The role of Rhox homeobox factors in tumorigenesis

Abstract

Homeobox genes encode transcription factors that have well-established roles in embryonic development. We recently discovered the Rhox genes, a new family of homeobox genes, which are selectively expressed in the developing embryo, postnatal and adult gonads, and accessory tissues associated with mammalian reproduction. The largest and best-characterized Rhox cluster is found in mouse. However, all mammals examined to date possess a set of Rhox genes that, while they may vary in number by species, appear relevant to reproduction and are located in the syntenic region of the X chromosome. Rhox5, the founding member of the family, was initially cloned from a screen to identify tumorigenic antigens from T-cell lymphomas, and was later found to be widely expressed in tumors from tissues of diverse origins that do not normally express the Rhox genes. This aberrant upregulation appears to be a general feature of many Rhox genes, but the implications of this misexpression remain largely uninvestigated. In this review, we will discuss the latest findings on the normal and abnormal roles of the Rhox genes and their potential contributions to the formation and progression of tumors.

2. INTRODUCTION

Homeobox genes encode transcription factors that are master regulators of developmental programs. The homeodomain is a 60-amino acid DNA-binding motif that confers the primary function of homeobox proteins, namely binding to the promoters of downstream target genes and either activating or repressing their transcription (1). However, as described in more detail later, homeodomains can be involved in specific protein-protein interaction and are typically found in the context of larger polypeptides (mostly near the C-terminus) that possess multiple functional domains that influence transcription factor recruitment or interaction with cofactors influencing non-transcription related functions (2, 3). More than 170 homeobox genes exist in higher eukaryotic genomes; they are grouped in subfamilies based on homeodomain sequence and gene structure (Figure 1). Best known is the highly conserved Hox subfamily, which governs diverse embryonic developmental processes, including body-axis formation, organogenesis, and limb development (4).

Figure 1.

Phylogenetic Comparison of mouse homeodomain proteins. The 60 amino acid homeodomain portions of all described mouse homeobox genes were aligned and compared via an unrooted phylogenetic tree using the neighbor joining method. Branch lengths represent the extent of divergence. Homeobox genes with links to cancer are indicated by blue dots and genes expressed in the reproductive tract are indicated by red dots.

This review focuses on the recently discovered Rhox genes, a family of homeobox genes clustered on the X chromosome, which are preferentially expressed in reproductive tissues and are thus thought to contribute to fertility (5). Similarly to other homeobox genes, the Rhox genes are expressed during embryonic development (5–9). However, unlike most other homeobox gene families, several members of the Rhox cluster remain expressed at high levels after birth and thus are candidates to control postnatal and adult developmental events, particularly those essential for male and female germ cell production (5, 10).

In contrast to their well established roles in controlling development, very little is known about the normal functions of homeobox genes in postnatal tissues, especially within the reproductive tract where the Rhox genes are predominantly expressed. This is somewhat surprising, as over 40 homeobox genes are expressed in postnatal and adult testes and ovary (indicated by red dots in Figure 1). Efforts to clarify the roles of some of these homeobox genes have been clouded by putative functional redundancies and embryonic lethality. Knockout mice for many homeobox genes (mainly Hox genes) do not exhibit any obvious defects in the reproductive tract, including spermatogenesis. It is not known whether this is because of functional redundancy (Hox genes are known to compensate for each other) or because some homeobox genes have little or no role in reproduction (that is, their expression in the reproductive tract is superfluous). In other cases, targeted disruption of homeobox genes, including Pou5f1, Pou3f1, Hoxa5, and Alx1, causes embryonic lethality, thereby precluding analysis of their role in postnatal and adult events, including spermatogenesis (11, 12).

Because homeobox genes are master regulators of development, and govern the processes of cellular proliferation, differentiation, and reorganization of tissues, it’s not surprising to find that many have links to cancer, hie misregulation of at least 35 mouse homeobox genes has been demonstrated in cancers of blood, breast, ovary, skin, prostate, gut and others (indicated by blue dots in Figure 1). In some instances, the homeobox gene exhibits tumor suppressor activity which is lost such as Cdx2 in the colon (13) or PDX1 in the stomach (14). Some homeobox genes possess proliferative activity such as Cdx1 in colorectal cancer (15) and Irx5 in prostate (16). hie role of some homeobox genes, such as Nkx3-1, in cancer is somewhat controversial as evidence for both of these simple cases exists (17, 18). hie pleiotropic effects of homeobox genes may result from stage-dependent differences in their regulation by different hormones or shifts in the cell-type distribution within a tissue.

The majority of the published work to date centers on regulation of the Rhox genes and their function in fertility, which for summaries and perspectives we direct readers to prior reviews (7, 8, 10, 11, 19–22). However, historically there have been many anecdotal lines of evidence to suggest that the Rhox genes are relevant in cancer biology, which has been borne out by recent directed studies. Here, we update what is known about the reproductive functions, regulation, and evolution of the Rhox gene cluster since our last review (10), and for the first time review emerging studies that abnormal expression and actions of the Rhox genes may contribute to timiorigenesis.

3. CHARACTERISTICS OF THE RHOX HOMEOBOX CLUSTER

3.1. Organization and evolution of the Rhox genes

The largest Rhox gene cluster is found in mice and has been subdivided into three subclusters: alpha, beta, and gamma with each gene receiving a numeric designation based on relative position on the X chromosome with respect to the centromere (Figure 2). Hie alpha subcluster originally was thought to contain four genes (5), but later it was discovered that a tandem duplications of three of the alpha subcluster genes, Rhox2, Rhox3, and Rhox4, had occurred, resulting in 24 genes in the alpha subcluster (23–26). These paralogs are nearly identical in DNA and protein sequence (greater than 92%) and are largely coexpressed in reproductive tissues, with the first set of copies being most highly detected as determined by qPCR using paralog-specific primers (10, 26, 27). Rhox5, the founding member of the Rhox family, is the first gene in the beta subcluster, and it is followed by three genes that had appeared in the literature as independent homeobox genes prior to discovery of the cluster: Rhox6 (Psx1), Rhox8 (Tox), and Rhox9 (Psx2 and Gpbox) (28–30). Hie gamma subcluster contains four genes including the most recently discovered member of the family, Rhox13 (5, 31). To date, the mouse Rhox cluster contains 33 homeobox genes, and there have been few changes in Rhox gene number in the past few builds of the mouse genome suggesting that the cluster is now complete.

Figure 2.

Organization of the Rhox cluster in mammalian species. The syntenic region of the X chromosome containing the Rhox orthologs and conserved flanking genes is shown. The rodent Rhox gene subclusters are indicated by red (alpha), green (beta), and blue (gamma). Established orthologus genes are indicated by dotted lines. The orthologus relationship between human RHOX genes and rodent Rhox genes cannot be clearly assigned because of the rapid evolution of Rhox genes. The map positions shown are according to builds 37.1 (mouse), 3.4 (rat), 3.1 (Dog), 6.1 (Cattle), and 37.1 (human).

Our lab and others have previously demonstrated that Rhox gene sequences are rapidly diverging, exhibiting signs of positive selection for amino acid changes (24, 26, 32). This is not surprising as it is a common feature of X-linked genes (with respect to their autosomal copies), large gene families (regardless of chromosomal location), and duplicated genes involved in reproduction (33–36). The significant divergence in sequence identity made the discovery of orthologus Rhox genes in different species difficult. For example, mouse Rhox5 shares only 72% identity with its counterpart in rat, and is only 42% identical with a putative human ortholog, RHOXF1 (37). However, recent advances in genome mapping, coupled with EST sequencing projects, have confirmed the existence of Rhox clusters in most mammalian species (and some fish) which have been investigated. These findings suggest that the rates of duplication (or selective pressures to maintain multiple Rhox gene copies) have varied between species.

