HOPX regulates bone marrow-derived mesenchymal stromal cell fate determination via suppression of adipogenic gene pathways

Previous studies of global binding patterns identified the epigenetic factor, EZH2, as a regulator of the homeodomain-only protein homeobox (HOPX) gene expression during bone marrow stromal cell (BMSC) differentiation, suggesting a potential role for HOPX in regulating BMSC lineage specification. In the present study, we confirmed that EZH2 direct binds to the HOPX promoter region, during normal growth and osteogenic differentiation but not under adipogenic inductive conditions. HOPX gene knockdown and overexpression studies demonstrated that HOPX is a promoter of BMSC proliferation and an inhibitor of adipogenesis. However, functional studies failed to observe any affect by HOPX on BMSC osteogenic differentiation. RNA-seq analysis of HOPX overexpressing BMSC during adipogenesis, found HOPX function to be acting through suppression of adipogenic pathways associated genes such as ADIPOQ, FABP4, PLIN1 and PLIN4. These findings suggest that HOPX gene target pathways are critical factors in the regulation of fat metabolism.


Retroviral transduction.
Full-length human coding sequence for HOPX (NCBI RefSeq: NM_001145459.1) was subcloned into the pRUF-IRES-GFP vector (Kind gift by Paul Moretti, University of South Australia, Australia). Retroviral transduction of HOPX/pRUF-IRES-GFP or empty vector control pRUF-IRES-GFP into human BMSC was performed as previously described 15 . Stably transduced BMSC expressing high levels of GFP were selected by FACS, using a BD FACSAria Fusion flow cytometer (https ://www.bdbio scien ces.com). Overexpression of HOPX was confirmed by qPCR analysis. siRNA knock-down transfections. Human BMSC were seeded at 10 4 cells/cm 2 and siRNA knockdown was performed on the following day. Sequence specific siRNAs against HOPX (ThermoFisher Scientific, https :// www.therm ofish er.com/) were used at 12 pMol to achieve a > 90% knockdown of transcript levels. The siRNA used in this study were: HOPX s39106 and s39107 and Silencer Select Negative Control #1 siRNA. The procedure was performed in accordance with manufacturer's instructions with a 72 h incubation period before performing functional assays. in vitro differentiation assays. Human BMSC (10 4 cells/cm 2 ) were cultured in either normal growth conditions (αMEM supplemented with 10% FCS, 2 mM l-glutamine, and 100 μM l-ascorbate-2-phosphate); or osteogenic inductive conditions (αMEM supplemented with 5% FCS, 2 mM l-glutamine, 50 U/mL penicillin-streptomycin, 10 mM HEPES buffer, 1 mM sodium pyruvate, 0.1 mM dexamethasone, 100 μM l-ascorbate-2-phosphate and 2.6 mM KH 2 PO 4 ); or adipogenic inductive conditions (αMEM supplemented with 10% FCS, 2 mM l-glutamine, 50 U/mL penicillin-streptomycin, 10 mM HEPES buffer, 1 mM sodium pyruvate, 120 mM indomethacin and 0.1 mM dexamethasone) for up to 3 weeks as previously described 15,16 . Mineralized bone matrix was assessed with Alizarin red (Sigma-Aldrich, Inc.) staining 15 . Extracellular calcium was measured as previously described 15 . Lipid formation was identified by Nile-red (Sigma-Aldrich, Inc.) staining as previously described 15  Differential gene expression and pathway analysis. Gene expression analyses were carried out in R using mostly Bioconductor packages EdgeR 22,23 and Limma 24 . Gene counts were filtered for low expression counts by removing genes with less than 1 count per million (cpm) in more than two samples and then normalised by the method of trimmed mean of M-values 25 . Differential gene expression was carried out on log-CPM counts and precision weights available from the Voom function in Limma 26 , with linear modelling and empirical Bayes moderation. Annotation of results were carried out using Ensembl annotations (https ://grch3 7.ensem bl.org) available in BiomaRt 27 , and expression results displayed in heatmaps using the Pheatmap package 28 .
Statistics. Generation of graphs and data analysis was performed using GraphPad Prism 7 (GraphPad Software, LA Jolla, CA, https ://www.graph pad.com/). Statistical significance (*) of p < 0.05 between samples are shown based on Student's t-test and One-way ANOVA as indicated.

HOPX expression is directly repressed by EZH2. Previous studies of global ChIPseq analyses found
that the H3K27 methyltransferase, EZH2, regulates HOPX expression during BMSC osteogenic differentiation 8 . Enforced expression of EZH2 in cultured human BMSC resulted in a decrease in HOPX gene expression levels ( Fig. 1A,B). Manual ChIP analysis was used to assess the binding of EZH2 to putative DNA binding sites on HOPX, using genomic DNA isolated from cultured human BMSC. The data showed preferential binding of EZH2 to the S3 binding region of the HOPX promoter region in BMSC cultured under normal growth conditions and osteogenic inductive conditions (Fig. 1C,D). However, EZH2 enrichment on all HOPX binding sites (S1, S2 and S3) was greatly diminished when BMSC were cultured under adipogenic inductive conditions (Fig. 1E).
