disease. In parallel, approaches such as methylated DNA immunoprecipita- there is an intrinsic elevation of anti-Mullerian hormone tion (MeDIP) [16] and Illumina platform-based methyla- (AMH) due to the presence of cystic follicles arrested in tion microarray [17] in peripheral blood leukocytes; preantral to antral stages [5]. These key events dictate MeDIP coupled with methyl promoter enrichment micro- the principal dogma behind the pathophysiology of array [18], as well as methylation microarray combined PCOS understood so far. with microarray-based transcriptome analysis [19]inovar- Although genetic factors impacting the development ian tissue biopsies and adipose tissue samples [20]; total and progression of PCOS have been amply investigated, RNA sequencing (RNA-seq) coupled with methylation identification and experimental corroboration of cognate measured by base cleavage and mass spectrometry (Epi- epigenetic factors that may contribute to the pathophysi- TYPER) [21]; and lastly, methylation microarray in mural ology of this multifaceted disorder remain enigmatic. Both granulosa cells (MGCs) collected from women undergo- environmental and physiological factors serve as strong ing controlled ovarian hyperstimulation (COH) [22]. determinants for epigenetic alterations. Factors such as in- However, next generation sequencing (NGS)-based trinsic hormonal aberrations, dysregulation of intrauterine methylome studies spanning individual CpG sites, specif- milieu by endocrine disruptors (EDCs) during gestational ically in CGCs which participate in extensive cross talk be- periods, and lifestyle modifications in subsequent phases tween the developing oocyte and surrounding follicular of growth and development have been recently implicated milieu while facilitating meiotic maturation and ovulation in epigenetic predisposition to this disease. Pioneering in- of competent oocytes, are yet to be reported in women vestigations on epigenetic alterations in PCOS began with with PCOS. In this study, we have carried out comparative studies on peripheral blood leukocytes (PBLs) [6]. How- genome-wide bisulfite sequencing of CGCs obtained from ever, tissue specificity of epigenetic modifications renders women with PCOS and normovulatory healthy controls, it difficult to extrapolate epigenetic data derived from cir- using a NGS-based, multiplexed Methyl-Capture Sequen- culating cells like PBLs, to organs such as ovaries or adi- cing (MC-Seq) approach. MC-Seq is advantageous over pose tissues that are highly affected in PCOS [7]. This reduced representative bisulfite sequencing (RRBS) and necessitates the undertaking of clinical epigenetic studies MeDIP-Seq, in that it avoids the over-representation of at a tissue-specific level. Ovary withstands most of the recurring reads and also over Infinium 450 K hormonal assaults triggered by systemic aberrations in the methyl-microarray wherein it enables the user to opt for neuroendocrine-ovarian axis. It is therefore a primary hot- much greater coverage of the epigenome (3.4 million CpG spot for epigenetic perturbations, which may contribute to sites and 20.8 million non-CpG sites, as opposed to the multiple follicular and oocyte defects observed in 4,50,000 CpG sites and 3091 non-CpG loci, respectively) women with PCOS. Defects related to steroidogenesis, [23]. Also, Infinium microarrays target only those genes Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 3 of 16

that are known to be differentially methylated in some DMGs, 438 genes harbored both hyper- and hypomethy- cancers while MC-Seq differentiates between any genomic lated CpG sites. Additionally, many noncoding RNAs in- region bearing altered methylation marks. Additionally, cluding 44 microRNAs (miRs) and 121 pseudogenes also compared to whole-genome bisulfite sequencing (WGBS), showed differential methylation in women with PCOS MC-Seq is cost effective and can identify novel genomic (Additional files 1 and 2). Of the 44 differentially methyl- loci while reducing the processing time associated with ated miRs, several miRs (miR23A, miR127, miR10B, WGBS. This study provides insights on genome-wide data miR193A mir200B, miR182, and miR140) have been re- on CpG sites that show altered methylation at a single ported to be implicated in impaired follicle growth and base resolution in CGCs of women with PCOS. steroidogenesis, anovulation, obesity, glucose metabolism, and so on [24]. In pathway enrichment and gene ontology Results (GO) analyses of the hypomethylated, hypermethylated, Hormonal characterization and oocyte quality parameters and combined gene lists, 227 (11%), 210 (11.82%), and 396 of the study participants (11.64%) genes from the three respective categories could All study participants were subjected to stringent anthropo- not be annotated. Among the identified pathways, those metric, hormonal, and biochemical characterization prior for Wnt signaling (Panther-GO-ID, P00057), integrin sig- to further investigation (Table 1). Baseline hormonal and naling (P00034), endothelin signaling (P00019), and cad- biochemical profiling of 20 controls and 20 women with herin signaling (P00012) were enriched in both the hypo- PCOS (days 3–5 of the follicular phase of menstrual cycle) and hypermethylated gene sets (Fig. 1). Other prominent showed that the baseline levels (day 3 follicular phase esti- pathways included the platelet-derived growth factor mates in serum) of luteinizing hormone (LH), LH/follicle-s- (PDGF) signaling (P00047), inflammation mediated by timulating hormone (FSH) ratio, and of AMH were high, chemokine and cytokine signaling (P00031), angiogenesis while the levels of FSH were low in women with PCOS, (P00005) and vascular endothelial growth factor (VEGF) compared to controls. Prolactin and thyroid-stimulating signaling (P00056), fibroblast growth factor (FGF) signal- hormone (TSH) levels were similar in both groups. Serum ing (P00021), G protein signaling (P00026, P00027), T cell levels of estradiol (E2) and progesterone (P4)measuredbe- activation (P00053), and nicotinic acetylcholine receptor fore the administration of recombinant human chorionic signaling pathways (P00044). gonadotropin (rhCG) were comparable between controls and PCOS. Serum E2 levels on the day of rhCG administra- Validation of genes showing differential methylation in tion were high in women with PCOS while P4 levels were regions upstream to transcription start sites similar between these groups. Analysis of oocyte parame- Among the DMGs selected for validation, the upstream ters revealed that the total numbers of follicles, oocytes CpG sites of five genes, namely aldo-keto reductase 1 fam- (mature + immature), mature MII oocytes, %MII oocytes, ily C3 (AKR1C3), calcium-sensing receptor (CASR), growth and number of fertilized oocytes were unchanged between hormone-releasing hormone receptor (GHRHR), resistin controls and women with PCOS. However, the rates of (RETN), and mastermind-like domain 1 (MAMLD1)were fertilization of MII oocytes were low in women with PCOS hypomethylated while those of transferrin (TF)andtumor compared to controls. In serum and follicular fluid (FF) necrosis factor (TNF) were hypermethylated in PCOS in samples collected on the day of oocyte pick up (d-OPU), our methylome analysis. We first investigated whether total testosterone (TT) levels were high and sex hormone these seven genes are expressed in CGCs, and whether they binding globulin (SHBG) levels were low in PCOS. are differentially expressed in women with PCOS using Androgen-excess indices such as free testosterone (Free-T), qPCR (Fig. 2). Apart from GHRHR,whichwasexpressed bioavailable testosterone (Bio-T), and free androgen index only at baseline levels, the remaining six genes were abun- (FAI) were high in FF, while in serum, only Bio-T and FAI dantly expressed in CGCs. Transcripts of AKR1C3, CASR, were high in PCOS. GHRHR, RETN,andMAMLD1 were upregulated while those of TF and TNF were downregulated in CGCs of Identification of differentially methylated targets and PCOS women (Fig. 2). To verify whether our NGS findings their gene ontology analysis were replicative in a larger study cohort, we performed py- MC-Seq of CGCs of women with PCOS and controls rosequencing to assess the average percent methylation of identified a total of 6486 differentially methylated CpG these genes in 17 controls and 17 women with PCOS. Hy- sites associated with 3403 unique genes across the gen- pomethylation of AKR1C3, CASR, GHRHR, RETN,and ome, of which 2977 CpG sites were hypomethylated and MAMLD1 genes at the indicated CpG sites was confirmed 2509 CpG sites were hypermethylated. Hypomethylated by pyrosequencing (Fig. 3). Decreased methylation in these CpG sites were representative of 2063 (Additional file 1) genes was consistent with the upregulation of their respect- genes in all, while the hypermethylated sites were linked ive transcripts in women with PCOS (Fig. 2), and the two to a total of 1777 genes (Additional file 2). Of the total variables showed an inverse correlation with one another Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 4 of 16

Table 1 Clinical characteristics of study participants undergoing controlled ovarian hyperstimulation (COH) assessed before and after oocyte retrieval Parameters assessed Control (n = 20) median (IQR) PCOS (n = 20) median (IQR) P value Age in years 30.0 (27.0–31.0) 31.5 (28.0–33.0) 0.124 BMI (kg m−2) 23.03 (20.97–25.06) 24.65 (22.24–27.76) 0.173 #Basal FSH (μU/mL) 7.64 (5.52–9.39) 5.71 (4.39–6.61) 0.037* #Basal LH (μU/mL) 6.19 (3.0–7.66) 9.13 (5.58–12.24) 0.02* #LH:FSH 0.75 (0.55–1.04) 1.74 (1.15–2.04) 0.0001*** #Prolactin (ng/mL) 14.2 (12.0–18.27) 14.48 (11.72–24.3) 0.43 #TSH (mIU/mL) 1.8 (1.14–2.75) 2.36 (1.55–4.08) 0.0577 #AMH (ng/mL) 3.4 (1.94–7.86) 7.99 (5.21–9.74) 0.0094** Regular cycle 20 (100%) 3 (15%) Oligomenorrhea 0 (0%) 15 (80%) <0.0001*** Secondary amenorrhea 0 (0%) 2 (5%) rFSH (IU) 1600 (1528–2313) 1706 (1350–2025) 0.788

E2 (ng/mL) before hCG administration 1.49 (1.247–2.24) 2.03 (1.22–2.47) 0.367

E2 (ng/mL) on hCG administration day 1.71 (1.436–2.32) 2.56 (1.82–3.72) 0.038*

P4 (ng/mL) before hCG administration 0.3 (0.2–0.55) 0.25 (0.17–0.525) 0.522

P4 (ng/mL) on hCG administration day 2.65 (2.07–4.05) 4.6 (1.95–6.32) 0.200 Total follicles (n) 15.0 (12.0–20.5) 18.5 (15.0–29.25) 0.099 Total oocytes (n) 13.5 (10.0–17.25) 16.5 (14.5–26.75) 0.059 MII oocytes (n) 10.5 (8.0–13.5) 15.0 (9.0–19.75) 0.057 MII oocytes (%) 82.84 (70.24–90.71) 86.85 (74.2–95.31) 0.366 Total fertilized oocytes (n) 6.0 (4.5–9.25) 6.5 (3.75–14.5) 0.464 Rate of MII oocyte fertilization (ROF) 65.63 (51.92–88.13) 52.27 (34.8–67.) 0.039* $ E2 (ng/mL) Serum 0.2 (0.15–0.6) 0.43 (0.15–0.76) 0.418 $ E2 (ng/mL) FF 374 (178.6–541.7) 280.3 (125–819) 0.586 $ P4 (ng/mL) Serum 1 (3–9) 3.9 (0.3–7.) 0.903 $ P4 (ng/mL) FF 8170 (4760–15,880) 6000 (3280–10,140) 0.325 $TT (ng/dL) Serum 126 (89–178.5) 172.6 (140–311.3) 0.020* $TT (ng/dL) FF 300 (140.2–372.6) 405.8 (222.6–639.3) 0.030* $SHBG (nmol/L) Serum 130 (103.0–150.0) 95 (79–134.5) 0.041* $SHBG (nmol/L) FF 158 (127.8–205.5) 127.7 (103.3–148.8) 0.024* $Free T (pmol/L) Serum 3.53 (2.07–4.85) 6.03 (3.93–9.73) 0.066 $Free T (pmol/L) FF 4.24 (1.55–7.87) 7.5 (5.28–11.74) 0.050* $Bio-T (nmol/L) Serum 0.82 (0.48–1.14) 1.09 (0.8–1.9) 0.033* $Bio-T (nmol/L) FF 0.99 (0.58–1.84) 1.75 (1.24–2.74) 0.050* $FAI serum 3.03 (2.38–6.48) 6.03 (3.93–9.73) 0.035* $FAI FF 3.7 (2.32–7.17) 6.93 (5.48–13.42) 0.050* Data are represented as median (inter-quartile range) for anthropometric and hormonal characteristics compared been controls and women with PCOS using Mann-Whitney U tests. Parameters marked with asterisk (#) denote those measured between days 3–5 of the menstrual cycle (early follicular phase) before initiating the controlled ovarian hyperstimulation (COH) procedure. Parameters marked by “$” were measured in sera and follicular fluids obtained on the day of ovum pick up (d-OPU). Menstrual characteristics were assessed using the chi-square analysis. P values < 0.05 are considered significant for all statistical tests. *P < 0.05, **P < 0.01, ***P

(Table 2). Also, hypermethylation of TNF was correlated hypermethylation in NGS data was found to be hypo- with downregulation of its transcript expression in women methylated upon pyrosequencing, though its transcript was with PCOS (Table 2). However, TF which showed downregulated in PCOS (Figs. 2 and 3). Also, there was no Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 5 of 16

Fig. 1 Pathway analysis of differentially methylated genes identified by MC-Seq of cumulus granulosa cells (CGCs). Pie charts displaying gene ontology (GO) and pathway analyses for the lists of hypomethylated genes (a), hypermethylated genes (b), and all differentially methylated genes (DMGs) (c) between CGCs of controls (C-CC, n = 3) and women with PCOS (P-CC, n = 3) have been shown in the figure. Analysis was carried out using the GeneCodis3 web tool. Each chart represents the top ten pathways enriched in each of the three datasets

correlation between TF transcript levels and its methylation Discussion status (Table 2). As supporting evidence, we also evaluated Tissue-specific DNA methylation changes sired by alter- the transcript expression of additional genes such as prosta- ations in the environmental or physiological milieu of an glandin E receptor (PTGER1), leukemia inhibitory factor individual can bring about significant changes in gene and (LIF), and hyaluronan and proteoglycan link protein 1 protein expression, and therefore predispose them to (HAPLN1) which showed hypermethylation in NGS ana- disease development. Alterations in both transcriptome and lysis. Transcripts of these genes were downregulated in proteome profiles of ovarian cells/tissues and FF have been PCOS (Fig. 2). previously reported in PCOS [25–28]. With this Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 6 of 16

A

B

Fig. 2 Transcript expression profiling of few differentially methylated genes identified by MC-Seq. Bar graphs representing transcript expression profiles of 10 differentially methylated genes (DMGs) between CGCs (C-CCs) of controls (n = 17) and women with PCOS (P-CCs, n =17)areshownin the figure. Panel (a) represents transcript expression data of genes identified as hypomethylated in PCOS by NGS analysis while (b) shows transcript profiling of hypermethylated genes. Fold change was evaluated using the 2-ΔΔCt method, where the expression was normalized to 18s levels, using a CGC calibrator sample. Data are presented as “mean +SEM.” *P <0.05,**P < 0.01, 'ns' denotes no significant change. Data are analyzed using the Mann-Whitney U test. background information, we had initially conducted a of differential methylation observed in few of these identi- pilot study to screen for the tissue-specific global DNA fied genes, which may contribute to hyperandrogenism, methylation changes in PBLs and CGCs of controls and defects in COC expansion, oocyte maturation/and ovula- women with PCOS. Here, subtle alterations were detected tion, premature luteolysis, and oxidative stress observed in in CpG methylation profiles of long interspersed nucleo- PCOS, have been discussed here. tide element 1 (LINE1) in PCOS, and these changes were found to be more prominent in CGCs of women with Androgen overproduction PCOS, compared to PBLs [8]. Since LINE1s are CpG hypomethylation in genes such as AKR1C3, GHRHR, self-replicating transposons and occupy ~ 17% of the hu- MAMLD1 and RETN, and hypermethylation in TNF,which man genome, even slight changes in their methylation can indirectly contribute to androgen excess, were consistent patterns can be reflective of genomic dysregulation. with increased and decreased levels of the respective gene Therefore, the primary goal of this study was to identify transcripts (Figs. 2 and 3). AKR1C3 is a steroidogenic en- the genome-wide methylation differences in CGCs of zyme that converts androstenedione (A4) to biologically ac- women with PCOS, at a single base resolution. GO ana- tive testosterone in non-testicular tissues [29]. In PCOS, the lysis of the current methylome data revealed that genes increased expression of AKR1C3 and AKR1C3-mediated an- regulating cell growth, adhesion, differentiation, prolifera- drogen production, have been reported in the adrenal cortex tion, cell polarity and fate determination, apoptosis, signal and visceral adipose tissues [29, 30]. This enzyme is also transduction, transcription, post-translational modifica- expressed by GCs (both mural and cumulus granulosa cells) tions, protein binding, metal and nonmetal ion binding, of periovulatory follicles [31]. Since AKR1C3 expression ATP binding, vesicular transport, etc., were differentially was found to be high in CGCs of women with PCOS, it methylated in PCOS (Additional file 3). The implications may contribute to the high androgen production in Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 7 of 16

Fig. 3 Validation of CpG methylation status of a few differentially methylated genes identified by MC-Seq. The figure depicts “box and whisker” plots for genes whose single CpG site was validated or “bar graphs” for genes in which validation was performed for CpG sites > 1. The plots compare percent (%) methylation for the stated CpG sites between cumulus granulosa cells (CGCs) of controls (C-CC, n = 17) and women with PCOS (P-CC, n = 17). Data for box plots are presented as whiskers ranging from minimum to maximum values and data for bar graphs are presented as “mean + SEM” using the Mann-Whitney U test. *P < 0.05, **P < 0.01, 'ns' denotes no significant change.

