A far-upstream (-70 kb) enhancer mediates Sox9 auto-regulation in somatic tissues during development and adult regeneration.

SOX9 encodes a transcription factor that presides over the specification and differentiation of numerous progenitor and differentiated cell types, and although SOX9 haploinsufficiency and overexpression cause severe diseases in humans, including campomelic dysplasia, sex reversal and cancer, the mechanisms underlying SOX9 transcription remain largely unsolved. We identify here an evolutionarily conserved enhancer located 70-kb upstream of mouse Sox9 and call it SOM because it specifically activates a Sox9 promoter reporter in most Sox9-expressing somatic tissues in transgenic mice. Moreover, SOM-null fetuses and pups reduce Sox9 expression by 18-37% in the pancreas, lung, kidney, salivary gland, gut and liver. Weanlings exhibit half-size pancreatic islets and underproduce insulin and glucagon, and adults slowly recover from acute pancreatitis due to a 2-fold impairment in Sox9 upregulation. Molecular and genetic experiments reveal that Sox9 protein dimers bind to multiple recognition sites in the SOM sequence and are thereby both necessary and sufficient for enhancer activity. These findings thus uncover that Sox9 directly enhances its functions in somatic tissue development and adult regeneration through SOM-mediated positive auto-regulation. They provide thereby novel insights on molecular mechanisms controlling developmental and disease processes and suggest new strategies to improve disease treatments.

INTRODUCTION

Heterozygous mutations within and around SOX9 cause campomelic dysplasia (CD), a severe skeletal malformation syndrome that is often lethal at birth (1–3). Moreover, two-thirds of XY CD patients show sex reversal or abnormal genitalia (4,5), and surviving patients have been reported to experience hearing loss, seizures, global developmental delay and heart and pancreas malformations (6). These clinical features and mouse genetic and molecular studies have proven that SOX9 specifies cell fate and differentiation in many lineages, including chondrocytes, Sertoli cells, neural stem cells, pancreas progenitor cells and neural crest, neuronal, glial, heart valve, gut and kidney cells (7–14). While CD and associated malformations are due to SOX9 haploinsufficiency, SOX9 duplication causes XX sex reversal (15), and increased or ectopic expression of SOX9 has been linked to liver fibrosis, melanoma and colon, pancreas and prostate cancer (16–19). To better understand and manage SOX9-dependent diseases, it is imperative to uncover the mechanisms that underlie SOX9 qualitative and quantitative expression. Currently, however, our knowledge of these mechanisms is still in its infancy.

Sequence alterations in the 2-Mb gene desert that surrounds SOX9 have been shown to cause CD and have thereby implied that critical SOX9 cis-regulatory elements reside within or beyond this region (1–3). Accordingly, experiments with yeast (YAC) and bacteria (BAC) artificial chromosomes have revealed that the ∼70-kb sequence lying 5′ of Sox9 is sufficient to mimic most of the Sox9 expression pattern in mouse embryos, but more robust expression is achieved with 350 kb of 5′ sequence (20,21). Experiments with smaller transgenes have shown that the Sox9 promoter can drive gene expression in the spinal cord and hindbrain (22); a −1.5-Mb enhancer in the mandible region (23); a −251-kb enhancer in the cranial neural crest and inner ear; a −28-kb enhancer in the node, notochord, gut, bronchial epithelium and pancreas; a +95-kb enhancer in the telencephalon and midbrain (22) and a −14-kb enhancer, named TES, in the undifferentiated embryonic gonad and testis (21). The transcription factor Sf1 was proposed to cooperate with Sry to activate TES in the undifferentiated gonad, and with Sox9 to maintain TES activity in Sertoli cells after SRY expression has ceased. While these enhancers can activate transgenes, their actual contribution to endogenous Sox9 expression remains unknown, as are the factors that mediate the activity of most of them.

We newly uncover here an enhancer that is highly conserved and located 70-kb upstream of Sox9 and we show that this enhancer mediates a feedback loop of Sox9 auto-regulation in multiple somatic tissues. The increase in Sox9 expression generated by this feedback loop results in optimization of Sox9 function in development and adult organ regeneration. These findings provide novel insights on molecular mechanisms controlling major processes and suggest new strategies to improve disease treatments.

