The RHOX Homeodomain Proteins Regulate the Expression of Insulin and Other Metabolic Regulators in the Testis*

Background: RHOX5 is a transcription factor important for spermatogenesis. Results: Using Rhox5-null mice, we demonstrate that RHOX5 regulates key metabolism genes, including insulin (Ins), resistin (Retn), and adiponectin (Adipoq) in specific cell types and developmental stages in the testis. Conclusion: RHOX5 is a crucial regulator of metabolism genes in the testis. Significance: This work provides a framework for understanding why metabolism defects cause male infertility. Defects in cellular metabolism have been widely implicated in causing male infertility, but there has been little progress in understanding the underlying mechanism. Here we report that several key metabolism genes are regulated in the testis by Rhox5, the founding member of a large X-linked homeobox gene cluster. Among these Rhox5-regulated genes are insulin 2 (Ins2), resistin (Retn), and adiponectin (Adipoq), all of which encode secreted proteins that have profound and wide-ranging effects on cellular metabolism. The ability of Rhox5 to regulate their levels in the testis has the potential to dictate metabolism locally in this organ, given the existence of the blood-testes barrier. We demonstrate that Ins2 is a direct target of Rhox5 in Sertoli cells, and we show that this regulation is physiologically significant, because Rhox5-null mice fail to up-regulate Ins2 expression during the first wave of spermatogenesis and have insulin-signaling defects. We identify other Rhox family members that induce Ins2 transcription, define protein domains and homeodomain amino acid residues crucial for this property, and demonstrate that this regulation is conserved. Rhox5-null mice also exhibit altered expression of other metabolism genes, including those encoding the master transcriptional regulators of metabolism, PPARG and PPARGC1A, as well as SCD1, the rate-limiting enzyme for fatty acid metabolism. These results, coupled with the known roles of RHOX5 and its target metabolism genes in spermatogenesis in vivo, lead us to propose a model in which RHOX5 is a central transcription factor that promotes the survival of male germ cells via its effects on cellular metabolism.

Normal energy metabolism is critical for male fertility. Perturbations in glucose metabolism cause defects in sperm devel-opment in experimental animals (1), and diabetes is strongly associated with abnormal spermatogenesis in humans (2,3). Indeed, evidence suggests that diabetes perturbs fertility in men of reproductive age (4), and human male infertility is linked to poor semen quality resulting from hypoglycemia-induced oxidative stress damage to sperm. Direct evidence for a role of glucose metabolism in male fertility is the finding that streptozotocin-induced disruption of pancreatic beta cell function reduces the fertility of male mice (5). Further support for this notion comes from analysis of Akita mutant mice, which exhibit a defect in insulin production and are hypofertile (5,6). In addition to roles in spermatogenesis, glucose metabolism is also important for other aspects of reproduction. For example, sperm must utilize glucose efficiently in order for them to fertilize eggs (7,8).
A key cell type that controls metabolic events crucial for spermatogenesis is the Sertoli cell. The Sertoli cell maintains tight associations with germ cells throughout their development and thereby provides the structural framework, nutritional support, cofactors, and an immune-privileged barrier that establishes the critical germ line stem cell niche required for successful spermatogenesis (9). Indeed, the proliferation, differentiation, and survival of germ cells depends on the delivery of amino acids, lipids, carbohydrates, vitamins, and metabolic cofactors produced in Sertoli cells (10). Sertoli cells are also essential for the proper regulation of glucose metabolism during spermatogenesis, both for the production of lactate, the preferred energy substrate for spermatocytes and spermatids, and for aerobic glycolysis, which is required for spermiation and capacitation (11). Recently, it has come to light that Sertoli cells also produce a plethora of metabolic hormones, including insulin, adiponectin, and resistin (12)(13)(14). Sertoli cells secrete these hormones into the adluminal space, allowing them to act on germ cells that are protected from the general circulatory system as a result of the blood-testis barrier. This barrier normally excludes molecules greater than 1 kDa (15), thereby pre-venting most pancreas-and adipocyte-derived hormones from germ cells within the adluminal compartment of seminiferous tubules. In addition to generating hormones and cytokines that influence cellular metabolism, Sertoli cells respond to such secreted proteins. For example, treatment of Sertoli cells with insulin and testosterone increases the production of acetate, which evidence suggests is important to maintain the synthesis of lipids and their byproducts, which, in turn, are essential for the normal development of germ cells (16). Conversely, depriving Sertoli cells of insulin results in down-regulation of several metabolism-associated genes, including LDHA and MCT4, as well as alterations in GLUT transporter levels, leading to shifts in lactate secretion and glucose uptake (17).
Although the roles of metabolism in male fertility are becoming well established, the regulatory events that act upstream of the metabolic enzymes and other metabolism factors in the male reproductive tract are poorly understood. In this paper, we report evidence that a homeobox transcription factor highly expressed in Sertoli cells, RHOX5, is a regulator of several genes encoding key metabolic factors. Homeobox genes encode evolutionarily conserved transcription factors harboring a 60amino acid DNA-binding motif called a homeodomain. Rhox5 is part of a large X-linked homeobox gene cluster that harbors at least 33 genes and thus is the largest homeobox gene cluster so far identified in any species (18 -24). All of the members of the Rhox gene cluster are selectively expressed in testis, epididymis, ovary, and/or placenta, suggesting that they encode a large set of transcription factors devoted to regulating and supporting fertility (19 -23). This role is likely to be conserved in other mammals, because the rat and human RHOX genes are also selectively expressed in reproductive tissues (19,(25)(26)(27)(28)(29)(30). Targeted deletion of the founding member of the Rhox cluster, Rhox5, in mice results in male subfertility, marked by increased germ cell apoptosis, reduced sperm number, and a reduced proportion of sperm with normal motility (18,19,28,31). Because RHOX5 is not detectably expressed in germ cells but instead is expressed in the directly adjacent Sertoli nurse cells (32)(33)(34)(35), we have previously hypothesized that the increased germ cell apoptosis in Rhox5-null mice is most likely due to the loss of one or more Sertoli cell-expressed RHOX5-dependent survival factor (31,36).
In this paper, we identify INS2 (insulin 2) as a candidate to be such a RHOX5-regulated germ cell survival factor. We demonstrate that the Ins2 gene and insulin protein are expressed in the testis in a developmentally regulated manner and that Rhox5null mice have a dramatic defect in their expression in the testes. We showed that Ins2 is a direct target of RHOX5 and identified specific RHOX5 domains and amino acids required for Ins2 induction. As part of this analysis, we identified other RHOX family members that share with RHOX5 the ability to induce Ins2 and determined the specific domains and homeodomain residues dictating this ability. We showed that this is a conserved response, and we showed that Rhox5-null testes have defects in insulin signaling and increased germ cell apoptosis during the postnatal period when Ins2 is expressed. Last, we showed that Rhox5 regulates several other metabolism genes, including those that either promote or antagonize insulin action. Together, our analysis suggests that Rhox5-null mice may be a useful model for providing insights into the link between metabolic syndromes and male infertility.