Table 2 Correlation analysis of CpG methylation levels of selected genes with their transcript expression profiles Gene name CpG site/s P value/FDR cutoff Methylation and transcript status by Correlation coefficient (R2); P value pyrosequencing and qPCR AKR1C3 CpG-880 < 0.05 Hypomethylated upregulated − 0.53; 0.02* CASR CpG-680 < 0.05 Hypomethylated upregulated − 0.74; < 0.0001* GHRHR CpG-809 < 0.025 Hypomethylated upregulated − 0.669; 0.002* CpG-721 − 0.501; 0.034 RETN CpG-177 < 0.05 Hypomethylated upregulated − 0.603; 0.029* MAMLD1 CpG-129 < 0.016 Hypomethylated upregulated − 0.759; < 0.0001* CpG-116 − 0.596; 0.007* CpG-113 − 0.612; 0.005* TNF CpG-242 < 0.025 Hypermethylated downregulated − 0.582; 0.011* TF CpG-358 < 0.05 Hypomethylated downregulated − 0.162; 0.507 CpG sites that showed significant difference between controls and PCOS in Fig. 3 were analyzed. P values < FDR cutoff values are marked as significant (*) Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 8 of 16

their ovaries. Next, GHRHR is a gene encoding the class B HEK293T cells and adipocytes [50]. Thus, the overex- GPCR subfamily receptor, which regulates the release of pression of RETN due to the hypomethylation of its pro- somatotropin (GH) in the brain and other tissues includ- moter may be an important factor contributing to the ing the ovary, via binding to the hypothalamic neuropep- androgen excess in PCOS (Fig. 4). tide GHRH [32, 33]. Although the classical ovarian function of GH is to enhance sexual maturation at puberty Oocyte development, ovulation, and COC matrix via binding to its receptors (GHRs) [34], the GH-GHR expansion defects interaction also stimulates the release of insulin-like Enrichment of calcium (Ca2+) signaling pathway and growth factor (IGF1) via transcriptional activation [34]. pathways for regulation of cytoskeletal and focal adhe- IGF1, like insulin, can increase LH production from the sion elements in our NGS analysis indicated impairment pituitary and augment ovarian androgen synthesis in of calcium homeostasis and cellular architecture in PCOS [35, 36]. Additionally, hyperinsulinemia in PCOS is CGCs of women with PCOS (Fig. 1, Additional file 3). known to increase the bioavailability of IGF1 via the Ca2+ signaling pathways are crucial for the development downregulation of its carrier protein, i.e., IGFBP1 [25, 34]. and maturation of healthy oocytes and CASR, which High levels of GH and IGF1 also increase the sensitivity of is an important mediator of this pathway, responds to developing follicles to gonadotropins [37, 38], and PCOS subtle changes in extracellular Ca2+ concentrations and follicles have been reported to exhibit increased sensitivity activates or ameliorates the mobilization of intracellular and responsiveness to FSH [39]. As a result, the expres- Ca2+ stored in tissues [51]. Expression and localization sion of LHCGR which is under the direct control of FSH of CASR have been reported in human oocytes and is found to be elevated in the follicles of women with CGCs, wherein it supposedly facilitates bidirectional PCOS [13]. This can further augment LH-mediated ovar- communication between these cells to either keep oo- ian androgen production in PCOS. Thus, hypomethyla- cytes arrested in MI phase, or assist in the full re- tion and overexpression of GHRHR observed in CGCs can sumption of their cytoplasmic and nuclear maturation be an indirect mediator of androgen excess in PCOS uponenteringtheMIIphase[52]. Therefore, alter- (Fig. 4). Further, the pro-inflammatory cytokine TNF, ations in CASR expression may affect oocyte matur- which suppresses FSH-induced LHCGR promoter activa- ation and yield poor quality oocytes as seen in PCOS tion via NF-κB p65, was hypermethylated and low in (Fig. 4).InPCOS,sofar,asinglereportexistsonthe PCOS [40], thus also making it an additional factor con- association of a CASR polymorphism (Hin1I) with al- tributing to hyperandrogenemia. tered global calcium homeostasis [53]. Altered methy- Mutations in MAMLD1, a transcriptional coactivator, lation has been previously reported in human CASR have been reported to result in compromised androgen promoter in a few cancer conditions [54–56]. Our re- synthesis during male fetal sexual development [41, 42]. sults indicate that altered CASR expression in CGCs However, information regarding a definite role of of PCOS women can be also influenced by altered MAMLD1, its regulation and mechanisms of action in methylation. the context of other reproductive functions, is limited. TNF expression in CGCs has been reported to be ei- In a murine study, MAMLD1 knockout male mice ther unchanged or reduced in CGCs of women with showed a reduced testosterone production in Leydig PCOS [57, 58]; however, its circulating levels have been tumor cells while the activation of MAMLD1 promoter found to be high in their serum and FF [59, 60]. Few by the transcription factor, SF1, augmented testosterone studies demonstrated that treatment with high levels of production via transactivation of the hairy/enhancer of TNF increased GC apoptosis, impaired P4 production split 3 HES3 promoter [43]. We therefore propose that from GCs, and caused other steroidogenic defects in MAMLD1 hypomethylation and upregulation of its tran- these cells [61–63]. However, these studies utilized TNF script in CGCs may be important in contributing to an- at 10–20-fold higher doses relative to its physiological drogen excess in PCOS ovaries (Fig. 4). Next, increased levels. In alternate studies, optimal TNF levels have been circulatory levels of the adipokine, RETN, which modu- reported to impart a protective function in the mainten- lates glucose tolerance and insulin action, have been ance of bovine GCs and oocytes [64], facilitate ovulation linked to a higher incidence of insulin resistance and [65], and increase the GC proliferation in animal models PCOS [44, 45]. However, there is some ambiguity re- [66]. Decreased levels of endogenous TNF in GCs has garding the role of RETN in PCOS, since its serum and been attributed to diminished oocyte competence due to FF levels have been either found to be high or un- a reduction in its downstream effector, i.e. tumor necro- changed in PCOS [46–48]. In theca cells, the sis factor-inducible gene 6 (TNFAIP6) [25], as well as dose-dependent increase in RETN showed augmented compromised ovulation, and GC proliferation in ovarian androgen production [49], while a RETN-like molecule follicles [67]. Thus, lowered TNF in CGCs of women β impaired the glucose tolerance and insulin actions in with PCOS owing to hypermethylation, may hamper Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 9 of 16

Fig. 4 Altered gene methylation can contribute to ovarian dysfuncion. The figure summarizes some of the processes that could be dysregulated in the follicular compartment of women, due to hypomethylation (indicated by green boxes) or hypermethylation (indicated by red boxes) of genes in cumulus granulosa cells (CGCs). AKR1C3, aldo-keto reductase family 1 member C3; AR, androgen receptor; CASR, calcium-sensing receptor; COC, cumulus-oophorus complex; FF, follicular fluid; FSH, follicle-stimulating hormone; FSHR, follicle-stimulating hormone receptor; GC, granulosa cells; IGF1; insulin-like growth factor 1; INS, insulin; MGC, mural granulosa cells; GHRHR, growth hormone-releasing hormone receptor; HAPLN1, hyaluronan and proteoglycan link protein 1; LH, luteinizing hormone; LHCGR, luteinizing hormone chorionic gonadotropin receptor; LIF, leukemia inhibitory factor; MAMLD1, mastermind-like domain containing 1; PTGER1, prostaglandin receptor E1; RETN, resistin; TC, theca cells; TF, transferrin; TNF, tumor necrosis factor alpha

COC expansion and compromise ovulation. LIF, which amphiregulin, TNFAIP6, and bikunin, whose diminished was hypermethylated and downregulated in our study, expression is implicated in COC matrix expansion defects has been reported to be low in FF and serum of women [25]. Supporting these findings, we also observed hyper- with PCOS [68, 69]. LIF has demonstrated embryo- methylation and downregulation of HAPLN1,whichis trophic effects in mice and humans [70, 71], and de- also a COC matrix-associated protein. HAPLN1 facilitates crease in its levels has shown a positive association with the expansion of COC matrix and imparts stability to the low implantation rates and poor IVF outcome [68, 69]. COC complex by binding to other matrix proteins and Since dose-dependent administration of LIF also proteoglycans like hyaluronic acid, versican, aggrecan, and showed induction of COC expansion and improvement IαI[72]. Exogenous treatment with HAPLN1 increased in oocyte competence in humans, it is imperative to in- the CGC viability in vitro and enabled their transform- vestigate the role of epigenetics in the regulation of LIF ation into granulosa lutein cells, while knockdown of expression in women with PCOS. Our earlier data on HAPLN1 decreased the cell viability [72]. Therefore, al- proteomics of FF in women with PCOS demonstrated tered HAPLN1 methylation may contribute to changes in the downregulation of COC matrix proteins including COC expansion dynamics in PCOS (Fig. 4). Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 10 of 16

Luteal insufficiency/premature luteolysis and proteins; and contribute to oxidative stress, which is Apart from being an androgen synthesizing enzyme, reported in PCOS ovaries [85]. Figure 5 depicts a sum- AKR1C3 also acts as a prostaglandin F synthase (PGFS), mary of biological processes which may be affected in which catalyzes the conversion of the luteotrophic pros- PCOS due to the altered methylation of genes in follicles taglandins, PGD and PGE2, to a luteolytic form, i.e., of women with PCOS. PGF2α to facilitate luteolysis [29]. Luteal insufficiency and premature luteolysis are frequent occurrences in Limitations PCOS [73], and these have been linked to aberrations in The primary limitation of this study was the low sample angiogenic mechanisms at the ovarian level [74]. Angio- size. A number of controls and women with PCOS ori- genesis is largely under the control of prostaglandins ginally recruited for the study had to be excluded due to [75, 76], and AKR1C family of enzymes are some major the presence of confounding factors such as hyperpro- regulators of prostaglandins [77]. Disparities in levels of lactinemia or thyroid dysfunction, and contamination of pro and anti-angiogenic factors in the ovary are largely FF with blood, or due to recent treatment with metfor- responsible for defects in follicle development, prema- min or thyroid medications. Further, women having low ture degeneration of oocytes, and regression of CL due CGC counts were also excluded as the quantities of nu- to an inefficient supply of oxygen and nutrients to the cleic acids were insufficient for methylation and expres- growing follicles [74]. Both, our present data and FF sion analyses. Also, since PCOS is a heterogeneous proteome study provide compelling evidence supporting disorder, investigation of epigenetic changes in a large angiogenic dysregulation in the ovarian compartment population based on phenotypic subgrouping of women [25]. Since AKR1C3 has been implicated as a trigger for with PCOS (A, B, C, and D phenotypes) as per the Rot- premature luteolysis [78] and was found to be hypo- terdam consensus may provide more accurate informa- methylated and high in PCOS follicles, it may serve as tion on the effect of epigenetic components on PCOS an important epigenetic target to investigate this development. Due to our limited sample size, it was not phenomenon in PCOS, Till now, no clear evidence exists possible to carry out such subgrouping and subsequent on whether PGE2 or PGF2α levels are altered in PCOS. analyses. However, our data shows that the receptor for PGE2 Our findings highlight that MC-Seq could identify sev- gene, i.e., PTGER1, was hypermethylated in NGS and its eral functionally important loci in CGCs, many of which transcript was low in CGCs of women with PCOS. Since were either known to be functionally dysregulated in activation of PTGER1 by specific agonists has demon- PCOS ovary or have a compelling potential to be estab- strated increased sprout formation in capillaries, both in lished as novel candidates influenced by epigenetic vivo and in vitro [79, 80], it can be an important factor changes. The molecular sequelae of these alterations lead- for the restoration of follicular angiogenesis. Also, ing to ovarian dysfunction need to be further addressed PTGER1 has been shown to stimulate progesterone bio- via robust functional studies in women with PCOS. synthesis in human GCs [81]. Since the follicles of PCOS women lack optimum levels of progesterone required Materials and methods for maintenance of CL, low levels of PTGER1 in PCOS Study design, participants, sample collection, and caused by promoter hypermethylation may explain this estimated parameters shortcoming. This study was carried out at the ICMR-National Insti- tute for Research in Reproductive Health (NIRRH) as Oxidative stress per ethical norms. All participants were recruited from Lastly, TF, which maintains the oxido-reductive homeo- the “Fertility Clinic and IVF Center” (Mumbai) after stasis in proliferating cells and showed hypermethylation obtaining written informed consents and underwent IVF in our NGS data, was found to be hypomethylated by using a long, GnRH agonist protocol as reported earlier pyrosequencing, though its mRNA was downregulated [8]. We initially recruited 35 women with PCOS and 38 in PCOS. Therefore, methylation changes at other CpG age-BMI-matched, healthy, and regularly menstruating sites in TF promoter need to be analyzed. TF primarily controls as per the Rotterdam consensus criteria [86]. transports iron released from hepatic, intestinal, and re- Women showing normal ovarian morphology on ultra- ticuloendothelial stores to the target tissues via its recep- sound with no signs of hyperandrogenism or insulin re- tor endocytosis while also alleviating local oxidative sistance, and undergoing COH strictly owing to stress, acting as a growth factor, and promoting follicle indications of male factor infertility in their spouses were and oocyte maturation [82, 83]. Low levels of TF in fol- recruited as controls. Women recruited as PCOS had at licular fluid of PCOS women have been previously re- least 2 of the following 3 features, i.e., polycystic ovaries ported [84], which may lead to a higher prevalence of (PCO) on ultrasound, irregularity/absence of menses, unbound Fe2+; trigger oxidative damage to DNA, lipids, and/or signs of hyperandrogenism during clinical Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 11 of 16