MATERIALS AND METHODS

Genome analyses and construction of plasmids

Sox9 sequence conservation and histone modifications were analyzed using the University of California, Santa Cruz (UCSC) genome browser (http://genome.ucsc.edu/). Sox9 elements were amplified using mouse 129 DNA and specific primers (Supplementary Table S1A) and sequence-verified on cloning. Upstream elements were multimerized using BamHI and BglII flanking sites. Reporters were assembled in pBluescript (Stratagene) using Sox9 sequences, a pACT intron (Promega), the βGeo and enhanced green fluorescent protein gene (EGFP) reporters separated by internal ribosome entry site (IRES; Clontech) and followed by a bovine Gh1 polyadenylation site. SOM was mutated using Quick Change Mutagenesis (Stratagene). Expression plasmids for Sox2, Sox8 and Sox10 were made as reported for SOX9 (24) and Sox11 (25).

Transfection and protein assays

Rat chondrosarcoma (RCS), 10T1/2 and COS1 cells were cultured and transfected as described (26). Transfection mixtures contained 1.2 µl FuGENE6 (Roche), 200 ng Sox9 reporter, 80 ng pGL3 control plasmid (Promega) and 120 ng empty or Sox expression plasmid. Cells extracts were prepared 40 h later and assayed for luciferase and β-galactosidase activities (Applied Biosystems). Reporter activities were normalized for transfection efficiency and are presented as the average with standard deviation of triplicates in one experiment representative of all other (two to five) independent experiments. Western blots of cell extracts were hybridized with anti-FLAG antibody as reported (25).

Mice

Mice were used according to federal guidelines and as approved by the Cleveland Clinic Institutional Animal Care and Use Committee. Transgenic founders were generated by DNA injection into fertilized mouse eggs. Founders and progeny were identified by lacZ polymerase chain reaction (PCR). Sox9-null embryos were generated using Sox9 conditional null alleles (27), as published (28). Data were reproduced with at least two pairs of control and tester littermates. SOM mice were generated as described (Supplementary Figure S1). Acute pancreatitis was induced in 7–13-week-old mice by intraperitoneal injection of 100 μg/kg body weight of cerulein (Sigma) eight times at hourly intervals for 2 days (29). The second day of injection was defined as day 0. Buprenorphine (Reckitt Benckiser Healthcare) was injected subcutaneously at 30 μg/ml (100 μl) as an analgesic twice a day for 3 days during and after cerulein injections.

X-gal staining and histology analysis

Standard protocols were used to stain whole embryos with X-gal, make frozen sections of paraformaldehyde-fixed embryos and tissues, stain sections with X-gal and nuclear fast red or with hematoxylin and eosin and process sections for immunostaining and counterstaining with 4',6-diamidino-2-phenylindole (DAPI). Rabbit Sox9 antibody (1:500; Millipore AB5535) was detected with Alexa Fluor® 594-conjugated goat anti-rabbit antibody (Invitrogen); goat Sox10 antibody (1:200; Santa Cruz Biotechnology, sc-17342) with Alexa Fluor® 488-conjugated donkey anti-goat antibody (Invitrogen); guinea pig insulin antibody (1:500; Dako) with Texas red-conjugated donkey anti-guinea pig antibody (1:100; Jackson ImmunoResearch) and mouse glucagon antibody (1:500; Sigma) with Cy3-conjugated donkey anti-mouse antibody (1:100; Jackson ImmunoResearch). Data were visualized with Leica DM2500 microscope, captured with Qimaging Micropublisher 5.0 RTV digital camera, and processed with Adobe Photoshop 7.0 software. Pancreas parameters, including percentages of Sox9-positive and EGFP-positive cells, were determined using NIH ImageJ software. Unpaired two-tailed Student’s t-tests were used to assess statistical significance.

RNA assays

Samples were isolated in RNAlater (Invitrogen), transferred to TRIzol (Invitrogen) and homogenized (Power Gen 125; Fisher Scientific). Total RNA was further purified using chloroform and Rneasy Mini Kit (Qiagen), treated with Rnase-Free Dnase (Qiagen) and reverse transcribed (SuperScript III First Strand, Invitrogen). Real-time quantitative reverse transcription (qRT)-PCR was performed using SYBR Green and StepOne Plus (Applied Biosystems). Primers are listed (Supplementary Table S1B). Data were calculated by the delta-delta-Ct method using Gapdh or Actb values as references. Unpaired, two-tailed Student’s t-tests were used to assess statistical significance.

