Identification of LMX1B As a Novel Oncogene in Human Ovarian Cancer

ORIGINAL ARTICLE Identiﬁcation of LMX1B as a novel oncogene in human ovarian cancer

LHe1,2, L Guo3, V Vathipadiekal2,4, PA Sergent3, WB Growdon2,3, DA Engler5, BR Rueda2,3, MJ Birrer2,4,7, S Orsulic6,7 and G Mohapatra1,2,7

Ovarian cancers are thought to result from the accumulation of multiple genetic aberrations that transform ovarian and/or fallopian tube surface epithelial cells, allowing for their abnormal growth, proliferation and metastasis. In the report presented here, we carried out genome-wide copy-number analysis using comparative genomic hybridization on a panel of mouse ovarian cancer (OVCA) cell lines previously established in our laboratory. We identified a recurrent focal amplification on mouse chromosomal region 2qB, which contains the LIM-homeodomain-containing transcription factor 1B (Lmx1b) gene. LMX1B is not expressed in normal human ovary, but is expressed in many human OVCA cell lines and primary tumors. High expression of LMX1B correlates with poor outcome. To clarify the role of LMX1B in ovarian carcinogenesis, we transduced LMX1B into a panel of mouse and human OVCA cell lines and demonstrated that LMX1B strongly promotes migration of cancer cells in culture and promotes xenograft growth in nude mice. Conversely, knockdown of LMX1B in a human cell line with endogenous high expression of LMX1B inhibits cell migration in vitro and tumor growth in vivo. Microarray analysis of cells overexpressing LMX1B identified the nuclear factor (NF)-kB pathway as a potential mediator of tumor progression and subsequent treatment of NFkB inhibitor decreased the migratory capacity of these cells. Thus, our data demonstrate that LMX1B is a novel oncogene in OVCA pathogenesis.

Oncogene (2014) 33, 4226–4235; doi:10.1038/onc.2013.375; published online 23 September 2013 Keywords: ovarian cancer; aCGH; LMX1B; tumor progression; transcription factor

INTRODUCTION several megabases of DNA sequences.1 The high degree of Human ovarian cancers display a multitude of genetic abnorm- genomic complexity of human ovarian tumors highlights the need alities including deletions, amplifications and structural rearrange- for biological systems capable of distinguishing ‘driver’ cancer ments as exemplified by the results of The Cancer Genome Atlas genes that are responsible for the neoplastic transformation (TCGA) project.1 Discovery of these aberrations and understanding process from ‘passenger’ genes.10 how they contribute to the pathophysiology of ovarian cancer are Molecular mechanisms that control important biological pro- necessary for improvements in diagnosis and therapy.2,3 Genomic cesses are often preserved during evolution. Thus, cross-species aberrations arise recurrently on some chromosomal locations in genomic analysis can be used as a biological filter to identify cancer cells, indicating that these locations may contain genes relevant genomic alterations that lead to cancer development. The that are important for cancer progression. The development of utility of comparative oncogenomic approaches for efficient array-based comparative genomic hybridization (aCGH) has identification of driver cancer genes was recently demonstrated provided genome-wide high-resolution detection of alterations by the identification of several new oncogenes in regions that are of DNA copy number.4 This approach has enabled the commonly amplified in mouse and human melanomas11 and identification of a large number of genomic loci that undergo hepatocellular carcinomas.12 It is expected that cross-species amplifications and deletions during the process of tumorigenesis comparison will greatly facilitate identification of other relevant in various cancer types, including human ovarian carcinomas.5–7 genes that are involved in human cancer progression. Several genes, including MYC, AKT2, PIK3CA and ERBB2, which are Mouse ovarian epithelial tumor cell lines containing various known to contribute to the pathophysiology of ovarian cancer combinations of genetic alterations in the p53, c-myc, K-ras, Akt (OVCA) are located in regions of recurrent DNA copy number and BRCA1 genes have been shown to produce ovarian tumors aberrations, suggesting that other regions of recurrent in vivo resembling the serous histology.13–15 These cell lines are amplifications and deletions contain putative oncogenes and ideal for dissecting pathways for targeted therapy as well as useful tumor suppressor genes, respectively. Indeed, recent efforts to for identifying novel genes that are involved in cancer characterize some of these alterations have lead to identification progression. We hypothesized that besides the retrovirally of novel cancer genes, such as RAB258 and MDS1/EVI1,9 that are transduced initiating genetic events that were used to generate associated with OVCA. However, copy number alterations this model, additional genetic events presumably occurred during obtained by aCGH analysis of human primary tumors are often tumor development in nude mice. In line with this hypothesis, the large, involving complex structural rearrangements and span secondary cell lines derived from the xenograft tumors exhibited

