The Homeobox Transcription Factor HOXA9 Is a Regulator of SHOX in U2OS Cells and Chicken Micromass Cultures

The homeobox gene SHOX encodes for a transcription factor that plays an important role during limb development. Mutations or deletions of SHOX in humans cause short stature in Turner, Langer and Leri-Weill syndrome as well as idiopathic short stature. During embryonic development, SHOX is expressed in a complex spatio-temporal pattern that requires the presence of specific regulatory mechanisms. Up to now, it was known that SHOX is regulated by two upstream promoters and several enhancers on either side of the gene, but no regulators have been identified that can activate or repress the transcription of SHOX by binding to these regulatory elements. We have now identified the homeodomain protein HOXA9 as a positive regulator of SHOX expression in U2OS cells. Using luciferase assays, chromatin immunoprecipitation and electrophoretic mobility shift assays, we could narrow down the HOXA9 binding site to two AT-rich sequences of 31 bp within the SHOX promoter 2. Virus-induced Hoxa9 overexpression in a chicken micromass model validated the regulation of Shox by Hoxa9 (negative regulation). As Hoxa9 and Shox are both expressed in overlapping regions of the developing limb buds, a regulatory relationship of Hoxa9 and Shox during the process of limb development can be assumed.


Introduction
The human pseudoautosomal gene SHOX encodes for a homeodomain transcription factor with a crucial role during limb development and growth regulation [1,2,3]. Mutations or deletions of SHOX have been identified as the primary cause of several disorders characterized by reduced body height and skeletal deformities including the short stature associated with Turner Syndrome, Léri-Weill Dyschondrosteosis and Langer Syndrome [4,5]. In these syndromes, the skeletal malformations manifest as a mesomelic shortening of the long bones of the limbs, particularly affecting the middle portion of the upper limbs, where a shortening and bowing of the radius leads to a dorsal subluxation of the ulna (Madelung deformity). The distinctly localized clinical symptoms are explained by the specific SHOX expression pattern in the developing limbs that is seen during embryonic and fetal development and later on during childhood. In human embryos, the most striking expression is seen in the middle part of the limb buds, where SHOX is initially expressed in the undifferentiated mesenchymal tissue. At later stages, when the mesenchyme condenses and endochondral ossification takes place, SHOX is mainly found in the perichondrial layer surrounding the forming bone [6]. Using immunohistochemical methods, the SHOX protein was also detected in the chondrocytes of fetal and childhood growth plates [7,8]. These observations have implied a role of SHOX in bone development and offer an explanation for the localized symptoms seen in SHOX-deficient patients.
As SHOX is not existent in rodent genomes, developing chicken embryos present an important model system for the analysis of SHOX during limb development. In chicken, the Shox expression pattern corresponds very well to the expression in human embryos. In early stages, Shox is uniformly expressed in the central mesoderm of the limb bud. In later stages, expression is restricted to the proximal two thirds of the developing limb bud [9]. In human and chicken, SHOX expression in limb buds exhibits a small overlap with the expression of SHOX2, a highly related SHOX paralog. Whereas SHOX expression is restricted to the middle part of the limb bud in later stages, SHOX2 expression is mainly seen in more proximal regions [6,9].
The generation of such a distinct and restricted expression pattern as it is seen for SHOX requires specific regulatory input. In recent years, several cis-regulatory elements of SHOX have been discovered. SHOX expression is controlled by two alternative promoters that generate transcripts with identical coding capacity but different 59UTRs leading to different translational efficiencies of the transcript [10]. In addition, enhancer elements residing upand downstream of the gene were found to control SHOX expression [11,12,13,14,15]. However, the regulatory mechanisms that control SHOX and the molecular pathways in which it is involved during limb development remain elusive. So far, no regulators have been identified that activate or repress SHOX transcription by binding to its regulatory elements.

Generation of HOX Expression Constructs
Human HOX genes and the cofactors PBX1 and MEIS1 were amplified out of cDNA from U2OS cells (human osteosarcoma cells, ATCC) using a Flag-tagged reverse primer and were then cloned into the expression vector pcDNA4/TO (Invitrogen) via the multiple cloning site.
