Introduction

The neural crest comprises a multi-potent stem cell population that contributes to formation of diverse tissues. Through an epithelial to mesenchymal transition (EMT) that permits delamination from the neural plate, the acquisition of mesenchymal characteristics enables neural crest cells to transition towards a migratory phenotype. During a frequently lengthy path, neural crest cells proliferate, generating precursor cells that populate targets in multiple tissues, and so contribute to the formation of diverse organs. Such attributes essential to embryonic morphogenesis, are reminiscent of the invasive steps efficiently co-opted by cancer cells to drive malignancy and metastasis.

The lineage-specific transcription factors that regulate neural crest development represent frequent drivers of malignancy. An example of this paradigm is the forkhead box transcription factor FOXC1, that influences the migration and differentiation of neural crest cells. In addition to key roles in brain, ocular, cardiac, renal and skeletal development [1], FOXC1 is aberrantly expressed in diverse cancers. This overexpression was first identified in basal like breast cancer [2, 3], the subtype with the worst prognosis, and subsequently in at least fifteen additional tumor types, encompassing both solid tumors and hematological malignancies, such as AML [4, 5]. FOXC1 is also a prognostic factor for metastasis and survival, with its expression increasing tumour proliferation, invasion and dissemination [6]. Roles that include EMT induction, cell cycle control, angiogenesis, and regulation of stem cell populations are thought to promote FOXC1’s adverse oncological effects. However, despite such insights, the precise mechanisms behind FOXC1’s contributions to diverse malignancies remain incompletely defined.

Dysregulation of Hedgehog (Hh) signaling occurs frequently in sporadic and inherited cancers [7–10] and is estimated to underlie some 30% of malignancies. Hedgehog’s critical role in malignancy is linked to the pathway’s regulation of the Glioma (Gli) transcription factors and the requirement of Hh activity to maintain normal and cancer cell stemness. Notably, FOXC1 and several of its paralogs are able to induce Hh expression and Hh pathway activity [11–15]. The output of the Hh pathway is tightly regulated by three equipotent inhibitors: the Patched transmembrane receptors, Suppressor of Fused (Sufu), and Protein Kinase A (PKA). Of these, PKA is positioned comparatively late in the signal transduction cascade and, by phosphorylating the GLI oncoproteins, directs their conversion from full length active to truncated repressor forms [8, 16, 17]. As a result, PKA controls the transcriptional outcome of Hh signaling. In recent years, a tissue-specific PKA antagonist has been identified. This Rho GTPase activating protein, Arhgap36, alters the balance between activator and repressor categories of GLI by inhibiting PKA, thereby inducing strong Hh pathway activation [18–20].

In this study, RNA- and ChIP-sequencing coupled with CRISPR interference were used to demonstrate that Foxc1 binds a locus located in a region of closed chromatin to induce Arhgap36 expression. In turn, Arhgap36 depletes the levels of PKA and its catalytic subunit PKAC, both strongly activating Hh signaling and making signal transduction less dependent on regulation via Smoothened. The oncological relevance of the Foxc1-Arhgap36 relationship is demonstrated in neuroblastoma, a neural crest derived pediatric malignancy where ARHGAP36 levels predict five-year mortality. Consequently, our study identifies a mechanism through which overexpression of FOXC1 dysregulates the Hedgehog pathway and contributes to malignancy.