HOPX is a promoter of BMSC proliferation. In order to determine if HOPX regulates BMSC proliferation, HOPX was overexpressed in BMSC using retroviral transduction ( Fig. 2A). Cell proliferation was assessed by BrdU incorporation under normal growth conditions. The data showed a significant increase in the proliferation rates of BMSC following enforced expression of HOPX (Fig. 2B). To further confirm that HOPX regulates BMSC proliferation, HOPX expression was knocked down using two independent siRNA molecules targeting HOPX transcripts (Fig. 2C). Knock down of HOPX in BMSC resulted in a significant decreased in proliferation rates (Fig. 2D). These data suggest that HOPX is a positive regulator of BMSC proliferation.
HOPX is an inhibitor of BMSC adipogenesis. We next explored the role of HOPX during human BMSC differentiation. Functional studies were carried out using retroviral transduced HOPX overexpressing constructs or empty vector infected BMSC, cultured in control or adipogenic inductive media (Fig. 3A). Overexpression of HOPX resulted in decreased Nile-red-positive lipid producing adipocytes compared with empty vector control Scientific RepoRtS | (2020) 10:11345 | https://doi.org/10.1038/s41598-020-68261-2 www.nature.com/scientificreports/ To verify these findings, siRNA-mediated knockdown using two independent siRNA targeting HOPX transcripts in BMSC was performed (Fig. 4A). The data showed a dramatic increase in Nile-red-positive lipid-producing adipocytes following adipogenic induction, compared with BMSC treated with control scramble siRNA ( Fig. 4B-E). Furthermore, siRNA knockdown of HOPX resulted in an increase in C/EBPα (Fig. 4F) and ADIPSIN (Fig. 4G) transcript levels compared with scramble siRNA-treated cells following adipogenic induction. Overall, these data demonstrate that HOPX is a repressor of adipogenesis. www.nature.com/scientificreports/ To identify the function of HOPX in BMSC osteogenic differentiation, HOPX overexpressing BMSC or empty vector infected BMSC were cultured under control or osteogenic inductive media (Fig. 5A). Assessments of extracellular calcium levels found no difference between HOPX overexpressing BMSC and vector control BMSC (Fig. 5B). Similarly, mineralized deposits were stained with Alizarin Red after 3 weeks under osteogenic growth conditions with no observable differences (Fig. 5C). In accord with these findings, HOPX overexpressing BMSC (Fig. 5D) showed no significant difference in the transcript levels of the osteogenic master regulator, RUNX2 (Fig. 5E) and the mature bone marker, OSTEOPONTIN (OPN) (Fig. 5F), compared to the vector control cells. Confirmatory studies employing siRNA-mediated knockdown of HOPX in BMSC (Fig. 5G) found no significant differences in the levels of Alizarin positive mineral and extracellular calcium levels compared with scramble siRNA-treated BMSC (Fig. 5H-K). Overall, these findings demonstrate that HOPX has no direct effect on the osteogenic capacity of BMSC.

HOPX inhibits BMSC adipogenic differentiation via suppression of adipogenic associated genes.
We next explored potential mechanisms of HOPX action during BMSC adipogenic differentiation.
Total RNA was collected from HOPX overexpressing and vector control BMSC cultured for 2 weeks under adipogenic inductive conditions, then processed for RNA-sequencing to identify novel HOPX-regulated genes during BMSC adipogenic commitment [79]. Due to the variable gene expression patterns between different individ-  www.nature.com/scientificreports/ uals (n = 3 donors), the P-value significance was excluded as a criteria to select for differentially expressed genes (DE). Therefore, the top 50 DE (Fig. 6) were selected based on the fold change (a log fold change (logFC) ≥ |1| or ≥ |− 1|). To validate the RNA-sequencing results, confirmatory qPCR was performed on a number of genes that appeared to change expression in HOPX overexpressing BMSC under adipogenic conditions. HOPX transcripts were found to be elevated in HOPX overexpressing BMSC compared to vector control BMSC, which were relatively higher during adipogenesis compared to normal growth conditions for the respective population. From the transcriptional expression heat map (Fig. 6), we observed a number of genes that were upregulated during adipogenesis but suppressed in HOPX overexpressing cells. Table 1 indicates the functional role of these genes following Gene Ontology (GO) enrichment analysis, with 188 genes involved in EMT, 185 genes in adipogenesis and 127 genes in fatty acid metabolism. The differential gene expression levels of representative upregulated genes, HOPX, ADIPOQ, AOC3, FABP4, G0S2, GPD1, PLIN1 and PLIN4 were confirmed by qPCR ( Fig. 7A-H).