After mouse, the next largest Rhox cluster is found in rat (Figure 2) (25, 27). While all three subclusters are present, the alpha subcluster has undergone an inversion (possibly resulting in the deletion of Rhox1) and there is no evidence from genomic mapping or cDNA screening to support the tandem duplication of Rhox2, Rhox3, and Rhox4 in rat (5, 23, 25, 26). Additionally, in the beta subcluster there is only one copy of the highly similar genes, Rhox6 and Rhox9, found in mice. Sequence analysis and genomic position suggests that the rat ortholog exhibits features more consistent with the designation of Rhox9 than Rhox6 (5, 25, 26). Hie gamma subcluster appears conserved in both structure and organization between rodent species. cDNA encoding Rhox orthologs have been identified in two other rodent species, guinea pig and hamster, but the preliminary builds of these two genomes are not refined enough to determine the size and organization of their Rhox clusters.

It appears that the majority of the other mammals have a much smaller set of RHOX genes with the syntenic portion of their X chromosomes (Figure 2). Examination of cDNA libraries typically yields two distinct genes, RHOX family 1 (RHOXF1) and RHOXF2 (which may be present in multiple copies). These genes all possess the characteristic exon/intron structure of Rhox members (5), but as the gene number and sequences are highly variable between species, assignment of true orthologs has remained elusive. In humans, RHOXF1 (originally called OTEX and hPEPP1), has traditionally been most associated with being the Rhox5-ortholog because it was first discovered, is androgen regulated, and selectively expressed in human testes (37, 38). Humans also have two recently duplicated genes, RHOXF2A (originally called hPEPP2) and RHOXF2B, which cannot be distinguished by qPCR analysis as they only vary in 2 positions in their 1291 nucleotide coding sequences (10, 37). Rhesus macaque, chimpanzee, and orangutans all share the precise structure of the human RHOX cluster and flanking genes as shown in Figure 2, only varying in MB numbering ranges (i.e. RHOXF1 in chimpanzee is at 120.5 with respect to the centromere). Smaller primates including marmosets and gibbons possess genes encoding RHOXF1 and RHOXF2, but the assembly of their genomes (build 1.1) is not yet sufficient to determine whether multiple copies of RHOXF2 are present. Final refinement of the primate genomes will likely look much different than the scenarios presented above. While the number of unique genes in nonrodents will probably not significantly change, there is evidence that the number of copies of RHOXF2 is under evolutionary selection (39). Niu et al. sequenced the RHOXF2 genomic locus for 111 human individuals and found that 13% had either only one copy of RHOXF2 or two identical copies, whereas the others had clear RHOXF2A and RHOX2B encoding genes. Similarly, 11 non-human primate species were found to have single or two duplicate RHOXF2 genes. However, 4 old world monkey species all possessed at least 2 distinct RHOXF2 genes, with cDNA evidence suggesting the number of copies in chimpanzee may be as large as 6 RHOXF2 genes. It is not clear what reproductive advantages having multiple RHOXF2 genes would provide from a Darwinian selection perspective, however, it appears that like the mouse Rhox genes they are under positive selection for amino acid changes (39). Given that RHOXF2 may have a causative role in cancer development and progression (see section 4.2), assessment of the number of copies (if truly variant in the human population) of RHOXF2 could provide a genetic screen for the risk of developing certain cancers.

The presence of a RHOXF2B gene in dog, which diverged from a common ancestor prior to the appearance of primates (40), suggests that duplication and loss of RHOXF2 copies is an ancient event, not restricted to primates (Figure 2). The bovine RHOX cluster contains a bona fide ortholog of RHOXF1 (Figure 2). However, the current build of the Bos Taurus genome (6.1) does not possess an obvious second copy of RHOXF2B in between NKAP and RHOXF1 as observed in primates and canines. At the RHOXF2A location, there is an ESX1-like gene, which likely will ultimately be determined to be a true RHOXF2 ortholog. ESX1 (discussed below) undoubtedly derived from the same primordial gene that gave rise to the RHOX genes as it shares the characteristic splicing pattern in its homeodomain (5, 37), but it resides ~20 MB outside of the Rhox cluster in all species examined in Figure 2. Most species share remnant RHOX sequences (i.e. partial exons/introns, but no full coding sequences) flanking their complete RHOX genes, but it is not yet clear if additional genes exist awaiting discovery, if they are leftovers from duplicated genes which were not maintained, or whether they are sequencing artifacts (JAM unpublished observations).

Evolutionarily, the RHOX family is most closely related to the paired-like subfamily of homeodomain proteins, POC1A, POC1B, and PIX3 (Figure 1). The genes encoding these proteins were probably derived from the same precursor that yielded the adjacent paired class of homeobox genes, most likely the ancient aristaless gene first characterized in drosophila (12, 41). However, it is currently believed that the Rhox-precursor and two paired class genes, Esx1 and Arx, likely resulted from duplication of the same primordial gene around the time of the divergence of mammals (19). There are several lines of evidence to suggest a tight relationship between the Rhox genes, Esx1, and Arx. Namely, all are on the X chromosome, are similar in homeodomain sequence, have amino terminal domains that while highly divergent are more similar to each other than those of other paired genes, and finally all possess the hallmark exon-intron structure specifically characteristic of Rhox genes (i.e., introns located at two particular sites within the homeodomain-encoding region) that are not present in other paired genes (5, 26, 37). For the purposes of this discussion, Esx1 and Arx are additionally similar to the Rhox genes in that they are normally restricted to reproductive tissues, but are misexpressed in tumors of divergent cellular origin.

3.2. Expression of the Rhox genes in reproductive tissues

The normal expression of RHOX transcription factors is restricted to reproductive organs and thus they are likely to control genes that modulate the production and differentiation of germ cells as well as governing the early development of the embryo (11, 25, 42). In our initial report detailing the discovery of the Rhox cluster (5), we demonstrated that all Rhox genes are selectively expressed in the ovary, testis, epididymis, and placenta in mice. Individual accounts of a few rat Rhox orthologs indicate that they exhibit conserved expression in reproductive tissues (25, 42–44). Recently, a formal panel screen of the rat Rhox cluster confirmed their selective expression in the ovary, testis, epididymis, and placenta (27). We previously reported that three mouse genes, Rhox4, Rhox7, and Rhox8 each exhibited expression in a single non-reproductive tissue; thymus, stomach, and intestine, respectively (5). However, the isolated expression of these Rhox genes in non-reproductive tissues did not appear to be conserved in rats (25, 27). While expression of mouse and rat Rhox genes is observed in similar tissues, there are some differences of note. For example, while Rhox5 and Rhox8 are the most highly expressed genes in the mouse testes, in the rat they are more highly expressed in the epididymis than testes (27). Rat Rhox2 is most highly expressed in the testis and relatively minor in other tissues, whereas in the mouse the converse is true with ovary, placental and epididymis exhibiting high expression. While rat and mouse Rhox3 are highest in the testes, rat Rhox3 is nearly as highly expressed in placenta, but barely detected in placenta in mouse. At present, it is not clear if differences between mouse and rat Rhox2, Rhox3, and Rhox4 are due to different relative expressions of the 7 duplicated mouse genes which may be independently regulated, in comparison to the rat where a single gene/promoter contributed to expression.