Results

To discern pathways by which Foxc1 may contribute to tumorigenesis, Foxc1 was stably overexpressed in NIH3T3 fibroblasts. A retroviral construct was used that recapitulated the high levels of FOXC1 expression observed in a range of malignancies [15]. After RNA-sequencing, analysis using a stringent false discovery rate (q ≤ 0.01; log₂fold change ≥ 1.0) defined a set of differentially expressed genes that were each distinguished by two independent bioinformatic workflows (DEseq2, and Cufdiff; Table S1). The majority were up-regulated by Foxc1 expression (n=192 genes; total 285; Fig. 1A). Gene-set enrichment analysis demonstrated significant enrichment for cancer invasiveness phenotypes, and malignancies, that importantly included neural crest associated malignancies such as melanoma and neuroblastoma (EMT P = 4.8 × 10⁻¹¹; Malignancies P = 10⁻³ - 10⁻⁴; Table S2). Also enriched were phenotypes either induced by FOXC1 mutation, (glaucoma and abnormal intraocular pressure, P ≤ 4.7 × 10⁻⁴) [21, 22] or present in Foxc1^−/− murine mutants (abnormal skeletal morphology and renal anomalies, P = 0.05 - 10⁻³) [1, 23, 24]. Finally, genes relevant to Foxc1’s biological functions (angiogenesis, endochondral ossification, and hair follicle development) [13, 14, 25–30] were also enriched in relevant cell types (vasculature P = 2.1 × 10⁻⁵; neural crest P = 3.0 × 10⁻⁴; Table S2). Together, these results support the differentially expressed genes being representative of Foxc1 functions in development and disease.

Figure 1.

20% of the significantly dysregulated genes are implicated in Hh signalling (62 of 285) either as regulators or targets of the pathway (Table S1). Besides a cytokine that regulates inflammation and immunity (Interleukin Receptor 1 antagonist), Arhgap36 mRNA was the most upregulated in the RNA-sequencing dataset (Fig. 1A). To test whether Foxc1 consistently induces Arhgap36 expression in mesenchymal-derived cells, qPCR measurements were performed and strong induction of Arhgap36 mRNA observed after Foxc1 overexpression in several relevant cell lines [C2C12 (murine myoblast) 10⁵-fold, ATDC5 (chondrogenic) 10²-fold, NIH3T3 10³-fold increase; Fig. 1B and Fig. S1]. Moreover, in a Hh signalling reporter cell line (NIH3T3-Gli2-mGFP) [15, 31], Foxc1 overexpression induced Arhgap36 protein expression at levels comparable to ectopically expressed Myc-FLAG tagged Arhgap36 (Fig. 1C). Both Foxc1 overexpression and ectopically expressed Arhgap36 elevated protein expression of Gli1, a terminal effector and major readout of Hh pathway activity (Fig. 1C). Notably, the cellular localisation patterns of Foxc1-induced endogenous and ectopically expressed Arhgap36 are similar (Fig. 1D, Fig. S2A). Collectively, these data demonstrate Foxc1’s ability to induce Arhgap36 expression and activate Hedgehog signalling.

To characterize the functional interaction between the transcription factor Foxc1 and Arhgap36, whole genome ChIP-sequencing (ChIP-seq) was performed on Gli2-mGFP NIH3T3 cells stably expressing Foxc1. Two independent Foxc1 antibodies were used, together with an isotype-matched control immunoglobulin as control for Foxc1 peak specificity. The number of uniquely mapped 60 bp reads (Input > 5.3 × 10⁸, Foxc1-ChIP 4.8 × 10⁸) corresponds with deep sequencing coverage. Examination of chromatin accessibility demonstrates that, with the exception of embryonic stem cells, the Arhagp36 locus is inactive (Fig. S3), an observation consistent with Arhgap36’s highly restricted tissue expression. Within a 200 kb interval encompassing the Arhgap36 locus, five regions displayed strongly overlapping ChIP-seq peaks with both anti-Foxc1 antibodies (Fig. 2A). Three (Prox-1 to Prox-3) were 0.2 - 2.5 kb from the transcription start sites, while two (Dist-1, Dist-2) were 55 and 71 kb upstream (Fig. 2A). Two amplicons were selected within each peak, and ChIP-qPCR demonstrated strongly increased Foxc1 signal at each peak, relative to DNA precipitated using a non-specific normal IgG control (Fig. S4). Two groups of position weight matrices were significantly enriched in the bulk ChIP-seq peak data (STREME sequence motif discovery algorithm) (Fig. 2B, Fig. S4). The first comprised consensus sequences closely resembling known Foxc1-binding motifs (P = 10⁻⁴¹ - 10⁻⁶⁵), while the second corresponded closely with that of Fos-Jun transcription factor dimers (P = 10⁻¹³ - 10⁻²⁰, Fig. 2B, Fig. S5). Each of the five ChIP-seq peaks contains the identified Foxc1 motifs, supporting Arhgap36 being a direct transcriptional target of Foxc1.