Chromatin immunoprecipitation and electrophoretic mobility shift assay

Chromatin immunoprecipitation (ChIP) was performed as described (30) using RCS cells, rabbit non-immune IgG (Sigma) and Sox9 antibody (Millipore AB5535). Electrophoretic mobility shift assay (EMSA) and SOX9-expressing COS1 cell extracts were prepared as described (25). ChIP primers and EMSA probes are listed (Supplementary Table S1C and D).

RESULTS

Identification of a Sox9 enhancer at −70 kb

The goal of this study was to delineate the enhancer(s) that activate a −70/0-kb BAC transgene in Sox9-expressing somatic tissues in mouse embryos (21). We hypothesized that these enhancers have been evolutionarily preserved, and we therefore used the UCSC genome browser to identify regions conserved over >500 bp in vertebrate genomes and exhibiting logarithm-of-the-odds (LOD) scores >200. We found such regions at −14 (TES), −19, −64 and −70 kb (Figure 1A). To test these regions for regulatory function, we cloned 1 copy or 4 tandem copies 5′ of the mouse Sox9 −307/+364-bp sequence in a βGeo-IRES-EGFP reporter (Figure 1B). We transiently transfected the reporters in RCS cells, which highly express Sox9, and in mouse mesenchymal 10T1/2 cells, which weakly express Sox9 (24). The Sox9 promoter-only reporter was minimally active in both cell lines, but was robustly activated in RCS cells by a cartilage-specific Col2a1 enhancer used as a positive control (24,31) (Figure 1C). The −14, −19 and −64 kb elements displayed weak if any activity in either cell line, even as tetramers. Interestingly, the −70 kb element activated the Sox9 promoter 10-fold as a monomer and 70-fold as a tetramer in RCS cells, but was inactive in 10T1/2 cells (Figure 1C and D). This element is conserved from human to lizard and its location varies between −18 and −102 kb (Supplementary Figure S2A). In addition, a 44-bp sequence located 270 kb 5′ of the zebrafish Sox9a gene shows 73% of identity with the enhancer segment that is most conserved in higher vertebrates (Supplementary Figure S2B). ENCODE/LICR high-throughput ChIP data showed histone modifications indicative of enhancer activity at −70 kb in mouse limb and brain at embryonic day 14.5 (E14.5) (Supplementary Figure S3). We thus concluded that the −70 kb element could be a potent enhancer of Sox9.

Figure 1.

Identification of a Sox9 enhancer. (A) UCSC genome browser analysis of the 85-kb sequence upstream of mouse Sox9 and first 2 kb of transcribed sequence (chr11:112 559 200–112 645 199). The upper schematic shows peaks of conservation in vertebrate genomes relative to the mouse genome. The most conserved regions are highlighted in red, and segments with a LOD score >200 in brown. The lower schematic is an alignment of conserved regions (black boxes) in 25 genomes. (B) Schematic of Sox9 transgenes. Up to 4 tandem copies of any conserved element were cloned 5′ of the Sox9 proximal promoter and 5′ untranslated region (−307/+364 bp). The latter was linked to an exogenous intron, a reporter encoding a fusion protein of Escherichia coli β-galactosidase and neomycin resistance βgeo), an IRES, the EGFP and a bovine polyadenylation site (pA). (C) Activity of Sox9 reporters in transiently transfected RCS and 10T1/2 cells. Reporters contained no upstream element (none) or 4 copies of a 48-bp Col2a1 enhancer or Sox9 conserved element. Normalized reporter activities are presented relative to the activity of the Sox9 promoter-only reporter. (D) Activity of reporters harboring 0, 1 or 4 copies of the −70-kb element in RCS cells.