1Department of Pathology, Massachusetts General Hospital, Boston, MA, USA; 2Harvard Medical School, Boston, MA, USA; 3Vincent Center for Reproductive Biology, Department of Vincent Obstetrics and Gynecology, Massachusetts General Hospital, Boston, MA, USA; 4Division of Hematology-Oncology and Center for Cancer Research, Massachusetts General Hospital, Boston, MA, USA; 5Department of Statistics, Brigham Young University, Provo, UT, USA and 6Women’s Cancer Research Institute, Cedars-Sinai Medical Center, Los Angeles, CA, USA. Correspondence: Dr G Mohapatra, Department of Pathology, Massachusetts General Hospital, 70 Blossom Street, Boston, MA 02114, USA. E-mail: [email protected] 7These authors contributed equally to this work. Received 11 September 2012; revised 15 July 2013; accepted 19 July 2013; published online 23 September 2013 LMX1B as an oncogene in ovarian cancer LHeet al 4227 an enhanced tumorigenic potential both in vitro and in vivo.14,15 amplification of this gene in cancer cells is likely to be of We therefore reasoned that high-resolution aCGH analysis of these functional significance. (3) Although the possible involvement of secondary tumor cell lines would be ideal for identifying novel LMX1B in cancer has never been reported, it is becoming genetic events associated with OVCA progression and metastasis. increasingly appreciated that many developmentally important In the report presented here, we carried out a high-resolution transcription factors, such as FOXC2, HOXB13 and Twist exhibit genome-wide copy-number analysis using aCGH on a panel of strong protumorigenic or prometastatic function in breast and mouse OVCA cell lines with known initiating genetic events.14,15 ovarian cancers.17,24,25 We thus hypothesized that similar As a result of this analysis, we identified a novel recurrent focal paradigm may also hold true for LMX1B. amplification on mouse chromosome 2B containing a single gene, aCGH results were validated using two-color fluorescence in situ the LIM domain containing transcription factor 1B (LMX1B); the hybridization. BAC clone RP23–65G6 containing the entire mouse syntenic region of human chromosome 9q33.3. Transcription Lmx1b gene and BAC clone RP23–60M5 mapping to 2qH4 used as factors have been known to contribute to tumorigenesis owing to control probe confirmed amplification of Lmx1b gene (Figure 1c). their role in regulation of gene expression.16 In ovarian cancer, Previous studies have shown that Lmx1b is highly expressed transcription factors that are commonly mutated and amplified during mouse embryogenesis in the limb buds, isthmus organizer include p53, RB1 and MYC.1 Developmental transcription factors and kidney.26,27 In adult mice however, its expression is low in that are involved in ovarian cancers are HOXB13,17 SHH18 and the almost all the tissues (including ovaries) except for podocytes in NOTCH superfamily19 among others. In the current study, we focus the kidney. We confirmed this by quantitative RT–PCR using RNA on LMX1B, which is one of the essential transcription factors from multiple adult mouse tissue types (Figure 1d). At the DNA involved in the development of midbrain.20,21 level, there is a high degree of homology between human and We found that LMX1B was not expressed in normal ovarian mouse LMX1B genes; and at the protein level, human and mouse surface epithelium, but showed moderate to high level of LMX1B proteins are almost identical (1 amino acid difference out expression in several human OVCA cell lines and primary tumors. of 300 amino acids). To test the role of LMX1B in ovarian carcinogenesis, we ectopically expressed LMX1B and found that LMX1B overexpression strongly Analysis of LMX1B expression in human OVCA cell lines and induces migration of cancer cells in vitro and accelerates tumor primary tumors growth in vivo. Conversely, knockdown of LMX1B in human OVCA To determine whether LMX1B is expressed in human OVCA cell line with endogenous high expression of LMX1B inhibits cell lines and primary tumors, we analyzed by qRT–PCR a panel of 15 migration in vitro and tumor growth in vivo. Thus, our results OVCA cell lines and 45 primary tumor samples with clinical follow demonstrate that LMX1B is a novel oncogene which has up information (Massachusetts General Hospital cohort). Com- important roles in ovarian cancer progression. pared with T29, an immortalized human ovarian surface epithelial cell line, more than twofold expression of LMX1B was found in 7/ 15 cell lines (Figure 2a) and relative to the tumor with the lowest RESULTS expression, more than 15-fold expression of LMX1B was detected aCGH analysis of mouse ovarian cancer cell lines in 29/45 primary tumors (Figure 2b). Notably, four primary tumors Ten genetically defined mouse OVCA lines T1, T11, T111, T2, T22, exhibited more than 100–800-fold increase in LMX1B expression. T222, T3, TBR2, TBR5 and TBR6 (Supplementary Table 1) were used This observation was further validated in an independent for copy number alteration analysis. The status of the Brca1 and microarray dataset generated using 53 micro-dissected OVCA the p53 gene in these genetically defined cancer cell lines tissues and 9 ovarian surface epithelial cells (P ¼ 0.02) provided excellent internal controls for aCGH data validation. (Figure 2c).28 Of the 53 samples, 5 also exhibited more than Mouse OVCA cell lines TBR2, TBR5 and TBR6 were derived from 100-fold increase in LMX1B expression. We then sought to identify mouse cell lines with a conditional deletion of the Brca1 exon if there was any correlation between LMX1B expression level and 11.15 Consistent with this, aCGH probe that mapped to the Brca1 clinical outcome. Primary tumor samples from the Massachusetts exon 11 detected decreased copy number (Supplementary General Hospital cohort were divided into low- and high Figure 1A) in OVCA cell lines. The OVCA cell lines T1, T11, T111, expressing subgroups using the median expression fold change T2, T22, T222 and T3 were derived from mice with a conventional (Figure 2d). We found that patients with higher LMX1B expression knockout of the p53 exons 2–5,22 whereas cancer cell lines TBR2, had a significantly poor overall survival compared with the low TBR5 and TBR6 were derived from mice with a conditional expressing group (Po0.05) (Figure 2e). These results indicate that deletion of the p53 exons 2–1023 (Supplementary Figure 1B). The LMX1B may be an important player in tumor progression in a p53 probe in the array falls outside the deletion in the subset of OVCA patients. conventional p53 knockout mouse; hence, the deletion of p53 was detected in the conditional p53 knockout only. Ectopic expression of LMX1B promotes tumorigenesis After successful validation of the aCGH data, we analyzed individual chromosomal aberrations in each sample. aCGH analysis Based on LMX1B expression analysis in primary tumors, we identified large regions of chromosomal losses and gains. It also reasoned that ectopic expression of LMX1B gene in OVCA cells will readily detected focal amplifications and deletions, including render them more malignant. To test this hypothesis, the coding homozygous deletions (Supplementary Figures 2A and B). sequence of human LMX1B was cloned into a retroviral expression vector. Virus particles containing LMX1B were generated to infect various cancer cell lines, including mouse OVCA cell lines C1, C2, Amplification of Lmx1b gene in mouse OVCA cell lines BR5 and human OVCA cell lines OVCAR5 and OV90. In addition, We observed a focal amplification in Lmx1b locus in two cell lines non-tumorigenic NIH3T3 mouse fibroblast cells were included to (log2 ratio 42.5) and copy number increase in three cell lines directly test the transforming ability of LMX1B. The ectopic (log2 ratio 41.5) (Figures 1a and b). In human serous ovarian expression of LMX1B after puromycin selection was confirmed by tumors, LMX1B is also amplified, although less frequently RT–PCR (data not shown). (Figure 1a). We decided to further study this gene for the To test whether LMX1B affects tumor cell biology, we following reasons: (1) this was a focal and recurrent amplification performed proliferation, soft agar colony formation and cell locus containing a single gene. (2) LMX1B is expressed during migration assay on the mouse and human cell lines expressing embryonic development but its expression is turned off in most LMX1B along with their corresponding GFP-expressing cells as adult tissues, except for podocytes in the kidney (Figure 1d). Thus, control. No obvious difference was observed in cell proliferation

& 2014 Macmillan Publishers Limited Oncogene (2014) 4226 – 4235 LMX1B as an oncogene in ovarian cancer LHeet al 4228

Figure 1. Identification of LMX1B gene amplification in ovarian cancer by high-resolution aCGH analysis. (a) Upper panel: aCGH profile of mouse chromosome 2 for the T2 cell line indicates the focal amplification at 2qB region, containing a single gene Lmx1b (LIM domain- containing transcription factor 1B); lower panel: corresponding region along 9q33.3 in human is amplified in a small subset of tumors and includes a large region containing many genes. aCGH profile of human chromosome 9 shown for a primary high-grade serous ovarian cancer; (b) table listing LMX1B log2 ratio for mouse OvCa lines screened by aCGH; (c) Fluorescence in situ hybridization using BAC clones RP23–65G6 (Lmx1b, in red) and RP23–60M5 (control, in green) confirms amplification of Lmx1b gene in mouse ovarian cancer cell lines; (d) qRT–PCR analysis of LMX1B in adult mouse tissues.