Figure 2.

We next investigated in a Hh reporter line [15] if Foxc1 overexpression recapitulated Arhgap36’s augmentation of Hh signaling, via down-regulation of PKA. As expected, ectopic expression of Foxc1 depleted the level of PKAC, PKA’s catalytic subunit (63% reduction, P = 1.4 × 10⁻⁷; Fig. 3A). Foxc1 overexpression also reduced phosphorylation of threonine 197 in PKAC’s activation loop, a residue essential to PKA enzymatic function and protein stability [32, 33]. This ~70% depletion of pT197 PKAC (P = 1.3 × 10⁻⁸) is analogous to the effect of ectopic Arhgap36 expression in parental NIH3T3 cells (Fig. 3A). Immunofluorescent staining demonstrated in cells with ectopic Foxc1 or Arhgap36 expression the near complete absence of PKAC and pT197 PKAC in the cytoplasm, especially at the basal body, where compartmentalized PKA-dependent control of Hh pathway occurs (Fig. 3B). Notably, in a transduced cell population with heterogeneous Foxc1 expression, PKAC signal is almost entirely depleted from cells with a high Foxc1 nuclear signal, demonstrating an inverse correlation between levels of PKAC and nuclear Foxc1 immunostaining (Fig. S2B). Similar results were observed in a second cell model, pre-adipocyte cells (3T3-L1) where expression of Foxc1 induced expression of Arhgap36 (albeit at lower levels) and a 40-50% reduction in protein levels of PKAC and pT197 PKAC (Fig. S2C,D). Direct overexpression of Arhgap36 led to strong downregulation of PKAC. These data, from two cell-based systems, demonstrate that Foxc1 expression depletes PKAC, the enzyme essential for promoting repressor forms of Gli, which regulate Hh target gene expression.

Figure 3.

We next examined using CRIPSR interference whether steric repression of the five Foxc1 binding loci affected Arhgap36 transcription. Two clones of NIH3T3-Gli2-mGFP cells that overexpress Foxc1 were created that each stably expressed an inactive form of Cas9 fused with the Krüppel-associated box (KRAB) repressor domain of ZIM3 protein. Recruitment of ZIM3 KRAB-dCas9 to each of the five Foxc1-binding regions by retroviral delivery of five separate pools each comprising 3-4 single guide RNAs, altered Arhgap36 expression, although to varying degrees (Fig. 4A, Fig. S6). The sgRNA pools targeting Prox-3 and Dist-2 reduced Arghap36 mRNA expression by 98-99% and 67-76% respectively (Figure 4A). In both cases, reduced expression of Arhgap36 was associated with lower levels of Gli1 mRNA, and the degree of Arhgap36 mRNA reduction correlated with the decrease in Gli1 mRNA levels. Western immunoblot analysis confirmed that silencing either the Prox-3 or Dist-2 regions decreased Arhgap36 protein expression. This in turn correlated with an increase in PKAC and pT197 PKAC levels and a reduction in Gli1 protein levels (Fig. 4B). Since these data support Prox-3 and Dist-2 functioning, respectively, as promoter and distal enhancer for murine Arhgap36, evolutionary conservation of the Foxc1-binding loci was assessed. Multiple alignment across 60 vertebrates (Multiz) and measurements of evolutionary conservation (PhastCons) demonstrated that Prox-3 and Dist-2 are both conserved in placental, but not in marsupial mammals, nor in other vertebrates (Fig. S7). Multiple potential binding motifs for Foxc1 and Fos-Jun family transcription factors were identified in both regions; as illustrated by 16 Foxc1-binding motifs in both Prox-3 (8 motifs) and Dist-2 (8 motifs) of the murine locus, that were conserved or partially conserved in human (Fig. S7; Table S3). Taken together, the data presented support the Prox-3 and Dist-2 regions functioning as cis-regulatory elements for murine Arhgap36, and other placental mammals, including human.