In vitro mutagenesis of the HOXA9and HOXD9 constructs was performed with the QuikChange Multi Site-Directed Mutagenesis Kit (Stratagene). All primers used are listed in Table S1.
For overexpression experiments, 1-2610 6 cells were transfected with 1 mg of the respective Flag-tagged HOX expression constructs cloned into pcDNA4/TO. Transfections were carried out using either the Cell Line Nucleofector Kit V (Lonza) or Lipofecta-mine2000 Transfection reagent (Invitrogen) according to the manufacturer's instructions. Medium was changed six hours after transfection.
Preparation and Reverse Transcription of RNA RNA from cell lines and chicken micromass cultures was prepared using the illustra RNA spin Mini Kit (GE Healthcare) according to the manufacturer's protocol. Reverse transcription of 1 mg RNA was performed with Superscript III Reverse Transcriptase (Invitrogen) using random hexamer and oligo dT primers.

Quantitative Real Time RT-PCR Analysis
Quantitative real time RT-PCR (qRT-PCR) was carried out using the Applied Biosystems 7500 Real-Time PCR System and Absolute SYBR Green ROX Mix (Abgene). Each sample was run in duplicates. Relative levels of mRNA expression were calculated according to the delta-delta Ct method by normalization to the expression of two different housekeeping genes (succinate dehydrogenase complex subunit A (SDHA) and peptidylprolyl isomerase A (PPIA). All primers used are listed in Table S1.

Reporter Constructs and Luciferase Assays
SHOX enhancers or different regions of the two SHOX CpG Islands were cloned upstream of a firefly luciferase reporter gene (Promega) into the pGL3-Promoter or pGL3-Basic vector, respectively. Primers used for cloning are listed in Table S1. For luciferase reporter gene assays, U2OS cells were seeded into 24 well plates and transfected with 200 ng firefly reporter construct and 200 ng of pcDNA4/TO-HOXA9 or pcDNA4/TO-HOXD9 or the respective mutants. Luciferase activity was measured in triplicates 24 h after transfection using the Dual-Luciferase Assay Kit (Promega) according to the manufacturer's protocol. This kit uses two different types of luciferase vectors, one of which serves as a control for normalization. In the case of HOXA9 however, several luciferase control vectors reacted with an unspecific increase of luciferase activity upon the overexpression of HOXA9 and were therefore not usable for normalization. Therefore, the presented data refer to the absolute activity of firefly luciferase. Experiments were repeated at least three times in triplicates with consistent results and representative data were shown.

Chicken in situ Hybridization
Whole mount in situ hybridizations of chicken embryos were performed as described previously [17]. Riboprobes were generated and digoxigenin-labeled by in vitro transcription (DIG RNA Labeling Mix, Roche) of PCR products amplified out of chicken cDNA, using the primers listed in Table S1.

Chicken Micromass Culture (chMM)
Limb buds of 30 chicken embryos (HH stage 24) were removed, pooled and washed twice with Hanks Balanced Salt Solution (HBSS; Gibco). To get a single cell suspension, the limb buds were digested with dispase (Gibco, 3 mg/ml in HBSS) for 15  and Penicillin/Streptomycin [Gibco]), the resulting suspension was passed through a 40 mm nylon filter (BD) to remove cell aggregates. Cells were counted, and the suspension was adjusted with chMM medium to a concentration of 2610 7 cells/ml. Aliquots of this suspension were treated with 50 ml of RCAS(BP) (replication-competent ASLV [avian sarcoma-leukosis virus] long terminal repeat [LTR] with a splice acceptor and Bryan polymerase) virus solution (5610 7 virus particles/ml) RCAS(BP)-Shox, RCAS(BP)-Hoxa9, RCAS(BP)-GFP or 50 ml of virus-free medium. 10 ml drops were seeded into 24 well plates and incubated for 2 h at 37uC, 5% CO 2 and 95% humidity. Subsequently, 2 ml of chMM medium were added and cells were cultivated, forming round cultures which were harvested after 3, 6, 9 and 12 days with medium change every two days. For qRT-PCR, RNA was isolated and processed as described above. For in situ hybridizations, cultures were fixed with 4% paraformaldehyde for 30-60 min and subjected to whole mount in situ hybridization as described above with probes for Shox or Hoxa9. For Alcian blue staining, chMM cultures were washed with PBS and fixed in Kahles fixing solution (30% EtOH, 0.4% PFA, 4% glacial acetic acid) at RT for 15 min. After washing with PBS, cultures were incubated in 0.05% Alcian blue solution (0.05 Alcian blue in 0.1 M HCl) at RT for 24 h and then rinsed with distilled water.