rate and soft agar colony forming ability between LMX1B- pathology of the control GFP-expressing tumors versus LMX1B- expressing and GFP-expressing control cells (data not shown). expressing tumors were similar, resembling human serous However, LMX1B-expressing cells had a higher migratory ability type epithelial cancer (except for NIH3T3 cells) (Supplementary (Figure 3a). To test the hypothesis that expression of LMX1B in vivo Figures 3A–D). Based on these results, we conclude that LMX1B may lead to a higher rate of tumor growth, we injected cells acts as an oncogene and that it promotes tumor progression. expressing either LMX1B or the GFP control into ﬂanks of nude mice or NOD/SCID mice. In all cell lines tested, including NIH3T3 cells, LMX1B overexpression accelerated tumor growth compared LMX1B knockdown abrogates tumor growth in vivo with GFP-expressing control cells within 2–8 weeks post injection As overexpression of LMX1B accelerated tumor growth in vivo,we (Figures 3b–e), regardless of the species and genetic make-up of reasoned that knocking down LMX1B in OVCA cells with high the cell lines. This observation was reproduced by intraperitoneal endogenous LMX1B expression would negatively affect their injections of OVCAR5 cells expressing GFP and LMX1B (Figure 3f). tumorigenic potential. To test this hypothesis, we knocked down Notably, in BR5 and NIH3T3 cells (Figure 3c), no viable tumor could endogenous LMX1B using lentiviral-mediated shRNA constructs in be detected except in one mouse in the GFP-expressing control TOV112D OVCA cells as these cells were among the high lines within the window of time of our experiments (up to 10 expressor lines determined by qRT–PCR analysis (Figure 2a). Two weeks). Further histological analysis showed that the overall shRNA constructs, sh-LMX1B-B10 and sh-LMX1B-C01 produced

Oncogene (2014) 4226 – 4235 & 2014 Macmillan Publishers Limited LMX1B as an oncogene in ovarian cancer LHeet al 4229

Figure 2. qRT–PCR analysis of LMX1B mRNA in human serous ovarian cancers. (a) qRT–PCR analysis of LMX1B in human OVCA lines; (b) qRT– PCR analysis of LMX1B in primary high-grade serous ovarian tumors; LMX1B expression was normalized to GAPDH expression. The x axis denotes sample ID and y axis denotes relative expression level of LMX1B; the sample with lowest expression was set to 1; (c) validation of LMX1B expression in an independent microarray dataset compared with normal ovarian surface epithelial cells shows significant difference in expression of LMX1B in tumor cells (P ¼ 0.02) (data used from Mok et al.28); (d) data from Figure 2b, used for box plot, indicates low and high expressors among the primary ovarian cancers with a significant difference (Po0.00001) using median fold-change as cutoff; (e) Kaplan–Meier survival analysis showing high LMX1B expressors correlate with poor overall survival (Po0.05); x axis: survival time. efficient knockdown, as confirmed by qRT–PCR (Figure 4a). In nuclear factor (NF)-kB pathway and identified the NFkB pathway order to confirm LMX1B knockdown at the protein level, LMX1B as a potential mediator of tumor progression. To confirm the complementary DNA was cloned into a pCMV6-Entry (C-terminal accuracy of microarray results, we validated 16 genes by qRT– Myc and FLAG-tagged) vector and co-transfected with the hairpins PCR (Table 2, Figure 5b and Supplementary Figure 4A). Of the 16 that produced efficient knockdown. Western blotting using a genes, 13 were validated in OVCAR5 (81%) and 15 were validated monoclonal anti-FLAG antibody showed that the exogenous in OV90 cells (94%) indicating a significant correlation between tagged LMX1B protein was much lower in cells co-transfected microarray and qRT–PCR data. Of these 16 genes, 11 belong to with either of the two hairpins (Figure 4b). Inhibition of LMX1B the NFkB pathway and are known to have altered expression in expression in TOV112D cells using either of the two shRNA high-grade serous tumors.29 The NFkB activity is known to constructs resulted in lower cell migration capacity (Figure 4c), increase the transcription of cytokines such as interleukin (IL)-6 although no significant difference was observed in proliferation and IL8. In OVCAR5-LMX1B cells, both IL6 and IL8 were expressed and colony forming ability (not shown). Furthermore, when more than 75-fold and 35-fold, respectively, compared with planted into the flanks of NOD/SCID mice, tumor growth was OVCAR5-GFP cells (Figure 5b). In OV90-LMX1B cells, IL8 was almost completely abrogated for TOV112D cells transduced with expressed 14-fold more compared with OV90-GFP cells either of the two shRNA constructs, compared with the control (Figure 5b). In order to functionally validate the involvement of shRNA against GFP (Figures 4d–e). Thus, suppression of LMX1B in NFkB activity in the production of IL6 and IL8, we used a NFkB TOV112D cells inhibits tumor growth in vivo. We conclude that inhibitor on OVCAR5-LMX1B and OV90-LMX1B cells. Cells were endogenous LMX1B has an essential role in tumor progression. treated with 20 nM QNZ, an NFkB transactivation inhibitor for 24 h. The RNA was isolated and qRT–PCR was performed. As shown in Figure 5c, QNZ treatment reversed LMX1B-induced Global gene expression analysis of cells overexpressing LMX1B upregulation of the NFkB pathway genes IL6, IL8, NFKB1, PTGS2 identifies the NFkB pathway as a potential mediator of tumor and TNFAIP6 but failed to alter the expression of the non-NFkB progression pathway genes DTL, ERBB3 and FOXC1, suggesting that the Expression profiling of OVCAR5 and OV90 cells overexpressing production of cytokines is dependent on NFkB transcriptional LMX1B compared with the GFP control cells identified 863 activation. In addition, OVCAR5-LMX1B and OV90-LMX1B cells differentially expressed genes (P ¼ 0.001) (Figure 5a, Supplementary treated with QNZ showed a significant reduction in their Table 2). In order to identify molecular and biological processes migratory ability (Figure 5d). At this concentration and time mediated by LMX1B, gene set enrichment analysis of the 863 scale, no alteration in cell viability was observed (data not differentially expressed genes was performed using PathwayS- shown). Finally, western blotting analysis of the canonical NFkB tudio software version 7.0. (http://www.ariadnegenomics.com/ components p65 and p50 showed a significant reduction of both company/news/all-news/article/ariadne-launches-pathway-studioR- proteins in the nuclear fraction in the LMX1B-expressing cells version-7/). Gene set enrichment analysis linked these genes into (Figure 5e, Supplementary Figure 4C). Taken together, these 28 different biological processes listed in Table 1. Among these results demonstrate that overexpression of LMX1B induces gene sets, response to cytokine stimulus included 36 genes and transcriptional activation of the NFkB pathway, consequently had the highest median fold-change of 15.25. Further examina- increases the migratory ability of OVCA cells and promotes tumor tion revealed majority of the genes in this set belong to the progression.