Figure 4.

Since PKA depletion de-represses Hh signaling, the pathway activation is predicted to become less dependent on Hh ligand, which pharmacologically would manifest as tumour resistance to inhibition of Smoothened (Smo). To test whether Foxc1 expression induces such ligand-independent Hh signal transduction, Gli1 levels were measured with and without two Smoothened-specific antagonists (Sonidegib and Cyclopamine). The results revealed that the Foxc1-induced increase of Gli1 mRNA and protein levels is resistant to inhibition by either antagonist (Fig. 5A,B). Furthermore, the Foxc1-induced resistance to Sonidegib is phenocopied by the ectopic expression of Arhgap36 (Fig. S8A). In both cases Gli1 upregulation, whether by Foxc1 or by Arhgap36 overexpression, is not significantly reduced by co-treatment with Sonidegib. Indeed, Foxc1 overexpression leads to a similar level of Gli1 upregulation in NIH3T3 cells as treatment with a Smoothened agonist (SAG), and co-treatment of cells with SAG does not significantly enhance Gli1 levels beyond Foxc1-induced upregulation alone (Fig. 5B). Consequently, these data demonstrate that Foxc1 induces strong non-canonical activation of Hh signaling, a characteristic of multiple malignancies, including advanced and metastatic tumors.

Figure 5.

PKA’s roles regulating the Hh pathway extend beyond repressing Gli activity. PKA also phosphorylates key serine and threonine residues within Sufu. This inhibits ciliary trafficking of Sufu-Gli complexes, resulting in retention of Sufu at the tip of the primary cilium [34, 35]. We found that in addition to substantially increased axonemal tip accumulation of Gli2-mGFP (Fig. 6A, Fig. S8B,C), expression of Foxc1 also increases Sufu levels at the tip of the axoneme 3-fold (P ≤ 2 × 10⁻¹⁰) while a similar 2-fold increase is induced by ectopic expression of Arhgap36 (Fig. 6A-C). The status of Sufu’s PKA-dependent phosphorylation sites is known to influence Sufu activity. Two of these adjacent serine residues comprise a classical dual phosphorylation site for PKA (pS346) and GSK3β (pS342) [34]. Consequently, the ~50% reduction in levels of Sufu pS342 after Foxc1 (P = 6.4 × 10⁻⁵) or Arhgap36 expression (Fig. 6D), provides further support for Hh pathway regulation by Foxc1. Together these data demonstrate that the effects of Foxc1 expression on Hh signal transduction involve several core regulators of Hh pathway activity.

Figure 6.

The presented data support a model by which Foxc1-dependent Arhgap36 expression attenuates PKA activity and Sufu function. To verify that the observed changes in Sufu and Gli2 ciliary accumulation induced by Foxc1 overexpression are mediated by Arhgap36, shRNA silencing of Arhgap36 was performed in Foxc1-expressing Gli2-mGFP cells (Fig. 7). Two different Arhgap36-targeting shRNAs each substantially diminished ciliary Gli2-mGFP accumulation (shRNA1: 35%, shRNA2: 64% reduction; P = 4.9 × 10⁻⁷ and 4.5 × 10⁻¹⁰; Fig. 7A-C) while Gli2-mGFP levels in control cells that express very low levels of Arhgap36 were unchanged. The decrease in axonemal Gli2-mGFP correlated with the reduction in Gli1 mRNA (Fig. 7D), and the shRNA that more efficiently inhibited Arhgap36 induced a greater decrease in both Gli2-mGFP accumulation and Gli1 expression. The specificity of Arhgap36 inhibition is apparent from marked reductions in Gli1 and Arhgap36 mRNA expression, while levels of Foxc1 were unaltered (Fig. 7D). Analysis of Gli1 expression in parental Foxc1-expressing NIH3T3 cells treated with the same Arhgap36-targeting shRNA recapitulated the prior findings (Fig. S9). Taken together, these data demonstrate that the effects of Foxc1 expression on Hh pathway activity are Arhgap36-mediated.