Other genes were found to be downregulated during adipogenesis and promoted by HOPX expression such as CNN1 (Fig. 7I). The RNA-sequencing analysis provides insight into putative targets of HOPX during BMSC adipogenesis.

Discussion
Our studies suggest that HOPX mediates postnatal BMSC proliferation and lineage determination. Protein structural studies have demonstrated that HOPX is unable to bind to DNA, suggesting that HOPX functions through protein-protein interaction with partner proteins. A number of HOPX partner proteins have been identified, including Hdac1, Hdac2, MTA 1/2/3, MBD3 and Rbbp4/7 29 . HOPX has been shown to be a key factor in cardiac development, where it regulates cell proliferation and differentiation at different stages during murine cardiac development 5,6 . The present study found that human BMSC express low levels of HOPX during normal growth yet expression is dramatically increased during osteogenesis and adipogenesis in agreement with previous observations 8 . Studies of homozygous mutations of the Hopx gene (loss-of-function mutations) in mouse showed partial penetrant embryonic lethality due to heart deformation during embryo development. However, those that survived display no gross deformities. Hopx heterozygous mutated mice are viable and comparable to wild type mice 5,6 . This suggests that Hopx is important for cardiac development. On the other hand, the incomplete penetrance of HOPX mutation indicates that there are other compensatory mechanisms that rescue part of the phenotype. However, no bone or fat-associated phenotypes have been reported in Hopx knockout studies. Functional studies using siRNA-mediated knockdown of HOPX did not affect BMSC osteogenesis but did alter the cellular proliferation and adipogenic potential of the cells. Our studies showed that siRNA-mediated HOPX knockdown in human BMSC decreased proliferation and increased adipogenic potential of these cells. This was demonstrated by an increase of lipid formation and increased expression of early adipogenic marker C/EBPα and mature marker ADIPSIN, when compared to the siRNA scramble controls. Conversely, these results were confirmed by enforced expression of HOPX in BMSC using retroviral transduction. HOPX overexpressing BMSC demonstrated decreased lipid formation and decreased expression of adipogenic associated markers. Our data suggest that HOPX is a novel molecular inhibitor of BMSC adipogenesis, which may have implications in the regulation of fat metabolism. Furthermore, HOPX overexpression or knockdown studies, failed to demonstrate any effects on the osteogenic potential of BMSC. We predict that HOPX acts by inhibiting adipogenesis via suppression of C/EBPα by potentially binding to adipogenic suppressor proteins that act on its promoter region as a complex. Given that EZH2 inhibits BMSC osteogenic differentiation but allows adipogenesis to proceed 11,15 , implicates HOPX as a potential counter balance to regulate BMSC adipogenesis.
HOPX is known to repress transcription by direct interaction with co-repressors such as HDAC2, which consequently inactivate GATA6/Wnt7 pathway important in development and differentiation 7 . However, conflicting data in the literature demonstrate the duo-functions of HOPX in promoting and inhibiting proliferation and differentiation at different developmental stages, suggesting the importance of HOPX in maintaining the balance between growth and differentiation in various tissues based on in vitro and in vivo systems 5,6 . Our data suggests that in humans, HOPX is likely to play a role in fat metabolism. vector only (Vector) BMSC were cultured in either control (Ctrl) or osteogenic inductive (Osteo) conditions and HOPX expression levels were determined relative to β-ACTIN using qPCR, n = 6 donors. Error bars represent mean ± S.E.M, One-way ANOVA p < 0.05(*). (B) Extracellular calcium levels were quantitated and normalized to total DNA content per well, n = 3 donors. (C) Vector control and HOPX OE BMSC stained with Alizarin red. Total RNA was harvested at 7-14 days post induction (n = 6 donors) from Vector and HOPX OE BMSC. Gene expression levels were measured by qPCR for (D) HOPX, (E) RUNX2, (F) OPN relative to β-ACTIN. (G) siScram, siHOPX1 and siHOPX2 BMSC were incubated in control (Ctrl) or osteogenic inductive (Osteo) conditions for 3 weeks, and HOPX expression levels were determined relative to β-ACTIN using qPCR, n = 8 donors. Extracellular calcium levels were quantitated in siScram, (H) siHOPX1 and (I) siHOPX2 BMSC and normalized to total DNA content per well, n = 4 donors. (J, K) (I, III) siScram, (II) siHOPX1 and (IV) siHOPX2 BMSC were stained with Alizarin red. Representative of one donor is shown. Error bars represent mean ± S.E.M, Student's t-test p < 0.05 (*), n.s. represents non-significant. Scale bar (20 μm). www.nature.com/scientificreports/ www.nature.com/scientificreports/ In bone marrow, the differentiation of MSC into osteoblasts and adipocytes is competitively balanced. The commitment of BMSC to the adipogenic lineage may result in increased adipocyte formation and decreased osteoblast numbers as observed in age-related bone loss 30 . Numerous in vitro experiments performed on BMSC have revealed various factors that promote adipocyte formation inhibit osteogenesis, and conversely, many factors that promote osteoblast formation inhibit adipogenesis 31,32 . This occurs through the interaction between different signaling pathways such as Wnt, Bmp, TGF-β, Notch, mTOR [33][34][35][36][37] . Previous findings implicate the Bmp/Wnt signaling pathways in regulating HOPX family members 29 . Inhibition of HOPX in mouse and zebrafish results in disruption of cardiac development and lethality. HOPX is found to be expressed in cardiomyoblasts, which interacts physically with activated Smad4 and functions to coordinate local Bmp signals to inhibit Wnt pathway, promoting cardiomyogenesis 29 . However, little is known about the biological function of HOPX in BMSC during postnatal skeletal development and homeostasis.