EXPERIMENTAL PROCEDURES
Plasmids and RHOX Antibodies-Full-length and domainspecific RHOX expression constructs were amplified by PCR with the primer sets described in Table 1 and cloned into the pIRES-hrGFP2a vector (Stratagene, La Jolla, CA). Expression vectors encoding mouse and human RHOX factors have been described previously (37). The Ins2 5Ј-flanking sequences in the pGL3-Ins2 promoter construct (G-507) (38) were shortened by deletion PCR and restriction enzyme digestion (39). Site-directed mutagenesis of both the Ins2 promoter and Rhox gene constructs was performed as described previously (40). The polyclonal rabbit antisera against RHOX peptides were kindly provided by IMGENEX (San Diego, CA). Insulin signaling was assessed by Western blot analysis using the Cell Signaling Technology Phospho-Akt antibody sampler kit (catalog no. 9916), GSK3␤ 27C10 (catalog no. 9315), and PTEN (phosphatase and tensin homolog) (catalog no. 9552) under the recommended conditions. Analysis was performed on total testes lysate from four individual WT or Rhox5-null post-partum day 16 (P16) 2 animals, assayed in duplicate.
Cell Culture, Transfection, and Cell Fractionation-The 15P-1, TM3, and TM4 cell lines were obtained from ATCC and maintained in DMEM supplemented with penicillin/streptomycin and 10% FBS. Cells were grown at 37°C in a 5% CO 2 atmosphere and split when ϳ80% confluent. Plasmid DNA concentrations were independently determined using spectrophotometry and analytical gel electrophoresis. Cells were transfected with 1-3 g of plasmid DNA (pGEMT was used as filler to normalize DNA mass) using Lipofectamine 2000 (Invitrogen), following the manufacturer's protocol. Cells were harvested 36 h post-transfection to measure them for luciferase activity using the Dual-Luciferase Assay (Promega, Madison, WI), following the manufacturer's instructions. For protein expression vectors, equivalent transfection efficiency and plasmid activity were monitored by expression of GFP from an internal ribosome entry site polycistronic message downstream of the Rhox protein expression cassette (37). The data presented in the figures are the mean Ϯ S.E. of at least three independent transfection experiments. Data from transient transfection assays were statistically analyzed by analysis of variance, and differences between individual means were tested by a Tukey multiple-range test using Prism version 4.0 (GraphPad Software). Differences were defined as significant if the p value was less than 0.05. We purified Sertoli cells and Leydig cells from P12 testes using a protocol involving sequential trypsin, collagenase, and hyaluronidase digestion, followed by hypotonic shock and differential gravity sedimentation (19,43). The purity of cell fractions was assessed by quantitative polymerase chain reaction (qPCR) analysis for established Leydig, Sertoli, and germ cell markers. Somatic cell preparations that exhibited significant contamination with germ cell markers were discarded and excluded from the analysis.
RNA Purification, qPCR, and EMSA-Total cellular RNA was isolated as described previously (19,41). Reverse transcription-PCR analysis was performed by first generating cDNA from 1 g of total cellular RNA using iSCRIPT (Bio-Rad) and performing qPCR analysis using SYBR Green incorporation and the ⌬⌬Ct method (with ribosomal L19 for normalization), as described previously (19). For EMSA, 32 P-labeled blunt-end double-stranded probes (5 ϫ 10 4 cpm) were incubated for 30 min at room temperature in 20 l of binding buffer (10 mM Tris (pH 7.9), 50 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, 5% glycerol, and 1 g of poly(dI:dC)) containing 2 g of mouse testes extract prepared as described previously (42). For anti- body supershift and blocking assays, the reaction mixtures were preincubated with polyclonal antisera at room temperature for 20 min before the addition of hot probe. The DNA-protein complexes were resolved in 4.5-5% nondenaturing polyacrylamide gels at 150 V for 3-4 h at 4°C. Microarray Analysis-Total cellular RNA prepared from the testes of six P12 mice (three Rhox5-null and three control) was treated with DNase and provided to the University of Iowa DNA Core Facility for quality testing using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA) and microarray analysis (20 g/mouse). RNA was processed for microarray analysis by using the Affymetrix GeneChip onecycle target labeling kit (Affymetrix, Santa Clara, CA), following the manufacturer's recommended protocols. The resultant biotinylated cRNA was fragmented and then hybridized to the Affymetrix murine genome MG-U74A array, which contains probes for 12,489 genes and expressed sequence tags. The arrays were washed, stained, and scanned using the Affymetrix model 450 Fluidics Station and the Affymetrix Model 3000 scanner using the manufacturer's recommended protocols. Expression values were generated using Microarray Suite version 5.0 software (Affymetrix) and PartekGS (Partek Inc., St. Louis, MO). The relative amount of transcript was determined by comparing the perfect match and mismatch signals from all probe pairs, which generates a weighted mean value, relatively insensitive to outlier values, calculated using the one-step Tukey biweight estimate. Determination of whether the difference in transcript intensity between two arrays was statistically significant was accomplished using the Wilcoxon signed rank test. A complete list of differentially expressed genes and gene ontology enrichment analysis in Rhox5-null mice is available upon request.
ChIP Analysis-ChIP assays were carried out according to the manufacturer's instructions (Upstate Biotechnology, Inc.) with some modifications as described previously (44,45). Briefly, ϳ10 6 15P-1 cells (ϳ70% confluent in a 100-mm dish) were transfected with 1 g of the FLAG-Rhox5 expression construct (RHOX5, R-177). After 36 h, the cells were incubated with 1% formaldehyde for 10 min at room temperature, followed by incubation with 0.1 M glycine for 5 min at 37°C to stop the fixation. The cells were then washed three times with cold PBS containing 1 mM PMSF and 1% protease inhibitor mixture (Sigma Inc.), harvested, and lysed with cell lysis buffer (5 mM PIPES (pH 8.0), 85 mM KCl, and 0.5% Nonidet P-40) at 4°C for 10 min. The debris was pelleted, and the supernatant was resuspended in nuclear lysis buffer (50 mM Tris-Cl, pH 8.0, 10 mM EDTA, and 1% SDS) for 10 min, and the DNA was sheared by sonication. The chromatin preparations were precleared by incubating with a Protein A-agarose/salmon-sperm DNA slurry (Upstate) and then immunoprecipitated overnight at 4°C with an antiserum against FLAG (Sigma). Protein A-agarose/salmon-sperm DNA slurry was added to immunoprecipitate the antibody-protein complexes, and these complexes were washed following the manufacturer's instructions. Immune complexes were eluted with a buffer containing 1% SDS and 0.1 M NaHCO 3 . Reverse cross-linking was performed by incubating with 200 mM NaCl at 65°C for 4 h. Samples were proteinase K-treated for 1 h, and the DNA was purified by phenol/chloro-form extraction, ethanol-precipitated, and resuspended in 30 l of H 2 O. PCR amplification was done with 3 l of each DNA sample, 200 M dNTPs, 1.5 mM MgCl 2 , 1 unit of AmpliTaq DNA polymerase (Roche Applied Science), 1ϫ buffer (supplied), and 0.2 M of each primer. The percentage of DNA immunoprecipitated was quantified by dividing the immunoprecipitated signal over the input signal (measured using dilutions of the input) detected as ethidium bromide-stained bands on agarose gels.