Figure 7.

To test the oncologic relevance of the Foxc1 - Arhgap36 relationship, ARHGAP36 mRNA levels were first surveyed in cancer expression datasets. This revealed high expression in a common pediatric malignancy, neuroblastoma, as well as specific CNS, breast, lung, and neuroendocrine tumors. Due to neuroblastoma’s neural crest origin and Foxc1’s roles in this stem cell population, we focused on neuroblastoma data. Survival was evaluated in three independent neuroblastoma patient cohorts using Kaplan-Meier analyses after stratification into terciles of high, medium and low ARHGAP36 mRNA expression. The highest tercile of ARHGAP36 expression correlated with a favorable overall survival compared to the lowest: GSE49711, 91 vs 63% 5-year survival; E-MTAB-178191, 91 vs 61%; TARGET study, 60 vs 31%, (Fig. 8A-C) [36–38]. The effect was consistent across all three cohorts, despite differences in cancer staging - the first two cohorts primarily comprise patients with early disease, while TARGET mainly consists of advanced or metastatic neuroblastoma cases. Merging the data into a single dataset of 1348 patients narrowed the confidence intervals to provide more precise estimates. The highest tercile of ARHGAP36 expression was associated with 87% five-year survival (range 86-92%) in contrast to 58% (53-63%) for the lowest tercile (P = 1.7 × 10⁻¹⁹). Illustrating these findings in the context of MYCN, whose amplification represents the primary genetic marker for high-risk and prognostically poor neuroblastoma, high ARHGAP36 expression was associated with an 89% five-year survival compared to 37% with MYCN amplification (Fig. S10).

Figure 8.

Discussion

Here we show that the PKA antagonist Arhgap36 is a central gene dysregulated by Foxc1 overexpression, and that this linkage provides new perspectives in Foxc1’s contributions to malignancy. Foxc1’s transcriptional regulation of Arhgap36, and the consequent inhibition of PKA, alters the equilibrium between full length transcriptional GLI activators and truncated transcriptional repressors, weakening regulation and increasing activity of a pathway fundamental to development. It is therefore logical that a transcription factor identified to enhance Arhgap36 expression is one that induces Hh activity in multiple tissues during development [13–15]. In some contexts, Hedgehog functions as a morphogen that confers positional information, with a gradient of activity specifying individual cell types and behaviors required for establishment of complex tissue patterns [39–41]. Increased gene dosage from FOXC1 copy number variation causes phenotypes indicative of impaired Hedgehog signaling, including cerebellar hypoplasia and anomalous facial skeletal development [42–45]. Consequently, Foxc1’s ability to induce Arhgap36 expression establishes a new mechanism by which the transcription factor may modulate Hh pathway activity. This has particular relevance for oncology, where dysregulated Hh expression makes major contributions and larger fold increases in FOXC1 expression occur.