& 2014 Macmillan Publishers Limited Oncogene (2014) 4226 – 4235 LMX1B as an oncogene in ovarian cancer LHeet al 4230

Figure 3. Ectopic expression of LMX1B promotes tumor progression. (a) LMX1B enhances the migration of OVCAR5 cells. GFP- and LMX1B- expressing cells were plated in the presence of mitomycin C to inhibit proliferation and Boyden chamber-based migration assay was performed to assess the migratory ability of cells. Histogram shows that LMX1B cells had higher migratory ability compared with GFP- expressing cells (P ¼ 0.016), error bars represent mean±s.d. from triplicate; (b) tumors from nude mice 15 days after C1 cells (5 Â 105) transduced with GFP or LMX1B were subcutaneously injected into the left and right flanks of nude mice, respectively. Table shows tumor weight; (c) mice injected with C1, C2, BR5 and NIH3T3 cells were euthanized 14–42 days post injection and tumors were weighed. One-tailed paired t-test was performed to calculate the P-values; (d) representative tumors after subcutaneous injection of human ovarian OVCAR5 and OV90 cells (5 Â 105) expressing GFP or LMX1B into NOD/SCID mice; (e) tumors were harvested 4 weeks post injection and weighed. LMX1B- expressing cells produced significantly larger tumors compared with GFP-expressing cells; (f) representative tumors after intraperitoneal injection of OVCAR5 cells expressing GFP or LMX1B and histogram showing average tumor weight (n ¼ 4) was significantly higher for LMX1B- expressing cells compared with GFP control cells (P ¼ 0.0013).

DISCUSSION dominant condition characterized by hypoplastic nails, hypoplas- The complexity of genomic aberrations in human cancers tic or absent patellae, joint abnormalities with contractures, iliac hampers identification of genes that drive tumorigenesis. Using horns (70–90% of patients), nephropathy (40% of patients) and comparative oncogenomic analysis, we have identified LMX1B as glaucoma (10% of patients).33 Mutations in LMX1B concentrate a novel oncogene and have provided functional and mechanistic within the exons encoding the LIM domains (LIM1 44%, LIM2 38%) evidence that supports the role of LMX1B in OVCA progression. In and the HD (18%).34 Functional characterization of the LMX1B agreement with our in vitro and in vivo data, expression analysis in mutants has been reported to have diminished transcriptional primary serous human ovarian cancers revealed that higher levels activity and loss of DNA-binding ability.26,35 However, in cancer, it of LMX1B expression correlate with poor outcome. is likely that enhanced expression of LMX1B is achieved via LMX1B belongs to the LIM-homeodomain (LIM-HD)-containing alternative mechanisms including epigenetic alterations and is family of transcription factors. To date, 12 LIM-HD transcription sufficient to promote cancer progression. factors have been characterized in mammals (LHX1–9, ISL1–2, Activation of the NFkB pathway has been known to have a LMX1A and LMX1B). They have key regulatory functions in cell-type critical role in OVCA.29,36 In our study, gene set enrichment specification during development and many of them are associated analysis of differentially expressed genes in LMX1B-overexpressing with human diseases.30,31 LIM-HD proteins feature two LIM domains cells identified response to cytokine stimulus gene set with the in their amino termini and a central HD. The HD is a highly highest median fold-change. This gene set included 36 genes conserved 60-amino-acid domain that mediates the binding of including IL6, IL8, ILF3, IL11, IL24, NFKBIA, NFKB1, NFKBIE, INSIG1, specific DNA elements within target genes.32 The LIM domain is a SOD2, STIL, PTGS1, PTGS2 and BCL6, which are associated with the conserved cystine- and histidine-rich domain of 50–60 amino acids, NFkB pathway (Supplementary Table 2).37,38 To further evaluate consisting of two tandemly repeated zinc fingers. The LIM domain is the LMX1B-induced NFkB activity in the production of a multifunctional protein–protein interaction domain that mediates proinflammatory cytokines, including IL6 and IL8, we blocked interactions with other transcription factors as well as transcriptional the NFkB activity in OVCAR5 and OV90 cells using QNZ (an NFkB cofactors that do not bind directly to DNA.30 transactivation inhibitor). QNZ treatment resulted in significantly In humans, LMX1B heterozygous mutations are associated with reduced levels of IL6 and IL8 after 24 h compared with control nail–patella syndrome. Nail–patella syndrome is an autosomal cells (Figure 5c, Supplementary Figure 4C) as well as decreased

Oncogene (2014) 4226 – 4235 & 2014 Macmillan Publishers Limited LMX1B as an oncogene in ovarian cancer LHeet al 4231

Figure 4. LMX1B knockdown leads to reduced cancer cell migration in vitro and abolishes tumor growth in vivo.(a) Efficiency of LMX1B knockdown level in TOV112D ovarian cancer cells by shRNAs. shRNA constructs against GFP and LMX1B were introduced into TOV112D cells by lentiviral infection and the relative expression of LMX1B mRNA was measured by qRT–PCR. Error bars represent mean±s.d. of triplicate samples; (b) cotransfection of 293T cells with Flag-tagged LMX1B plasmid along with pLKO, shLMX1B-1 (B10) or shLMX1B-2 (C01), respectively. Western blot analysis using anti-FLAG antibody (Sigma) confirms the qRT–PCR results and shows significant knockdown of LMX1B at the protein level; (c) LMX1B knockdown results in two- to three-fold reduction in migration of TOV112D cells (P ¼ 0.0003 and 0.0023, respectively). Error bars represent mean±s.d. of quadruplicate experiments; (d) representative image of TOV112D-GFP and TOV112D-shLMX1B (B10) tumors. TOV112D-GFP cells produced large tumors in 4 weeks compared with LMX1B knockdown cells; (e) TOV112D-GFP, TOV112D-shLMX1B- 1 (B10) and TOV112D-shLMX1B-2 (C01) cells (5 Â 105) were injected into opposite flanks of NOD/SCID mice. Tumors were harvested 4 weeks post injection and weighed. In all mice, TOV112D-GFP cells produced considerably larger tumors compared with TOV112D cells with LMX1B knockdown (P ¼ 0.0005 and 0.003, respectively).