In silico Analysis of SHOX CpG Islands and Enhancer Sequences for Transcription Factor Binding Sites
To identify potential transcription factors that can bind to SHOX regulatory elements, we carried out in silico transcription factor binding site predictions using MatInspector (Genomatix Software GmbH) [18,19]. The analysis of regulatory sequences comprised the SHOX upstream enhancers CNE-5, CNE-3 and CNE-2 [12], the three downstream enhancers CNE+4, CNE+5 and CNE+9 [13] and two CpG islands in the 59 region of SHOX which contain the two identified SHOX promoters (CpG 1 chrX/ Y:504,564-505,326; CpG 2 chrX/Y:510,430-512,197; as annotated by the UCSC browser; NCBI36/hg18) [10]. For the binding site analyses, we only considered sequences within the CpG islands or the enhancers that were 100% conserved (identical) among human, dog, opossum, chicken, frog and zebrafish. Analysis of these conserved sequences of about 20-30 bp yielded predictions for 10-20 binding partners each. For five out of six analyzed enhancer sequences (CNE-3, CNE-2, CNE+4, CNE+5 and CNE+9) and for the CpG island 2, homeobox transcription factor binding sites were predicted. These predictions included binding site predictions for HOX proteins in general and in particular HOXA9, together with the typical HOX cofactors PBX1 and MEIS1. This accumulation of predicted HOXA9 binding sites together with the fact that certain HOX genes (similar to SHOX) are well-known to play an essential role during limb development (e.g. [20,21,22,23,24]), rendered HOXA9, PBX1 and MEIS1 as the most interesting binding candidates.
HOX genes encode for a highly conserved family of closely related transcription factors playing an essential role during the formation of the main body axes [25] and the axes of appendicular structures such as limbs and genital buds [26,27]. During limb development, especially the 59 genes of the HOXA and HOXD cluster (HOXA9-HOXA13 and HOXD9-HOXD13) are of special importance for the patterning of the proximo-distal axis of the developing limbs and are expressed in well-defined segmental domains along the limbs (reviewed in [20]). Here, the expression pattern of Hoxa9 overlaps with the Shox expression [9,21,28] Luciferase assays for the SHOX enhancers. SHOX enhancers were cloned upstream of a firefly luciferase into the vector pGL3-Promoter and cotransfected with a HOXA9 expression vector or the empty or mutant control vectors, respectively. Overexpression of HOXA9 or its mutants produced only low increases of comparable levels of luciferase activity arguing for a HOXA9 independent effect. (C-D) Deletion analysis of SHOX CpG Islands 1 and 2 to narrow down the site of regulatory HOXA9 activity by luciferase assays. (C) SHOX CpG Islands (schematically drawn as green bars) were cloned upstream a firefly luciferase into the vector pGL3-Basic and cotransfected with a HOXA9 overexpression vector or the empty or mutant control vectors, respectively. Upon HOXA9 expression, luciferase activity increases for CpG1 (8 fold) and for CpG2 (70 fold) (left and middle panel). As a control, CpG luciferase vectors were also cotransfected with HOXD9 expression vectors and the respective control vectors. HOXD9 was not able to evoke an increase of luciferase activity as seen for HOXA9 (right panel). (D) Subdivision of CpG2 (as indicated by green bars). Upon HOXA9 overexpression, a stronger increase of luciferase activity was seen for CpG2 part 2 than for part 1. (E) Subdivision of CpG2 part 2. CpG2 part 2a was able to evoke stronger luciferase activity compared with CpG2 part 2b. This region is therefore considered to inherit the main sites that are important for the HOXA9 mediated regulatory activity. doi:10.1371/journal.pone.0045369.g002 Overexpression Studies Identify HOXA9 as a Potential Regulator of SHOX Expression HOXA9 and the genes encoding for the HOX cofactors PBX1 and MEIS1 were PCR-amplified out of human U2OS cell line cDNA, Flag-tagged C-terminally and cloned into the eukaryotic expression vector pcDNA4/TO (Invitrogen).