Identification of Genes Exhibiting Altered Expression in
Rhox5-null Testes-We used expression profiling to identify genes differentially expressed in Rhox5-null versus control mice. We chose P12 testes for microarray analysis because the levels of Rhox5 mRNA and protein are high at this time point of testes development (19,33). Furthermore, testes at this postnatal stage contain a high proportion (ϳ40%) of Sertoli cells, the cell type that expresses Rhox5 (46). A robust multichip matrix comparison analysis of three Rhox5-null and three control testis samples revealed that 316 of the known genes and 59 of the expressed sequence tags exhibited Ն2-fold altered expression that was statistically significant (p Ͻ 0.05) when examined across all nine possible Rhox5-null versus control data set comparisons. The altered expression of a subset of these genes was confirmed by qPCR using six different P12 testes samples for each genotype (Table 2).

TABLE 2 Genes exhibiting altered expression in Rhox5-null testes
The "Array ⌬" column indicates the average change in the mRNA level of selected genes in P12 testes from Rhox5-null mice relative to littermate control mice, as judged by microarray analysis. A negative value means the gene was down-regulated in response to loss of Rhox5. The values are based on robust multichip analysis of three Rhox5-null and three control testis samples, as described under "Experimental Procedures." The "qPCR ⌬" column provides analogous data from qPCR analysis of six additional sets of Rhox5-null and control testes assayed two independent times in duplicate. In most cases, the -fold change in mRNA level determined by qPCR was remarkably close to that determined by microarray analysis. A complete list of all differentially expressed genes and expressed sequence tag sequences identified by our microarray analysis is available upon request. Developmental Expression of RHOX5-regulated Metabolism Genes in the Postnatal Testis-Inspection of the list of genes exhibiting altered expression in Rhox5-null testes revealed that the most highly down-regulated gene was the insulin gene Ins2 (Table 2). We were surprised that Ins2 is expressed in the testes, because it, along with its paralog, Ins1, both of which encode proteins with identical known functions, are best known for their expression in the pancreas (47,48). During the process of preparing this manuscript, confirmation that Ins2 is expressed in the testes was reported (14). We also found that Rhox5-null mice testes exhibited altered expression of two other genes encoding secreted metabolic regulators: RETN and ADIPOQ ( Table 2). Both of these secreted metabolic regulators were initially thought to be exclusively expressed by adipocytes but were recently shown to be expressed by Sertoli and Leydig cells in the testes (12,13). Rhox5-null mice also expressed altered levels of two genes encoding two master transcriptional regulators of metabolism ( Table 2). One of these is Pparg (peroxisome proliferator-activated receptor-␥), a nuclear hormone receptor that regulates the transcription of a large battery of genes involved in metabolism (49). The other is Ppargc1a (PPARG coactivator 1␣), a transcriptional coactivator that regulates the expression of a large subset of genes that together promote energy expenditure (50) and modulate the function of steroid receptors (51). Last, we found that Rhox5-null mice expressed altered levels of Scd1 (stearoyl-CoA desaturase-1) mRNA (Table 2), which encodes the rate-limiting enzyme for the synthesis of monounsaturated fatty acids (52).

Gene
To determine the developmental context in which Rhox5 regulates these metabolism genes, we compared their expression during postnatal development in Rhox5-null and control mice testes. Fig. 1 shows relative mRNA expression levels (determined by qPCR analysis, expressed as -fold level above background for each primer pair) in testis from P7-P25 mice. This analysis showed that the two metabolism genes identified by microarray analysis as positively regulated by Rhox5 at P12 (Ins2 and Ppargc1a; Table 2) were also positively regulated by Rhox5 at other early postnatal time points, as indicated by their reduced expression in Rhox5-null mice testes between P7 and P16. Together with the finding that Ins2 and Ppargc1a exhibit a temporal expression pattern in wild-type mice that is strikingly similar to that of Rhox5, this strongly suggests that RHOX5 has a major role in dictating the developmental pattern of expression of these two genes during the first wave of spermatogenesis. The three metabolism genes identified by microarray analysis as negatively regulated by Rhox5 at P12 (Retn, Scd1, and Adipoq; Table 2) were also negatively regulated by Rhox5 at most other postnatal time points, as indicated by their up-regulated expression in Rhox5-null mice. Retn and Scd1 mRNA levels were most elevated in Rhox5-null mice testes at the time points when Rhox5 is normally most highly expressed, suggesting that RHOX5 has a major role in shaping their developmental expression pattern at these postnatal time points. Note that we detected some Rhox5 mRNA in Rhox5-null mice testes (ϳ5% of wild type), as expected given that these mice have an insertion in exon 6 that disrupts the homeodomain and probably destabilizes the mRNA but would not be expected to completely prevent mRNA accumulation (33).
To assess which of these metabolism genes may be direct targets of Rhox5, we determined whether they are expressed in Sertoli cells, the site of Rhox5 expression (32,35). To do this, we purified Sertoli cells from P12 testes using a protocol involving gravity sedimentation and enzymatic treatments (19,43). This procedure also generates purified interstitial (mainly Leydig) cells, which allowed us to also determine whether this somatic cell subset expresses any of the Rhox5-regulated metabolism genes. The Sertoli cell and interstitial cell fractions were each greater than 95% pure, based on microscopic morphological assessment of cell smears, the expression of the Leydig cell marker Cyp11a1 (53), and the Sertoli cell markers Gata1 and Rhox5, and the absence of the germ cell marker Adam2 (32,33,35). qPCR analysis of the two metabolism genes positively regulated by RHOX5 (Ins2 and Ppargc1a) indicated that although they were expressed in both the Sertoli and interstitial cell fractions, RHOX5 significantly up-regulated their expression only in the Sertoli cell fraction (Fig. 2). This indicates that these two genes are candidates to be direct RHOX5 targets. The regulation of the other three metabolism genes was more complex, because we found that the loss of Rhox5 significantly altered their expression in both the Sertoli and interstitial cell fractions. Scd1 was up-regulated in both cell fractions in Rhox5-null testes, whereas Retn and Adipoq were up-regulated in the interstitial cell fraction but down-regulated in the Sertoli cell fraction (Fig. 2). These results indicate that although these three metabolism genes have the potential to be direct RHOX5 targets (in Sertoli cells), they must also be indirectly regulated by RHOX5 (in interstitial cells) (see "Discussion").