their migratory ability (Figure 5d). Inhibition of the NFkB activity manner. Thus explains the discrepancy between in vitro and resulted in the reduction in NFkB components p65 and p50 in the in vivo results. nucleus (Figure 5e). These results demonstrate that LMX1B In addition, we identified many other gene sets that belong to mediates its oncogenic effect partly by inducing the NFkB different pathways such as cell motion (CFL1, ACTR2 and ABI2), pathway and accelerating tumor progression. Further work is G-protein-coupled receptors (RSG4 and ARRB1), peptidase activity necessary to show whether LMX1B mediates its effect through (MMP3, CD44, SPARC and TIMP3), PPAR signaling pathway (CD36), promoter binding or through cofactors. cell–cell signaling (STC1)44 and brain development (BMP4, DLX2, Cytokines are critical components of inflammation participating BPTF and FOXC1).45 Further work is necessary to evaluate the in the interaction between the tumor cells and the micro- potential role of these pathways in LMX1B-mediated tumor environment. The mechanisms of inflammation-mediated tumor progression. promotion involve secretion of specific cytokines by both To date, B400 cancer genes, corresponding to 2% of protein- inflammatory and tumor cells as well as activation of transcription coding genes in the human genome have been identified factors NFkB and STAT3. These transcription factors activate primarily through cytogenetic analyses (fusion genes), systematic genes responsible for cell survival, proliferation, angiogenesis, sequencing of exons (oncogenes, protein complexes involved in invasiveness and production of cytokines.39–41 NFkB belongs chromatin restructuring and maintenance of cytosine methyla- to a family of transcription factors that regulate the secretion of tion) and high-throughput DNA copy number analysis (amplifica- many inflammatory cytokines and the prostaglandin biosynthetic tions and homozygous deletions).46 The process of identifying pathway. It also regulates the expression of antiapoptotic proteins ‘driver genes’ is complicated partly due to the complex structural and angiogenic factors.42,43 In our study, LMX1B-expressing cells rearrangements in ovarian cancers and the involvement of large did not show enhanced proliferation or anchorage-independent genomic segments spanning megabases of DNA sequences growth compared with GFP-expressing cells in vitro; however, and multiple genes. We believe that comparative oncogenomic there was a significant enhancement in tumor growth in vivo.We approach is useful as a cancer-gene discovery tool for believe, that in vivo, there was a considerable interaction between identification of genes that are altered at the DNA copy number the tumor and the stroma and the proinflammatory cytokines IL6, level. Using this approach, we have for the first time demonstrated IL8 and IL11 along with PTGS1, PTGS2 and BCL6 create a that LMX1B is a novel oncogene and that it has key roles in OVCA microenvironment ideal for tumor growth in an accelerated progression.

& 2014 Macmillan Publishers Limited Oncogene (2014) 4226 – 4235 LMX1B as an oncogene in ovarian cancer LHeet al 4232

Figure 5. Identification of NFkB as the key downstream pathway mediated by LMX1B. (a) Heatmap showing 863 differentially expressed genes in LMX1B-overexpressing cells versus control GFP-expressing cells. These putative LMX1B responsive genes were subjected to gene set enrichment analysis for identification of cellular/biological pathways (Table 1); (b) qRT–PCR results for selected NFkB and non-NFkB genes in LMX1B-overexpressing cells compared with GFP control; (c) qRT–PCR was performed on the same genes as shown in Figure 5b using RNA from OVCAR5 and OV90 cells with ectopic expression of LMX1B or GFP treated with NFkB inhibitor QNZ for 24 h. Expression of the NFkB pathway genes was severely affected, whereas expression of non-NFkB genes DTL, ERBB3 and FOXC1 was not affected confirming that LMX1B-induced NFkB leads to the production of inflammatory cytokines; (d) Boyden chamber migration assay of cells after 24 h of 20 nM QNZ treatment showed up to 50% reduction in their migratory ability, confirming that LMX1B-induced NFkB activation accelerates tumor cell invasion (*P ¼ 0.003), error bars represent mean±s.d. from triplicate experiments; (e) western blot analysis showing reduction of nuclear NFkB components p65 and p50 after blocking NFkB activity with QNZ. Cross-examination of compartment protein extraction as well as loading is indicated by GAPDH (cytoplasmic marker) and Histone H3 (nuclear marker).

Oncogene (2014) 4226 – 4235 & 2014 Macmillan Publishers Limited LMX1B as an oncogene in ovarian cancer LHeet al 4233 Table 1. Gene set enrichment analysis of differentially expressed genes

Gene set name Number of genes Median fold-change P-value Gene set category

Response to cytokine stimulus 36 15.25 0.0076419 Biological_process Cell motion 6 5.44 0.0290921 Biological_process Brain development 6 4.80 0.0143631 Biological_process Protein dimerization activity 8 4.37 0.0021692 Molecular_function Regulation of cell cycle 58 3.76 0.0324208 Biological_process Cell cycle arrest 5 3.72 0.0100680 Biological_process Neuron migration 5 3.55 0.0076542 Biological_process Response to nutrient 6 3.51 0.0312133 Biological_process Protein heterodimerization activity 13 3.35 0.0273994 molecular_function Cell–cell signaling 11 3.31 0.0028182 Biological_process Positive regulation of cell proliferation 13 3.31 0.0370128 Biological_process Endocytosis 11 3.28 0.0429402 Biological_process Extracellular space 52 3.22 0.0002939 Cellular_component Meiosis 6 3.21 0.0266751 Biological_process Fatty acid biosynthetic process 6 3.16 0.0255559 Biological_process Inﬂammatory response 19 3.15 0.0085627 Biological_process Regulation of cell growth 5 3.14 0.0203792 Biological_process Lipid biosynthetic process 8 3.00 0.0217810 Biological_process Transcription corepressor activity 7 2.95 0.0058287 Molecular_function Transcription factor binding 11 2.95 0.0110869 Molecular_function G-protein-coupled receptor protein 5 2.68 0.0138521 Biological_process signaling pathway Transcription factor complex 312 2.65 0.0021869 Cellular_component Apoptosis Regulation 23 2.41 0.0335813 Ariadne signaling pathways Lipid metabolic process 9 2.35 0.0007067 Biological_process Response to stress 21 2.35 0.0079548 Biological_process Peptidase activity 15 2.35 0.0194074 Molecular_function Metabolic process 31 À 1.96 0.0044231 Biological_process Ser/Gly/Thr/Cys metabolism 6 À 2.50 0.0284851 Ariadne metabolic pathways