HOXA9 or the cofactors, respectively, were then individually transfected into U2OS cells and expression was verified on RNA (qRT-PCR) as well as protein level (Western Blot using an a-Flag antibody; data not shown). The influence of the overexpression of these transcription factors on SHOX expression was then analyzed by qRT-PCR. The transcription factor SHOX is generally only expressed at very low levels in U2OS cells (like in all other cell lines and cultured primary cells) [29]. Overexpression of MEIS1 and PBX1 did not change the SHOX expression level compared to that of the control transfection with the empty overexpression vector. In contrast, a consistent increase of SHOX expression was detected after HOXA9 overexpression for all time points analyzed (12 h, 24 h and 48 h after transfection).
We therefore focused on the analysis of the regulatory potential of HOXA9 on SHOX expression. To rule out unspecific protein mass effects, which might have elicited the observed increase of SHOX expression after HOXA9 overexpression, we generated two different HOXA9 mutants by introducing point mutations into the most conserved parts of the homeodomain (HOXA9 mut1: K223E; HOXA9 mut2: K223E, N256del, R257P, R258G). In subsequent time course experiments, only HOXA9 wild type but not the two mutants were shown to increase SHOX expression, arguing for a HOXA9-dependent upregulation of SHOX expression ( Figure 1).

HOXA9 Binds to Distinct Sites within SHOX Promoter 2 and Activates SHOX Expression
To determine which of the SHOX regulatory sequences are targeted by HOXA9, we performed luciferase reporter gene assays. For that purpose, we cloned the known six limb specific SHOX enhancer elements (CNE-5, CNE-3, CNE-2, CNE+4, CNE+5 and CNE+9) [12,13] into the pGL3-Promoter vector and the two SHOX CpG islands into the vector pGL3-Basic. An overview of the genomic location of the SHOX regulatory elements is given in Figure 2A. The constructs were then cotransfected in U2OS cells together with HOXA9 wild type or HOXA9 mutant expression constructs or the empty expression vector, respectively. The results in Figure 2B indicate that no specific increase of firefly luciferase activity was detected in cells cotransfected with the enhancer constructs. However, the analysis of the CpG islands revealed a specific and very strong increase of firefly luciferase activity for CpG2 and a specific but mild increase for CpG1 ( Figure 2C, left and middle panel), suggesting a regulatory role of CpG2 in the HOXA9-mediated activation of SHOX expression. . qRT-PCR of precipitated DNA of a chromatin immunoprecipitation (ChIP) experiment. ChIPs were performed from U2OS cells transfected with HOXA9-wt-Flag using an a-Flag-Antibody or mouse IgG as control, respectively. Samples of immunoprecipitated DNA were checked for an enrichment of the putative binding sites compared to randomly selected sequences residing in 0.8 to 2.5 kb distance. In total, four primer pairs were established, two of which reside within SHOX promoter 2 containing potential HOXA9 binding sites, and two of which reside outside that region. For better comparability, the amount of DNA that was amplified out of the control sample (IgG precipitation) was set to 1. The two PCR products amplifying the potential HOXA9 binding sites (ChIP HOXA9 Amp1 and Amp2) show a higher enrichment of immunoprecipitated DNA compared to the control regions (ChIP HOXA9 Contr 1 and 2). ChIP HOXA9 Contr 1 and 2 both are residing more than 2 kb from the promoter. doi:10.1371/journal.pone.0045369.g004 As a control, we also investigated whether the paralog HOXD9 can also affect SHOX expression and carried out luciferase reporter assays using HOXD9 and two HOXD9 mutants (HOXD9 Mut1: K292E; HOXD9 Mut2: K292E, R326P, R327G). As shown in Figure 2C (right panel), HOXD9 was not able to increase luciferase activity comparable to the effect caused by HOXA9 overexpression, arguing for a specific effect of HOXA9 on SHOX.