Ins2 Is a Direct Target of RHOX5-The finding that Ins2 is regulated by Rhox5 specifically in Sertoli cells (Fig. 2) indicated that it was a good candidate to be directly regulated by Rhox5. As a first step toward addressing this question, we sought to identify Sertoli cell lines that exhibit increased Ins2 gene expression in response to Rhox5. We found that the 15P-1 and TM4 Sertoli cell lines, neither of which express detectable levels of Rhox5 mRNA (26), responded to forced Rhox5 expression by dramatically up-regulating Ins2 mRNA levels (Fig. 3A). Forced expression of either rat or mouse RHOX5 up-regulated Ins2 mRNA levels, indicating that this is a conserved response.
To determine whether this regulation is conferred via a RHOX5-responsive element in the Ins2 promoter, we transfected a luciferase reporter construct harboring 375 nt of Ins2 5Ј-flanking sequence (38) into both Sertoli cell lines. Luciferase expression from this reporter was increased in response to forced mouse and rat RHOX5 expression in both cell lines (Fig.   3B), indicating that the Ins2 promoter harbors a RHOX5-responsive element. To identify the specific Ins2 promoter sequences responsible for RHOX5 inducibility, we generated deletion promoter constructs and analyzed their ability to respond to RHOX5. We used TM4 cells for these studies, because we found that this cell line exhibited higher transient transfection efficiency than did 15P-1 cells. Deletions leaving as little as 103 nt of Ins2 5Ј-flanking sequence (Ϫ103) had no significant effect on either RHOX5-mediated induction or basal promoter activity, whereas deletions to Ϫ72 abolished RHOX5-mediated induction, and deletion to Ϫ18 abolished basal transcriptional activity (Fig. 3C). The basal transcription region defined by this analysis (between Ϫ18 and Ϫ50) had consensus-binding sites for the testis-determining factor SRY and the ubiquitously expressed transcription factors SP1 and MZF1 (Fig. 4). The region responsible for RHOX5 responsiveness (between Ϫ72 and Ϫ103) contained the sequence CCT-TAATGG, which was previously shown to bind the homeobox transcription factors CDX1 and NKX2 (the core element bound by most homeobox proteins is underlined) (54,55). To test whether this consensus site is a RHOX5-responsive element, we mutated two nucleotides in its TAAT core. This mutation abolished RHOX5-mediated induction of Ins2-driven reporter expression (Fig. 3D). Transcriptional induction was similarly reduced regardless of whether the mutation was introduced in the Ϫ375, Ϫ260, or Ϫ103 construct, indicating that this site is not redundant with other putative homeobox binding sites in the Rhox5 promoter (such as the TCTAATTA sequence between Ϫ205 and Ϫ196; Fig. 4).
The finding that a homeobox consensus-binding site in the Ins2 promoter is required for RHOX5-mediated inducibility strongly suggested that Ins2 is a direct target of RHOX5. To more definitively test this possibility, we used electromobility shift analysis (EMSA) to determine whether RHOX5 binds to the Ins2 RHOX5-responsive element. Incubation of a 22-nt radiolabeled probe corresponding to this Ins2 region with total testis protein extracts produced two bands that were both dramatically reduced in level by wild-type (WT) cold competitor probe (Fig. 5A, lane 3). While the two bands were accompanied by a smear, the two bands were reproducibly present. A mutant probe harboring the 2-nt mutation that eliminated RHOX5 responsiveness in transfected cells (Fig. 3D) did not detectably form the lower band, and it generated much reduced levels of the upper band (Fig. 5A, lane 2). In addition, co-incubation of a cold version of this mutant probe with labeled wild-type probe did not inhibit the formation of these two bands (Fig. 5A, lane  4). A polyclonal antiserum raised against a mouse RHOX5-GST fusion protein (56) inhibited complex formation when incubated with testis extract for 20 min prior to the addition of probe (Fig. 5A, lane 7). The RHOX5 antiserum shifted the migration of both bands when antibody was added after stable complexes were allowed to form (Fig. 5A, lane 8). Control (anti-GST) antiserum did not significantly alter band migration, whether it was added before or after complex formation (Fig.  5B, lanes 5 and 9, respectively). Collectively, these results suggest that RHOX5 forms two protein complexes that specifically bind to the RHOX5-responsive element in vitro.
To determine whether RHOX5 interacts with the Ins2 promoter in cells, we transiently transfected a plasmid encoding FLAG-tagged RHOX5 into the 15P-1 Sertoli cell line and examined binding of this tagged version of RHOX5 to the endogenous Ins2 promoter using chromatin immunoprecipitation (ChIP) analysis. This analysis showed that RHOX5 was recruited to the Ins2 promoter, whereas a negative control, the RNA-binding protein UPF3B fused to FLAG (57), was not (Fig.  5B). As a further control for specificity, we used primers specific for a region ϳ4 kb upstream of the Ins2 promoter and observed a signal only with input DNA, not anti-FLAG-immunoprecipitated DNA (Fig. 5B). These results demonstrate that RHOX5 is specifically recruited to the Ins2 promoter region in Sertoli cells.
Homeodomain and Amino-terminal Domains Specifying Ins2 Transcriptional Induction-To define the domains in RHOX5 required for it to induce Ins2 expression, we generated and analyzed RHOX5 mutants (Fig. 6A). A mutant lacking the car-FIGURE 2. Cell type-specific regulation of metabolism genes. qPCR analysis of total cellular RNA from Sertoli and interstitial cell fractions purified from P12 Rhox5-null (KO) and control (WT) testes, quantified and normalized as described in the legend to Fig. 1 from four independent experiments. Gata1 and Cyp11a1 serve as markers for Sertoli cells and Leydig cells, respectively. We assessed potential germ cell contamination using the germ cell-specific marker, Adam2, which we found was expressed Ͼ21.2 Ϯ 3.4-fold higher in the total testis input than in the enriched Leydig and Sertoli cell fractions (only a background signal was detected in these latter two samples; data not shown). Comparison of mean expression levels between WT and KO samples was performed using unpaired Student's t test. Different letters denote mean expression levels significantly different (p Ͻ 0.001) from each other and those not marked. Error bars, S.E.
boxyl-terminal domain of RHOX5 exhibited Ins2-inducing activity, albeit less than that of full-length RHOX5, when tested for its ability to up-regulate either Ins2 promoter-driven luciferase reporter activity (Fig. 6A) or the endogenous Ins2 promoter in two Sertoli cell lines (Fig. 6B). This indicated that the carboxyl-terminal domain (CTD) of RHOX5 promotes its transcriptional activity but is not absolutely required for this action. In contrast, a mutant lacking not only the CTD but also the homeodomain did not significantly up-regulate either Ins2 promoter-driven reporter expression or endogenous Ins2 expression (Fig. 6, A and B). This indicated that the homeodomain is required for RHOX5 activity, as expected. Both of these mutants were expressed at similar levels as the full-length RHOX5, but unlike full-length RHOX5, which was primarily nuclear, the amino-terminal domain (ATD) mutant proteins accumulated in both the nucleus and the cytoplasm (Fig. 7). We also generated constructs designed to express RHOX5 mutants lacking either the ATD or both the ATD and CTD but found that their steady-state levels were very low, as assessed by immunofluorescence, suggesting that they were unstable; neither of these mutants measurably up-regulated either Ins2 promoter-driven reporter expression or endogenous Ins2 expression (data not shown).