Fig. 5 Summary of the genes and their associated processes found to be affected in PCOS due to altered DNA methylation. The figure represents sa preovulatory follicle of women with PCOS, showing a few differentially methylated genes involved in the perpetuation of androgen excess, premature luteolysis, impaired calcium signaling, COC defects, oxidative stress and angiogenic defects in the follicles of women with PCOS as observed in our study. AKR1C3, aldo-keto reductase family 1 member C3; CASR, calcium-sensing receptor; COC, cumulus-oophorus complex; FF, follicular fluid; FSH, follicle-stimulating hormone; FSHR, follicle-stimulating hormone receptor; GC, granulosa cells; IGF1; insulin-like growth factor 1; INS, insulin; MGC, mural granulosa cells; GHRHR, growth hormone-releasing hormone receptor; HAPLN1, hyaluronan and proteoglycan link protein 1; LH, luteinizing hormone; LHCGR, luteinizing hormone chorionic gonadotropin receptor; LIF, leukemia inhibitory factor; MAMLD1, mastermind-like domain containing 1; PTGER1, prostaglandin receptor E1; RETN, resistin; TC, theca cells; TF, transferrin; TNF, tumor necrosis factor alpha screening, however the presence of PCO morphology suspended in FF aspirates were separated from FF and was used as a mandate for all women recruited as PCOS. manually stripped off to dissociate the cumulus granu- These women were carefully screened for their baseline losa cells (CGCs) from their oocytes. CGCs were hormonal estimates and biochemical characteristics washed, resuspended in ovum buffer, and transported using serum and follicular fluid samples (Table 1). Fur- from IVF center to the lab at 37 °C for further process- ther, we had to exclude samples of women who were on ing. The numbers of total retrieved oocytes and mature metformin or thyroid medications, whose follicular fluid oocytes (in the MII phase of meiosis) were obtained had blood contamination and whose CGC counts were from clinical records, and rates of fertilization of MII oo- low, since substantial amounts of DNA were required cytes were calculated. for high-throughput methylome sequencing and bisulfite PCRs. After this, 20 controls and 20 women with PCOS Methyl-capture sequencing (MC-Seq) by NGS approach were selected for the study (Table 1). Baseline hormonal and analysis of DMRs estimates (between days 3–7 of menstrual cycle) for LH, From among the 20 women with PCOS and 20 controls FSH, prolactin, TSH, and AMH could be obtained from selected for the study, individual DNA samples of CGCs IVF clinical records, while fasting serum and macroscop- from 3 PCOS women and 3 age-BMI-matched controls ically clear FF collected on d-OPU were assayed for E2, having a total yield of > 2 μg were subjected to MC-Seq P4, TT, and SHBG using commercial ELISA kits (Diag- using the NGS approach. Whole genomic DNA was ex- nostics Biochem Canada Inc., Dorchester, Ontario, tracted from CGCs of these women using QIAamp Canada). Androgen excess indices were calculated using DNA mini kit (Qiagen, Hilden, Germany) and processed TT and SHBG values [8]. Upon follicle maturation, the for library preparation. 1 μg of genomic DNA per sample levels of E2 and P4, which are routinely measured 1 day was sheared in fragments of approximately 150 bp using prior to and 1 day after rhCG administration (10,000 IU) the Covaris S220 ultrasonicator and further processed to monitor ovarian response, were also recorded. COCs for DNA end repair, 3′-adenylation, methylated adapter Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 12 of 16

ligation, and hybridization to the Methyl-Seq library identified. OKdb is an online search tool based on probes using the SureSelect Methyl-Seq Library Prep microarray-based transcriptome profiling and independ- Kit (Agilent Technologies, CA, USA) as per the manu- ent study reports on genes and proteins identified in ovar- facturer’s instructions. The hybrids were captured using ian tissues and cells [88]. Upon identification of genes streptavidin beads (Dynabeads) and subjected to bisulfite common to our NGS dataset and OKdb, and a careful re- conversion using the EZ-DNA Methylation Gold Kit view of literature, 7 genes participating in the perpetration (Zymo Research, CA, USA). The library was PCR ampli- of androgen excess, impaired angiogenesis, luteal insuffi- fied, indexed using SureSelectXT Methyl-Seq indexing ciency, COC matrix defects, and oocyte defects, namely primers (Agilent Technologies, CA, USA), and pooled AKR1C3, CASR, GHRHR, MAMLD1, RETN, TF, and TNF, for 100 bp paired-end multiplexed sequencing on Illu- were selected for validation by both pyrosequencing and mina HiSeq 2500 platform. qPCR. Additionally, the transcript expression profiles of prostaglandin E receptor (PTGER1), leukemia inhibitory Processing and alignment of reads factor (LIF), and hyaluronan and proteoglycan link protein Whole-genome targeted methylation sequencing reads 1(HAPLN1), were evaluated in CGCs of controls and of the above six samples were aligned to human refer- PCOS women as supporting evidence for pathways associ- ence genome hg19 assembly. The hg19 reference gen- ated with the above genes. ome was converted to a DNA methylation reference genome, and genome indexing was performed using Bis- Extraction of nucleic acids from CGCs for qPCR and mark genome preparation utility (v.14.3). The adapter pyrosequencing trimmed sequencing reads were aligned to the converted Total DNA and RNA were extracted from CGCs of 17 methylation reference genome using Bismark tool with controls and 17 women with PCOS using the NucleoS- two allowed mismatches. The remaining parameters pin TriPrep kit (Macherey-Nagel, Düren, Germany) for from Bismark were used as default. Further, deduplica- validation of the NGS data. The quality and yield of nu- tion of aligned reads was performed using “deduplicate_- cleic acids were assessed by agarose gel electrophoresis bismark” utility of Bismark tool. and by evaluating their spectrophotometric ratios at 260 and 280 nm. DNA samples were stored at − 20 °C while Methylation analysis and functional annotation RNA was stored at − 80 °C until further use. Methylation extractor utility of Bismark tool was used to extract the methylation call for every methylated cyto- cDNA synthesis and quantitative real-time PCR sine (C) in all three contexts: CpG, CHG, and CHH cDNA was synthesized from 500 ng of RNA (n = 34) using (where H is A or C or T). Differential methylation ana- a first strand cDNA synthesis kit (Takara Bio USA Inc.). lysis was performed using “calculateDiffMeth” function Transcript levels of DMGs selected for validation of R-based methylKit package at q value cutoff threshold (Additional file 4) were assayed by TaqMan chemistry of ≤ 0.01 and methylation difference ≥ 25%. Hierarchical using the TaqMan™ Universal Master Mix II with UNG, clustering of samples based on the similarity of their and FAM-labeled probes (ThermoFisher Scientific, MA, methylation profile was performed using the euclidean USA). Assay containing VIC-labeled 18s rRNA probe was distance metric, and ward method clustering approaches used as the housekeeping control. qPCR was carried out of methylKit package. Further, enrichment and annota- using cDNA dilutions ranging between neat to 1:100. Fold tion of both hypo- and hypermethylated sites within up change in gene expression between controls and PCOS ΔΔ 10 kb of annotated transcription start site (TSS) to tran- was evaluated using the 2- Ct method, where the expres- scription end site (UCSC hg19) were performed using sion was normalized to 18s levels, using a CGC calibrator in-house perl scripts. GeneCodis3 web-based tool [87] sample. was used for gene ontology (GO) and pathway enrich- ment analysis of the DMGs. Bisulfite primer design and pyrosequencing Primers for the validation of selected genes were de- Selection of genes for validation studies signed on the Pyromark Q96 ID machine using the Genes showing differential methylation at CpG sites up- PyroMark® Assay Design SW 2.0 software (Qiagen), the stream up to 1000 bases relative to their TSS in our NGS list of which has been provided in Additional file 4. data were selected for validation. These included a total of Primers were procured from Sigma-Aldrich. The reverse 354 hypermethylated and 397 hypomethylated genes (total pyrosequencing primers were tagged with biotin at the n = 735 unique genes). The lists of these 735 genes, and 5′-end and HPLC purified. Approximately 300–500 ng the 4354 genes enlisted in Ovarian Kaleidoscope (OKdb), DNA from CGCs of 17 controls and 17 women with were compared using a Venn analysis, and a total of 132 PCOS was bisulfite converted using the MethylCode genes that were common to both datasets could be bisulfite conversion kit (Invitrogen-ThermoFisher Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 13 of 16

Scientific, MA, USA). The region of interest was ampli- dehydrogenase; HAPLN1: Hyaluronan and proteoglycan link protein 1; fied by two rounds of PCR. PCR product (1–3 μL from hCG: Human chorionic gonadotropin; IGF1R: Insulin-like growth factor μ receptor; IGFBP: Insulin-like growth factor binding proteins; INS: Insulin; IRS- total volume of 20 L) from the first round was used as 2: Insulin receptor substrate 2; IVF: In vitro fertilization; LH: Luteinizing nd template for the 2 PCR round, which was scaled up to hormone; LHCGR: Luteinizing hormone chorionic gonadotropin receptor; 45 μL. For both PCRs, initial denaturation was per- LIF: Leukemia inhibitory factor; LINE1: Long interspersed nucleotide element 1; MAMLD1: Mastermind-like domain containing 1; MC-Seq: Methyl-capture formed at 95 °C for 15 min followed by 40 rounds of sequencing; MeDIP: Methylated DNA immunoprecipitation; MGC: Mural amplification at 94 °C for 30s, annealing at the respective granulosa cells; NGS: Next generation sequencing; OKdb: Ovarian optimized temperatures for 10s, and 72 °C for 60s with a Kaleidoscope; P4: Progesterone; PBL: Peripheral blood leukocytes; P- CC: Cumulus granulosa cells of PCOS; PCOS: Polycystic ovary syndrome; final extension at 72 °C for 10 min. Product amplification PDGF: Platelet-derived growth factor; PGD: Prostaglandin D; after both rounds was confirmed by agarose gel electro- PGE2: Prostaglandin E2; PGF: Prostaglandin F; PGFS: Prostaglandin F synthase; phoresis. For pyrosequencing, 40 μL of the PCR product PRKCB: Protein kinase C Beta; PTGER1: Prostaglandin receptor E1; RETN: Resistin; SHBG: Sex hormone binding globulin; T2DM: Type 2 diabetes was subjected to clean up on the Pyromark Q96ID se- mellitus; TES: Transcription end site; TF: Transferrin; TNF: Tumor necrosis quencing workstation as per the manufacturer’s instruc- factor alpha; TNFAIP6: Tumor necrosis factor-inducible gene 6; TSH: Thyroid tions and sequenced using 1.6 μL (16 picomoles) of each stimulating hormone (TSH); TSS: Transcription start site; TT: Total testosterone; VEGF: Vascular endothelial growth factor; WGBS: Whole of the gene-specific sequencing primers. genome bisulfite sequencing; YAP1: Yes-associated protein 1

Statistical analysis Acknowledgements The authors thank all the consenting study participants and collaborators Mann-Whitney U tests were employed for all univariate who became a part of this study and Ms. Kavita Suri (Fertility Clinic and IVF assessments of continuous variables including hormonal Center, Mumbai) who collected the clinical samples of our study participants and biochemical parameters, DNA methylation, and and provided us with their IVF data records. transcript expression levels assessed between CGCs of Disclosure summary controls and women with PCOS. Correlation between The authors have nothing to disclose. the CpG methylation status of genes and their respective Funding transcripts was determined using a two-tailed Spear- This work is partially supported by Department of Science and Technology man’s correlation coefficient. FDR cutoffs were applied (DST) (EMR/2015/001014), NIRRH (NIRRH/RA/623/04–2018) and Indian for genes showing multiple CpG sites as significant after Council of Medical Research (ICMR), New Delhi, India. Financial assistance provided by DST and Department of Atomic Energy-Board of Research in Mann-Whitney testing. Nuclear Sciences (DAE-BRNS), Government of India; to PS for pursuing her doctoral studies is acknowledged.

Additional files Availability of data and materials Datasets analyzed during the current study will be made available from the Additional file 1: Hypomethylated CpG sites were representative of corresponding author upon reasonable request. Supplementary data related to 2063 genes. (XLSX 408 kb) the methylome analysis of cumulus granulosa cells is available on Figshare, an Additional file 2: Hypermethylated sites were linked to a total of 1777 online digital data repository https://doi.org/10.6084/m9.figshare.7370438 genes (XLSX 354 kb) Authors’ contributions Additional File 3: Figure includes horizontal bar charts showing The study design was conceived by SM and PS. PS collected the patient components such as A) molecular functions, B) biological processes, and C) samples, processed the samples, carried out all the experiments, performed cellular components that were most highly enriched in datasets obtained data interpretation, and wrote the paper. PK carried out the bioinformatic for all differentially methylated, hypomethylated, and hypermethylated analysis and interpretation of the NGS data. VM and SD are clinical X genes identified in the NGS analysis. -axis represents the number of genes collaborators on this manuscript who supervised the recruitment of study present within each annotated category. (TIF 901 kb) participants and collection of their samples at the Fertility Clinic and IVF Additional File 4: The table enlists pyrosequencing primer sets Center, Mumbai. SM critically revised the manuscript for important designed to evaluate upstream/promoter CpG methylation in intellectual content. All authors read and approved the final manuscript. differentially methylated genes (DMGs) selected for validation in controls (C-CC, n = 17) and women with PCOS (P-CC, n = 17) and commercial Ethics approval and consent to participate TaqMan Assay IDs used for validating the transcript expression levels of Written informed consent was obtained from all study participants who gave selected DMGs in controls and women with PCOS. (DOCX 13 kb) blood, follicular fluid and cumulus granulosa cell samples, in accordance with current edition of the NIRRH Ethics Committee for Clinical Studies (Ref: D/ICEC/ Sci-54/56/2018 dated May 29, 2018). This committee is recognized by Strategic Abbreviations Initiative for Developing Capacity in Ethical Review (SIDCER) and Forum for A4: Androstenedione; AKR1C3: Aldo-keto reductase family 1 member C3; Ethics Review Committees in Asia and the Western Pacific Region (FERCAP). AMH: Anti-Mullerian hormone; AMHR2: Anti-Mullerian hormone receptor 2; AR: Androgen receptor; CASR: Calcium-sensing receptor; C-CC: Cumulus Consent for publication granulosa cells of controls; CGC: Cumulus granulosa cell; CL: Corpus luteum; Not applicable. COC: Cumulus-oophorus complex; COH: Controlled ovarian hyperstimulation; DMG: Differentially methylated gene; E2: Estradiol; EDC: Endocrine disruptor; Competing interests FAI: Free androgen index; FF: Follicular fluid; FGF: Fibroblast growth factor; The authors declare that they have no competing interests. FSH: Follicle-stimulating hormone; FSHR: Follicle-stimulating hormone receptor; FST: Follistatin; GC: Mural and cumulus granulosa cells; GH: Growth hormone; GHR: Growth hormone receptor; GHRHR: Growth hormone- Publisher’sNote releasing hormone receptor; GnRH: Gonadotropin-releasing hormone; Springer Nature remains neutral with regard to jurisdictional claims in GO: Gene ontology; H6PD: Hexose-6-phosphate dehydrogenase/glucose 1- published maps and institutional affiliations. Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 14 of 16