While less information is currently available for the non-rodent RHOX genes, it appears that like their rodent cousins they are expressed in reproductive tissues. RHOXF1 and RHOXF2 are expressed in the human testis and epididymis, as assessed by Northern blot and qPCR analyses (37, 38). In our initial report, we could not detect either gene in post-delivery placental tissues (37). However, subsequent analysis of flash frozen, near term placenta suggests that RHOXF1 may be present in trophoblast tissue. Preliminary immunohistochemical analyses with antisera generated against RHOXF1 peptides stains the same cells as androgen receptor-specific antibodies suggesting that it is produced in Sertoli cells. Furthermore, consensus androgen response elements have been identified in the putative promoters of both RHOXF1 and RHOXF2A gene (37). However, AR-dependent expression has only been demonstrated for the RHOXF1 gene (38). Thus the normal expression of RHOXF1 is most like that of genes in the rodent beta subcluster. Hie relatively strong expression of RHOXF2 in testis and epididymis, but absence in placenta and ovary, most resembles that of mouse Rhox11. However, absence of expression in the placenta is not consistent in general with genes of the rodent gamma subcluster suggesting that the evolution of Rhox cluster regulation between species is more complex that duplication of single co-regulated genes and promoter sequences. To the best of our knowledge, formal analysis of non-primate RHOX genes (guinea pig, cattle, dog, and chicken) in tissue panels has not been performed. However, EST and cDNA sequences representing these putative orthologs have been obtained from testis and placental sources, suggesting that they may be similarly restricted to reproductive tissues.

At present, it is not known whether a local enhancer exists that drives groups of Rhox genes to be expressed specifically in one reproductive tissue, or if their expression is entirely dependent on their individual promoters which have also been duplicated as the cluster proliferated. In support of the latter the Rhox-related genes, Arx which is present at the opposite end of the X chromosome from the Rhox cluster, and Esx1 which is located halfway between them, are both expressed in reproductive tissues. While Arx is highly expressed in developing male gonads, unlike the other Rhox genes, it is also found in several regions of the developing and adult brain (45, 46). Detailed analysis of Arx expression and function in fertility has been derailed by the finding that ablation of Arx results in severe neurological defects that lead to perinatal lethality (45–47). Esx1 and its human ortho log, ESX1, are both expressed in testes and placenta (48–51). In the mouse testis, Esx1 transcripts are specifically localized to spermatogonia, spermatocytes, and round spermatids in stages IV-VII of the seminiferous epithelial cycle (48). Interestingly, while Esx1 transcripts can be detected in pre-meiotic germ cells, the ESX1 protein appears to only be translated in post-meiotic round spermatids (51). The mechanism of Esx1 translational control has not been investigated, however, translation of Rhox13 transcripts is inhibited in prenatal gonads by NANOS2 (52). Translation of RHOX13 occurs when the retinoic acid signaling pathway is activated in neonatal male and female gonads, and can be induced earlier by exogenous RA treatment. It is not yet certain how many other Rhox genes may be similarly regulated as suitable anti-sera to compare Rhox mRNA and RHOX protein levels have not been developed for all of the germ cell-specific RHOX factors.

4. ABERRANT EXPRESSION OF THE RHOX CLUSTER IN CANCER

4.1. Rhox genes are broadly expressed in cancerous cells of diverse tissue origin

While the founding member of the reproductive homeobox cluster, Rhox5 (originally called Pem), is an important regulator of development and fertility, it actually has its roots in tumor biology. Rhox5 was identified in a screen designed to find novel genes functioning in the development of T-cell lymphoma. This assay identified Rhox5 as being differentially expressed between two T-cell lymphoma clones of common parental origin that exhibited different malignancy potential (53). Subsequent reports identified that the aberrant expression of Rhox5 is not unique to cancerous T-cells, but rather that it is widely distributed in tumors derived from many different tissues and cell types (37, 53, 54). As shown in a panel screen of 11 representative tumors, Rhox5 is highly expressed (> 100-fold above background) in 8 cell lines, and expressed in the other 3 (Figure 3). Conservative estimates suggest that ~75% of tumor lines screened to date express Rhox5 transcripts. Hie ubiquitous expression in tumors may derive, in part, from its ability to be induced by the Ras proto-oncogene (42, 55). Mutations that constitutively activate TP53 also lead to high expression of Rhox5 in mouse and rat derived cell lines. In the testes, Rhox5 mRNA and RHOX5 protein are restricted to Sertoli cells. Interestingly, Rhox5 mRNA is not detectable in most immortalized Sertoli cell lines (TM3, TM4, and 15P-1) and is only lowly expressed in MSC1 Sertoli cells. It is probable that although these cell lines maintain both their endogenous methylation and silencing (described in detail in the next section) of the Pd regulatory sequence, and hypomethylation of the Pp, that a necessary cofactor for androgen regulation is missing or androgen receptor levels are inadequate.

Figure 3.

Expression of the mouse Rhox genes in tumor and immortalized cell lines. Quantitative real-time RT-PCR (qPCR) was used to determine the relative expression of each Rhox gene using primer pairs specific for each unique Rhox gene and pan-paralog-specific primers that amplify total Rhox2, Rhox3, and Rhox4. Data is presented as mean +/− SEM fold above background after normalization with Rpl19 mRNA to account for differences in cDNA loading. The cell lines examined were TEL (thymic epithelial cell line), WEHI122 (lymphoma), PCC4 (embryonic carcinoma), I-10 (Leydig cell line), PS-1 (prostate mesenchymal cell line), N4TG1 (neuroblastoma), TS3.5 (a trophoblast stem cell line), MME (mouse mammary epithelial), SL12.4 (T-cell lymphoma, the original source of Rhox5), F9 (embryo carcinoma), and NIH3T3 (fibroblast cell line).

The other Rhox genes do not appear to be as broadly or as highly expressed in tumor cells. Transcripts for Rhox1, Rhox8, and Rhox11 were not detected in any of the cell lines examined in our panel screen (Figure 3 and data not shown). Two genes, Rhox6 and Rhox9, were previously shown to be expressed in a single cell line, the TS3.5 trophoblast stem cell line (37). Since these two genes are highly expressed in normal placental tissue, we believe that this represents activation of their endogenous control mechanism and not aberrant tumor-specific activity. However, Rhox6 and Rhox9 are induced by the DNA-methylation inhibitor 5AzaC in mouse cell lines (56), suggesting they could potentially be expressed in tumor cells where epigenetic silencing has gone awry. Hie remaining Rhox genes exhibit expression to varying degrees in multiple cell lines. In some cases, we can speculate why a Rhox gene may be present, for most it’s a complete mystery as regulatory elements within their 5’ flanking sequences have yet to be examined. For example, the promoter of Rhox4 contains an essential NF-Y transcription factor binding site (57). NF-Y has been shown to induce the expression of ATPase genes in N4TG1 neuroblastoma cells (58). Thus, the necessary elements for permissive transcription of Rhox4 are present. However, why Rhox4 would be present in one embryo carcinoma, PCC4, and not another, F9, cannot be predicted. The Rhox cousins, Arx and Esx1, are also expressed in cancerous cells. Mutation of ARX is linked to the formation of brain cysts (59), but few if any of the ~400 papers detailing Arx/ARX expression and function in developmental processes mention aberrant expression in tumor cell lines. Conversely, Esx1 was detected by northern blot in ~40% of mouse tumor lines (37), and is a significant player in human colorectal carcinomas (60). Interestingly, expression of Esx1 and Rhox5 were inversely correlated in the initial tumor panel screen, but whether this is significant to tumor biology or an artifact of sample size is not known (37).