Table 2. List of genes used for microarray data validation

Gene P-value False discovery Fold Description symbol rate change

FOXC1 0.0002872 0.00655 3.590 Forkhead box C1 DTL 0.000212 0.00543 3.360 Denticleless homolog (Drosophila) ERBB3 0.0007893 0.0123 À 6.667 v-erb-b2 erythroblastic leukemia viral oncogene homolog 3 (avian) PDK4 o1e–07 o1e–07 À 4.000 Pyruvate dehydrogenase kinase, isozyme 4 SOD2 9.98e–005 0.00355 5.930 Superoxide dismutase 2, mitochondrial INSIG1 6e–007 0.000189 4.200 Insulin induced gene 1 CTGF 0.0003222 0.00993 8.580 Connective tissue growth factor IL6 0.0001 0.0001 4.167 Interleukin 6 IL8 o1e–07 o1e–07 12.020 Interleukin 8 ILF3 0.0003106 0.00975 2.940 Interleukin enhancer binding factor 3, 90 kDa TNFAIP6 0.0003264 0.00999 3.390 Tumor necrosis factor, alpha-induced protein 6 PTGS1 0.0000313 0.00263 4.200 Prostaglandin-endoperoxide synthase 1 (prostaglandin G/H synthase and Cyclooxygenase) PTGS2 0.0000001 0.0000759 35.110 Prostaglandin-endoperoxide synthase 2 (prostaglandin G/H synthase and Cyclooxygenase) NFKBIA 0.0005363 0.0132 4.210 Nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, alpha NFKBIE 0.0005605 0.0134 2.540 Nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, epsilon NFKB1 0.0000708 0.00424 3.440 Nuclear factor of kappa light polypeptide gene enhancer in B-cells 1

MATERIAL AND METHODS aCGH analysis Cell lines Genomic DNA was isolated from 10 mouse ovarian cell lines using a Derivation of genetically defined mouse OVCA lines is described in the routine protocol. Samples were profiled on the 44K Agilent mouse genome study by Xing and Orsulic14 and summarized in Supplementary Table 1. oligonucleotide array. DNA labeling, array hybridization, washing and Mouse OVCA lines and NIH3T3 cells were maintained in DMEM/F12 image processing were performed following previously published proto- 48 containing 10% fetal bovine serum, 100 IU/ml penicillin and 100 mg/ml col. To facilitate identification of DNA copy number aberrations, a log2 streptomycin (Invitrogen, Grand Island, NY, USA). Human OVCA cell lines ratio of þ 1/ À 1 was used as a cutoff, which was based on control OVCAR5, OV90, TOV112D were obtained from the American Type Culture hybridizations. Collection (Manassas, VA, USA). Immortalized non-tumorigenic ovarian surface epithelial cell line T29 was provided by Jinsong Liu, MD Anderson Cancer Center.47 All human cells were maintained in RPMI medium Fluorescence in situ hybridization analysis for aCGH data validation supplemented with 10% fetal bovine serum, 100 IU/ml penicillin and To validate the DNA copy number alterations detected by aCGH, two-color 100 mg/ml streptomycin. fluorescence in situ hybridization was performed on cell lines following

& 2014 Macmillan Publishers Limited Oncogene (2014) 4226 – 4235 LMX1B as an oncogene in ovarian cancer LHeet al 4234 previously reported protocol.49 The BAC clone RP23–65G6 containing In vivo tumorigenicity assays Lmx1b was purchased from Invitrogen and validated by PCR screening. For All experiments utilizing mice were reviewed and approved by the fluorescence in situ hybridization, RP23–65G6 was labeled in Cy3-dUTP and Massachusetts General Hospital Institutional Animal Care and Use the control clone RP23–60M5 mapping to 2qH4 was labeled in FITC-dUTP. Committee and were performed in strict accordance with the NIH Guide for the Care and Use of Laboratory Animals. Six to eight week old female nude mice (Charles River Laboratories, Wilmington, MA, USA) were used for Expression profiling xenograft experiment of mouse cell lines. Six to twelve week old female RNA from human ovarian cell lines OVCAR5 and OV90, transduced with NOD/SCID mice (strain NOD.CB17-Prkdcscid/J, Jackson Laboratory, Bar either LMX1B cDNA or GFP control was isolated using Qiagen RNeasy kit Harbor, ME, USA) were used for xenograft experiment of human cell lines. (Qiagen, Valencia, CA, USA). Generation of complimentary ribo nucleic acid Briefly, a total of 5 Â 105 cells were suspended in 1:1 PBS/Matrigel (BD (cRNA) and expression profiling was performed following the published Biosciences, San Jose, CA, USA) and subcutaneously injected into the protocol.28 Differential expression was considered significant at a P-value dorsal side of mice. Control animals were simultaneously injected with 1:1 of p0.001. PBS/Matrigel only. Tumor development was assessed weekly. The same number of cells were used for intraperitoneal injections.