We next addressed whether other RHOX family members besides RHOX5 are able to induce Ins2 transcription. To do this, we used expression vectors containing the nine other known rat Rhox genes, all of which are expressed at similar levels in transfected cells (28,37,45). We chose to generate expression vectors encoding rat Rhox genes in order to match the rat Ins2 promoter-reporter construct used for our transfection analyses. We found that only RHOX8 and RHOX11 were able to induce Ins2 promoter activity, albeit more weakly than RHOX5 (Fig. 6C). RHOX8 and RHOX11 also up-regulated the endogenous mouse Ins2 promoter in the 15P-1 Sertoli cell line (Fig. 6D). Mouse Ins2 expression was up-regulated by both rat and mouse versions of these two RHOX proteins, which differ  Fig. 1, using total cellular RNA prepared from the indicated Sertoli cell lines transfected with an expression vector encoding the indicated species of RHOX5 or the empty pCMV expression vector (Ϫ). B, luciferase activity from the indicated cell lines cotransfected with (i) an Ins2 reporter containing 375 bp of rat Ins2 5Ј-flanking sequence driving firefly luciferase (38), (ii) the Renilla luciferase vector pRL-Tk as an internal control, and (iii) the pCMV-Rhox5 expression vector encoding RHOX5 from the indicated species or the empty pCMV expression vector (Ϫ). The luciferase activities shown are -fold above that of a promoterless firefly luciferase vector (pGL3-basic) from cell extracts from triplicate wells from at least six separate transfections. Prior to cell lysis, equivalent transfection efficiency was monitored by examining the level of GFP, which is expressed from the same transcript encoding RHOX5 in the pCMV vector. C, luciferase activity from TM4 cells cotransfected with (i) the rat RHOX5 expression vector or empty vector used in B and (ii) the indicated Ins2 promoter deletion constructs (length of 5Ј-flanking sequence present is shown), performed and quantified as in B. The Ϫ375 construct is the same as that used for B. D, luciferase activity of TM4 cells cotransfected as in C with Ins2 promoter constructs containing the indicated length of 5Ј-flanking sequence with (mutant: ccAT) or without (WT: TAAT) a mutation in the TAAT core homeobox consensus site in the Ϫ103 to Ϫ72 nt region, analyzed and quantified as in C. Intergroup comparisons were performed by unpaired Student's t test; *, p Ͻ 0.001.   (38). Gray boxes indicate putative transcription factor binding sites as determined by TFSEARCH version 1.3. The region responsible for RHOX5 responsiveness (between Ϫ72 and Ϫ103) contains the sequence CCTTAATGG, which was previously shown to bind the homeobox transcription factors CDX1 and NKX2 (54,55). The sequence used for the EMSA probe in Figs. 5A and 6E is underlined. The corresponding region of the mouse Ins2 promoter possesses 88% nucleotide identity, with 100% sequence identity in the homeobox binding site. The most notable difference is a 19-nt insertion in the mouse Ins2 promoter 3 nt upstream of the homeobox binding site.
considerably in sequence between mice and rats (28,31,58), providing evidence for conservation. Consistent with their modest ability to induce Ins2 transcription, RHOX8 and RHOX11 exhibited modest binding to the Ins2 element containing the RHOX5-binding site, as demonstrated by EMSA (Fig. 6E). We also examined the ability of human RHOX proteins to stimulate Ins2 promoter activity. There are only three human RHOX proteins, two of which are almost identical (RHOXF2 and RHOXF2B only differ by two amino acids (30,34)), and thus we only examined the activity of RHOXF1 and RHOXF2. We found that only RHOXF2, and not RHOXF1, activated the Ins2 promoter (Fig. 6F), suggesting that the conservation of RHOX-mediated regulation of testis-derived insulin extends to primates.
To understand the molecular basis for why RHOX proteins differ in their ability to induce Ins2 transcription, we made a series of constructs encoding chimeric and mutant proteins. We first determined whether the RHOX5 ATD could confer Ins2 transcriptional activity when it was substituted for the ATD in other RHOX proteins (Fig. 8A). Co-transfection analysis of expression vectors encoding RHOX5 ATD substitution chimeras corresponding to each of the RHOX proteins demonstrated that they could be divided into three functional classes: class I, those that either acquired or increased their ability to stimulate Ins2 transcription when provided with the RHOX5 ATD (RHOX8 and RHOX10); class II, those that lacked the ability to stimulate Ins2 transcription even when the RHOX5 ATD was provided (RHOX2, RHOX3, RHOX4, RHOX7, RHOX9, and RHOX12); and class III, a single RHOX protein (RHOX11) whose ability to stimulate Ins2 transcription was not significantly affected by introduction of the RHOX5 ATD in place of its own ATD (Fig. 8A). The simplest interpretation of these results is that class I RHOX proteins are capable of binding the Ins2 element but have a weak or non-functional activation domain (at least with regard to the Ins2 promoter), class II RHOX proteins are those that lack a homeodomain capable of binding to the Ins2 element, and class III RHOX proteins have the ability to both bind and transactivate the Ins2 promoter (albeit less strongly than RHOX5). A more detailed analysis of these mutants is provided under "Discussion." The third helix of the homeodomain harbors four amino acid positions responsible for making base-specific contacts with DNA (59). RHOX proteins mainly differ between each other in the first three of these residues (positions 47, 50, and 51; Fig.  8B), and thus we examined their functional role. Alteration of two of these three residues in RHOX5 so that they matched those in the non-Ins2-inducing RHOX9 protein (conversion of  5 and 9 and lanes 7  and 8, respectively. The star (lanes 8 and 9) indicates samples in which anti-bodies were added after stable complexes had been allowed to form, whereas the unlabeled lanes (lanes 5 and 7) have samples preincubated with antibody before the addition of probe. B, ChIP analysis of 15P-1 cells transfected with expression vectors encoding FLAG-RHOX5 or FLAG-UPF3B and immunoprecipitated with the antiserum indicated. The top and bottom panels show standard PCR analysis with the primers indicated. Quantification of RHOX5 recruitment to the Ins2 promoter (using primers c and d) was determined by qPCR as described above. The values shown are mean values from three different experiments. As a negative control, PCR was performed on sequences 4 kb upstream of the Ins2 promoter (using primers a and b), which has no consensus homeobox binding sites. Each bar represents the mean Ϯ S.E. (error bars) of five independent experiments. NOVEMBER 29, 2013 • VOLUME 288 • NUMBER 48

JOURNAL OF BIOLOGICAL CHEMISTRY 34817
Lys to Arg and Ile to Met at position 50 and 51, respectively; Fig.  8B) reduced the ability of RHOX5 to stimulate the Ins2 promoter activity by ϳ50% (Fig. 8C). Thus, one or both of these residues is necessary for maximal Ins2 induction by RHOX5. We next asked whether RHOX proteins with reduced ability or an inability to stimulate Ins2 transcription have non-optimal amino acids at base-specific contact residues. We first examined RHOX8, which is weaker than RHOX5 in its ability to induce Ins2 transcription (Fig. 8A) but only differs from RHOX5 in one of the base-specific contact residues (Ser at position 51; Fig. 8B). We found that conversion of this single residue to the one in RHOX5 did not significantly increase the ability of RHOX8 to stimulate Ins2 transcription (Fig. 8C), indicating that the deficiency of RHOX8 must lie elsewhere. Last, we examined RHOX10, which is not capable of significantly inducing Ins2 transcription (Fig. 8A). A substitution at position 47 to match that in the highly related Ins2-inducing RHOX11 protein (Fig. 8B), converted RHOX10 into an Ins2-inducing protein (Fig. 8C). This demonstrates that a single amino acid substitution in a homeobox protein can confer the ability to transcriptionally activate a new target gene.