Both groups that initially cloned the human RHOX orthologs were initially concerned with determining the androgen regulation of these novel genes in the male reproductive tract, rather than their potential role in tumor biology. However, one showed that RHOXF1 was not endogenously expressed the androgen-receptor negative PC-3 prostate cancer cell line (38). RHOXF1 was abundant in HPB-ALL (acute lymphocytic leukemia), LNCaP (prostate tumor), and Hec1A (endometrial adenocarcinoma) cells, but not detected in A375 (melanoma), Hela (cervical adenocarcinoma), and SW620 (colorectal adenocarcinoma) cells as assessed by RT-PCR / Southern blot panel screening (37). A larger screen including these cell lines and 8 additional lines demonstrated expression of RHOXF2 only in K562 (erythroleukemia) cells (37). Subsequent reports from many groups have demonstrated expression and regulation of both genes in tumors and immortalized cell lines. RHOXF1 displays variable expression in lingual squamous carcinomas (61) and colorectal cancers (62). An RT-PCR screen of 13 colon, breast, and pancreatic cancer cell lines revealed that RHOXF1 was broadly expressed, being absent in only the SW480 and SW620 colon cancer lines (56). Interestingly, SW480 cells were the only member of this panel to exhibit detectable expression of RHOXF2. However, RHOXF2 has been identified as a prominent cancer/testis antigen (63). As with Rhox5, RHOXF1 and RHOXF2B are induced by 5AzaC in human cell lines (56). Thus, it is likely that aberrant expression of the RHOX genes as cells transform is dependent on permissive hypomethylation and the presence of specific transcription factors which may vary by tissue type and mutation status. In support of this, examination of publically available normal vs. cancer expression array data using stringent parameters (threshold p-value 0.0001, 4-fold change, top 10% gene rank) in the ONCOMINE database, finds a mixture of both up and down-regulated outlier datasets. However, consistent with its putative role as a tumor suppressor (discussed below) RHOXF1 is consistently down-regulated 5-fold or greater in invasive, mucinous, and mixed lobular and ductal breast carcinomas. In contrast, RHOXF2 is up-regulated 20-fold or more in brain, head and neck, kidney, and pancreatic cancer datasets without any counterbalancing outliers exhibiting down-regulation, making RHOXF2’s potential role as a cancer promoting gene clearer. Thus, directed studies to examine the reciprocal status of RHOXF1/RHOXF2 gene expression on a tissue by tissue basis would seem in order.

4.2. Potential mechanisms of Rhox gene misregulation

More than 30 studies have been devoted to characterization of the transcriptional regulation of Rhox genes using both in vitro cell-based and in vivo transgenic models. The majority of these reports center upon androgen regulation of the founding member of the cluster Rhox5/Pem in epididymis and Sertoli cells of the testis. For an in-depth summary of Rhox gene regulation in reproductive tissues, we direct readers to our recent review in Reproduction (10). In this report, we will limit our discussion to recent findings expanding the mechanism of imprinting of the Rhox cluster and the exciting development that RHOX factors participate in cross-regulation of other Rhox genes. The role of DNA methylation in the development and progression of cancer is currently one of the largest growing topics in the life sciences. The aberrant regulation of methylation may contribute in multiple tumorigenesis pathways as there is potential for both the derepression of oncogenes and silencing of anticancer genes (64). If the gene that is aberrantly activated is a “master switch” for cancer development (65), then a single misregulation event may lead to the acceleration of cancer development. The Rhox genes may be relevant to such a scenario as their expression is governed by the epigenome, and one gene that is normally silenced in non-reproductive tissues has recently been shown to actively organize the expression of other members of the cluster.

Epigenetic mechanisms contribute to control of gene regulation in a wide array of developmental events including establishment of the placenta, embryonic growth, organ formation, and tissue differentiation (66, 67). Methylation is the major epigenetic modification in mammals and gene silencing, genomic imprinting, and X-chromosome inactivation all depend on its proper regulation (68, 69). Given that homeobox genes control developmental processes and that the Rhox genes reside on the X chromosome, it is not surprising that several reports have linked DNA methylation to control of Rhox gene (in particular Rhox5 and Esx1) regulation (21, 50, 56, 70, 71). Recently, we formally examined the imprinting status of the entire mouse Rhox cluster using four independent models of imprinting and impaired DNA methylation (72). Depletion of H1 linker histone, an abundant and essential component of chromatin, in mouse embryonic stem (ES) cells was predicted to result in the global misregulation of genes silenced by methylation (73). However, microarray analysis revealed that very few genes were upregulated, and that methylation of specific CpGs within the regulatory regions of H1-dependent genes was restricted to a small subset of genes, predominantly on the X-chromosome. Interestingly, Rhox5 was among these H1-regulated targets (73). Subsequent analysis of the Rhox cluster revealed that Rhox1, Rhox2, Rhox4, Rhox6, Rhox9, Rhox10, Rhox11, Rhox12 and Rhox13were all additionally derepressed in H1-depleted ES cells (72). However, the neighboring genes upstream (Sept6, Ndufa1) and downstream (Lamp2, Mcts1, Hprt1, and G6pdx) of the Rhox cluster remained silenced suggesting that regulation by H1 may be directed specifically towards the Rhox cluster. The aberrant upregulation of Rhox expression was reversible as transfection of knockdown cells with H1 expression vectors resulted in the rescue of Rhox gene silencing. As further evidence for the role of methylation, ES cells lacking the de novo methylation enzymes DNMT3A and DNMT3B were found to express the same subset of Rhox genes upregulated upon H1 -depletion (72). Conversely, the H1-indepenent Rhox3, Rhox7, Rhox8, and Rhox11 were not upregulated in Dnmt3A/Dnmt3B-null ES cells.

To assess whether the Rhox cluster is subject to Xp imprinting, i.e. the silencing of the paternal copy of an X-linked gene, we have examined the parent of origin of Rhox genes expressed in the placenta. For this analysis, we identified polymorphisms between Rhox gene sequences obtained from Mus musculus musculus and Mus musculus molossinus mouse subspecies. Analysis of placental cDNAs from hybrid animals demonstrated that indeed the maternal copy each Rhox gene was the one found in trophoblast (72). Three genes were excluded from the analysis as no polymorphisms were present for Rhox2, and neither Rhox7 nor Rhox11 are expressed in the placenta. Only Rhox3a and Rhox8 exhibited equivalent expression of either parent’s allele, suggesting that these genes escape imprinting and exhibit random inactivation of their X chromosomes (72). Neither of these genes were upregulated in H1-depleted and Dnmt3A/Dnmt3B-null ES cells. As a final test, the imprinting status of the Rhox cluster was examined in uniparental ES cells that are manipulated to possess two copies of either parental genome (74). Microarray and qPCR analyses indicated that Rhox1, Rhox2, Rhox4, Rhox5, Rhox6, Rhox9, Rhox10, Rhox12, and Rhox13 were all upregulated in maternally derived uniparental ES cells, consistent with expression of the maternal copy we observed in the placenta (72). In summary, the same subset of Rhox genes was misregulated in all four methylation/imprinting models examined. Why certain genes escape inactivation and ultimately what developmental events are affected by epigenetic regulation of RHOX factors is not yet known. However, perturbations in DNA methylation and genomic imprinting likely explain why Rhox genes that are normally expressed only in reproductive tissues are upregulated when cells become cancerous.