The −70-kb enhancer activates a Sox9 transgene in most Sox9-expressing somatic tissues

As a first test of the activity of the −70-kb element in vivo, we generated transgenic mice with the promoter-only and 4-copy-enhancer reporters. When stained with X-gal (β-galactosidase assay), E14.5 founder embryos harboring the promoter-only reporter showed no transgene activity in any tissue or showed weak activity in neuronal tissue (data not shown), as reported (22). In contrast, three E14.5 founder embryos (out of seven) and one mouse line (out of one) carrying the 4-copy-enhancer reporter showed identical transgene expression patterns (data not shown). We therefore analyzed in depth progeny from the mouse line. X-gal staining of E9.5 embryos revealed transgene activity in characteristic sites of Sox9 expression, including otic vesicles, notochord, neural crest, foregut and head and branchial arch mesenchyme (Figure 2A). At E14.5, both X-gal staining and EGFP fluorescence also revealed transgene activity in many sites of Sox9 expression, namely cartilage, pancreas, kidney, gut, lung, choroid plexus, hypothalamus, olfactory epithelium and salivary glands (Figure 2B and C). The transgene was inactive in heart valves and testis, despite expression of Sox9, and it was active in dorsal root ganglia, which express Sox9’s closest relative, Sox10, but not Sox9. Three-week-old pups expressed the transgene in the same tissues as embryos, as well as in liver bile ducts, which express Sox9 (Figure 2D). Hence, the −70-kb element is able to activate a Sox9 promoter transgene in most Sox9-positive somatic tissues. We therefore named this element SOM and the transgene TgSOM.

Figure 2.

Analysis of a TgSOM mouse line. (A) E9.5 TgSOM embryo stained with X-gal (blue). Left, side view of whole-mount-stained embryo. Double red arrows indicate the level of transverse sections (a and b) shown in the same panel, next to a sagittal section. Sections are counterstained with nuclear fast red. (B) X-gal staining of a mid-sagittal section through an E14.5 embryo. (C) High-magnification pictures of tissue sections from the same embryo stained with X-gal staining, analyzed for EGFP fluorescence and immunostained for Sox9 and Sox10. (D) TgSOM activity and Sox9 protein expression in P23 mouse tissues.

SOM enhances endogenous Sox9 expression in somatic tissues

We next used DNA homologous recombination to delete SOM from the mouse genome and assess its contribution to Sox9 expression (Supplementary Figure S1A and C). Although SOM and SOM mice were viable and developed normally (Supplementary Figure S1D), a significant reduction in the Sox9 RNA level was detected in the pancreas, lung, salivary gland, kidney, gut and liver of SOM fetuses (E14.5), newborns (postnatal day 0, P0) and weanlings (P23), but not in adults (P90) (Figure 3). The reduction range was 18–37%, with a mean of 23%. A similar reduction was observed in fetal brain and adult cartilage, but none in heart and testis. Thus, SOM enhances Sox9 expression in selective somatic tissues, primarily during development.

Figure 3.

Sox9 RNA levels in SOM and SOM mice. Various tissues were collected at E14.5 (n = 13 and 10, respectively), P0 (n = 7 and 11, respectively), P23 (n = 9 and 12, respectively) and P90 (n = 14 and 14, respectively). Sox9 RNA levels were measured by qRT-PCR and normalized to Gapdh RNA levels. For each tissue, data are presented as average with standard deviation relative to the value obtained in P23 SOM mice. The percentage of Sox9 RNA level in mutants compared with controls is indicated for each type of sample. *P < 0.05; **P < 0.01. ND, not determined.

SOM is necessary for normal pancreas development

Like Sox9 newborns, which are Sox9 haploinsufficient in the pancreas (32), SOM weanlings exhibited half-size pancreas endocrine islets, but no change in the pancreas’ overall size and number of islets (Figure 4A and B). Accordingly, the insulin and glucagon RNA levels were reduced by 15–20% at P0 and P23 (Figure 4C). The RNA levels of Pdx1 and Ngn3, which encode transcription factors essential for pancreas development downstream of Sox9, were similarly reduced in SOM fetuses and newborns, whereas the RNA level of Hnf6, which controls pancreas development upstream of Sox9, was unchanged (Figure 4D). Thus, SOM is necessary for normal pancreas development.

Figure 4.

Pancreas developmental defects in SOM pups. (A) Pancreas sections from P23 SOM and SOM mice were immunostained for insulin (red) and glucagon (green) to mark endocrine islets and counterstained with DAPI to identify cell nuclei (blue). (B) Number and size of endocrine islets in P23 SOM mice relative to controls. Data are presented as the average with standard deviation of values obtained for four non-adjacent tissue sections in each of four mice per genotype. Mutant values are indicated. (C) Relative levels of insulin (Ins1) and glucagon (Gcg) messenger RNA (mRNAs) at P0 (n = 7 and 10, respectively) and P23 (n = 9 and 12, respectively). (D) Relative mRNA levels of pancreas regulatory genes at E14.5 (n = 13 and 10, respectively), P0 (n = 7 and 10, respectively) and P23 (n = 9 and 12, respectively). *P < 0.05; **P < 0.01; ***P < 0.001.