Author details 19. Wang X, Wei J, Jiao J, Jiang S, Yu D, Li D. Genome-wide DNA methylation 1Department of Molecular Endocrinology, ICMR-National Institute for and gene expression patterns provide insight into polycystic ovary Research in Reproductive Health, J.M. Street, Parel, Mumbai, Maharashtra syndrome development. Oncotarget. 2014;5(16):6603–10. 400012, India. 2Colin Jamura Lab, Institute for Stem Cell Biology and 20. Kokosar M, Benrick A, Perfilyev A, Fornes R, Nilsson E, Maliqueo M, Johan Regenerative Medicine (inStem), National Centre for Biological Sciences Behre C, Sazonova A, Ohlsson C, Ling C, Stener-Victorina E. Epigenetic and (NCBS), GKVK, Bellary Road, Bangalore 560065, India. 3Fertility Clinic and IVF transcriptional alterations in human adipose tissue of polycystic ovary Center, 12-Springfield, 19-Vachha Gandhi Road, Gamdevi, Mumbai, syndrome. Sci Rep. 2016;6(22883). Maharashtra 400007, India. 21. Pan JX, Tan YJ, Wang FF, Hou NN, Xiang YQ, Zhang JY, et al. Aberrant expression and DNA methylation of lipid metabolism genes in PCOS: a new Received: 14 January 2019 Accepted: 27 March 2019 insight into its pathogenesis. Clin Epigenetics. 2018;10(1):6. 22. Xu J, Bao X, Peng Z, Wang L, Du L, Niu W, et al. Comprehensive analysis of genome-wide DNA methylation across human polycystic ovary syndrome ovary granulosa cell. Oncotarget. 2016;7(19):27899–909. References 23. Ling Teh A, Pan H, Lin X, Lim YI, Pawan C, Patro K, et al. Comparison of 1. Fauser BCJM, Tarlatzis BC, Rebar RW, Legro RS, Balen AH, Lobo R, et al. methyl-capture sequencing vs. Infinium 450K methylation array for Consensus on women’s health aspects of polycystic ovary syndrome methylome analysis in clinical samples. Epigenetics. 2016;11(1):36–48. (PCOS): the Amsterdam ESHRE/ASRM-Sponsored 3rd PCOS Consensus 24. Sagvekar P, Dadachanji R, Patil K, Mukherjee S. Pathomechanisms of Workshop Group. Fertil Steril. 2012;97(1):28–38 e25. polycystic ovary syndrome: multidimensional approaches. Front Biosci (Elite 2. Norman RJ, Dewailly D, Legro RS, Hickey TE. Polycystic ovary syndrome. Ed). 2018;10:384–422. Lancet. 2007;370(9588):685–97. 25. Ambekar AS, Kelkar DS, Pinto SM, Sharma R, Hinduja I, Zaveri K, et al. 3. Diamanti-Kandarakis E. Polycystic ovarian syndrome: pathophysiology, Proteomics of follicular fluid from women with polycystic ovary syndrome molecular aspects and clinical implications. Expert Rev Mol Med. suggests molecular defects in follicular development. J Clin Endocrinol 2008;10:e3. Metab. 2015;100(2):744–53. 4. Baptiste CG, Battista M-C, Trottier A, Baillargeon J-P. Insulin and 26. Choi DH, Lee WS, Won M, Park M, Park HO, Kim E, et al. The apolipoprotein hyperandrogenism in women with polycystic ovary syndrome. J Steroid A-I level is downregulated in the granulosa cells of patients with polycystic Biochem Mol Biol. 2010;122(1–3):42–52. ovary syndrome and affects steroidogenesis. J Proteome Res. 2010;9(9): 5. Wiweko B, Maidarti M, Priangga MD, Shafira N, Fernando D, Sumapraja K, et 4329–36. al. Anti-mullerian hormone as a diagnostic and prognostic tool for PCOS 27. Escobar-Morreale HF, Peral B, Corton M, Camafeita E, Lopez JA, San Millan patients. J Assist Reprod Genet. 2014;31(10):1311–6. JL, et al. Proteomic analysis of human omental adipose tissue in the 6. Xu N, Azziz R, Goodarzi MO. Epigenetics in polycystic ovary syndrome: a polycystic ovary syndrome using two-dimensional difference gel pilot study of global DNA methylation. Fertil Steril. 2010;94(2):781–3 e1. electrophoresis and mass spectrometry. Hum Reprod. 2007;23(3):651–61. 7. Sandovici I, Naumova AK, Greenwood CM, editors. Epigenetics and 28. Ma X, Fan L, Meng Y, Hou Z, Mao YD, Wang W, et al. Proteomic analysis of establishment of tissue-specific epigenetic states during development traits. human ovaries from normal and polycystic ovarian syndrome. Mol Hum Epigenetics complex trait; 2013. p. 35–62. Reprod. 2007;13(8):527–35. 8. Sagvekar P, Mangoli V, Desai S, Patil A, Mukherjee S. LINE1 CpG-DNA 29. Nakamura Y, Hornsby PJ, Casson P, Morimoto R, Satoh F, Xing Y, et al. Type hypomethylation in granulosa cells and blood leukocytes is associated with 5 17beta-hydroxysteroid dehydrogenase (AKR1C3) contributes to PCOS and related traits. J Clin Endocrinol Metab. 2017;102(4):1396–405. testosterone production in the adrenal reticularis. J Clin Endocrinol Metab. 9. Pruksananonda K, Wasinarom A, Sereepapong W, Sirayapiwat P, The Endocrine Society. 2009;94(6):2192–8. Rattanatanyong P, Mutirangura A. Epigenetic modification of long 30. O’Reilly MW, Kempegowda P, Walsh M, Taylor AE, Manolopoulos KN, interspersed elements in cumulus cells of mature and immature oocytes Allwood JW, et al. AKR1C3-mediated adipose androgen generation drives from patients with polycystic ovary syndrome. Clin Exp Reprod Med. 2016; lipotoxicity in women with polycystic ovary syndrome. J Clin Endocrinol 43(2):82–9. Metab. The Endocrine Society. 2017;102(9):3327–39. 10. Le Jiang L, Xie JK, Cui JQ, Wei D, Yin BL, Zhang YN, et al. Promoter 31. Dozier BL, Watanabe K, Duffy DM. Two pathways for prostaglandin F2 alpha methylation of yes-associated protein (YAP1) gene in polycystic ovary synthesis by the primate periovulatory follicle. Reproduction. 2008;136(1): syndrome. Med. 2017;96(2):e5768. 53–63. 11. Sang Q, Zhang S, Zou S, Wang H, Feng R, Li Q, et al. Quantitative analysis of 32. Matsoukas M-T, Spyroulias GA. Dynamic properties of the growth hormone follistatin (FST) promoter methylation in peripheral blood of patients with releasing hormone receptor (GHRHR) and molecular determinants of GHRH polycystic ovary syndrome. Reprod BioMed Online. 2013;26(2):157–63. binding. Mol BioSyst. The Royal Society of Chemistry. 2017;13(7):1313–22. 12. Yu Y-Y, Sun C-X, Liu Y-K, Li Y, Wang L, Zhang W. Promoter methylation of 33. Barabutis N, Schally AV. Growth hormone-releasing hormone: extrapituitary CYP19A1 gene in Chinese polycystic ovary syndrome patients. Gynecol effects in physiology and pathology. Cell Cycle. 2010;9(20):4110–6. Obstet Investig. 2013;76(4):209–13. 34. Sharara FI, Giudice LC. Role of growth hormone in ovarian physiology and 13. Wang P, Zhao H, Li T, Zhang W, Wu K, Li M, et al. Hypomethylation of the onset of puberty. J Soc Gynecol Investig. 1997;4(1):2–7. LH/choriogonadotropin receptor promoter region is a potential mechanism 35. Bergh C, Carlsson B, Olsson J-H, Selleskog U, Hillensjö T. Regulation of underlying susceptibility to polycystic ovary syndrome. Endocrinology. 2014; androgen production in cultured human thecal cells by insulin-like growth 155(4):1445–52. factor I and insulin. Fertil Steril. 1993;59(2):323–31. 14. Qu F, Wang F-F, Yin R, Ding G-L, El-Prince M, Gao Q, et al. A molecular 36. Poretsky L. The insulin-related ovarian regulatory system in health and mechanism underlying ovarian dysfunction of polycystic ovary syndrome: disease. Endocr Rev Oxford University Press. 1999;20(4):535–82. hyperandrogenism induces epigenetic alterations in the granulosa cells. J 37. van Dessel HJHMT, Lee PDK, Faessen G, Fauser BCJM, Giudice LC. Elevated Mol Med. 2012;90(8):911–23. serum levels of free insulin-like growth factor I in polycystic ovary syndrome 15. Zhu JQ, Zhu L, Liang XW, Xing FQ, Schatten H, Sun QY. Demethylation of 1. J Clin Endocrinol Metab. 1999;84(9):3030–5. LHR in dehydroepiandrosterone-induced mouse model of polycystic ovary 38. Bachelot A, Monget P, Imbert-BOLLORÉ P, Coshigano K, Kopchick JJ, Kelly syndrome. Mol Hum Reprod. 2009;16(4):260–6. PA, et al. Growth hormone is required for ovarian follicular growth. 16. Shen H-R, Qiu L-H, Zhang Z-Q, Qin Y-Y, Cao C, Di W. Genome-wide Endocrinology. 2002;143(10):4104–12. methylated DNA immunoprecipitation analysis of patients with polycystic 39. Franks S, Stark J, Hardy K. Follicle dynamics and anovulation in polycystic ovary syndrome. PLoS One. 2013;8(5):e64801. ovary syndrome. Hum Reprod Update. 2008;14(4):367–78. 17. Li S, Zhu D, Duan H, Ren A, Glintborg D, Andersen M, et al. Differential DNA 40. Nakao K, Kishi H, Imai F, Suwa H, Hirakawa T, Minegishi T. TNF-α suppressed methylation patterns of polycystic ovarian syndrome in whole blood of FSH-induced LH receptor expression through transcriptional regulation in Chinese women. Oncotarget. 2016;8(13):20656–66. rat granulosa cells. Endocrinology. 2015;156(9):3192–202. 18. Yu Y-Y, Sun C-X, Liu Y-K, Li Y, Wang L, Zhang W. Genome-wide screen of 41. Kalfa N, Fukami M, Philibert P, Audran F, Pienkowski C, Weill J, et al. ovary-specific DNA methylation in polycystic ovary syndrome. Fertil Steril. Screening of mamld1 mutations in 70 children with 46,xy dsd: identification 2015;104(1):145–53 e6. and functional analysis of two new mutations. PLoS One. 2012;7(3):e32505. Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 15 of 16

42. Kalfa N, Cassorla F, Audran F, Oulad Abdennabi I, Philibert P, Béroud C, et al. 64. Paulino LRFM, Cunha EV, Barbalho Silva AW, Souza GB, Lopes EPF, Donato Polymorphisms of MAMLD1 gene in hypospadias. J Pediatr Urol. 2011;7(6):585–91. MAM, et al. Effects of tumour necrosis factor-alpha and interleukin-1 beta 43. Fukami M, Wada Y, Okada M, Kato F, Katsumata N, Baba T, et al. on in vitro development of bovine secondary follicles. Reprod Domest Mastermind-like domain-containing 1 (MAMLD1 or CXorf6) transactivates Anim. 2018;53(4):997–1005. the Hes3 promoter, augments testosterone production, and contains the 65. Crespo D, Goetz FW, Planas JV. Luteinizing hormone induces ovulation via SF1 target sequence. J Biol Chem. 2008;283(9):5525–32. tumor necrosis factor α-dependent increases in prostaglandin F2α in a 44. Çapoğlu İ, Erdem F, Uyanık A, Turhan H. Serum levels of resistin and hsCRP nonmammalian vertebrate. Sci Rep. 2015;5:14210. in women with PCOS. Open Med SP Versita. 2009;4(4):428–32. 66. Son D-S, Arai KY, Roby KF, Terranova PF. Tumor necrosis factor α (TNF) 45. Benomar Y, Gertler A, De Lacy P, Crépin D, Hamouda HO, Riffault L, et al. increases granulosa cell proliferation: dependence on c-Jun and TNF Central resistin overexposure induces insulin resistance through toll-like receptor type 1. Endocrinology. 2004;145(3):1218–26. receptor 4. Diabetes. 2013;62(1):102–44. 67. Yuan Y, Ida JM, Paczkowski M, Krisher RL. Identification of developmental 46. Seow KM, Juan CC, Hsu YP, Ho LT, Wang YY, Hwang JL. Serum and follicular competence-related genes in mature porcine oocytes. Mol Reprod Dev. resistin levels in women with polycystic ovarian syndrome during IVF- 2011;78(8):565–75. stimulated cycles. Hum Reprod. 2005;20(1):117–21. 68. Li Z, Zhu Y, Li H, Jiang W, Liu H, Yan J, et al. Leukaemia inhibitory factor in 47. Panidis D, Koliakos G, Kourtis A, Farmakiotis D, Mouslech T, Rousso D. Serum serum and follicular fluid of women with polycystic ovary syndrome and its resistin levels in women with polycystic ovary syndrome. Fertil Steril. 2004; correlation with IVF outcome. Reprod BioMed Online. 2018;36(4):483–9. 81(2):361–6. 69. Lédée-Bataille N, Laprée-Delage G, Taupin JL, Dubanchet S, Taieb J, Moreau 48. Dinka P, Baldani L, Skrgatic M, Kasum G, Zlopasa S, Kralik O, et al. Altered JF, et al. Follicular fluid concentration of leukaemia inhibitory factor is leptin, adiponectin, resistin and ghrelin secretion may represent an intrinsic decreased among women with polycystic ovarian syndrome during assisted polycystic ovary syndrome abnormality. Gynecol Endocrinol. 2019:1–5. In reproduction cycles. Hum Reprod. 2001;16(10):2073–8. press 70. Cheung LP, Leung HY, Bongso A. Effect of supplementation of leukemia 49. Munir I, Yen HW, Baruth T, Tarkowski R, Azziz R, Magoffin DA, et al. Resistin inhibitory factor and epidermal growth factor on murine embryonic stimulation of 17α-hydroxylase activity in ovarian theca cells in vitro: relevance development in vitro, implantation, and outcome of offspring. Fertil Steril. to polycystic ovary syndrome. J Clin Endocrinol Metab. 2005;90(8):4852–7. 2003;80(SUPPL. 2):727–35. 50. Kushiyama A, Shojima N, Ogihara T, Inukai K, Sakoda H, Fujishiro M, et al. 71. De Matos DG, Miller K, Scott R, Tran CA, Kagan D, Nataraja SG, et al. Resistin-like molecule β activates MAPKs, suppresses insulin signaling in Leukemia inhibitory factor induces cumulus expansion in immature human hepatocytes, and induces diabetes, hyperlipidemia, and fatty liver in and mouse oocytes and improves mouse two-cell rate and delivery rates transgenic mice on a high fat diet. J Biol Chem. 2005;280(51):42016–25. when it is present during mouse in vitro oocyte maturation. Fertil Steril. 51. Hofer AM, Brown EM. Extracellular calcium sensing and signalling. Nat Rev 2008;90(6):2367–75. Mol Cell Biol. 2003;4(7):530–8. 72. Liu J, Park E-S, Curry TE, Jo M, Jo M. Periovulatory expression of hyaluronan 52. Elena Dell’Aquila M, De Santis T, Cho YS, Reshkin SJ, Caroli AM, Maritato F, and proteoglycan link protein 1 (Hapln1) in the rat ovary: hormonal et al. Localization and quantitative expression of the calcium-sensing regulation and potential function. Mol Endocrinol The Endocrine Society. receptor protein in human oocytes. Fertil Steril. 2006;85:1240–7. 2010;24(6):1203–17. 53. Ranjzad F, Mahban A, Shemirani AI, Mahmoudi T, Vahedi M, Nikzamir A, et 73. Shah D, Patel S. Polycystic ovarian syndrome as a cause of recurrent al. Influence of gene variants related to calcium homeostasis on pregnancy loss. J Obs Gynecol India. 2007;57(5):391–7. biochemical parameters of women with polycystic ovary syndrome. J Assist 74. Di Pietro M, Pascuali N, Parborell F, Abramovich D. Ovarian angiogenesis in Reprod Genet. 2011;28(3):225–32. polycystic ovary syndrome. Reproduction. 2018;155(5):R199–209. 54. Hizaki K, Yamamoto H, Taniguchi H, Adachi Y, Nakazawa M, Tanuma T, et al. 75. Karnezis T, Shayan R, Fox S, Achen MG, Stacker SA. The connection between Epigenetic inactivation of calcium-sensing receptor in colorectal lymphangiogenic signalling and prostaglandin biology: a missing link in the carcinogenesis. Mod Pathol. 2011;24(6):876–84. metastatic pathway. Oncotarget. 2012;3(8):893–906. 55. Fetahu IS, Höbaus J, Aggarwal A, Hummel DM, Tennakoon S, Mesteri I, et al. 76. Mantel A, Carpenter-Mendini AB, Vanbuskirk JB, De Benedetto A, Beck LA, Calcium-sensing receptor silencing in colorectal cancer is associated with Pentland AP. Aldo-keto reductase 1C3 is expressed in differentiated human promoter hypermethylation and loss of acetylation on histone 3. Int J epidermis, affects keratinocyte differentiation, and is upregulated in atopic Cancer. 2014;135(9):2014–23. dermatitis. J Invest Dermatol. 2012;132(4):1103–10. 56. Casalà C, Gil-Guiñón E, Luis Ordóñez J, Miguel-Queralt S, Rodríguez E, 77. Rižner TL, Penning TM. Role of aldo-keto reductase family 1 (AKR1) enzymes Galván P, et al. The calcium-sensing receptor is silenced by genetic in human steroid metabolism. Steroids. 2014;79:49–63. and epigenetic mechanisms in unfavorable neuroblastomas and its 78. Nio-Kobayashi J, Kudo M, Sakuragi N, Iwanaga T, Duncan WC. Loss of reactivation induces ERK1/2-dependent apoptosis. Carcinogenesis. luteotropic prostaglandin E plays an important role in the regulation of 2013;34(2):268–76. luteolysis in women. MHR Basic Sci Reprod Med. 2017;23(5):271–81. 57. Lee JY, Tae JC, Kim CH, Hwang D, Kim KC, Suh CS, et al. Expression of the 79. Trau HA, Davis JS, Duffy DM. Angiogenesis in the primate ovulatory follicle genes for peroxisome proliferator-activated receptor-γ, cyclooxygenase-2, is stimulated by luteinizing hormone via prostaglandin E2. Biol Reprod. and proinflammatory cytokines in granulosa cells from women with 2015;92(1):15. polycystic ovary syndrome. Clin Exp Reprod Med. 2017;44(3):146. 80. Trau HA, Brännström M, Curry TE, Duffy DM, Duffy DM. Prostaglandin E2 58. Schmidt J, Weijdegård B, Mikkelsen AL, Lindenberg S, Nilsson L, Brännström M. and vascular endothelial growth factor A mediate angiogenesis of human Differential expression of inflammation-related genes in the ovarian stroma ovarian follicular endothelial cells. Hum Reprod. 2016;31(2):436–44. and granulosa cells of PCOS women. Mol Hum Reprod. 2014;20(1):49–58. 81. Chandras C, Harris TE, López Bernal A, Abayasekara DRE, Michael AE. PTGER1 59. Amato G, Conte M, Mazziotti G, Lalli E, Vitolo G, Tucker AT, et al. Serum and and PTGER2 receptors mediate regulation of progesterone synthesis and follicular fluid cytokines in polycystic ovary syndrome during stimulated type 1 11β-hydroxysteroid dehydrogenase activity by prostaglandin E2in cycles. Obstet Gynecol. 2003;101(6):1177–82. human granulosa-lutein cells. J Endocrinol. 2007;194(3):595–602. 60. Gao L, Gu Y, Yin X. High serum tumor necrosis factor-alpha levels in women 82. Aleshire SL, Osteen KG, Maxson WS, Entman SS, Bradley CA, Parl FF. with polycystic ovary syndrome: a meta-analysis. Atkin SL, editor PLoS One. Localization of transferrin and its receptor in ovarian follicular cells: 2016;11(10):e0164021. morphologic studies in relation to follicular development. Vol. 51. Fertil 61. Yamamoto Y, Kuwahara A, Taniguchi Y, Yamasaki M, Tanaka Y, Mukai Y, et Steril. 1989;51(3):444–9. al. Tumor necrosis factor alpha inhibits ovulation and induces granulosa cell 83. Gomme PT, McCann KB. Transferrin: structure, function and potential death in rat ovaries. Reprod Med Biol. 2015;14(3):107–15. therapeutic actions. Elsevier Current Trends. 2005;10(4):267–73. 62. Sakumoto R, Shibaya M, Okuda K. Tumor necrosis factor-alpha (TNF alpha) 84. Wu Y-T, Wu Y, Zhang J-Y, Hou N-N, Liu A-X, Pan J-X, et al. Preliminary inhibits progesterone and estradiol-17beta production from cultured proteomic analysis on the alterations in follicular fluid proteins from women granulosa cells: presence of TNFalpha receptors in bovine granulosa and undergoing natural cycles or controlled ovarian hyperstimulation. J Assist theca cells. J Reprod Dev. 2003;49(6):441–9. Reprod Genet. 2015;32(3):417–27. 63. Morrison LJ, Marcinkiewicz JL. Tumor necrosis factor α enhances oocyte/ 85. Chattopadhayay R, Ganesh A, Samanta J, Jana SK, Chakravarty BN, follicle apoptosis in the neonatal rat ovary1. Biol Reprod. 2002;66(2):450–7. Chaudhury K. Effect of follicular fluid oxidative stress on meiotic spindle Sagvekar et al. Clinical Epigenetics (2019) 11:61 Page 16 of 16