The Rhox5 gene possesses two promoters, an androgen-dependent proximal promoter (Pp) and androgen-independent distal promoter (Pd) (10). The Pp directs expression of Rhox5 in the testis and epididymis in mice and rats (43, 75–77), whereas the Pd is responsible for ES cell, ovarian, placental, and tumor cell expression (42, 43, 55, 72). Both of Rhox5’s promoters are methylated in tissues in which the gene is not expressed, i.e. the Pd is hypermethylated in male tissues (21). The loss of methylation is a first step to the expression of Rhox5 as it is permissive to the recruitment of essential transcription factors which drive tissue-specific expression. This mechanism appears to be in play for both promoters. For example, in ES cells demethylation of the Pd, specifically at cytosine −1644, unmasks an ETS factor binding site which allows the recruitment of transcription factors that promote expression (72). The ETS factor GABP (42, 72) is sufficient to drive expression of Rhox5 in cultured ES and ovarian granulosa cells, but the full array of endogenous ETS factors that contribute to tissue-specific expression in vivo remain to be determined. Transcription of Rhox5 in the testis and epididymis depends on the recruitment of both androgen receptor and GATA transcription factors to the Rhox5 Pp (78). Four tandem androgen response elements contribute to Rhox5 regulation. Interestingly three contain CpG sites that are differentially methylated in expressing and non-expressing tissues (79). The recruitment of AR and GATA is dependent upon demethylation of specific ARE and GATA binding elements that was correlated to the region-specific and temporal patterns of Rhox5 expression in the epididymis (79). The factors that control demethylation of the Pp predominantly in only one region of the epididymis (the caput) and at a specific time in postnatal development (~P25) remain to be determined. However, activation of this single gene has recently been shown to have profound effects on regulation of the Rhox cluster in the epididymis.

As in the reproductive tract, the expression of Rhox5 in cancer cells is dependent on both the local action of transcription factors binding its promoter and epigenetic mechanisms. It was established by ribonuclease protection assay and RT-PCR analysis that the Rhox5 Pd is the primary determinant of aberrant Rhox5 expression in tumor cells (10, 43). Dissection of this promoter revealed a minimal element required for Pd transcription, located ~100-nt upstream of the transcription start site, which is only 22-nt in length (42, 55). Within this element are two ETS family-binding sites and one SP1 family-binding site. Mutation of any of these three significantly impaired transcription of Rhox5 and combinations of mutations abolished activity in both tumor cell lines and primary ovarian granulosa cells (42, 55). The ETS factors GABP and ELF1 are sufficient to drive expression of Rhox5 in SL12.4 and 10T1/2 tumor cells (55), but the full array of endogenous ETS factors that contribute to tumor-specific expression in vivo remain to be determined. However, upregulation of ELF1 has been associated with gynecological cancers of the cervix, uterus, and breast (80, 81). SP1 and SP3 factors are both capable of binding and activating the Pd, and both are widely known to participate in the regulation of genes in normal and cancerous tissues (82, 83). Because expression of these genes is thought to be ubiquitous and constitutive, it is not clear how they may differentially regulation Rhox5 in cancer cells, although different cell types have been shown to vary greatly in their turnover rate suggesting that the kinetics of SP1/SP3 activation may be altered when cell become cancerous. A more likely explanation for the selective expression of the Pd is that epigenetic mechanisms preclude binding of ETS and SP1 in normal male tissues and non-cancerous cells. In support of this, Rhox5 can be induced by the DNA-methylation inhibitor 5AzaC in mouse tumor cell lines (56). Subsequent analyses by this group discovered a correlation between low levels of active histone marks (H3ac, H4ac, and H34me2), coupled with hypermethylation and abundant repressor H3K9me2 histone marking, in tumor cell lines and primary tumors in which Rhox5 expression is low (62). Conversely, hypomethylation and active marks were correlated to high expressing cells. Low expressing cells could be converted to active status by treatment with epigenetic drug MS-275 (62), and in agreement with previous observations, treatment of F9 embryonic carcinoma cells with retinoic acid induced Rhox5 expression (84) and induced dynamic changes in epigenetic status (62).

We recently demonstrated that RHOX5 is a master regulator of Rhox cluster expression in the epididymis (27) and related commentary (85). In the mouse, expression of Rhox5 has previously been localized to the caput epididymis (76, 79, 86). Subsequent, in situ hybridization, immunohistochemistry, and qPCR analyses indicate that the other mouse Rhox genes are also predominantly localized to the caput epididymis (27, 87). In Rhox5-null animals, the expression of Rhox1, Rhox3, Rhox4, Rhox8, Rhox10, and Rhox11 was diminished in the caput epididymis, suggesting that RHOX5 (or a RHOX5-dependent factor) positively regulates the expression of these genes (27). Expression of these genes was not significantly altered in the corpus and cauda epididymis, indicating caput-specific regulation. In the rat, Rhox5 makes a switch to primarily caudal expression (27, 44). Interestingly, rat orthologs to the Rhox genes that were shown to be regulated by RE10X5 in Rhox5-null mice, also make a transition to primarily caudal localization in the rat (27). Conversely, Rhox2, Rhox7, and Rhox9, which were not misregulated in Rhox5-null animals, were equally highly expressed in all three regions of the rat epididymis. These findings support a model in which the alteration of the regional-specific regulation of a single gene could result in regional misexpression of a cadre of subordinate genes. This may be relevant to the formation and progression of tumors as Rhox5 is widely expressed in tumors of diverse tissue of origin and its upregulation may result in the abnormal expression of other Rhox genes that may also have oncogenic potential.

5. CELLULAR PROCESSES GOVEREND BY RHOX FACTORS

The function of most Rhox genes is not yet known. However, it is likely that some or all are all involved in reproduction, as all Rhox genes are selectively expressed in reproductive tissues including placenta, epididymis, testis and ovary in adult mice (5, 10). It is probable that they also have roles in the fetal development of the gonad, as most Rhox genes are also expressed in the primordial germ cells and one (Rhox8) is expressed in somatic cells in the developing gonad (6). Several Rhox genes are induced by androgens through androgen receptor-mediated activation at their promoters indicating that they are good candidates to regulate male fertility through induction of downstream genes that support spermatogenesis (11, 37, 38, 88). These RHOX factors are likely to govern distinct subsets of genes as they display unique temporal and stage-specific expression patterns in the testes and exhibit variations in the key contact residues that interact with target promoters (5).

Transgenic mouse models have been used to examine the function of Rhox5, Rhox9, Esx1, and Arx. Rhox5’s region-specific expression in the epididymis and stage-specific expression in the testis suggested that Rhox5 may be involved in regulating both spermatogenesis and sperm maturation. Rhox5’s role in these events is supported by our recently published studies showing that Rhox5-null male mice are subfertile (5). Subfertility in Rhox5-null animals results from a combination of insufficient germ cell output and developmental defects resulting in poor motility of spermatozoa that do survive. Rhox9 is expressed in placenta and fetal germ cells, but testicular, ovarian and placental histology all appear normal in Rhox9-null animals (89). Testicular development is compromised in Arx-null mice and humans with ARX mutations, specifically failure to develop Leydig cells (46). However, unlike the Rhox genes, Arx is also expressed in the developing CNS resulting in lethality and/or mental retardation precluding further characterization of reproductive phenotype (46, 47, 90, 91). Esx1 is an imprinted gene when ablated or misexpressed in extraembryonic tissues leads to fetal growth retardation (50). However, the smaller pups recover after birth to match their littermates and are fertile. Thus, single gene knockouts in mice have not produced animals which are completely infertile indicating overlapping function between multiple genes. In some cases the gene responsible can be predicted. In Rhox9-null animals, expression of the nearly identical (in sequence and expression pattern) Rhox6 gene is unaltered in the knockout, suggesting compensation by RHOX6 is responsible for the lack of RHOX9 phenotype. For other genes, such as the case for Rhox5, the redundant partner is not so obvious.