Insulin Signaling Is Impaired in Rhox5-null Testes-The dramatic induction of Ins2 in response to RHOX5 during the first wave of spermatogenesis (Fig. 1) suggested the possibility that RHOX5 is essential for insulin signaling in the testes during this time period. To test this hypothesis, we examined the abundance of insulin and the phosphorylation status of molecules involved in the insulin-signaling cascade in Rhox5-null versus control P16 testes. We first examined insulin protein level and found that it was dramatically reduced by ϳ90% (p Ͻ 0.001) in Rhox5-null animals (Fig. 9A). We then examined thymoma viral proto-oncogene 1 (AKT), a protein kinase that responds to FIGURE 6. RHOX5 functional domains and RHOX family members that induce Ins2. A, luciferase activity from TM4 cells cotransfected with (i) expression plasmids encoding full-length or truncated versions of rat RHOX5 (HD, homeodomain), (ii) the Ϫ103 nt Ins2 firefly luciferase reporter (Fig. 3B), and (iii) the Renilla luciferase pRL-TK vector as an internal control, performed and quantified as in Fig. 3B from at least four separate transfections. Significance was determined by one-way analysis of variance followed by Tukey's multiple comparison post hoc test. Different letters denote mean expression levels significantly different (p Ͻ 0.001) from each other and those not marked. Error bars, S.E. B, endogenous levels of Ins2 mRNA were determined in TM4 cells as in Fig. 3A. Similar results were obtained in 15P-1 Sertoli cells (data not shown). C, luciferase activity from TM4 cells transfected and analyzed as in A except that expression plasmids encoding the indicated RHOX proteins were cotransfected with the luciferase plasmids. D, endogenous levels of Ins2 mRNA were determined in TM4 cells as in Fig. 3A. Similar results were obtained in MSC1 and 15P-1 Sertoli cells (data not shown). E, EMSA analysis performed as in Fig. 5A with the wild-type oligonucleotide probe, adult testes nuclear extracts, and antisera against the indicated RHOX proteins, normal rabbit serum (NRS), or no serum (none). The arrow indicates supershifted complexes. The lanes shown were excised from a single film exposed under identical conditions. F, luciferase activity from the indicated cell line transfected as in B except that expression plasmids encoding human RHOXF1 or RHOXF2 proteins were cotransfected with the promoter plasmids. Intergroup comparisons were performed by unpaired Student's t test; *, p Ͻ 0.01; **, p Ͻ 0.001. insulin signaling by activating multiple cell proliferative and survival pathways (60). In particular, the phosphorylation of Thr-308 in the AKT kinase domain by phosphoinositide-dependent protein kinase-1 (PDK1) and Ser-473 in the AKT hydrophobic motif by mTOR are both critical for AKT activation (61). We found that Thr-308 and Ser-473 AKT phosphorylation was reduced by ϳ90% (p Ͻ 0.001) and ϳ20% (p Ͻ 0.01), respectively, in Rhox5-null testes (Fig. 9A). This indicated that loss of Rhox5 reduced insulin-mediated signaling in the testes, and it suggested that this was mainly the result of defective signaling via PDK1-medated phosphorylation. Because PDK1 can be activated by an insulin-independent pathway through the action of PTEN, we also examined PTEN phosphorylation but observed no appreciable change in its status in Rhox5-null testes (Fig. 9A). Rhox5-dependent activation of AKT is likely to be physiologically relevant, because we found that a downstream mediator of AKT cell proliferation-promoting and prosurvival activity, GSK3␤, exhibited significantly reduced levels of phosphorylation (at Ser-9) (p Ͻ 0.001) in Rhox5-null testes. Taken together, these results indicate that a major axis of insulin receptor signaling depends on Rhox5 in the testis.
Because insulin is known to promote cell survival of many cell types, including germ cells (3), we postulated that one function of Rhox5-induced insulin signaling during postnatal development of the testes is to increase germ cell survival. Because the tools to directly determine whether insulin signaling has such a role in the testes are not available, we elected to instead determine whether ablation of Rhox5, which in turn would prevent Ins2 from being up-regulated during the first wave of spermatogenesis, had a negative consequence on germ cell survival. Using terminal dUTP nick-end labeling (TUNEL) analysis on P12 and P16 testes from Rhox5-null and control mice, we observed that loss of Rhox5 significantly increased the number of TUNEL-positive tubules (Fig. 9B) and the number of TUNEL-positive cells per positive tubule (Fig. 9C). These data clearly demonstrate that RHOX5 promotes germ cell survival during the first wave of spermatogenesis.