Due to the strong upregulation of luciferase activity for CpG2 upon HOXA9 overexpression, we focused on CpG2. Transfection of serial deletion constructs of CpG2 enabled us to narrow down the strongest HOXA9 responsive element of CpG2 to an interval of 465 bp (CpG2 part 2a) containing the known SHOX promoter 2 (298 bp) (Figure 2 D, E). We then used three DNA sequences of equal lengths (75-77 bp) within SHOX promoter 2 and performed electrophoretic mobility shift assays (EMSA). Two of these oligonucleotides (oligo 2 and oligo 3) showed a retarded gel migration after incubation with purified GST-tagged HOXA9, indicating a binding of HOXA9 protein to these sequences (Figure 3 A, left panel). A further subdivision into three partially overlapping oligonucleotides of the same lengths (oligos 2a to 2c and oligos 3a to 3c) revealed that the oligos 2b and 3b (each 31 bp) were sufficient for GST-HOXA9 binding (Figure 3 A, middle and right panel). Oligo 2b and 3b both comprise AT-rich palindrome sequences which represent classical HOX binding sites [30]. Mutation of these AT motifs prevented binding of HOXA9 to the oligos (Figure 3 B), thus confirming that these palindromic sequences are essential for HOXA9 binding in this region. The specific binding of HOXA9 to SHOX promoter 2 in U2OS cells Figure 5. In situ hybridizations for Hoxa9 and Shox in chicken embryos (d3-d7). The whole body is imaged for d3 to d4 embryos. Emerging limb buds are marked by an asterisk, pharyngeal arches are pointed by an arrow (A-D, A'-D'). For d5-d7 embryos, the right wing bud is presented to provide a detailed view of expression in the limb bud only (E-G, E'-G'). Hoxa9 is expressed very early during embryonic development: expression is seen in d3 embryos along the vertebral axis of the posterior part of the body. In limb buds, expression starts at d3.25 (B) and persists until d6 (C-F). Hoxa9 is expressed uniformly in the mesenchyme of the limb buds (A-G). Shox is also expressed during early embryonic stages and is already visible in the pharyngeal arches of d3 embryos (A'). With the outgrowth of the limb buds at d3.25 (B'), expression is also seen in wing and leg buds. Until stage d4, expression is seen in the whole limb bud (C'-D'); in later stages, expression is restricted to the middle segments of the limb buds (E'-G'). By stage d7, expression also begins to appear along the digital rays of the autopod (G'). Expression in the pharyngeal arches persists during all developmental stages analyzed (A'-G'). doi:10.1371/journal.pone.0045369.g005 HOXA9 as Regulator of SHOX PLOS ONE | www.plosone.org was also validated by chromatin immunoprecipitation (ChIP) where the sequences identified by luciferase assays and EMSA (SHOX promoter 2) were enriched in HOXA9-immunoprecipitated DNA (Figure 4).
Usage of SHOX promoter 2 has been previously demonstrated to lead to a high translational activity of SHOX mRNA lacking the non-coding exon 1, whereas usage of SHOX promoter 1 gives rise to an mRNA with a long 59UTR containing exon 1 [10]. We therefore carried out qRT-PCR experiments on HOXA9-transfected U2OS cells to discriminate between SHOX mRNA isoforms that have been transcribed from promoter 1 or 2, respectively, using a primer pair spanning from exon 1 to exon 2. No increase of the SHOX exon 1 containing isoform was found using exon 1 specific primers ( Figure S2), validating that HOXA9-mediated activation of SHOX transcription is accomplished via promoter 2.
Taken together, these results show that HOXA9 can activate SHOX expression in U2OS cells by binding to distinct sites within SHOX promoter 2.