Patient sample acquisition and processing Statistical analysis Human serous type OVCA samples were obtained from the Massachusetts Differences between LMX1B expression levels in subgroups based upon General Hospital Gynecologic Tissue Repository. Use of tissues for this T29 cells were assessed using two-sample t-tests. Kaplan–Meier survival study was approved by the Institutional Review Board. RNA/DNA was analysis was used to estimate the survival difference between low and isolated from frozen tumor samples with 480% tumor cells. high expressors defined using the median expression level as cutoff. Values were considered statistically significant (Po0.05), except for the microarray analysis, where significance was set at Pp0.001. Tumor weight Quantitative real-time PCR comparison was carried out by paired Student’s t-test. For quantitating mRNA expression, mouse and human tumors and human cell lines were lysed in Trizol reagent (Invitrogen). RNA was isolated using QIAGEN RNeasy kit and qRT–PCR was done in triplicate using gene-specific CONFLICT OF INTEREST primers and the SYBR Green PCR Master Mix and SYBR Green RT–PCR Reagents (Applied Biosystems, Grand Island, NY, USA). The authors declare no conflict of interest. For human samples, GAPDH was utilized as internal control; the level was then normalized against T29 cells. To calculate the relative expression for each gene, the 2 À DDCT method was used. The qRT–PCR primer ACKNOWLEDGEMENTS sequences for LMX1B are as follows: forward primer CACAGCAGCA- We thank Jinsong Liu (MD Anderson Cancer Center) for the T29-immortalized human GATCGTGG and reverse primer AGCAGTCGCTGAGGCTGG. The same ovarian surface epithelial cell line. This work was supported by the American Cancer primer set was used for both human and mouse qRT–PCR. Society Grant 118900-RSG-10–252–01-TBG (SO), Advanced Medical Research Foundation (BRR) and Vincent Memorial Research Funds (BRR) and the Michael Wall Gynecologic Oncology Research Fund (GM). Cell migration assay Boyden chamber-based migration assay was performed in a 24-well cell culture insert with 8.0 mm pore size PET membrane (BD Biosciences, San REFERENCES Jose, CA, USA). Cells were resuspended at a density of 100 000 cells per ml 1 Bell D, Berchuck A, Birrer M, Chien J, Cramer DW, Dao F et al. Integrated genomic in basal medium (RPMI1640 þ 0.1% fetal bovine serum) and plated in analyses of ovarian carcinoma. Nature 2011; 474: 609–615. triplicate in 350 ml into the upper chamber. The bottom chamber contained 2 Cannistra SA. Cancer of the ovary. N Engl J Med 2004; 351: 2519–2529. 500 ml of basal media with chemoattractant (10% fetal bovine serum). After 3 Ozols RF, Bookman MA, Connolly DC, Daly MB, Godwin AK, Schilder RJ et al. Focus cells were plated for 4 h, the insert was removed and wiped with a cotton on epithelial ovarian cancer. Cancer Cell 2004; 5: 19–24. swab to dislodge the non-migrated cells. Migrated cells on the underside 4 Pinkel D, Albertson DG. Array comparative genomic hybridization and its appli- were fixed and stained with Diff-Quick Stain Set (Dade Behring, Deerfield, cations in cancer. Nat Genet 2005; 37(Suppl): S11–S17. IL, USA) and quantified by counting under Â 40 magnification (field covers 5 Fishman A, Shalom-Paz E, Fejgin M, Gaber E, Altaras M, Amiel A. Comparing the 480% of the membrane). genetic changes detected in the primary and secondary tumor sites of ovarian cancer using comparative genomic hybridization. Int J Gynecol Cancer 2005; 15: 261–266. LMX1B cloning and retrovirus-mediated gene transduction 6 Israeli O, Gotlieb WH, Friedman E, Korach J, Friedman E, Goldman B et al. Genomic Human cDNA clone for LMX1B was obtained from Open Biosystems analyses of primary and metastatic serous epithelial ovarian cancer. Cancer Genet (Pittsburgh, PA, USA), and the coding region was PCR-amplified and cloned Cytogenet 2004; 154: 16–21. into pWZL retroviral vector. Final constructs were verified by sequencing. 7 Mayr D, Kanitz V, Anderegg B, Luthardt B, Engel J, Lohrs U et al. Analysis of gene C1, C2, BR5, NIH3T3, OVCAR5 and OV90 cells were transduced with LMX1B amplification and prognostic markers in ovarian cancer using comparative or GFP via retrovirus-mediated gene transfer as previously described.15 genomic hybridization for microarrays and immunohistochemical analysis for Subsequently, cells were drug selected for 10–15 days. LMX1B over- tissue microarrays. Am J Clin Pathol 2006; 126: 101–109. expression was confirmed by qRT–PCR. 8 Cheng KW, Lahad JP, Kuo WL, Lapuk A, Yamada K, Auersperg N et al. The RAB25 For western blot analysis of exogenous LMX1B protein, LMX1B ORF was small GTPase determines aggressiveness of ovarian and breast cancers. Nat Med cloned into pCMV6 vector with C-terminal FLAG tag from OriGene 2004; 10: 1251–1256. Technologies (Rockville, MD, USA). This construct was used for transient 9 Nanjundan M, Nakayama Y, Cheng KW, Lahad J, Liu J, Lu K et al. Amplification of expression of tagged LMX1B protein along with short hairpins against MDS1/EVI1 and EVI1, located in the 3q26.2 amplicon, is associated with favorable LMX1B (by cotransfection with sh-LMX1B-B10 or sh-LMX1B-C01) to patient prognosis in ovarian cancer. Cancer Res 2007; 67: 3074–3084. monitor the knockdown effect at the protein level using a monoclonal 10 Futreal PA, Coin L, Marshall M, Down T, Hubbard T, Wooster R et al. A census of anti-FLAG antibody (Sigma, St Louis, MO, USA). human cancer genes. Nat Rev Cancer 2004; 4: 177–183. Lentiviral shRNA constructs against human LMX1B were obtained from 11 Kim M, Gans JD, Nogueira C, Wang A, Paik JH, Feng B et al. Comparative onco- the Broad RNAi Consortium in pLKO-puro vector. Viral production and genomics identifies NEDD9 as a melanoma metastasis gene. Cell 2006; 125: infection was carried out according to the instructions provided by the 1269–1281. Consortium. The knockdown efficiency was tested by qRT–PCR. Two 12 Zender L, Spector MS, Xue W, Flemming P, Cordon-Cardo C, Silke J et al. Identi- constructs, sh-LMX1B-B10 (target sequence: GAACCAAAGAGCAAAGAT fication and validation of oncogenes in liver cancer using an integrative onco- GAA) and sh-LMX1B-C01 (target sequence: GCTACGCAAGGGCGACGAATT), genomic approach. Cell 2006; 125: 1253–1267. that showed significant knockdown effect were chosen for further studies. 13 Orsulic S, Li Y, Soslow RA, Vitale-Cross LA, Gutkind JS, Varmus HE. Induction of Transduced cells were passaged no more than two times before injection ovarian cancer by defined multiple genetic changes in a mouse model system. into NOD/SCID mice, expression analysis or in vitro assays. Cancer Cell 2002; 1: 53–62.