While it has been difficult to assign a definitive functions to each Rhox gene, data from transgenic mice has provided clues to their potential roles in governing cellular processes. To date, cell line studies have primarily been employed to investigate the regulation of Rhox genes, or to identify components in the downstream signaling pathways that they govern. However, recent in vitro studies have uncovered potential roles for Rhox2, Rhox4, Rhox5, Rhox6, Esx1, RHOXF1 and RHOXF2 in cell proliferation, differentiation, and survival.

5.1. RHOX factor regulation of proliferation

Historically, anecdotal evidence from our lab and others has suggested that RHOX5 may play a role in cellular senescence (JAM, M Wilkinson, M Rao and C Wayne, unpublished observations). This prediction was based on transfection studies in which introduction of plasmids encoding Rhox5 seemed to increase the time required for immortalized cell lines to become confluent. However, no formal investigation into this phenomenon was undertaken as to goal of those experiments was to monitor the effect of RHOX5 on target gene promoters or to examine biochemical effects of mutant RHOX5 proteins. RHOX5 was predicted to have a causal role in tumor formation, as it induces an immune response in mouse tumors (92) and it has been shown to interact with proteins involved in malignancy (93, 94). Recently, this prediction was proven correct by Li et al. who demonstrated a role for RHOX5 in proliferation and migration of colon cancer cells in vitro and tumor growth in vivo (62). In their model, lentivirus-mediated knockdown of Rhox5 in CT26 cells (a colon cancer cell line which is high in endogenous Rhox5) resulted in reduced proliferation compared to parental and control lentivirus-treated cells. Additionally, knockdown of Rhox5 resulted in reduced cellular migration, implying that RHOX5 could mediate tumor proliferation and invasiveness. To test this hypothesis, control and Rhox5-knockdown CT26 cells were examined for tumor forming capability in nude mice where the absence of RHOX5 resulted in slower tumor growth (62). At the time of sacrifice, 19 days after inoculation, tumors derived from Rhox5-knockdown cells were 20-30% the size of control tumors. However, the in vivo study was not extended to examine the dissemination potential of these cells. At present, whether aberrant expression of RHOX5 functions universally to accelerate tumor growth in tumors derived from different cellular sources. For example, our prior observations were primarily based in cell lines mimicking Sertoli cells, the site of normal Rhox5 expression. Thus, it’s possible that Sertoli cells may possess a unique subset of RHOX5 interacting proteins (some of which are described in the next section) that are capable of handling high Rhox5 expression without deleterious effects. Whereas, cell types where Rhox5 is normally epigenetically silenced may be respond differently to the interloping homeobox factor. While we have not yet examined this hypothesis, differential expression of RHOX5-regulated genes in tumor cells has been observed suggesting that all cell types do not equivalently respond to RHOX5. For example, the RHOX5 downstream gene Unc5c (95), a tumor suppressor frequently silenced in colon cancer, is absent normally in CT26 cells and was not upregulated in Rhox5-knockdown CT26 cells (62). Thus, additional studies are required to identify novel downstream factors and events controlled by RHOX5 that lead to tumorigenesis.

Shortly after the human RHOX orthologs were first identified (37, 38), they were identified as cancer testis antigens (63). This heterogeneous group of genes (members encode transcription factors, structural proteins, and enzyme, etc.) are normally expressed in trophoblast and germ cells, but are aberrantly expressed in ~40% of tumors of diverse origin (96), much like had been observed for Rhox5. As described previously, the regulation of RHOXF1 and RHOXF2 in cancer cells has received some attention, but few studies have focused on the impact of RHOX misexpression in cancer cells. Current data suggests that the two human orthologs may be functionally at odds in cancer cells. A screen to discover potential oncogenes expressed in gastric, pancreatic, and glioma cell lines identified RHOXF2 as a candidate cancer promoting gene (97). RHOXF2 was subsequently found to be highly expressed in a variety of cancer cell lines. Knockdown of RHOXF2 in HGC27 cancer cells, which express RHOXF2 highly, resulted in inhibition of cell growth. Conversely, overexpression of RHOXF2 in HF6 cells, which normally lack RHOXF2, resulted in the rapid development of leukemia in irradiated mice (97). RHOXF2 cDNA obtained from tumors was free of mutation suggesting that the native function of RHOXF2 protein is to induce cell proliferation and cancer development.

In contrast, RHOXF1 is implicated to be a tumor suppressing factor. A microarray screen designed to identify differentially expressed homeobox genes between lingual squamous carcinomas and surrounding normal tissue showed that RHOXF1 was downregulated in 5 of 7 samples compared (61). However, to the best of our knowledge, no active assays have been performed to examine the impact of modulation of RHOXF1 levels and tumorigenicity. This is the case, however, for the RHOX-related gene, ESX1. One study has demonstrated that a naturally produced 20-kDa C-terminal fragment of ESX1 inhibits the degradation of cyclins and leads to cell-cycle arrest (98). Another study by this group characterizing the homeodomain containing N-terminal 45-kDa region of ESX1, demonstrated that ESX1 slows tumor growth by homeodomain-dependent inhibition of K-ras transcription and amino-terminal domain-dependent reduction in tumorigenicity (60). The bivalent growth inhibition properties of the ESX1 protein observed in human tumor cells contrasts with the phenotype of feri-null mice. In those mice, ablation of Esx1 resulted in hyperplasia of the placenta and vascular abnormalities that resulted indirectly in impaired fetal growth (50). The homeodomains of mouse, rat, and human ESX1 are well conserved, in particular Helix III that primarily determines target gene specificity, so it is unlikely that the different functions observed for ESX1 are due to regulation of different sets of genes using different promoter elements. Although differences in epigenetic status of embryonic cells and cancer cells that allow permissive regulation by ESX1 cannot be ruled out. As with the majority of the molecules discussed in this review, additional experiments are required to characterize the overlapping and unique functions of the RHOX domains, in both the same cells/tumors and between tumors of different origins.

5.2. RHOX factors govern differentiation events

As previously mentioned, Jackson et al. showed that Rhox4b is expressed in ES cells and that introduction of an antisense RNA blocked RHOX4B action resulting in the inhibition of ES cell differentiation in vitro (99). Transfection with plasmids to overexpress RHOX4B resulted in advanced ES cell differentiation, as assessed by the appearance of hematopoietic, endothelial, and cardiac differentiation markers upon the removal of LIF. Additional assays for self-renewal and differentiation indicated that inhibition of RHOX4B results in the maintenance of the stem cell phenotype under low LIF availability. It is likely that the episomal vectors used to manipulate Rhox4b were equally effective in depleting all 7 paralogous copies of Rhox4. Rhox4’s function in ES cell development is consistent with a role in embryonic development where Rhox4 transcripts exhibit a developmentally regulated pattern of expression in the early embryo (100). In situ hybridization initially detected Rhox4 in the extraembryonic endoderm of E6.5 embryos, increasing between E8.5-10.5, and finally in the embryo proper, first in the anterior foregut endoderm and then in the pharyngeal pouches (100). Rhox2 has been proposed to be most similar to Rhox4 in structure and function, thus the function of RHOX2 was assessed in the same episomal system used to characterize Rhox4b (23). Depletion of Rhox2 (presumably multiple paralogs) and overexpression of RHOX2A, resulted in the same affects as manipulation of RHOX4B. The addition of Rhox2 antisense RNA did not deplete RHOX4 protein, indicating that alterations in ES differentiation were specific to RHOX2A. At present, it is still not clear why two large sets of homeobox genes overlapping in function would have been evolutionarily maintained. It’s likely that these genes display differences in relative expression during differentiation of specific populations of cells in the developing embryo. Unfortunately, clues to what structures might be involved are speculative as their relative expression in both ES cells and post-natal reproductive associated tissues is very similar (5, 10, 23, 72).