Analyses in Chicken Micromass Cultures Reveal a Negative Regulation of Shox by Hoxa9 during Cartilage Differentiation
To further evaluate the regulation of SHOX by HOXA9 in vivo, we used chicken embryos as a model since SHOX has no ortholog in mouse or other rodents. We first compared the expression patterns of Hoxa9 and Shox in the developing embryo, especially focusing on the expression in developing limb buds.
Whole mount in situ hybridization showed that Hoxa9 expression is already present in day3 (d3) embryos along the vertebral axis of the posterior part of the body ( Figure 5). In limb buds, Hoxa9 expression starts at d3.25, is uniformly expressed in the limb bud mesenchyme and weakens after d6. Shox expression is first seen in the pharyngeal arches and with the outgrowth of the limb buds at d3.25, expression is also detected in wing and leg buds. Until d4, expression broadens to the whole limb bud and then gets confined to the middle limb segments during later stages. By d7, Shox expression is also found along the digital rays of the autopod ( Figure 5). Hence, Hoxa9 and Shox are coexpressed in the limb buds, which is a requirement for a regulatory relationship of the two genes.
To study the HOXA9/SHOX regulatory mechanisms during cartilage formation, we used chicken micromass cultures (chMM) as previously described [16]. ChMM are an in vitro culture system that simulates the processes occurring during endochondral ossification and limb development. In this system, cultures of mesenchymal cells that were isolated from embryonic chicken limbs at HH-stage 24 spontaneously differentiate into chondrocytes and connective tissue [31,32]. The cultures can be infected with RCAS, a replication-competent retroviral vector system, enabling overexpression of a gene of interest in chicken tissue. In our study, chMM were infected with RCAS Hoxa9, RCAS Shox, RCAS GFP or were left uninfected.
To analyze the effect of Hoxa9 overexpression on Shox, we performed whole mount in situ hybridization on d3, d6 and d9 cultures. A strong Hoxa9 expression was observed in all Hoxa9infected cultures, indicating good transduction efficiencies (data not shown). After 6 and 9 days of differentiation, Shox expression was reduced in Hoxa9-infected cultures compared to control cultures (Figure 6 A) pointing to a negative regulation of Shox by Hoxa9. To quantify this effect, RNA from d3, d6, d9 and d12 cultures was isolated and gene expression levels were determined by qRT-PCR. For all time points analyzed, a 2 to 3-fold lower Shox expression in the Hoxa9-infected cultures compared to control cultures was detected (Figure 6 B). In addition, morphological analyses of the differentiation status of the chMM cultures were carried out to further evaluate the observed negative regulation of Shox by Hoxa9. After 3 days of cultivation, all cultures had formed aggregates of undifferentiated chondrocytes. Morphological differences between the control cultures and the Hoxa9or Shoxinfected cultures became apparent after 6 days of cultivation. While Shox-infected cultures grew more compactly and higher, Hoxa9-infected cultures were flattened as compared to control cultures ( Figure S3). These opposing effects of Hoxa9 and Shox on ChMM differentiation again argue for a negative regulation.
In a last step, we wanted to know if the regulation in human and chicken is accomplished by the same binding sites. Interestingly, the two identified HOXA9 binding sites within human SHOX promoter 2 reside in a highly conserved region. Compared to chicken, the identified target sequences (oligo 2b and 3b) exhibit a conservation of 74% and 77%, respectively, with the palindromic AT-rich region being conserved to 100%. We therefore decided to carry out additional EMSA experiments with purified chicken Hoxa9 protein and the homologous chicken Shox promoter 2 target sequences (chOligo 2b and 3b). Here, we could show that cHoxa9 is able to bind to the homologous DNA sequences within the promoter whereas mutated chicken oligos largely prevented the binding ( Figure 3C). Thus, it is very likely that the regulation of Shox by Hoxa9 in chMM cultures is accomplished by the same binding sites as in human U2OS cells.
In summary, in situ hybridization, qRT-PCR and differentiation studies of Hoxa9and Shoxinfected cultures indicate a negative regulation of Shox by Hoxa9 in chicken micromass cultures. EMSA experiments have shown that the same binding sites within the human or chicken promoter can be used.