A second finding concerns the breadth of alteration to Hh signaling. Foxc1’s induction of Arhgap36 depletes an evolutionary conserved negative regulator of Hh signaling (PKA) from the cytoplasm and base of the primary cilium. Such wide loss of PKA activity should release Gli transcription factors from phosphorylation-induced direct inhibition (in case of Gli2) and proteolytic processing (in case of Gli2 and Gli3). In parallel, Foxc1-induced depletion of PKAC predictably dysregulates the PKA substrate Sufu: a second major negative regulator of the Hh pathway, whose roles encompass development, disease, and notably tumour suppression. Sufu acts as a scaffold between PKA and Gli proteins, that facilitates their phosphorylation and proteolytic processing, while also directly inhibiting transcriptional activity of active Gli forms [35]. The reduced phosphorylation of Sufu and increased translocation of Sufu and Gli2 to the tip of the cilium represent alterations to fundamental features of Hh signal transduction. As Arhgap36 acts independently of Smo [18], another consequence of Foxc1-driven Arhgap36 expression observed in our experiments is acquired resistance of Hh pathway activity to Smo antagonists. Consequently, by altering activity of PKA, Foxc1 overexpression impairs central elements of the Hh pathway resulting in increased transcription of Hh-target genes and, via augmented Smo-independent activity, tumor resistance to Smoothened antagonists. Collectively, such properties are consistent with the overexpression of FOXC1 in multiple tumors being associated with more aggressive cancer phenotypes, drug resistance and metastasis.

Forkhead box transcription factors’ ability to access areas of closed chromatin is mediated by a conserved winged-helix DNA binding domain that is structurally similar to histones H1 and H5 [46]. This capacity to directly bind DNA targets in condensed chromatin enables pioneer factors to activate enhancers and, by opening condensed chromatin, provide access to transcription factors that lack pioneering ability. The closely correlated ChIP-seq peaks identified in this study, which contain consensus sequences for Foxc1 and Fos-Jun dimers, support coordinated transcription factor binding at the Arhgap36 locus. The consistent co-localisation of these transcription factors’ binding motifs, including in other placental mammals, accords with enrichment of FOSL2 and JUNB motifs at the FOXC1 ChIP-seq peaks noted in a previous study [3]. Together, such data support a model in which the overexpression of FOXC1, by initiating local opening of chromatin [47], makes the ARHAGP36 locus accessible to other transcription factors to induce gene expression. Such combinatorial control of Arhgap36’s transcriptional regulation is predicted to have heterogeneous effects on Hh pathway activity, varying in tissue and cell-specific manners and being determined by levels of FOXC1 and FOS-JUN dimers.

FOXC1’s induction of Hh signaling via ARHGAP36 and PKA inhibition is expected to modify processes in both development and disease. Since our data were derived with retroviral constructs, where 10-20 fold increases in Foxc1 protein levels reiterate overexpression in FOXC1-associated tumours [2], we evaluated the relevance of the Foxc1-Arhgap36 interaction in human malignancy. Analyses in neuroblastoma, a heterogeneous, life-threatening and common pediatric malignancy that originates from sympathoadrenal neural crest progenitors, revealed higher levels of ARHGAP36 expression were comparatively protective. The 2.8 to 8-fold greater five-year survival identified in three independent patient cohorts suggests ARHAGP36 may represent an informative prognostic factor. Because neuroblastoma treatment regimens frequently impact the development of surviving children [48], the ability to stratify by prognosis may facilitate treatment choice and minimize morbidity for a portion of patients. Although the mechanism(s) mediating this effect remain to be explored, this observation is consistent with the requirement of Hh activity for many tumours to proliferate [49, 50] and for maintenance of stem cell self-renewal, including that of cancer stem cells [51].

In summary, dysregulated expression of FOXC1 is well-implicated in malignancy, with expanding evidence that elevated FOXC1 levels are associated with increased mortality in breast and hepatic cancer and malignancies as diverse as AML and low grade glioma [4, 52–54]. In contrast with such progress, determining how FOXC1 promotes malignancy has proved more elusive. This work establishes that overexpression of FOXC1, at levels consistent with FOXC1-associated malignancies, transcriptionally activates Arhgap36 expression which in turn dysregulates multiple fundamental aspects of Hh pathway activity. The presented data further demonstrate that ARHGAP36 levels are predictive of five-year survival in a common pediatric malignancy and identify a biological characteristic that merits further investigation. Overall, this study provides additional mechanistic insight into FOXC1’s capacity for stimulating malignancy.