formation in infertile women with polycystic ovarian syndrome. Gynecol Obstet Investig. 2010;69(3):197–202. 86. Revised 2003 consensus on diagnostic criteria and long-term health risks related to polycystic ovary syndrome. Fertil Steril. 2004;81(1):19–25. 87. Tabas-Madrid D, Nogales-Cadenas R, Pascual-Montano A. GeneCodis3: a non-redundant and modular enrichment analysis tool for functional genomics. Nucleic Acids Res. 2012;40(W1):W478–83. 88. Leo CP, Vitt UA, Hsueh AJW. The ovarian kaleidoscope database: an online resource for the ovarian research community. Endocrinology. 2000;141(9): 3052–4. Gene Bionetwork Analysis of Ovarian Primordial Follicle Development

Eric E. Nilsson1, Marina I. Savenkova1, Ryan Schindler1, Bin Zhang2, Eric E. Schadt2, Michael K. Skinner1* 1 Center for Reproductive Biology, School of Molecular Biosciences, Washington State University, Pullman, Washington, United States of America, 2 Sage Bionetworks, Seattle, Washington, United States of America

Abstract Ovarian primordial follicles are critical for female reproduction and comprise a finite pool of gametes arrested in development. A systems biology approach was used to identify regulatory gene networks essential for primordial follicle development. Transcriptional responses to eight different growth factors known to influence primordial follicles were used to construct a bionetwork of regulatory genes involved in rat primordial follicle development. Over 1,500 genes were found to be regulated by the various growth factors and a network analysis identified critical gene modules involved in a number of signaling pathways and cellular processes. A set of 55 genes was identified as potential critical regulators of these gene modules, and a sub-network associated with development was determined. Within the network two previously identified regulatory genes were confirmed (i.e., Pdgfa and Fgfr2) and a new factor was identified, connective tissue growth factor (CTGF). CTGF was tested in ovarian organ cultures and found to stimulate primordial follicle development. Therefore, the relevant gene network associated with primordial follicle development was validated and the critical genes and pathways involved in this process were identified. This is one of the first applications of network analysis to a normal developmental process. These observations provide insights into potential therapeutic targets for preventing ovarian disease and promoting female reproduction.

Citation: Nilsson EE, Savenkova MI, Schindler R, Zhang B, Schadt EE, et al. (2010) Gene Bionetwork Analysis of Ovarian Primordial Follicle Development. PLoS ONE 5(7): e11637. doi:10.1371/journal.pone.0011637 Editor: Hugh Clarke, McGill University, Canada Received March 18, 2010; Accepted June 11, 2010; Published July 16, 2010 Copyright: ß 2010 Nilsson et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: National Institutes of Health (NIH) R01 HD043093-01. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * E-mail: [email protected]

Introduction lifespan of a female, follicles gradually leave the arrested pool to undergo a primordial to primary follicle transition. A follicle An emerging concern in the field of biomedical research is that undergoing follicle transition has an increase in oocyte diameter the common reductionist approach to studying biological and the associated granulosa cells proliferate and change from a processes may not be adequate to fully understand the complex flattened to cuboidal in shape. Once primordial to primary follicle interplay of cellular signaling, gene expression, and other complex transition has occurred the follicle either continues to develop to molecular processes that occur within a tissue or organ. Examples the point of ovulation or undergoes atresia [1,2,3,4]. Previously of reductionist studies that have driven much of our understanding cell-to-cell communication with extra-cellular growth factors has of biological processes associated with complex phenotypes like been shown to regulate the initiation of primordial follicle disease include the knockout mouse experiments and in vitro development. These studies have primarily used a reductionist cytokine treatment to assess the effects of gene-specific perturba- approach to test candidate growth factors one at a time for their tions on cell or tissue biology. Results from these types of studies ability to affect follicle transition. A number of paracrine growth provide information on candidate regulatory factors, but typically factors have been identified as having a role in early follicle do not elucidate the network of factors or processes required for a development (reviews [4,5]). normal developmental biology or pathobiology. A holistic, systems To move beyond examining single gene effects on this biology, approach to studying normal developmental processes development process, gene network analysis can be employed to can be a powerful tool that is complementary to the more identify groups (e.g. modules) of genes whose expression is reductionist experiments. In the spirit of a systems-based approach regulated in a coordinated manner (gene network) [6,7,8]. In this to development, the current study was designed to identify gene type of analysis, a biological system is surveyed in the context of networks involved in ovarian primordial follicle development and disease (or other interesting phenotypes) with microarrays multiple to characterize critical regulatory factors involved in this times with and without perturbations that cause the system to development process. change. A novel bioinformatics analysis is used to identify modules In mammals, all the oocytes (eggs) that will be used over a of genes associated with biological systems (bionetwork). The great female’s lifetime are present in the ovary at birth in a finite pool. majority of network analyses have focused on disease states and These oocytes are arrested in prophase of the first meiotic division been used to better understand the systems biology of disease and are each surrounded by flattened pre-granulosa cells to form a processes and identify potential therapeutic targets [9,10,11, structure called a primordial follicle [1]. During the reproductive 12,13,14,15]. The current study was designed to determine if

PLoS ONE | www.plosone.org 1 July 2010 | Volume 5 | Issue 7 | e11637 Primordial Follicle Bionetwork network analysis can be applied to study a normal development process. The current study used whole rat ovaries cultured in vitro in a manner that allowed primordial to primary follicle transition. The ovaries were treated with one of eight different growth factors previously shown to regulate primordial follicle transition in comparison to untreated control cultures. The mRNA was isolated from the ovaries and used for microarray transcriptome analysis to globally survey gene expression under these different treatment conditions. The effects of each growth factor on gene expression were analyzed to determine similarities and differences in gene expression between the different growth factor treatments. Those genes whose mRNA expression changed with any treatment were subjected to network analysis to identify pathways and genes with a high degree of connectivity between other genes and pathways. From the networks constructed from these data we identified a list of critical modules of regulated genes forming gene sub-networks that were used to identify regulatory genes involved in primordial follicle development. Not only were previously identified regula- tory factors/genes associated with this process identified from this Figure 1. Number of genes and pathways overlapped between network analysis, but a number of putative regulators of follicle signature (growth factor treatment group) lists. Total number of differentially expressed genes for each growth factor is shown in dark transition not previously associated with this process were also yellow column, number of genes overlapped between each pair of determined. One of the new candidate genes, connective tissue signature lists – in light yellow columns. Total number of KEGG growth factor (Ctgf) [16], was tested experimentally and found to pathways affected by each growth factor is shown in dark green row; promote primordial to primary follicle transition. Observations number of KEGG pathways overlapped for each pair of growth factors is demonstrate the utility of this network analysis to be used as a shown in light green row. CTGF analysis separate from the network analysis. systems biology approach to study normal developmental doi:10.1371/journal.pone.0011637.g001 processes in complex systems. heavily impacted by genes whose expression altered in response to Results the growth factor treatments (Table 1) included pathways involved Primordial Follicle Transcriptome Analysis in cell surface and extracellular matrix regulation (cell adhesion A number of regulatory factors have been shown to affect molecules, adherens junction, focal adhesion, tight junction, gap primordial to primary follicle transition, including Amh [17,18], junction, regulation of actin cytoskeleton), known signaling Fgf2 [19,20,21], Bmp4 [22,23], Gdnf [24], Fgf7/KGF [25], Kitlg pathways (MAPK, notch, B-cell receptor, adipocytokine, toll-like [19,26,27], Lif [28] and Pdgfa [29]. In order to determine the receptor, ErbB, GnRH, Wnt, hedgehog, VEGF, Jak-STAT, TGF- underlying gene networks and processes involved in primordial beta, p53, insulin, PPAR), the complement cascade, axon follicle development, microarray analysis was performed on RNA guidance, glycan structure biosynthesis and pathways listing cell from whole rat ovaries treated for two days in vitro with each of communication ligand-receptor interactions (cytokine-cytokine the above listed growth factors independently. There were three receptor, neuroactive ligand-receptor, ECM-receptor). There independent RNA samples of pooled ovaries for each growth was a high degree of overlap of affected pathways between factor treatment (except for GDNF, which had only two sample different growth factor treatments (Figure 1). For the list of replicates), and corresponding control samples for a total of 38 pathways containing altered genes, from 70% to 82% of those RNA samples. These were evaluated using 38 Affymetrix Rat pathways are shared with at least one other treatment. Application Gene 1.0 ST microarrays. The array data were analyzed together of the hypergeometrical Fisher Exact Test to assess whether the using normalization and pre-processing described in the Methods. number of overlapped pathways was significantly greater than Each growth factor treatment resulted in 79 to 349 genes with expected by chance, revealed that the majority were statistically altered expression compared to controls (Figure 1). The lists of the significant. The pathways containing altered genes from several genes affected by each treatment are presented in Table S1. There growth factor treatments are presented in Table 1. Although few were relatively few genes with altered expression in common altered genes were found to overlap between different treatments between the different treatments (Figure 1). Less than 10% of the (Figure 1), each growth factor treatment influenced similar genes changed by any one growth factor treatment were found to pathways, Table 1 and Table S1. Therefore, each growth factor be changed in any other treatment. The exception was Fgf7/KGF, affects similar pathways via different genes. which had a more than 30% overlap of altered genes with Amh. There were no individual genes that changed expression levels Bionetwork Analysis in response to more than three of the eight original treatments The complete list of genes whose expression levels changed with (Table S1). any growth factor treatment was subjected to a network analysis as The complete list of genes whose expression levels changed with described in Methods. Potential batch effects for culture date, any of the treatments was compared to curated lists of genes from RNA processing data and microarray performance date were the KEGG database to identify processes that may be important corrected during the analysis, with no major effect on the analysis. for primordial follicle development. Automated unbiased match- The data were fit using a robust linear regression model (rim ing of lists of affected genes to KEGG pathways was performed function from R statistical package), and then the residuals with with Pathway Express (Intelligent Systems and Bioinformatics respect to the model fit were carried forward in all subsequent Laboratory; http://vortex.cs.wayne.edu/ontoexpress/). Pathways analysis. The network analysis scores each gene according to how

PLoS ONE | www.plosone.org 2 July 2010 | Volume 5 | Issue 7 | e11637 LSOE|www.plosone.org | ONE PLoS Table 1. Number of Regulated Genes in Pathways.