Oncogene (2014) 4226 – 4235 & 2014 Macmillan Publishers Limited LMX1B as an oncogene in ovarian cancer LHeet al 4235 14 Xing D, Orsulic S. A genetically defined mouse ovarian carcinoma model for the 31 Hunter CS, Rhodes SJ. LIM-homeodomain genes in mammalian development and molecular characterization of pathway-targeted therapy and tumor resistance. human disease. Mol Biol Rep 2005; 32: 67–77. Proc Natl Acad Sci USA 2005; 102: 6936–6941. 32 Gehring WJ, Affolter M, Burglin T. Homeodomain proteins. Annu Rev Biochem. 15 Xing D, Orsulic S. A mouse model for the molecular characterization of brca1- 1994; 63: 487–526. associated ovarian carcinoma. Cancer Res 2006; 66: 8949–8953. 33 McIntosh I, Dunston JA, Liu L, Hoover-Fong JE, Sweeney E. Nail patella syndrome 16 Lee TI, Young RA. Transcriptional regulation and its misregulation in disease. Cell revisited: 50 years after linkage. Ann Hum Genet 2005; 69(Pt 4): 349–363. 2013; 152: 1237–1251. 34 Bongers EM, Gubler MC, Knoers NV. Nail-patella syndrome. Overview on clinical 17 Miao J, Wang Z, Provencher H, Muir B, Dahiya S, Carney E et al. HOXB13 promotes and molecular findings. Pediatr Nephrol 2002; 17: 703–712. ovarian cancer progression. Proc Natl Acad Sci USA 2007; 104: 17093–17098. 35 Sato U, Kitanaka S, Sekine T, Takahashi S, Ashida A, Igarashi T. Functional char- 18 McCann CK, Growdon WB, Kulkarni-Datar K, Curley MD, Friel AM, Proctor JL et al. acterization of LMX1B mutations associated with nail-patella syndrome. Pediatr Inhibition of Hedgehog signaling antagonizes serous ovarian cancer growth in a Res 2005; 57: 783–788. primary xenograft model. PLoS One 2011; 6: e28077. 36 White KL, Rider DN, Kalli KR, Knutson KL, Jarvik GP, Goode EL. Genomics of the NF- 19 Ivan C, Hu W, Bottsford-Miller J, Zand B, Dalton HJ, Liu T et al. Epigenetic analysis kappaB signaling pathway: hypothesized role in ovarian cancer. Cancer Causes of the Notch superfamily in high-grade serous ovarian cancer. Gynecol Oncol Control 2011; 22: 785–801. 2013; 128: 506–511. 37 Johnson JD, Zhang W, Rudnick A, Rutter WJ, German MS. Transcriptional synergy 20 Burbach JP, Smits S, Smidt MP. Transcription factors in the development of between LIM-homeodomain proteins and basic helix-loop-helix proteins: the midbrain dopamine neurons. Ann NY Acad Sci 2003; 991: 61–68. LIM2 domain determines specificity. Mol Cell Biol 1997; 17: 3488–3496. 21 Guo C, Qiu HY, Huang Y, Chen H, Yang RQ, Chen SD et al. Lmx1b is essential for 38 Rascle A, Neumann T, Raschta AS, Neumann A, Heining E, Kastner J et al. The LIM- Fgf8 and Wnt1 expression in the isthmic organizer during tectum and cerebellum homeodomain transcription factor LMX1B regulates expression of NF-kappa B development in mice. Development 2007; 134: 317–325. target genes. Exp Cell Res 2009; 315: 76–96. 22 Jacks T, Remington L, Williams BO, Schmitt EM, Halachmi S, Bronson RT et al. 39 Grivennikov SI, Greten FR, Karin M. Immunity, inflammation, and cancer. Cell 2010; Tumor spectrum analysis in p53-mutant mice. Curr Biol 1994; 4: 1–7. 140: 883–899. 23 Jonkers J, Meuwissen R, van der Gulden H, Peterse H, van der Valk M, Berns A. 40 Yu H, Pardoll D, Jove R. STATs in cancer inflammation and immunity: a leading Synergistic tumor suppressor activity of BRCA2 and p53 in a conditional mouse role for STAT3. Nat Rev Cancer 2009; 9: 798–809. model for breast cancer. Nat Genet 2001; 29: 418–425. 41 Grivennikov SI, Karin M. Dangerous liaisons: STAT3 and NF-kappaB collaboration 24 Mani SA, Yang J, Brooks M, Schwaninger G, Zhou A, Miura N et al. Mesenchyme and crosstalk in cancer. Cytokine Growth Factor Rev 2010; 21: 11–19. Forkhead 1 (FOXC2) plays a key role in metastasis and is associated with 42 Karin M. Nuclear factor-kappaB in cancer development and progression. Nature aggressive basal-like breast cancers. Proc Natl Acad Sci USA 2007; 104: 2006; 441: 431–436. 10069–10074. 43 Mantovani A, Garlanda C, Allavena P. Molecular pathways and targets in cancer- 25 Taube JH, Herschkowitz JI, Komurov K, Zhou AY, Gupta S, Yang J et al. Core related inflammation. Ann Med 2010; 42: 161–170. epithelial-to-mesenchymal transition interactome gene-expression signature is 44 Liu G, Yang G, Chang B, Mercado-Uribe I, Huang M, Zheng J et al. Stanniocalcin 1 associated with claudin-low and metaplastic breast cancer subtypes. Proc Natl and ovarian tumorigenesis. J Natl Cancer Inst 2010; 102: 812–827. Acad Sci USA 2010; 107: 15449–15454. 45 Pressman CL, Chen H, Johnson RL. LMX1B, a LIM homeodomain class transcription 26 Dreyer SD, Zhou G, Baldini A, Winterpacht A, Zabel B, Cole W et al. Mutations in factor, is necessary for normal development of multiple tissues in the anterior LMX1B cause abnormal skeletal patterning and renal dysplasia in nail patella segment of the murine eye. Genesis 2000; 26:15–25. syndrome. Nat Genet 1998; 19: 47–50. 46 Stratton MR. Exploring the genomes of cancer cells: progress and promise. Science 27 Chen H, Lun Y, Ovchinnikov D, Kokubo H, Oberg KC, Pepicelli CV et al. Limb and 2011; 331: 1553–1558. kidney defects in Lmx1b mutant mice suggest an involvement of LMX1B in 47 Liu J, Yang G, Thompson-Lanza JA, Glassman A, Hayes K, Patterson A et al. human nail patella syndrome. Nat Genet 1998; 19: 51–55. A genetically defined model for human ovarian cancer. Cancer Res 2004; 64: 28 Mok SC, Bonome T, Vathipadiekal V, Bell A, Johnson ME, Wong KK et al. A gene 1655–1663. signature predictive for outcome in advanced ovarian cancer identifies a survival 48 Gabeau-Lacet D, Engler D, Gupta S, Scangas GA, Betensky RA, Barker II FG et al. factor: microfibril-associated glycoprotein 2. Cancer Cell 2009; 16: 521–532. Genomic profiling of atypical meningiomas associates gain of 1q with poor 29 Hernandez L, Hsu SC, Davidson B, Birrer MJ, Kohn EC, Annunziata CM. Activation clinical outcome. J Neuropathol Exp Neurol 2009; 68: 1155–1165. of NF-kappaB signaling by inhibitor of NF-kappaB kinase beta increases aggres- 49 Mohapatra G, Kim DH, Feuerstein BG. Detection of multiple gains and losses of siveness of ovarian cancer. Cancer Res 2010; 70: 4005–4014. genetic material in ten glioma cell lines by comparative genomic hybridization. 30 Bach I. The LIM domain: regulation by association. Mech Dev 2000; 91: 5–17. Genes Chromosomes Cancer 1995; 13: 86–93.

Supplementary Information accompanies this paper on the Oncogene website (http://www.nature.com/onc)

& 2014 Macmillan Publishers Limited Oncogene (2014) 4226 – 4235