Discussion
The homeobox gene SHOX is known to play a key role during limb development, and mutations or deletions lead to the limb malformations seen in LWD and Langer syndrome or to short stature without limb anomalies in patients with idiopathic short stature [3,4,5,33]. Limb malformations can be due to defects in the coding region of the gene but also to deletions of regulatory elements on either side of the SHOX gene [11,12,13,34,35]. These findings demonstrate that the correct function of SHOX does not only depend on the correct protein sequence and folding but also on accurate regulatory processes that direct the appropriate expression of SHOX. The aim of this study was to identify regulators of SHOX expression that interact with SHOX enhancer or promoter elements. These studies remain challenging as most of the standard techniques for the analysis of DNA-protein interactions such as ChIP-Seq (chromatin immunoprecipitation) or SELEX (Systematic Evolution of Ligands by EXponential Enrichment) are directed to identify DNA sequences where the protein of interest can bind to. However, for the identification of proteins that bind to a given DNA sequence, as in our case, no standard high throughput methods have been developed. We therefore initially used an in silico approach and identified HOX proteins as putative binding partners of SHOX regulatory elements. Subsequent overexpression experiments then revealed that HOXA9 activates SHOX expression by binding to promoter 2 in U2OS cells (Due to the very low expression level of SHOX in cell lines and tissues, a resulting down-regulation of SHOX after HOXA9 depletion is impossible to detect with confidence. This low expression level is a characteristic for some transcription factors, including SHOX and is a limiting factor for certain experimental approaches, including knock-down studies). Also in chicken micromass cultures, Shox expression was regulated by Hoxa9. In contrast to the findings in U2OS cells, Hoxa9 acted as a repressor of Shox expression in chMM, suggesting that, dependent on the cellular environment, HOXA9 can either function as an activator or a repressor of SHOX.
All 39 human HOX proteins share a highly conserved homeodomain which recognizes the same core DNA consensus sequence [30]. The individual binding specificity of each HOX protein is accomplished through cooperative binding with cofactors, whose abundance is dependent on the cellular environment and subject to spatio-temporal regulation. Depending on the activation of different cofactors and signaling pathways in different tissues, individual HOX genes can switch from activators to repressors of gene transcription [30,36]. The most common HOX cofactors are the TALE-homeodomain proteins PBX1 and MEIS1, but further cofactors have been identified recently and are expected to be identified in the future [37]. We also have examined the influence of cofactors on HOXA9mediated SHOX regulation in U2OS cells by coexpressing HOXA9 with PBX1 and/or MEIS1 but no differences in the SHOX expression level were detected compared to when solely HOXA9 was expressed (data not shown). Thus, we speculate that other, additional cofactors may modulate the effect of Hoxa9 in U2OS and chMM cultures. We also cannot exclude that other HOX proteins might influence SHOX expression levels in cells or model systems other than the ones assessed in our study.
By luciferase assays, we have shown that HOXA9 has an activating effect on SHOX promoter 2 in U2OS, which was not seen in control experiments where HOXA9 mutants or HOXD9 were used, confirming the specificity of the binding. In EMSA experiments, the HOXA9 binding site could be narrowed down to two AT-rich palindrome sequences of 31 bp within promoter 2. These sequences were specifically bound by HOXA9 but not by other homeodomain proteins such as SHOX. The identification of two HOXA9 binding sites within SHOX promoter 2 is concordant with recent findings where homotypic clusters of transcription factor binding sites for the same transcription factor within promoter regions were identified [38]. This clustering of transcription factor binding sites is especially prevalent in the proximal promoter regions of transcription factor genes with nearly two thirds of these genes exhibiting multiple binding sites for other transcription factors.
A similar clustering of binding sites is also seen in the chicken Shox promoter where the two binding sites identified in human are highly conserved. In total, the identified target sequences (oligo 2b and 3b) exhibit a conservation of 74% and 77%, respectively, with the palindromic AT-rich region being conserved to 100%. As we could show a binding of cHoxa9 protein to the homologous DNA sequences within the chicken Shox promoter, it is very likely that the regulation of Shox by Hoxa9 in chMM cultures is accomplished by the same binding sites as in human U2OS cells.