Total List Pathway Name (1540) Module Lists Signature Lists Short/Connected List turquoise blue brown yellow green red black pink magenta purple green-yellow tan salmon cyan grey AMH bFGF BMP4 GDNF KGF KL LIF PDGF CTGF Total Overlap Number Impact Factor

Number of genes in the list 55 194 182 158 150 139 112 99 85 68 45 32 29 28 22 157 268 248 79 148 123 271 349 275 155

Neuroactive ligand-receptor interaction 22 1.23 1 2 5213 1121 1345 145463 Focal adhesion 19 3.62 3 7 171 1 11 33 4831 MAPK signaling pathway 19 3.00 6 7 1341 1 116 4 546 Cell adhesion molecules 15 107 1 1 321 1 3 11 26232 3221 (CAMs) Axon guidance 15 6.34 1 4 1241 11 6 132333 3 Cytokine-cytokine receptor interaction 15 2.24 1 3 343 1 1214 165 1 Calcium signaling pathway 13 1.72 13 512 1 13212341 Regulation of actin cytoskeleton 12 2.18 2 5 142 42 54 Cell Communication 11 4.70 2 43 1 1 312 1411 Jak-STAT signaling pathway 11 1.88 3 131 2 1 11 144 1 Adipocytokine signaling pathway 10 2.94 2 22 1 11 1 112 711 ECM-receptor interaction 9 4.23 2 2 42 1 121 52 Complement and coagulation cascades 9 3.63 1 124 1 24 14 1

uy21 oue5|Ise7|e11637 | 7 Issue | 5 Volume | 2010 July Apoptosis 9 2.94 1 221 21 212 25 Toll-like receptor signaling pathway 9 2.92 3 1 1 1 1 1 12 1 1 12 1 1 Leukocyte transendothelial migration 9 2.74 1 231 1 131 51 Gap junction 9 2.73 1 2 1221 1 2 1313 rmrilFlil Bionetwork Follicle Primordial Insulin signaling pathway 9 1.35 11 12 12 1 21 14 21 Adherens junction 8 33.1 33 1 12 341 Glycan structures - biosynthesis 2 8 3.83 122 11 1113131 T cell receptor signaling pathway 8 2.71 2 111 1 23 1221 Tight junction 7 3.28 111111 11 24 GnRH signaling pathway 6 2.84 1 2 2 111 22 1 TGF-beta signaling pathway 6 1.40 1 2 12 1 212 111 LSOE|www.plosone.org | ONE PLoS Table 1. Cont.

Total List Pathway Name (1540) Module Lists Signature Lists Short/Connected List turquoise blue brown yellow green red black pink magenta purple green-yellow tan salmon cyan grey AMH bFGF BMP4 GDNF KGF KL LIF PDGF CTGF Total Overlap Number Impact Factor

Number of genes in the list 55 194 182 158 150 139 112 99 85 68 45 32 29 28 22 157 268 248 79 148 123 271 349 275 155

PPAR signaling pathway 6 1.24 1 1 111 1 1 1 1 12 Antigen processing and presentation 5 36.8 1 1 2 11111 1 Melanogenesis 5 3.34 12 11 1 11 121 B cell receptor signaling pathway 5 2.95 212 21 1 2 Wnt signaling pathway 5 2.84 31 12113 4 Hematopoietic cell lineage 5 1.28 1 1 2 1122 1 1 ErbB signaling pathway 4 2.89 2 1 11 13 Hedgehog signaling pathway 4 2.38 22 12 11 VEGF signaling pathway 4 2.17 12 1 1 111 p53 signaling pathway 4 1.37 11 1 1 1 121 Glycan structures - biosynthesis 1 4 0.85 1 1 1 11211 Renin-angiotensin system 3 3.21 12 1 2 Notch signaling pathway 3 2.95 111 1 11 1 uy21 oue5|Ise7|e11637 | 7 Issue | 5 Volume | 2010 July mTOR signaling pathway 3 2.82 1111111 Basal transcription factors 3 2.03 11 1 1 11 Fc epsilon RI signaling pathway 3 1.87 11 1 1 2 Nucleotide excision repair 3 1.49 1221 1 1 rmrilFlil Bionetwork Follicle Primordial Cell cycle 3 1.13 11 1 1 1 111 Ubiquitin mediated proteolysis 3 0.98 11 1 1 1 11 Phosphatidylinositol signaling system 2 28.0 112

Number of regulated genes in affected KEGG pathways for each of modules and signature (Growth Factor Treatment) lists. Pathway enrichment analysis was performed using KEGG database, current for July 2009, for which gene symbols were used as input data. From 1,540 probe sets, 1,137 were identified as annotated genes using Affymetrix RaGene- 1_0-st-v1 Transcript Cluster Annotations, release 28 (March, 2009). Top 44 pathways shown except for disease pathways (not shown). No KEGG pathways were affected by genes from light-cyan or midnight-blue modules. doi:10.1371/journal.pone.0011637.t001 Primordial Follicle Bionetwork

Figure 2. The ovary gene co-expression network and corresponding gene modules. A topological overlap matrix of the gene co- expression network consisting of the 1540 genes regulated by the various growth factors. Genes in the rows and columns are sorted by an agglomerative hierarchical clustering algorithm (see Methods). The different shades of color signify the strength of the connections between the nodes (from white signifying not significantly correlated to red signifying highly significantly correlated). The hierarchical clustering (top) and the topological overlap matrix strongly indicate highly interconnected subsets of genes (modules). Modules identified are colored along both column and row and are boxed. The number of genes in each module is listed as size of module. doi:10.1371/journal.pone.0011637.g002 well, under different treatments, its changes in gene expression are modules identified. The membership of each module can be found correlated with the changes in expression of every other gene. This in Table S1. The sixteen modules contained 1,383 genes with the gives a connectivity score for each gene. High connectivity scores remaining 157 genes (colored as gray) not failing into any module. indicate that expression of this gene changes in concert with that of The pathways containing genes whose expression changed with many other genes. In addition, the network analysis identifies gene growth factor treatment were compared to the genes from each modules in which the member genes have similar changes in module that were associated with specific pathways (Table 1 and expression in response to the various growth factor treatments. Table S1 for full list). For most pathways genes from several Gene modules are functional components of the network that are network modules were present. However, several pathways were often associated with specific biological processes. To identify associated with selected modules. For example, out of 19 altered modules comprised of highly interconnected expression traits genes present in the focal adhesion pathway, seven were from the within the co-expression network, we examined the topological turquoise module and seven from the brown. Similarly, of the five overlap matrix [30] associated with this network. The topological altered genes in the Wnt signaling pathway three were from the overlap between two genes not only reflects their more proximal turquoise module. For the fifteen changed genes in the cell interactions (e.g., two genes physically interacting or having adhesion molecule pathways three were from blue and three from correlated expression values), but also reflects the higher order magenta modules. Those genes whose expression changed with interactions that these two genes may have with other genes in the specific growth factor treatments were cross-matched with the network. Figure 2 depicts a hierarchically clustered topological genes assigned to each network module to determine if specific overlap map in which the most highly interconnected modules in modules were heavily influenced by particular growth factors. the network are readily identified. The specific details of the gene Interestingly, each module was biased toward having many genes co-expression network analysis (Figure 2) are presented in the in common with selected growth factors (Table S1). In contrast, Methods section. To identify gene modules (sub-networks) some growth factor treatments induced changes in genes that were formally from the topological overlap map, we employed a distributed among several different modules. previously described dynamic cut-tree algorithm with near optimal In order to identify genes that could be key regulators of performance on complicated dendrograms [31] (see Methods for primordial follicle development, a shorter list of candidate genes details). Figure 2 shows the topological overlap map of the co- was generated from the results of the network analysis. Six expression network with gene modules color-coded for the 16 modules were chosen for having the highest numbers of known

PLoS ONE | www.plosone.org 5 July 2010 | Volume 5 | Issue 7 | e11637 Primordial Follicle Bionetwork regulatory genes and pathways (yellow, turquoise, blue, brown, Discussion red, purple). The top 10% of most connected transcripts in each of the six modules were identified as potential important regulators A systems biology approach was used to elucidate the changes in [13,15], except for the blue module for which the top 20% were gene expression that are important for ovarian primordial to chosen since so many of that module’s most highly connected primary follicle transition. A gene network analysis was performed transcripts were not annotated as genes. The compiled list on the ovarian transcriptomes following treatment with 8 different included 55 transcripts annotated as genes (Table 2), and these growth factors. The rat ovary was used as a model system to test genes were subjected to more intensive investigation. the utility of this approach in investigating a normal developmen- An automated unbiased analysis of published scientific literature tal process. This is one of the first applications of network analysis was applied to the lists of differentially expressed genes described to a normal developmental process. The objective was to identify above using Genomatix/BiblioSphere software, as described in the critical regulatory factors and pathways in primordial follicle Methods. Figure 3 shows a small integrated gene network among development following a bionetwork analysis. the short list of the 55 candidate regulators (Table 2). These Microarray analysis determined the alterations in the ovarian relationships in literature raise the possibility that physiological transcriptome that occurred in response to treatment of ovaries with interactions exist between these genes. The gene Ctgf (Connective AMH,FGF2,BMP4,GDNF,FGF7,KITL,LIF,andPDGFB.Allof tissue growth factor) was seen to relate to several other genes with these have previously been shown to effect follicle transition high connectivity, Figure 3. Interestingly, two of the identified [17,18,19,20,21,22,23,24,25,26,27,28,29,34]. All these factors stimu- genes were previously shown to influence primordial follicle late primordial follicle development except AMH that inhibits follicle development, Pdgfa [29] and Fgfr2 [19,20,21]. development. The presence of both positive and negative factors The entire set of 1540 transcripts differentially expressed with provides a wider diversity of gene regulation to facilitate the network growth factor treatment was also subjected to analysis using analysis. As expected the AMH regulated gene set is more distinct BiblioSphere. Only 632 were recognized by BiblioSphere and 613 from the others. Surprisingly, there were few altered genes in were connected. A diagram of literature relationships between common between all these growth factors and there were no genes these genes is presented in Figure S1. Five major gene clusters that significantly changed in expression level in response to more than were identified as associated with Nfkb1, Vegfa, Gadd45a, Esr1 three of the eight growth factors. In contrast, the physiological and Egfr1. This analysis is useful to compare with the expression processes impacted by these altered genes were found to have a network analysis, but is biased toward the literature and finding higher level of overlap. Since a pathway includes groups of genes, it is relationships among more heavily studied factors. expected that the overlap of pathways between growth factor treatments will be higher. The overlap of pathways was markedly Analysis of CTGF Actions high (70% to 82%) and statistically different, suggesting pathway Critical regulatory candidates for primordial follicle develop- associations provide a predicted capacity to identify regulatory ment were selected due to their being differentially expressed in factors. Certain pathways were significantly over-represented in the response to treatment with growth factors, having a high pool of genes with changed expression. This suggests that there are connectivity score, and being related in literature to other highly selected physiological pathways that are influenced by all the different connected genes. For the purpose of the current study, candidate growth factors (Figure 5), but that each growth factor affects different regulatory genes that were also extracellular growth factors were genes at different points in these pathways (Table 1). Multiple input considered. Therefore, CTGF was selected based on all these points into these physiological pathways could allow for more precise criteria for further analysis. Experiments were performed to see if regulation and more effective compensation between the growth CTGF could regulate primordial to primary follicle transition. factors. Since many growth factors are acting in parallel to regulate Ovaries from four-day old rats were treated with 50ng/ml CTGF these pathways, any one pathway system is robust and maintains protein for ten days in an organ culture system as described in the function if one growth factor becomes inoperative. Since primordial Methods (Figure 4A and 4B). Transforming growth factor beta 1 follicle development is essential for female reproduction, a complex (TGFB1), which is known to interact with CTGF [32,33], was also network of regulatory factors influencing different aspects of critical tested. Untreated cultured ovaries were used as a negative control, signaling pathways has evolved. and ovaries treated with 50ng/ml each of Kit ligand (KITL) and For the eight growth factors evaluated the cellular processes Fibroblast growth factor 2 (FGF2) were used as a positive control. affected in common (Figure 5) included changes in cell contact, CTGF treatment resulted in a significant (p,0.05) increase in morphogenesis, and cell proliferation and differentiation. These developing follicles compared to untreated controls, as did are processes that are necessary for the morphological changes treatment with the combination of KITL and FGF2 (Figure 4C). that occur with primordial to primary follicle transition. During TGFB1 had no effect, either alone or in combination with CTGF. follicle transition granulosa cells change from squamous to RNA was collected from CTGF and control cultured ovaries as cuboidal and the oocyte starts to grow in diameter (Figure 4B). described in the Methods from three replicate experiments. The Unexpectedly, what was also seen as an important affected cellular RNA was used for microarray analysis using the same criteria as process was regulation of several key components of the for the other growth factors used in the network analysis. One complement and coagulation cascades (Figure S2). These genes hundred fifty-five transcripts were differentially expressed in are not known for having roles in ovary or follicle development, CTGF-treated ovaries, Table S1. As was seen for the other and merit further investigation. growth factors used in the network analysis, there was little overlap Gene networks provide a convenient framework for exploring the of these changed genes with the genes showing changed expression context within which single genes operate. For gene networks in response to any other growth factor treatment, Figure 1. associated with biological systems, the nodes in the network typically However, as seen among the other growth factors, there was a represent genes, and edges (links) between any two nodes indicate a high degree of overlap between the pathways impacted by CTGF relationship between the two corresponding genes. An important treatment and treatment with other growth factors (Figure 1). end product from the gene co-expression network analysis is a set of Therefore, a critical regulatory gene predicted from the network gene modules which member genes are more highly correlated with analysis was confirmed to regulate primordial follicle development. each other than with genes outside a module. It has been

PLoS ONE | www.plosone.org 6 July 2010 | Volume 5 | Issue 7 | e11637 Primordial Follicle Bionetwork

Table 2. List of 55 Candidate Regulatory Genes.

Regulation by Function GeneSymbol GeneBank Probeset k.in* Module Growth Factor GeneTitle

Apoptosis Bcl2l10 NM_053733 10911690 10.0 red KL-dwn BCL2-like 10 Cell Cycle Cdkn2c NM_131902 10878705 7.6 red KL-dwn CDK4 inhibitor 2C ECM Cdh3 NC_005118 10807525 28.0 turq AMH-dwn cadherin 3, type 1, P-cadherin Col11a1 AJ005396 10818502 20.6 brown KL-dwn collagen, type XI, alpha 1 Krt19 NM_199498 10747262 25.2 turq AMH-dwn keratin 19 Lama5 NC_005102 10852270 30.9 turq AMH-dwn laminin, alpha 5 Development Bnc2 NM_001106666 10877880 26.5 turq AMH-dwn basonuclin 2 Emx2 NM_001109169 10716454 27.5 turq KGF-dwn empty spiracles homeobox 2 Usmg5 NM_133544 10730633 13.5 blue AMH, GDNF-dwn muscle growth 5 homolog (mouse) Epigenetics Dnmt1 NM_053354 10915437 22.9 brown KL-dwn DNA (cytosine-5-)-methyltransferase 1 Golgi B4galt6 NM_031740 10803394 11.7 blue AMH-dwn, KGF-dwn galactosyltransferase, polypeptide 6 Growth Factors Ctgf NM_022266 10717233 8.2 blue KGF-up, LIF-dwn connective tissue growth factor Il16 NM_001105749 10723351 31.9 turq AMH-dwn, KGF-dwn interleukin 16 Pdgfa NM_012801 10757129 27.0 turq AMH-dwn, KGF-dwn platelet-derived growth factor Immune LOC287167 NM_001013853 10741765 21.6 brown KL-dwn globin, alpha Metabolism Cacna2d3 NM_175595 10789819 19.9 brown KL-dwn calcium channel Hbq1 XM_001061675 10741761 24.0 brown KL-dwn hemoglobin, theta 1 Hhatl NM_001106868 10914424 15.7 yellow PDGF-up hedgehog acyltransferase-like Hmgcs2 NM_173094 10817759 19.5 brown KL-dwn Coenzyme A synthase 2 Hsd11b1 NM_017080 10770795 17.0 yellow BMP4-dwn hydroxysteroid 11-beta dehydro Kirrel NM_207606 10824123 14.1 yellow AMH-dwn kin of IRRE like (Drosophila) Plod2 NM_175869 10912255 18.3 brown KL-dwn procollagen lysine, 2-oxoglutarate Podxl NM_138848 10861662 26.6 turq AMH-dwn, KGF-dwn podocalyxin-like Scn3a NM_013119 10845809 17.4 brown KL-dwn sodium channel, type III, alpha Slc4a4 NM_053424 10775997 29.0 turq AMH-dwn, KGF-dwn solute carrier family 4 Slc7a5 NM_017353 10811531 15.6 yellow PDGF-up solute carrier family 7 Slc29a1 NM_031684 10921833 15.4 yellow PDGF-up solute carrier family 29 Eno1 NM_012554 10874152 9.0 purple PDGF-up enolase 1, (alpha) Receptors Axl NM_031794 10719900 15.4 yellow BMP4-dwn Axl receptor Ednrb NM_017333 10785724 20.7 brown KL-dwn endothelin receptor type B Fgfr2 NM_012712 10726172 30.8 turq AMH-dwn, KGF-dwn fibroblueast growth factor receptor 2 Itgb3bp NM_001013213 10878272 11.0 blue GDNF-dwn integrin beta 3 binding protein Plxna4a NM_001107852 10861678 30.9 turq AMH-dwn, KGF-dwn A4, A Tmem151a NM_001107570 10727725 26.2 turq AMH-dwn, KGF-dwn transmembrane protein 151A Signaling Nrgn NM_024140 10916228 16.1 yellow AMH, KGF-dwn, PDGF-up neurogranin Dusp4 NM_022199 10792035 17.3 yellow AMH-dwn dual specificity phosphatase 4 Dusp6 NM_053883 10895144 16.6 yellow BMP4-dwn dual specificity phosphatase 6 Efna5 NM_053903 10930204 15.0 yellow AMH, GDNF-dwn ephrin A5 Map3k1 NM_053887 10821276 26.8 turq KGF-dwn mitogen activated protein kinase Pde7b NM_080894 10717069 17.8 brown KL-dwn phosphodiesterase 7B Rem1 NM_001025753 10840861 14.5 yellow PDGF-up RAS (RAD and GEM)-like Shc4 NC_005102 10849423 17.0 yellow AMH-dwn SHC family, member 4 Ubash3b AC_000076 10916476 16.9 yellow PDGF-up ubiquitin associated Transcription Btg4 NM_001013176 10909937 7.4 red KL-dwn B-cell translocation gene 4 Etv5 NM_001107082 10752034 14.4 yellow AMH-dwn ets variant 5 Fbxo15 NM_001108436 10803025 11.9 red KL-dwn F-box protein 15 Misc. & Unknown Depdc2 NM_001107899 10875023 29.5 brown KL-dwn DEP domain containing 2 Fam154a AC_000073 10877890 11.4 red KL-dwn similarity 154, member A LOC686725 AC_000076 10915208 25.9 turq AMH-dwn hypothetical protein LOC686725