DISCUSSION
Although it is clear that insulin signaling and proper glucose metabolism are crucial for normal male fertility, the molecular mechanisms responsible for its regulation in Sertoli cells have not been defined (14,16,17). Our study herein begins to provide insight into this issue by identifying a transcription factor, RHOX5, that regulates several key metabolism genes in the testis. We provide several lines of evidence that a key target of RHOX5 is the Ins2 insulin gene: (i) Rhox5 directly regulates Ins2 in Sertoli cell lines (Fig. 3); (ii) Rhox5 is required for the dramatic induction of Ins2 gene expression during the first wave of spermatogenesis in vivo (Fig. 1); (iii) loss of Rhox5 prevents insulin production by Sertoli cells in vivo (Figs. 2 and 9A); and (iv) Rhox5 is required for insulin signaling in the testis in vivo (Fig. 9A). Coupled with our discovery that the Rhox5 gene also regulates the genes encoding several other key metabolism factors, including the insulin action regulators, ADIPOQ and RESISTIN, and the energy metabolism transcriptional regulators, PPARG and PPARGC1A, this suggests that RHOX5 is a key transcription factor controlling cellular metabolism events important for spermatogenesis.
Intriguingly, a mouse model exists with a crippling mutation in the Ins2 gene that leads to similar defects in male germ cell survival and male fertility as in Rhox5-null mice (19). These  NOVEMBER 29, 2013 • VOLUME 288 • NUMBER 48

RHOX5-mediated Regulation of Metabolism Genes
Akita mice have a naturally occurring point mutation in the Ins2 gene that generates a misfolded INS2 protein that is unable to activate the insulin receptor (6). Secretion of INS2 is impaired in these mice, leading to insulin accumulation in the endoplasmic reticulum. This results in stress and death of the ␤-islet cells in the pancreas, thereby causing type-1 diabetes (62). Relevant to our study, male Akita mice are subfertile and exhibit defects in meiotic progression and spermatid maturation, a decrease in testis weight, increased apoptosis, and ϳ50% decline in both sperm count and the proportion of sperm with forward motility (5,14,63). The similarity of the male reproductive defects of these insulin-deficient mice to those of Rhox5-null mice, coupled with our demonstration that RHOX5 is responsible for driving insulin expression in the testis, supports a model in which RHOX5 acts through insulin signaling to promote male germ cell survival. Because RHOX5 is an androgen receptor-and androgen-dependent transcription factor (31,64,65), it is tempting to speculate that the increased male germ cell apoptosis observed in mouse and rat models in which androgens are inhibited or deprived (66) is partly the result of loss of Rhox5 expression and the consequent loss of Ins2 expression. Because androgens tightly regulate insulin action, glucose utilization, and lactate production in cultured Sertoli cells, RHOX factors could be a key mediator between hormone stimulation and transcriptional changes (11,16).
In addition to preventing apoptosis, Rhox5 and Ins2 may actively promote cell proliferation in the testes, because both Rhox5 and Ins2 are expressed at time points during the first wave of spermatogenesis when proliferating spermatogonia are abundant and Sertoli cells have not yet terminally differentiated and hence are still proliferating (Fig. 1). Although we are not aware of direct evidence that insulin promotes the proliferation of male germ cells and Sertoli cells in mice, this has been shown in the chicken (67), swine (68), and newt (69). Insulins have also been shown to have other roles in the reproductive tract. For example, insulin promotes the differentiation of male germ cells and increases the production of various gene products from testicular cells (69,70). Insulin signaling is necessary for specification of male gonads in mice (71) and contributes to reproductive longevity and life span in Caenorhabditis elegans (72).
Why is insulin expressed in the testis? One likely explanation is that the testis provides a constitutive source of insulin for spermatogenesis, whereas pancreatic insulin levels vary in response to diet and the circadian rhythm. Testicular insulin production may be particularly important during the first wave of spermatogenesis, because systemic insulin levels are much lower in postnatal mice than in adult mice (73,74). The secretion of insulin by Sertoli cells in the testis may also be essential to provide adequate levels of this hormone to germ cells that Empty vec tor   Fig. 6A except that expression plasmids encoding the indicated rat RHOX fusion proteins were cotransfected with the luciferase plasmids. Quantification was performed as in Fig. 3B from at least four separate transfections. B, alignment of the predicted amino acid sequences of the rat RHOX protein homeodomains. Yellow-shaded residues are absolutely conserved; green-shaded residues are the most common ones at each position; blue-shaded residues are similar to conserved residues. Hydrophobic amino acid residues essential for homeodomain packing are indicated with H, and amino acids that make base-specific contacts in the major groove of DNA are indicated with B. C, luciferase activity from TM4 cells transfected as in A with either mutant homeodomain residues (HD mut) or wild-type versions of the constructs indicated (the specific amino acid substitutions are indicated under "Results"). Significance was determined by one-way analysis of variance followed by Tukey's multiple comparison post hoc test. Different letters denote mean expression levels that elicit responses significantly different (p Ͻ 0.01) from other expression constructs.

RHOX5-mediated Regulation of Metabolism Genes
would not have access to plasma insulin as a result of the bloodtestis barrier created by the tight junctions between Sertoli cells during postnatal testes development (73,75,76). In support of this, recent studies suggest that insulin is an essential autocrine factor produced by Sertoli cells. For example, Sertoli-specific ablation of insulin and insulin-like growth factor (IGF) receptors results in severely diminished testis size and spermatogenic output (77). This decrease is largely attributed to the failure to establish a full complement of Sertoli nurse cells in the pubertal testis (77), which is essential to generate normal adult sperm counts (78). Loss of IGF signaling also disrupts androgen production and signaling, (79), which is intriguing in light of the fact that Rhox5 is highly induced by androgen signaling (32, 80 -82). This suggests the possible existence of a feed-forward circuit revolving around RHOX5 that promotes insulin and IGF signaling.
Our studies revealed that the homeodomain, ATD, and CTD regions of RHOX5 all have roles in Ins2 transcriptional induction ( Figs. 6 and 8). We identified key amino acids in the third "recognition" helix of the homeodomain essential for Ins2 transcription; substitutions at positions 50 and 51 disrupted the function of RHOX5, and a substitution at position 47 conferred Ins2 inducibility to RHOX10 (Fig. 8C). We also established that the RHOX5 ATD is sufficient to confer Ins2 transcriptional induction to RHOX10 (Fig. 8). Although the RHOX5 ATD does not contain recognizable motifs, it does have a long stretch of negatively charged amino acids (19), a common feature of activation domains that recruit transcriptional coactivators (83,84). ATDs rich in negatively charged amino acids are also present in other RHOX proteins that support Ins2 transcription (RHOX8, RHOX11, and mutant RHOX10 with an "improved" homeodomain) (19,20). Like the ATD, the CTD may also recruit coactivators essential for RHOX5 to regulate transcription. Interestingly, the CTD region of RHOX5 is unique among RHOX proteins in possessing a CXXC motif, which is the catalytic core for many redox enzymes (27). It is possible that this CXXC motif regulates the ability of RHOX5 to transcriptionally activate its targets by sensing redox potential.