Formation of limb buds and their continued outgrowth depend on an integrated highly conserved network of multiple different signaling molecules and gene regulators. The two major signaling centers during limb development are the Apical Ectodermal Ridge (AER) and the Zone of Polarizing Activity (ZPA), which mediate their activities by specific secreted signaling molecules in the limb buds. Fgfs (Fibroblast growth factors) are the main signaling factors produced by the AER [39,40], whereas the ZPA mainly acts via the secreted morphogen Shh [41]. Experiments in chicken have shown that Shox expression is regulated by different Fgf molecules but independent from Shh signaling [6,9]. Thus, Shox expression depends on AER but not on ZPA signaling.
Correct limb bud formation also requires specific expression of Hox genes of the A and D clusters, in particular those of the 59 end (i.e., Hoxa9 to Hoxa13 and Hoxd9 to Hoxd13). Studies in mice showed that loss-of-function mutations in these genes strongly impair limb morphology with patterning defects generally affecting particular regions of the developing limb (reviewed by [20]). Hox genes are therefore thought to determine the individual segment identity of the limb skeleton. Hoxa9 2/2/ Hoxd9 2/2 double knockout mice display a phenotype in the stylopod, which is shortened and malformed compared to wild type mice [21]. However, due to the lack of a SHOX ortholog in the mouse genome, these results cannot be directly transferred to the human system. In humans, only a few HOX genes have been associated with a disease phenotype. While it is known that mutations in HOXA13 or HOXD13 can cause specific limb phenotypes, such as hand-foot-genital-syndrome [42,43] or synpolydactyly [44,45], no HOXA9-related disease phenotype has been identified in patients so far, leaving the role of HOXA9 during human limb development unclear. SHOX mutations in humans mainly affect the bones of the zeugopod region (LWD, Langer syndrome) which is in conformity with the most prominent expression of the chicken and human SHOX gene in the middle part of the developing limb [6,9]. Due to the overlap of expression of Shox and Hoxa9 in the developing chicken limb it is conceivable that HOXA9 and SHOX are in a regulatory relationship during human limb development. The negative regulation of Shox by Hoxa9 observed in chMM (a model for chondrogenesis and bone formation) argues for a likewise negative regulation of these processes during limb development; however the expression data in chicken suggest that the regulation is not exhaustive, at least not at the surface of the limb.
The negative regulation of Shox by Hoxa9 observed in chMM is also supported by the model of Zakany and Duboule [20] where Hox genes are proposed to play fundamental roles in determining AER and ZPA functions during mammalian forelimb development. According to this hypothesis, Hox genes of group 10-13 are involved in the induction of the ZPA, whereas the lower Hox genes are important for AER formation where in particular Hoxa9 is known to be able to induce Fgf expression. Since it has been shown that Shox expression is downregulated by AER signaling molecules (Fgfs and Bmps), Shox might be negatively regulated by Hoxa9 in both a direct (by binding to the Shox promoter) and indirect manner (by inducing AER signaling).
In summary, we have successfully identified a regulator of SHOX expression and provide evidence for a direct regulation of SHOX by HOXA9. These findings may further contribute to unravel the diversified regulatory networks during limb development and may also help to improve our knowledge of the etiopathogenesis of short stature.

Web addresses
UCSC genome browser: http://genome.ucsc.edu/ Genomatix software suite with MatInspector program: http:// www.genomatix.de/en/produkte/genomatix-software-suite.html  Figure S2 RT-PCR using primer spanning SHOX exon 1 to 2 confirms the regulation via promoter P2. (A) Schematic representation of the two isoforms that are transcribed from promoter P1 or P2, respectively. With qPCR experiments using a primer pair spanning from exon 1 to 2 (depicted by the two opposing arrows) it is possible to discriminate between the two isoforms. (B) RT-PCR using primers spanning exon 1 to 2 using the same cDNA samples in which we had seen an increase of SHOX expression after HOXA9 overexpression (for comparison see Figure 1; here, a primer pair spanning exon 5 to 6 was used). No increase of SHOX expression can be seen for exon 1 to 2 confirming that HOXA9 mediated activation of SHOX transcription is accomplished via promoter 2.