PLoS ONE | www.plosone.org 7 July 2010 | Volume 5 | Issue 7 | e11637 Primordial Follicle Bionetwork

Table 2. Cont.

Regulation by Function GeneSymbol GeneBank Probeset k.in* Module Growth Factor GeneTitle

RGD1306186 BC090317 10881318 9.2 red KL-dwn similar to RIKEN cDNA 4930569K13 RGD1306622 XM_001074493 10728647 32.6 turq AMH-dwn, KGF-dwn similar to KIAA0954 protein RGD1308023 XR_006437 10850490 17.6 brown KL-dwn, LIF-dwn similar to CG5521-PA RGD1566021 AC_000086 10800122 19.6 brown KL-dwn similar to KIAA1772

Short list of 55 genes that are the most connected genes with known functions in the modules of interest. Selected from the top 10% most connected genes in each module (except blue module for which considered top 20% as many hubs are not annotated. Abbreviations used: dwn - down-regulated; up - up-regulated; (*)- k in. is connectivity coefficient obtained/calculated in network analysis. doi:10.1371/journal.pone.0011637.t002 demonstrate that these types of modules are enriched for known receptor) for Fgf2 and Fgf7 (KGF). PDGF, KGF/FGF7 and FGF2 biological pathways for genes that associate with disease traits and proteins have previously been shown to regulate primordial to for genes that are linked to common genetic loci [6,35]. primary follicle transition [29,34]. Therefore, the bionetwork The current study employed a weighted gene co-expression predicted to be involved in the regulation of primordial follicle network approach that has been extensively used for uncovering development identified two previously known regulatory factors biologically meaningful gene modules [7,13,15] to explore novel which validated the utility of the network analysis for identifying pathways involved in primordial follicle development. An unsuper- candidate regulatory genes, consistent with previous network studies vised and unbiased approach was used to nominate potential [13,15]. This sub-network also identified connective tissue growth regulatory candidates for these modules based on gene network factor (Ctgf) [16,36] as a putative regulator of primordial follicle connectivity. The connectivity score shows how well under different development. An ovarian organ culture experiment confirmed that treatments the changes in gene expression for a gene are correlated CTGF promotes primordial to primary follicle transition. There- with the changes in expression for every other gene. In the current fore, a regulatory factor predicted to be important for primordial study, the gene co-expression network analysis helped select 55 follicle development was confirmed to be involved which further highly connected genes for further functional analysis. An validated the bionetwork approach. A microarray analysis of automated literature search of these 55 genes revealed a sub- CTGF-treated ovaries showed an altered gene set similar to those of network relationship among them as presented in Figure 3. This the other growth factors known to regulate follicle transition. These sub-network suggested regulatory roles for Pdgfa and Fgfr2 (the observations validate the network-based systems biology approach to elucidate the regulation of a complex developmental process. Consideration of the 55 intra-module hub genes from critical regulatory modules revealed a number of signaling and cellular processes were influenced, Figure 5 and Figure S3. In the growth factor/chemokine family Pdgfa and Ctgf were confirmed to be involved. The IL16 identified is currently being investigated as a potential regulatory candidate. The specific genes identified in Table 2 and associated regulatory processes provide potential therapeutic targets to regulate primordial follicle development. The ability to inhibit or stimulate primordial follicle development with a therapeutic treatment has a number of clinical applications. A delay in primordial follicle development and maintenance of the primordial pool could delay the onset of menopause and extend the reproductive life span of a female. In addition, the ability to therapeutically inhibit primordial follicle development would provide a treatment for premature ovarian failure, a disease when the primordial pool is lost early in life causing female infertility. In contrast, the therapeutic stimulation of primordial follicle development could treat forms of female infertility [4]. The induction of primordial follicle development also could promote the loss of the primordial pool and induce female sterility. The bionetwork identified in the current study produced a number of potential therapeutic targets to manipulate primordial follicle development and female reproductive capacity. The systems biology approach taken with this network analysis Figure 3. Sub-network connection scheme for the most highly of primordial follicle development identified clusters and modules connected 55 candidate genes obtained by global literature of genes involved in this critical development process. A number of analysis. Only 15 connected genes from the list of 55 are shown, while the growth factors previously shown to be involved (e.g. PDGF the rest are not connected and not shown. Red and yellow colors represent up-regulated genes, blue and turquoise colors – down- and bFGF) were identified, but other factors known to be regulated genes. important for ovarian development were not identified. Often a doi:10.1371/journal.pone.0011637.g003 reductionist approach such as a knockout mouse model can

PLoS ONE | www.plosone.org 8 July 2010 | Volume 5 | Issue 7 | e11637 Primordial Follicle Bionetwork

Figure 4. Analysis of the role of CTGF in primordial follicle transition. A) Hematoxylin-eosin stained ovary sections showing a representative arrested primordial follicle (left), and a developing primary follicle after having undergone primordial to primary follicle transition (right). B) Graphic representation of primordial and primary follicles. C) Effect of CTGF on cultured ovaries. Ovaries were cultured for 10 days with the treatments indicated. Ovarian histological analysis determined the percentage of primordial versus developing follicles for each ovary. Bars are the mean percent developing follicles 6 SEM. N = 5–7 per treatment from four different replicate experiments. Asterisks indicate a significant (*p,0.05 or ***p,0.01) difference from control by Dunnet’s post-hoc test after a significant result of ANOVA. doi:10.1371/journal.pone.0011637.g004 identify a factor as being important for the maintenance of tissue mental biology of this process and provide potential therapeutic development or function, but this does not mean the factor is targets for ovarian disease and function. This sub-network was regulated during the process. In addition, critical developmental validated by the presence of two genes previously identified as processes such as primordial follicle development often have a set being important. A new gene identified, Ctgf, was tested and of compensatory factors that have evolved such that loss of any one found to regulate primordial follicle development. Therefore, the will still allow the process to proceed. Therefore, knockout models network based systems biology approach was partially validated often do not have phenotypes for these factors. This does not mean for a normal developmental process. This type of approach will the factor is not important, but instead that the developmental likely be invaluable to study development on a systems biology process is essential and thus multiple factors compensate to assure level in the future. the developmental process occurs. The current study takes a systems biology approach to identify networks of genes involved in Materials and Methods the process without the bias of a reductionist model. Therefore, novel groups of factors and cellular processes were identified that Ovarian organ culture now require further investigation. Four-day old female Sprague-Dawley rats (Harlan Laborato- The integrative analysis revealed a gene sub-network involved ries, Inc., USA) were euthanized according to the laboratories in primordial follicle development to elucidate the basic develop- Washington State University IACUC approved (#02568-014)

PLoS ONE | www.plosone.org 9 July 2010 | Volume 5 | Issue 7 | e11637 Primordial Follicle Bionetwork

Figure 5. Scheme of direct connections to cellular processes for the 55 candidate regulatory genes obtained by global literature analysis. Only 19 connected genes from the list of 55 are shown, the rest are not connected and not shown. Node shapes code: oval and circle – protein; crescent – protein kinase and kinase; diamond – ligand; irregular polygon – phosphatase; circle/oval on tripod platform – transcription factor; ice cream cone – receptor. Red color represents up-regulated genes, blue color – down-regulated genes, grey rectangles represent cell processes; arrows with plus sign show positive regulation/activation, arrows with minus sign – negative regulation/inhibition. doi:10.1371/journal.pone.0011637.g005 protocols and the ovaries removed and cultured whole as In order to determine the effect of CTGF on primordial to described previously [24]. Four-day old rat ovaries contain primary follicle development, ovaries were cultured as above for almost exclusively primordial follicles. For ovary culture ten days in the absence or presence of CTGF (50ng/mL), alone or experiments in which ovarian RNA was collected, 2–3 ovaries in combination with TGF-beta1 (50ng/mL). After culture ovaries per well were cultured with media changes every 24 hours for were fixed with Bouin’s solution, paraffin embedded, sectioned two days in the absence (controls) or presence (treated) of either onto microscope slides and stained with hematoxylin and eosin as AMH (human Anti-Mu¨lerian hormone)(50ng/mL, R&D Sys- described previously [24]. tems Inc., USA), FGF2 (rat Fibroblast growth factor 2)(50ng/ mL, R&D Systems Inc., USA), BMP4 (human Bone morpho- Morphometric Analysis genetic protein 4)(50ng/mL, R&D Systems Inc., USA), GDNF The number of follicles at each developmental stage was counted (rat Glial derived neurotrophic factor)(50ng/mL, Calbiochem, and averaged in two serial sections from the largest cross-section USA), FGF7 (human fibroblast growth factor 7/keratinocyte through the center of the ovary. Total follicle number has not been growth factor)(50ng/mL, R&D Systems Inc., USA), KITLG found to change between treatment groups. Rather, only the (mouse Kit ligand)(50ng/mL, R&D Systems Inc., USA), LIF (rat percentage of follicles at each developmental stage changes with leukemia inhibitory factor)(50ng/mL, Chemicon/Millipore, treatment [28,34]. KL was used as a positive control for the organ USA), PDGF-AB (rat platelet derived growth factor AB culture experiments. Follicles in ovarian cross sections were heterodimer)(50ng/mL, R&D Systems Inc., USA), TGFb1 classified as primordial (stage 0), or developing (stages 1–4: early (human transforming growth factor beta 1)(50ng/mL, R&D primary, primary, transitional and preantral) as previously de- Systems Inc., USA) or CTGF (human connective tissue growth scribed [37]. Primordial follicles consist of an oocyte arrested in factor)(50ng/mL, PeproTech Inc., NJ USA). After only two days prophase I of meiosis that is partially or completely encapsulated by of culture there are no morphological differences between flattened squamous pregranulosa cells. Early transition primary control and growth factor-treated ovaries, so measurements of follicles have initiated development (i.e., undergone primordial to whole-ovary gene expression reflect differences in RNA tran- primary follicle transition) and contain at least two cuboidal scription, rather than differing proportions of cell types due to granulosa cells. Primary and preantral follicles exhibit one or more differing cell proliferation between treatments. complete layers of cuboidal granulosa cells. Four-day old ovaries

PLoS ONE | www.plosone.org 10 July 2010 | Volume 5 | Issue 7 | e11637 Primordial Follicle Bionetwork contain predominately primordial follicles [26,38]. Hematoxylin/ previously established criteria: (1) fold change of group means eosin stained ovarian sections were analyzed at 4006magnification $1.2 or #0.83; (2) T test p-value #0.05; and (3) absolute using light microscopy. Follicles containing degenerating red eosin- difference of group means $10. The union of the differentially stained oocytes were not counted. expressed genes from the different treatments resulted in 1,540 genes being identified and used for constructing a weighted gene RNA preparation co-expression network [7,8]. Unlike traditional un-weighted gene RNA was isolated from whole rat ovaries after homogenization co-expression networks in which two genes (nodes) are either in one ml TrizolTM reagent (Sigma-Aldritch, USA), according to connected or disconnected, the weighted gene co-expression manufacturer’s instructions. Two or three ovaries from the same network analysis assigns a connection weight to each gene pair culture well (from different rat pups out of the same litter) and using soft-thresholding and thus is robust to parameter selection. receiving the same treatment were pooled and homogenized The weighted network analysis begins with a matrix of the Pearson together. On any given day a culture experiment was performed, correlations between all gene pairs, then converts the correlation b the treatment groups included untreated control ovaries and one matrix into an adjacency matrix using a power function f(x) = x . to three different growth factor treatments. Homogenized samples The parameter b of the power function is determined in such a were stored at 270 C until the time of RNA isolation. After way that the resulting adjacency matrix (i.e., the weighted co- isolation from Trizol, RNA was further purified using RNeasy expression network), is approximately scale-free. To measure how MinElute Cleanup Kits (Qiagen, USA) and stored in aqueous well a network satisfies a scale-free topology, we use the fitting solution at 270 C. index proposed by Zhang & Horvath [7] (i.e., the model fitting index R2 of the linear model that regresses log(p(k)) on log(k) where Microarray Analysis k is connectivity and p(k) is the frequency distribution of The microarray analysis was performed by the Genomics Core connectivity). The fitting index of a perfect scale-free network is Laboratory, Center for Reproductive Biology, Washington State 1. For this dataset, we select the smallest b (=7)which leads to an University, Pullman, WA using standard Affymetrix reagents and approximately scale-free network with the truncated scale-free 2 protocol. Briefly, mRNA was transcribed into cDNA with random fitting index R greater than 0.75. The distribution p(k) of the 21.29 primers, cRNA was transcribed, and single-stranded sense DNA resulting network approximates a power law: p(k),k . was synthesized which was fragmented and labeled with biotin. To explore the modular structures of the co-expression network, Biotin-labeled ssDNA was then hybridized to the Rat Gene 1.0 ST the adjacency matrix is further transformed into a topological microarrays containing 27,342 transcripts (Affymetrix, Santa overlap matrix [30]. As the topological overlap between two genes Clara, CA, USA). Hybridized chips were scanned on Affymetrix reflects not only their direct interaction, but also their indirect Scanner 3000. CEL files containing raw data were then pre- interactions through all the other genes in the network. Previous processed and analyzed with Partek Genomic Suite 6.3 software studies [7,30] have shown that topological overlap leads to more (Partek Incorporated, St. Louis, MO) using an RMA GC-content cohesive and biologically meaningful modules. To identify adjusted algorithm. Raw data pre-processing was performed in 2 modules of highly co-regulated genes, we used average linkage groups. The first group containing 38 samples CEL files were pre- hierarchical clustering to group genes based on the topological processed in Partek all together as one experiment. Comparison of overlap of their connectivity, followed by a dynamic cut-tree all array histogram graphs demonstrated the data for all 38 chips algorithm to dynamically cut clustering dendrogram branches into were similar and appropriate for further analysis. The second gene modules [39]. A total of sixteen modules were identified and group of samples for microarray analysis consisting of 3 CTGF- the module size was observed to range from 20 to 194 genes. treated and 3 corresponding control ovaries was run, pre- To distinguish between modules, each module was assigned a processed and analyzed post factum, separately from the rest of unique color identifier, with the remaining, poorly connected the samples as a result of a discovery from network analysis. Partek genes colored grey. The hierarchical clustering over the topolog- pre-processing algorithm for these 6 CEL files used the same ical overlap matrix (TOM) and the identified modules is shown. In criteria as used for the first group. this type of map, the rows and the columns represent genes in a The microarray quantitative data involves over 10 different symmetric fashion, and the color intensity represents the oligonucleotides arrayed for each gene and the hybridization must interaction strength between genes. This connectivity map be consistent to allow a statistically significant quantitative highlights that genes in the ovary transcriptional network fall into measure of gene expression and regulation. In contrast, a distinct network modules, where genes within a given module are quantitative PCR procedure only uses two oligonucleotides and more interconnected with each other (blocks along the diagonal of primer bias is a major factor in this type of analysis. Therefore, we the matrix) than with genes in other modules. There are a couple did not attempt to use PCR based approaches as we feel the of network connectivity measures, but one particularly important microarray analysis is more accurate and reproducible without one is the within module connectivity (k.in). The k.in of a gene was primer bias such as PCR based approaches. determined by taking the sum of its connection strengths (co- All microarray CEL files (MIAME compliant raw data) from expression similarity) with all other genes in the module which the this study have been deposited with the NCBI gene expression and gene belonged. hybridization array data repository (GEO, http://www.ncbi.nlm. nih.gov/geo) (GEO Accession number: GSE20324), all arrays Gene Co-expression Network Analysis Clarification combined with one accession number, and can be also accessed Gene networks provide a convenient framework for exploring through www.skinner.wsu.edu. For gene annotation, Affymetrix the context within which single genes operate. Networks are annotation file RaGene1_0stv1.na30.rn4.transcript.csv was used simply graphical models comprised of nodes and edges. For gene unless otherwise specified. co-expression networks, an edge between two genes may indicate that the corresponding expression traits are correlated in a given Network analysis population of interest. Depending on whether the interaction The network analysis was restricted to genes differentially strength of two genes is considered, there are two different expressed between the control and the treatment groups based on approaches for analyzing gene co-expression networks: 1) an