Three homeobox transcription factors have previously been shown to regulate the expression of insulin genes: PDX1, NKX2.2, and NKX6.1. PDX1 binds both the Ins1 and Ins2 promoters as well as the Ins1 enhancer (85,86) and promotes glucose-mediated up-regulation of Ins1 expression in pancreatic cells in vitro and in vivo (87). NKX2.2 and NKX6.1 both up-reg- FIGURE 9. Loss of Rhox5 impairs insulin signaling and triggers germ cell apoptosis in vivo. A, Western blot analysis of total testes protein extracts from P16 mice (n ϭ 4 from each genotype, mean Ϯ S.E.) using the indicated antibodies. Phosphoprotein levels in duplicated blots were quantified using the LI-COR Odyssey system by normalizing to total protein level and were expressed as a percentage relative to the control value, which was set to 100%. B and C, TUNEL analysis of testes from Rhox5-null (KO) and wild-type control (WT) mice of the indicated postnatal ages (4 mice/genotype at each time point; Ͼ400 round tubules were analyzed per mouse). Significance was determined using unpaired Student's t test; **, p Ͻ 0.001; *, p Ͻ 0.01. NOVEMBER 29, 2013 • VOLUME 288 • NUMBER 48 ulate Ins1 expression in pancreatic beta cells; NKX2.2 appears to act by directly binding the Ins1 promoter, whereas NKX6.1 indirectly activates Ins1 expression (88,89). Interestingly, RHOX5 binds a consensus NKX2 site in the Ins2 promoter (Fig.  4), suggesting the possibility that NKX2 family members and RHOX5 bind identical DNA sequences. This leads to the intriguing possibility that RHOX5 could substitute for NKX/ PDX homeobox transcription factors in promoting insulin expression in ␤-islet cells. Likewise, NKX and PDX factors may able to substitute for RHOX5 in inducing Ins2 transcription in Sertoli cells.

RHOX5-mediated Regulation of Metabolism Genes
Rhox5-null mice exhibit altered expression levels of not only Ins2 but also many other metabolism genes ( Figs. 1 and 2). The master regulators of metabolism, Pparg and Ppargc1a, are both down-regulated in Rhox5-null mice testes, suggesting that RHOX5 positively regulates these genes. PPARG is a nuclear hormone receptor that promotes insulin sensitivity and regulates the transcription of a battery of genes that elicit a "thrifty" response involving fat storage (50). The coactivator PPARGC1A was originally cloned as a binding partner of PPARG but has since been shown to interact with many other transcription factors whose common goal is to promote ␤-oxidation of fatty acids and the generation of energy (50). Of note, Ppargc1a mRNA is highly expressed in the genital ridge of XY embryos at embryonic day 11.5 (90), the same localization observed for Rhox5 (91), consistent with the possibility that RHOX5 also drives PPARGC1A expression and energy production in male primordial germ cells. Another energy metabolism gene that we found was regulated by RHOX5 is Scd1, which encodes the rate-limiting enzyme responsible for the synthesis of monounsaturated fatty acids that not only promotes fat storage but inhibits fatty acid ␤-oxidation (52). Rhox5-null mice also have altered expression of Retn and Adipoq, which encode the secreted proteins resistin and adiponectin, respectively. In agreement with our results in the mouse, Retn was recently shown to be developmentally regulated in the postnatal rat testes and is expressed highly in Leydig cells and weakly in Sertoli cells (13). Retn is most highly expressed in the early stages of spermatogenesis, when Rhox5 is expressed at lower levels ( Fig.  1), supporting our data suggesting that RHOX5 inhibits resistin expression in interstitial cells (Fig. 2). Originally discovered by virtue of its being down-regulated by PPARG agonists, resistin opposes insulin action (92), which is consistent with RHOX5 repressing resistin expression. ADIPOQ, on the other hand, promotes insulin action and is induced by both PPARG agonists and insulin (50), as well as RHOX5 (Figs. 1 and 2). The hormone-regulated and stage-specific expression of ADIPOQ in the rat testes was recently reported (12).
We do not know whether RHOX5 directly regulates these metabolism genes or whether instead it indirectly regulates them through other transcription factors. We suspect that the latter is the case for many metabolism genes, because many of them are known to regulate each other. For example, insulin is known to activate both PPARG expression and activity (93,94), and PPARG agonists regulate the expression of a large number metabolism genes, including Retn, Adipoq, and Scd1 (92,95,96), all of which we found are regulated by Rhox5. As another example, the repression of Scd1 expression by RHOX5 may be through the ability of RHOX5 to activate Ins2 transcription because insulin signaling is known to inhibit Scd1 expression in Sertoli cells (97,98).
The altered expression of several key metabolism genes in Rhox5-null mice testes raises the possibility that RHOX5 serves to regulate metabolism in the testis. Most of the gene expression changes in Rhox5-null mice support a model in which RHOX5 promotes energy production. One line of evidence is the down-regulation of Ppargc1a in Rhox5-null testes (Figs. 1 and 2) because this implies that RHOX5 up-regulates the expression of PPARGC1A, a master transcriptional regulator that promotes fatty acid oxidation and mobilization (50,99). Also consistent with this scenario is our finding that Rhox5 represses Scd1 expression (Figs. 1 and 2) because SCD1 promotes fat storage (50,99). Our finding that Ins2 is a direct positive target of Rhox5 (Figs. 1-4) and that Rhox5 promotes insulin signaling (Fig. 9A) also supports our model because insulin promotes energy expenditure (2,100). This may be further amplified by the ability of Rhox5 to positively regulate Adipoq ( Figs. 1 and 2), a well known insulin action stimulator that is known to be expressed in Sertoli cells (95). Finally, our finding that Rhox5 negatively regulates Retn expression (Figs. 1 and 2) also supports our model because the protein product of Retn, RESISTIN, is a potent inhibitor of insulin action (92).
The metabolic defects in Rhox5-null mice may extend beyond energy metabolism, based on our GO analysis of genes differentially expressed in Rhox5-null versus control testes (available upon request). This analysis revealed that Rhox5-null testes also had altered expression of genes involved in alcohol and lipid catabolism as well as sulfur, phenol, and catecholamine metabolic pathways. The defects in germ cell motility observed in Rhox5-null mice (19) may also directly result from altered expression of motility components, because six genes involved in cellular motility exhibited altered expression in response to loss of Rhox5. In the future, it will also be important to determine whether the ability of Rhox5 to regulate Ins2 and other metabolism genes is restricted to Sertoli cells or also occurs in other cell types, including Rhox5-positive cells in the epididymis, placenta, and ovary. We propose that an understanding of the regulatory networks downstream of Rhox5 and other Rhox cluster genes has the potential to provide insight into the molecular hardware linking metabolism and reproduction.