The potential function of Rhox6 has recently been examined using a different model of ES differentiation. Rhox6 is highly expressed in the placenta and post-migratory primordial germ cells (PGC) (5, 6, 28, 30). To investigate the role of RHOX6 in PGC differentiation, levels of Rhox6 were manipulated in a cell line in which EGFP under the control of the Oct3/Oct4 promoter could be used to monitor differentiation state (101). Rhox6 and the highly related (and putatively functionally redundant (30, 89)) Rhox9 were both detected in undifferentiated cells, with Rhox6 being more abundant (102). However, upon initiation of an established protocol to differentiate ES cells to PGC (103), Rhox6 levels transiently increased while Rhox9 dropped to near undetectable levels. The expression of Rhox6 correlated well with the transition from ES to PGC as assessed by several markers of PGC differentiation. Unfortunately, overexpression of Rhox6 alone had little or no significant affect on PGC differentiation (102). Ablation of Rhox6, Rhox9, or both using stably transfected anti-Rhox6/9 shRNAs (that achieved ~90% continuous knockdown) did not impair development of the epiblast. However, knockdown of Rhox6 was found to significantly impair subsequent attempts to differentiation cultured epiblasts to PGC-like cells (102). These results and the absence of germ cell phenotype in Rhox9-null animals (89), suggests that RHOX6 may be uniquely necessary for the determination of the germ cell lineage. Future studies introducing Rhox6-knockdown cells into host blastocysts may help address this, but the lack of quality probes to differentiate localization of Rhox6 and Rhox9 make the proposed studies technically challenging. The authors are not currently aware whether the previously described Rhox9-null animals are available to use as recipients and the generation of double knockouts through combination of future Rhox6-null lines would be difficult as these two genes reside in close proximity on the X chromosome.

5.3. RHOX factors promote cell survival

As described previously, Rhox5-null mice suffered from hypofertility and had reduced numbers of round and elongated spermatids in the testis (5). This depletion was due, at least in part, to increased apoptosis of meiotic germ cells in the testis. Increased numbers of TUNEL positive cells were observed as early as day 12 post partum (P12) in Rhox5-null animals, but Sertoli cell numbers in the adult were normal. Significant increases in apoptotic germ cells were observed at P17, P25, and in the adult. In the testis, apoptosis is a normal process associated with spermatogenesis, but can be aberrantly triggered by gonadotropin withdrawal, heat stress, torsion, and assault by many toxic biochemical agents. Interestingly, Rhox5-null testes exhibited an increased frequency of apoptosis in both germ cells that normally die (stage-I to -IV spermatogonia and stage-XII spermatocytes) as well as those that do not normally die (stage-V to -XI spermatocytes) (5). This apoptosis is presumed to be due to loss of a RHOX5-dependent survival factor presented by Sertoli cells to germ cells (5, 20, 21). However, characterization of RHOX5-regulated genes in 15P-1 Sertoli cells identified Unc5c, a pro-apoptosis inducing factor, as a target gene repressed by RHOX5 (95). Thus, RHOX5 may function to promote cell proliferation and survival through positive mechanism, while at the same time inhibiting cell death pathways. Alternatively, RHOX5 may normally function as a checkpoint surveillance transcription factor that, if absent, causes premature entry to the next stage and therefore an increased sensitivity to apoptosis (104). Loss of androgen signaling within the testis results in increased apoptosis beginning in mid stage VII and continuing through stage IX. Because Rhox5 is regulated by androgen, it is tempting to speculate that RHOX5 is one mediator of apoptotic survival lost when androgen is deprived. Investigation of the mechanisms of androgen-dependent survival of gonadal cells may be relevant to survival of cancer cells in androgen-responsive tumors. In support of this, prostate-targeted androgen receptor silencing constructs eradicate xenograft tumors in mice (105). However, additional studies are required to determine whether RHOX5 is a key mediator of this process and if so, to identify the key downstream factors controlled by RHOX5 that promote cell survival.

On this front, a recent study has identified ERK signaling as a target pathway for RHOX5 in cervical cancers (106). Exogenous expression of Rhox5 in TC-1 cells resulted in the downregulation of pro-apoptotic factors such as BCL-2 and upregulation of pro-survival factors BIM. The protective effect of RHOX5 is consistent with what we have previously observed in male germ cells (5) and Sertoli cell lines (20, 95). However, using a combination of phosphoantibody screens and chemical inhibitors, Kim et al. discovered that the mechanism behind the protective effects of RHOX5 lies in the regulation of the ERK1/2 signaling network and not AKT which exhibited no differences in control and Rhox5 overexpressing cells (106). As further evidence of the role of ERK signaling, while introduction of Rhox5 increased the tumor growth potential of TC-1 cells in vivo, an intra-tumor injection of ERK inhibitor stemmed tumor growth. Further insights gleaned from this study are that exogenous Rhox5 confers resistance to chemotherapeutic medications such as Paclitaxel and blocked the ability of immune cells to kill TC-1 cells. Additional studies are necessary to determine whether these protective properties are unique to RHOX5 or a general feature of the RHOX family. The latter may be the case as in addition to a potential role in proliferation, RHOXF2 may help tumor cells evade programmed cell death, as like RHOX5, it is able to repress the expression of the pro-apoptosis factor UNC5C in vitro (95). If these prosurvival properties extended to RHOXF1 or RHOXF2 in general, then it may be suitable to assess tumor biopsies for RHOX levels as it could have implications for treatment prognosis or the incidence of the chemoresistant recurrence of some cancers.

6. RHOX INTERACTING PROTEINS AND LINKS TO CANCER DEVELOPMENT AND PROGRESSION

Studies to characterize the potential biochemical actions of the Rhox genes have primarily focused on Rhox5. This is likely due in part to the fact that it apparently has the highest and broadest expression in immortalized cell models (Figure 3 and prior discussion) and also the fact its discovery had a 10-15 year head start on other members of the cluster (5, 10, 53). Homeodomain transcription factors rarely act alone to stimulate or repress their target genes and in most cases coordinate with other proteins to achieve transcriptional regulation. To date, four interacting proteins have been reported for RHOX5, some were generated in a directed search to identify RHOX5 partners, some serendipitously as the “bait” of interest “selected” RHOX5 from a complex protein mixture or expression library. These factors include menin (MEN1), prosaposin (PSAP), inhibitor of MyoD family (I-MFA; also known as MDFIC), and cell division cycle 37 (CDC37) (93, 94, 107, 108). In the final section of this review we will discuss the discovery of these factors and speculate on their potential role in complexes with RHOX5 as part of their normal function in reproductive tissues and/or abnormal roles in cancerous cells.

6.2. MEN1

The first and best-characterized protein identified best characterized RHOX5-interacting protein is MEN1, also known as MENIN, MEAI, and SCG2. MEN1 is a nuclear protein with established tumor suppressor activity that when mutated underlies the dominant familial cancer syndrome multiple endocrine neoplasia Type 1 (109). Two groups independently identified RHOX5 as a MEN1 interacting protein using co-immunoprecipitation (110) and GST-pulldown experiments in transfected cells (93). Transcripts for Men1 and Rhox5 colocalize in seminiferous tubules, suggesting thatMENl interaction is not superfluous and could contribute to RHOX5’s normal role in the testes (93). However, it is still not known whether the endogenous source of MEN1 in the testes are Sertoli cells where RHOX5 is made and functions. Although, this may be the case as upregulation of Rhox5 occurs after mutation or conditional knockout of Men1 (111, 112).