PLoS ONE | www.plosone.org 11 July 2010 | Volume 5 | Issue 7 | e11637 Primordial Follicle Bionetwork unweighted network analysis that involves setting hard thresholds Pathway Edition Software (Genomatix Software GmbH, on the significance of the interactions, and 2) a weighted approach Munchen, Federal Republic of Germany). Different node colors that avoids hard thresholds. Weighted gene co-expression represent different modules. A - the whole scheme clearly indicates networks preserve the continuous nature of gene-gene interactions 5 distinguished groups of genes (each group is shown separately on at the transcriptional level and are robust to parameter selection. pp. 2–6) connected to 5 central genes: Nfkb1 (B, page 2), Vegfa (C, An important end product from the gene co-expression network page 3), Egfr (D, page 4), and Gadd45a (F, page 6). Only analysis is a set of gene modules in which member genes are more connected genes are shown. highly correlated with each other than with genes outside a Found at: doi:10.1371/journal.pone.0011637.s001 (2.07 MB module. Most gene co-expression modules are enriched for GO PDF) functional annotations and are informative for identifying the functional components of the network that are associated with Figure S2 KEGG Pathway ‘‘Complement and Coagulation disease [6]. Cascades’’ enriched by regulated genes from 1,540 gene list. Red This gene co-expression network analysis (GCENA) has been nodes represent up-regulated genes, blue - down-regulated, green - increasingly used to identify gene sub-networks for prioritizing not affected genes. gene targets associated with a variety of common human diseases Found at: doi:10.1371/journal.pone.0011637.s002 (0.07 MB such as cancer and obesity [11,12,13,14,15]. One important end PDF) product of GCENA is the construction of gene modules comprised Figure S3 Scheme of shortest connections to cellular processes of highly interconnected genes. A number of studies have for 55 candidate regulatory genes, as obtained by global literature demonstrated that co-expression network modules are generally analysis using Pathway Studio 7.0 (Ariadne Genomics, Inc., enriched for known biological pathways, for genes that are linked Rockville, MD; trial version). Only 22 connected genes from the to common genetic loci and for genes associated with disease list out of 55 are shown, the rest from the list are not connected [6,7,11,13,14,15,40,41,42]. In this way, one can identify key and not shown. Node shapes code: oval and circle - protein; groups of genes that are perturbed by genetic loci that lead to crescent - protein kinase and kinase; diamond - ligand; irregular disease, and that define at the molecular level disease states. polygon - phosphatase; circle/oval on tripod platform - transcrip- Furthermore, these studies have also shown the importance of the tion factor; ice cream cone - receptor. Red color represents up- hub genes in the modules associated with various phenotypes. For regulated genes, blue color - down regulated genes, grey nodes example, GCENA identified ASPM, a hub gene in the cell cycle represent genes closely connected (next neighbor) to the list genes; module, as a molecular target of glioblastoma [15] and grey rectangles represent cell processes; arrows color: grey solid or MGC4504, a hub gene in the unfolded protein response module, dotted - regulation, blue - expression, green - promoter binding; as a target potentially involved in susceptibility to atherosclerosis arrows with plus sign show positive regulation/activation, arrows [13]. with minus sign - negative regulation/inhibition. Found at: doi:10.1371/journal.pone.0011637.s003 (0.35 MB Pathway Analysis PDF) Resulting lists of differentially expressed genes for each growth factor treatment as well as for each module generated in the Table S1 Rat Genes Expressed Differentially After Growth network analysis were analyzed for KEGG (Kyoto Encyclopedia Factor Treatment of Ovary. Legends: * - absolute value of for Genes and Genome, Kyoto University, Japan) pathway difference between means of Control and GF Treatment enrichment using Pathway-Express, a web-based tool freely expression values ** - abbreviations used for modules’ color: trq available as part of the Onto-Tools (http://vortex.cs.wayne.edu) -turquoise; brw - brown; blu- blue; ylw- yellow; prp - purple; gr - [43]. Global literature analysis of various gene lists was performed grey; grn - green; grlw - green-yellow; blc- black; mbl - midnight- using BiblioSphere PathwayEdition (Genomatix Software GmbH, blue; slm - salmon; lcn - light cyan; ***- k in. is connectivity Munchen, Federal Republic of Germany) software which performs coefficient determined in network analysis. pathway and interaction analysis and labels genes which belong to Found at: doi:10.1371/journal.pone.0011637.s004 (1.37 MB certain known metabolic and signal transduction pathways. A PDF) program based on literature analysis Pathway Studio (Ariadne, Genomics Inc. Rockville MD) was used to evaluate cellular Author Contributions processes connected to differentially expressed genes. Conceived and designed the experiments: EES MKS. Performed the experiments: EEN MIS RS BZ. Analyzed the data: EEN MIS RS BZ EES Supporting Information MKS. Wrote the paper: EEN MIS RS BZ EES MKS. Figure S1 Network scheme for 1540 differentially expressed genes obtained by global literature analysis using BiblioSphere

References 1. Hirshfield AN (1991) Development of follicles in the mammalian ovary. Int Rev identify core functional modules for obesity and diabetes. J Neurochem 97 Suppl Cytol 124: 43–101. 1: 50–62. 2. Rajah R, Glaser EM, Hirshfield AN (1992) The changing architecture of the 7. Zhang B, Horvath S (2005) A general framework for weighted gene co- neonatal rat ovary during histogenesis. Dev Dyn 194: 177–192. expression network analysis. Stat Appl Genet Mol Biol 4: Article17. 3. Peters H, Byskov AG, Himelstein-Braw R, Faber M (1975) Follicular growth: 8. Zhu J, Wiener MC, Zhang C, Fridman A, Minch E, et al. (2007) Increasing the the basic event in the mouse and human ovary. J Reprod Fertil 45: 559– power to detect causal associations by combining genotypic and expression data 566. in segregating populations. PLoS Comput Biol 3: e69. 4. Skinner MK (2005) Regulation of primordial follicle assembly and development. 9. Schadt EE, Lamb J, Yang X, Zhu J, Edwards S, et al. (2005) An integrative Hum Reprod Update 11: 461–471. genomics approach to infer causal associations between gene expression and 5. Pangas SA (2007) Growth factors in ovarian development. Semin Reprod Med disease. Nat Genet 37: 710–717. 25: 225–234. 10. Winrow CJ, Williams DL, Kasarskis A, Millstein J, Laposky AD, et al. (2009) 6. Lum PY, Chen Y, Zhu J, Lamb J, Melmed S, et al. (2006) Elucidating the Uncovering the genetic landscape for multiple sleep-wake traits. PLoS ONE 4: murine brain transcriptional network in a segregating mouse population to e5161.

PLoS ONE | www.plosone.org 12 July 2010 | Volume 5 | Issue 7 | e11637 Primordial Follicle Bionetwork

11. Chen Y, Zhu J, Lum PY, Yang X, Pinto S, et al. (2008) Variations in DNA 26. Parrott JA, Skinner MK (1999) Kit-ligand/stem cell factor induces primordial elucidate molecular networks that cause disease. Nature 452: 429–435. follicle development and initiates folliculogenesis. Endocrinology 140: 12. Emilsson V, Thorleifsson G, Zhang B, Leonardson AS, Zink F, et al. (2008) 4262–4271. Genetics of gene expression and its effect on disease. Nature 452: 423–428. 27. Hutt KJ, McLaughlin EA, Holland MK (2006) KIT/KIT ligand in mammalian 13. Gargalovic PS, Imura M, Zhang B, Gharavi NM, Clark MJ, et al. (2006) oogenesis and folliculogenesis: roles in rabbit and murine ovarian follicle Identification of inflammatory gene modules based on variations of human activation and oocyte growth. Biol Reprod 75: 421–433. endothelial cell responses to oxidized lipids. Proc Natl Acad Sci U S A 103: 28. Nilsson EE, Kezele P, Skinner MK (2002) Leukemia inhibitory factor (LIF) 12741–12746. promotes the primordial to primary follicle transition in rat ovaries. Mol Cell 14. Ghazalpour A, Doss S, Zhang B, Wang S, Plaisier C, et al. (2006) Integrating Endocrinol 188: 65–73. genetic and network analysis to characterize genes related to mouse weight. 29. Nilsson EE, Detzel C, Skinner MK (2006) Platelet-derived growth factor PLoS Genet 2: e130. modulates the primordial to primary follicle transition. Reproduction 131: 15. Horvath S, Zhang B, Carlson M, Lu KV, Zhu S, et al. (2006) Analysis of 1007–1015. oncogenic signaling networks in glioblastoma identifies ASPM as a molecular 30. Ravasz E, Somera AL, Mongru DA, Oltvai ZN, Barabasi AL (2002) target. Proc Natl Acad Sci U S A 103: 17402–17407. Hierarchical organization of modularity in metabolic networks. Science 297: 16. Leask A, Abraham DJ (2003) The role of connective tissue growth factor, a 1551–1555. multifunctional matricellular protein, in fibroblast biology. Biochem Cell Biol 31. Langfelder P, Zhang B, Horvath S (2008) Defining clusters from a hierarchical 81: 355–363. cluster tree: the Dynamic Tree Cut package for R. Bioinformatics 24: 719–720. 17. Durlinger AL, Kramer P, Karels B, de Jong FH, Uilenbroek JT, et al. (1999) 32. Cicha I, Goppelt-Struebe M (2009) Connective tissue growth factor: context- Control of primordial follicle recruitment by anti-Mullerian hormone in the dependent functions and mechanisms of regulation. Biofactors 35: 200–208. 33. Gressner OA, Gressner AM (2008) Connective tissue growth factor: a fibrogenic mouse ovary. Endocrinology 140: 5789–5796. master switch in fibrotic liver diseases. Liver Int 28: 1065–1079. 18. Nilsson E, Rogers N, Skinner MK (2007) Actions of anti-Mullerian hormone on 34. Nilsson E, Parrott JA, Skinner MK (2001) Basic fibroblast growth factor induces the ovarian transcriptome to inhibit primordial to primary follicle transition. primordial follicle development and initiates folliculogenesis. Mol Cell Reproduction 134: 209–221. Endocrinol 175: 123–130. 19. Nilsson EE, Skinner MK (2004) Kit ligand and basic fibroblast growth factor 35. Ghazalpour A, Doss S, Sheth SS, Ingram-Drake LA, Schadt EE, et al. (2005) interactions in the induction of ovarian primordial to primary follicle transition. Genomic analysis of metabolic pathway gene expression in mice. Genome Biol Mol Cell Endocrinol 214: 19–25. 6: R59. 20. Garor R, Abir R, Erman A, Felz C, Nitke S, et al. (2009) Effects of basic 36. Harlow CR, Hillier SG (2002) Connective tissue growth factor in the ovarian fibroblast growth factor on in vitro development of human ovarian primordial paracrine system. Mol Cell Endocrinol 187: 23–27. follicles. Fertil Steril 91: 1967–1975. 37. Oktay K, Schenken RS, Nelson JF (1995) Proliferating cell nuclear antigen 21. Matos MH, Lima-Verde IB, Bruno JB, Lopes CA, Martins FS, et al. (2007) marks the initiation of follicular growth in the rat. Biol Reprod 53: 295–301. Follicle stimulating hormone and fibroblast growth factor-2 interact and 38. Kezele PR, Ague JM, Nilsson E, Skinner MK (2005) Alterations in the ovarian promote goat primordial follicle development in vitro. Reprod Fertil Dev 19: transcriptome during primordial follicle assembly and development. Biol Reprod 677–684. 72: 241–255. 22. Nilsson EE, Skinner MK (2003) Bone morphogenetic protein-4 acts as an 39. Langfelder P, Horvath S (2007) Eigengene networks for studying the ovarian follicle survival factor and promotes primordial follicle development. relationships between co-expression modules. BMC Syst Biol 1: 54. Biol Reprod 69: 1265–1272. 40. Schadt EE, Molony C, Chudin E, Hao K, Yang X, et al. (2008) Mapping the 23. Tanwar PS, O’Shea T, McFarlane JR (2008) In vivo evidence of role of bone genetic architecture of gene expression in human liver. PLoS Biol 6: e107. morphogenetic protein-4 in the mouse ovary. Anim Reprod Sci 106: 232–240. 41. Zhu J, Zhang B, Schadt EE (2008) A systems biology approach to drug 24. Dole G, Nilsson EE, Skinner MK (2008) Glial-derived neurotrophic factor discovery. Adv Genet 60: 603–635. promotes ovarian primordial follicle development and cell-cell interactions 42. Zhu J, Zhang B, Smith EN, Drees B, Brem RB, et al. (2008) Integrating large- during folliculogenesis. Reproduction 135: 671–682. scale functional genomic data to dissect the complexity of yeast regulatory 25. Kezele P, Nilsson EE, Skinner MK (2005) Keratinocyte growth factor acts as a networks. Nat Genet 40: 854–861. mesenchymal factor that promotes ovarian primordial to primary follicle 43. Draghici S, Khatri P, Tarca AL, Amin K, Done A, et al. (2007) A systems transition. Biol Reprod 73: 967–973. biology approach for pathway level analysis. Genome Res 17: 1537–1545.

PLoS ONE | www.plosone.org 13 July 2010 | Volume 5 | Issue 7 | e11637