Supplementary figures

Figure S1.

Figure S2.

Figure S3.

Figure S4.

Figure S5.

Figure S6.

Figure S7.

Figure S8.

Figure S9.

Figure S10.

Acknowledgements

The authors are grateful to the patients who have contributed to this study. The authors thank Dr. Fred Berry for numerous fruitful discussions that facilitated and helped guide this research. We also thank multiple colleagues for critically reviewing earlier versions of the manuscript. Funding was provided by the Canadian Institutes of Health Research (CIHR) (MOP-133658). Experiments were performed at the Cell Imaging Core (RRID:SCR_019200), which is supported by the University of Alberta Faculty of Medicine & Dentistry, University Hospital Foundation, Striving for Pandemic Preparedness – The Alberta Research Consortium, and Canada Foundation for Innovation.

Additional Information

Data Availability Statement

All data generated and/or analysed in this study are available from the corresponding author upon reasonable request. The RNA-sequencing and ChIP-sequencing data have been deposited to the NCBI GEO database with the following accession numbers: GSE297719 (RNA-seq), GSE297865 (ChIP-seq).

Funding

Canadian Institutes of Health Research (MOP-133658)

Additional files

Additional experimental methods.

File S1. List of differentially expressed genes in NIH3T3 cells expressing Foxc1 or pLXSH empty vector control. Complete list of differentially expressed genes identified in NIH3T3 cells expressing Foxc1 or empty vector control [pLXSH] by RNA-sequencing (n = 3) and used for volcano plot (Figure 1A). n = 292 differentially expressed genes; 195 upregulated, 97 downregulated. Selection criteria: q-value (FDR-adjusted p-value) ≤ 0.01, absolute log2 ratio ≥ 1, signal cut-off ≥ 1 FPKM in either condition].

File S2. Functional profiling of the list of differentially expressed genes from File S1. Functional profiling of the list of genes differentially expressed in Foxc1-expressing NIH3T3 cells vs empty vector control [pLXSH] was performed using gene set enrichment analysis (GSEA) platform Enrichr [https://maayanlab.cloud/Enrichr/]. The list was analysed against multiple gene set libraries and databases. Results demonstrate significant associations between differentially expressed genes and molecular pathways / biological processes / pathological conditions with previous implication of Foxc1, as well as Foxc1-specific cell and tissue expression. Overall, these results indicate that ectopic expression of Foxc1 in NIH3T3 affects functionally relevant targets. Examples include association with ocular and vascular disease [glaucoma p = 5.8 × 10⁻⁴; intraocular pressure p = 5.9 × 10⁻³; hypertension p = 5.7 × 10⁻⁴; myocardial ischemia p = 2.4 × 10⁻⁴; coronary artery disease p = 0.01], vascular development [regulation of angiogenesis p = 9.4 × 10⁻⁵], bone development [endochondral ossification 2.8 × 10⁻⁴], a number of mouse skeletal and renal phenotypes [MGI phenotypes]; multiple cancers and cancer invasiveness phenotypes [epithelial mesenchymal transition p = 7.7 × 10⁻¹¹]. Relevant cell and tissue expression signatures [mesenchyme p = 3.4 × 10⁻⁸; mesenchymal stem cells p = 5.8 × 10⁻⁴; (mesenchymal) stromal cells p = 1.8 × 10⁻⁶; vasculature p = 2.7 × 10⁻⁵; neural crest p = 3.8 × 10⁻⁴; several skeletal tissues; differentiating osteoblasts p ≤ 1.4 × 10⁻⁶]. p-values are FDR-adjusted.

File S3. List of predicted Fox and Fos-Jun transcription factor binding sites in Prox 3 and Dist 2 regions of mouse Arhgap36.