In order to identify novel HOPX target genes during BMSC adipogenesis, RNA-seq analysis was performed on HOPX overexpressing and vector control BMSC cultured under normal growth or adipogenic inductive condition for 2 weeks. Differentially expressed genes were identified between normal growth and adipogenic inductive conditions. Survey of the literature identified 188 genes involved in EMT, 185 genes in adipogenesis and 127 genes in fatty acid metabolism. To identify possible signaling or molecular pathways involved in HOPX signaling, gene ontology (GO) enrichment analysis was performed. A heatmap was constructed according to the fold change of gene expression between HOPX overexpressing and vector control BMSC cultured under either normal growth or adipogenic conditions. Many of the top 50 differentially expressed genes were found to be associated with adipogenesis such as ADIPOQ, FABP4, PLIN1 and PLIN4, which generally showed a negative correlation with HOPX expression.
ADIPOQ is a cytokine secreted in various tissues including BMSC 38 . Adiponectin signals through its cell surface receptors adipoR1 (adiponectin receptor 1) and adipoR2 (adiponectin receptor 2) and can act in either endocrine, paracrine or autocrine pathway 39,40 . Upon ligand binding, distinct signaling pathways are initiated across tissues including PPARα, mTOR, AMPK [41][42][43] . On the other hand, the downstream signaling of adipoR1 can stimulate oxidative phosphorylation, which subsequently increases cell differentiation via suppression of the Wnt inhibitor, sclerostin 44,45 . Therefore, suppression of ADIPOQ by HOPX leads to termination of various pro-adipogenic signaling pathways and results in decreased adipogenic potential of BMSC.
Interestingly, CNN1 gene expression was increased in HOPX overexpressing BMSC compared to vector control BMSC, suggesting that CNN1 is positively regulated by HOPX. CNN1 is an actin binding protein (ABP) that regulates the dynamics of actin cytoskeleton by direct/indirect participating in the assembly/disassembly of actin filament, which in turn regulates the cell contraction and movement 46 . CNN1 has been shown to play a role in bone homeostasis, where high expression of CNN1 leads to delayed bone formation and decreased bone mass 47,48 . CNN1 is known to interact directly with activated or inactivated Smad1/5/8 protein and inhibit Bmp2-Smad1/5/8 signaling 49 . Although the function of CNN1 in the regulation of fat metabolism is unknown, it is involved in the Bmp/Smad pathway, which is a critical pathway in the crosstalk between BMSC osteogenesis and adipogenesis. However, more studies are needed to determine the effects of HOPX on the 'stemness' state of BMSC, and whether HOPX is dysregulated during skeletal aging and bone disease in vivo, which are often associated with increased marrow adipogenesis at the expense of bone formation.
Collectively, our findings suggest that HOPX promotes human BMSC proliferation and inhibits adipogenesis, and this is the first ever finding showing the importance of the HOPX in human BMSC self-renewal and cell fate determination as a possible counter balance to EZH2 function ( Supplementary Fig. 2), which normally represses HOPX gene expression in BMSC under normal growth conditions. HOPX appears to act by inhibiting BMSC adipogenesis via suppression of C/EBPα and potentially through co-factors binding to adipogenic suppressor proteins. Our future studies will employ co-immunoprecipitation and ChIPseq analyses to identify putative Table 1. Gene ontology annotations of differentially expressed genes from RNA-seq analysis of HOPX overexpressing and Vector only BMSC cultured under normal growth or adipogenic conditions. www.nature.com/scientificreports/ binding partners/co-factors of HOPX, the genome wide binding sites of HOPX protein complexes and the role of putative HOPX targets in human BMSC growth and lineage determination. This study lays the foundation for further research into the role of the homeobox family members in BMSC biology and fat metabolism.