MEN1 is thought to play a key role in the Gl-S transition, serving as a checkpoint control factor during the cell cycle (109, 113). In one study, introduction of antisense cDNA to block MEN1 expression resulted in increased IEC-17 cell proliferation (114). Thus, a potential mechanism by which RHOX5 could induce tumor cell proliferation (62) would be through the binding and sequestering of MEN1 so that it could not perform its proliferation inhibiting function. Presumably, RHOX5 could act by inhibiting interaction of MEN1 with cell cycle control factors, rather than through the use of its homeodomain to drive the transcription of growth promoting genes. It is not yet known whether there would be an advantage to RHOX5-MEN1 interaction in controlling proliferation events pertinent to germ cell development. This may be unlikely as such an interaction would probably occur in Sertoli cells which have completed division by post-natal day 12 in mice, although a role in establishment of Sertoli cells cannot be ruled out. MEN1-RHOX5 interaction is more likely to have an effect on germ cells via a signal that’s translated from Sertoli cell to germ cell.

This may be the case for some of the RHOX5-regulated genes which we have begun to characterize (MacLean et al., submitted and (20, 95)). For example, MEN1 has been shown to interact with the homeobox transcription factor PDX1 in the pancreas to modulate the expression of insulin and insulin-like growth factor binding protein 2 (109, 113). Thus it is likely that RHOX5-menin interaction might also influence the expression of Ins 2 and other metabolism-related genes regulated by RHOX5. In pancreatic cells, PDX1 induces insulin transcription. Like the insulin and insulin-like growth factor genes, the Ins2 promoter has a consensus PDX1 binding site (115). Because MEN1 inhibits PDX1 access to this site on the insulin promoter (116), by analogy in the testes, MEN1 may serve to inhibit access of RHOX5 to the Ins2 promoter. We currently believe that disruption of local insulin signaling in the testes is responsible in part for increased germ cell apoptosis in Rhox5-null testes (10, 20, 21). However, insulin signaling is also known to lead to increased cellular proliferation, including germ cells, through the AKT pathway activation (117–120). Thus it is possible one of the routes of RHOX5’s tumor growth promoting ability relies on overexpression that exceeds the capability for MEN1 to block the inhibition of growth promoting factors. Future studies will need to address whether RHOX5 and MEN1 compete directly the same target gene promoters, work together to stimulate a unique subset of genes that neither factor can regulate individually, or have pertinent actions other than transcriptional control in both normal cells and cancer cells.

6.2. CDC37

Using yeast two-hybrid analysis with RHOX5 as bait, CDC37 was identified as a molecular partner for RHOX5 (94). CDC37 is a cell-cycle regulator that facilitates formation of the CDC28-G1 cyclin complex (121). This suggests that RHOX5 and CDC37 might cooperate to promote cell proliferation either as part of its normal function in growing reproductive tissues or cancerous cells. CDC37 has been thought to function primarily as an accessory factor for HSP90, primarily in the shuttling of substrate kinases (122). Endogenous levels of CDC37 are typically low and targeted overexpression in tissues that lack it leads to transformation at the same rate as cyclin D1 transgenic mice (123). It is not known whether inappropriate co-regulation of both CDC37 and RHOX5 is necessary to elicit transformation or induce proliferation. If this is the case, the absence of CDC37 could explain why high expression of RHOX5 is tolerated in Sertoli cells, which do not proliferate in the testis after the second postnatal week. This may explain our anecdotal findings that transfection of Sertoli cells (and other selected cell lines) appears to result in slowing of their growth, rather than the advance in proliferation seen in colon cancer cells (62). It would be interesting to determine the time course and cellular localization of CDC37 expression in post-natal testes. Although, if CDC37 is an essential mediator of Sertoli proliferation, then it may likely interact with other RHOX factors, as ablation of Rhox5 does not alter Sertoli cell numbers (5). Alternatively, as part of their role in the cell cycle, CDC37 and RHOX5 interaction may also function as a DNA checkpoint regulator. CDC37 may serve as a chaperone to prevent RHOX5 from coordinating the advancement of germ cells to their next step in development prior to reaching competency to do so. In support of this, transgenic mice overexpressing RHOX5 in Sertoli cells have increased DNA strand breaks in the adjacent germ cells during their maturation into elongated spermatids (104). Thus, maintaining a proper CDC37/RHOX5 ratio may be necessary for cellular health and development.

6.3. PSAP

PSAP (also known as prosaposin and SGP1), is a lysosomal enzyme activator that is necessary for the normal development of the testes and male accessory organs (124). The finding that knockouts for both Rhox5 and Psap have phenotypes in the testis supports the notion that RHOX5 and PSAP protein complexes may collaborate in the testis. However, it is unlikely that the interaction found by yeast two hybrid analyses occurs frequently in vivo. While RHOX5 protein can occasionally be seen in the cytoplasm, it is by far a nuclear protein in Sertoli cells of the testes (76, 125). In contrast, PSAP is an almost entirely cytoplasmic and often secreted protein (126). Thus, it is unlikely that these factors meet often enough to have profound normal physiological effects. However, forced overexpression of Rhox5 in transgenic mice does result in slightly more RHOX5 protein localized to the cytoplasm (104). Since many tumor cells express significantly higher levels of Rhox5 than observed normally, it is possible that PSAP may have a role in these aberrantly expressing cells. In this case, PSAP may serve as a PSAP may serve as a negative regulator of RHOX5’s transcriptional functions in the nucleus by sequestering RHOX5 in the cytoplasm.

6.4. MDFIC

The final protein which has been shown to directly bind RHOX5 is MDFIC. The MyoD family inhibitor domain containing protein, MDFIC, is an established repressor of transcription (127) and has been identified as a differentially expressed gene in gastric cancer following loss of heterozygosity (128). Unfortunately, subsequent analyses did not support a role for MDFIC in tumor development or suppression (128). Currently, the literature does not provide sufficient evidence to speculate about the functional link between RHOX5 and MDFIC either in reproductive tissues or cancerous cells, other than possesses putative inhibitory activity to factors that govern muscle development which may or may not have homologous actions in the gonads.

7. DISCUSSION AND FUTURE DIRECTIONS

The discovery of the Rhox homeobox gene cluster has opened up a new frontier in reproductive physiology and cancer biology. While the size of the Rhox cluster has diverged during evolution, all 33 genes in the mouse, 10 in rat, and 2–6 in other mammals including humans, are selectively expressed in the male and female reproductive tract, suggesting that native function of the RHOX homeodomain transcription factors involves regulating genes that promote male and female fertility. However, while normally restricted to reproductive tissues, many genes in the Rhox cluster, and their relatives Esx1 and Arx, are commonly misexpressed in tumors and immortalized cells of diverse tissue origins.

Most work to date has focused on dissecting the molecular pathways that regulate RHOX factor expression, in particular androgen signaling, and their role in supporting male and female fertility in mice. While we intend to continue to investigate the redundant and unique functions of the Rhox genes in mice, there is a clear need to expand our functional studies to the human RHOX genes. The studies described in this review have likely only begun to scratch the surface on the potential events mediated by RHOX factors. However, evidence exists, perhaps as expected of homeobox genes, for a causative role in cellular processes including proliferation, differentiation, apoptosis, angiogenesis, resistance to chemotherapies, invasion, and metastasis. These qualities are typically used to define candidate factors to serve as a “Master Switch” for the development and progression of cancer (65). Whether one or more RHOX factors serve as such a switch remains to be seen. However, more studies are warranted to investigate the potential value of the RHOX genes as markers for tumor development (particularly for those cancers which early detection techniques are currently lacking) or as targets for therapeutic intervention.