SOM is necessary for prompt recovery from acute pancreatitis

To determine the importance of SOM in adult tissue regeneration, we administered SOM and SOM mice with cerulein, a cholecystokinin-like oligopeptide. As expected (29), both types of mice developed severe pancreatitis, by the end of a 2-day regimen of drug injections, as visible by loss of H&E staining in many acini (Figure 5A). Control tissues were already recovering one day later (day 1) and appeared normal at day 7, whereas mutant tissues were still severely damaged at day 1 and still recovering at day 7. Control pancreas increased its percentage of Sox9-positive cells by >7-fold by the end of injections (day 0) and returned to baseline by day 7 (Figure 5B and C). TgSOM mice subjected to the same treatment exhibited a 5- to 20-fold increase in transgene activity during this time period, suggesting a role for SOM in generating this effect (Supplementary Figure S4A–C). Proving this role, SOM mice increased their percentage of Sox9-positive cells in pancreas only half as much as control littermates by day 0, and maintained this reduced proportion stationary for 1 week (Figure 5B and C). Control pancreas raised its global Sox9 RNA level by 5.0-fold by day 1 and returned to normalcy by day 7, whereas SOM tissue raised this level by only 2.7-fold by day 1 and kept it stable through day 7 (Figure 5C). The fact that the percentage of Sox9-positive cells in control and mutant tissues was already maximal at day 0, whereas the global levels of Sox9 RNA peaked only at day 1, suggests that most cells initiated Sox9 expression by day 0 and then continued to accumulate Sox9 RNA and protein through day 1. The tempered upregulation of Sox9 expression in SOM pancreas was functionally consequential, as Pdx1 and Ngn3 were also mildly upregulated, and expression of the acinar tissue differentiation marker Amy1 (amylase 1) was more slowly recovered. Thus, SOM is necessary to forcefully upregulate Sox9 expression upon pancreas injury, and thereby allow timely regeneration of the tissue.

Figure 5.

Pancreas regeneration defects in SOM adult mice. (A) Pancreas sections from SOM and SOM mice untreated (none) or treated with cerulein and analyzed at days 0, 1, 4 and 7. Dark red H&E staining identifies healthy exocrine acini. Arrows, severely damaged acini. Ei, endocrine islets. Arrowheads, cell debris. (B) Sox9 immunostaining (red) in sections generated as in A and counterstained with DAPI (blue). (C) Quantification of the percentage of Sox9-positive cells in tissue sections similar to those shown in B. Data are presented as the average with standard deviation of values obtained for two to three non-adjacent tissue sections in each of three mice per genotype and time point. (D) qRT-PCR of Sox9, Ngn3, Pdx1 and Amy1 RNA levels (n = 3). *P < 0.05; **P < 0.01.

Sox9 is necessary and sufficient for TgSOM activity

The prevalent activity of TgSOM in somatic tissues that share few similarities beside Sox9 or Sox10 expression raised the possibility that SOM could be under the control of the Sox9 and Sox10 proteins themselves. Supporting this hypothesis, forced expression of human SOX9 protein in 10T1/2 cells led to 10- and 30-fold activation of the 1- and 4-copy-SOM reporters, respectively (Figure 6A). This effect was as robust as for the Col2a1 enhancer, a known target of SOX9 (24,33). Sox10 was half as potent as SOX9, whereas Sox8 (a SoxE protein, like Sox9 and Sox10), Sox2 (SoxB) and Sox11 (SoxC) showed little if any activity compared with SOX9 and Sox10 (Figure 6B). Thus, Sox9 and, to a lesser degree, Sox10 have the specific ability and are sufficient to activate SOM. We next generated Sox9 and Sox9 littermates carrying TgSOM and analyzed them at E11.5, just before Sox9 embryos die (28), to test whether Sox9 is necessary to activate SOM in vivo. Unlike Sox9 embryos, Sox9 embryos were unable to activate TgSOM in cartilage primordia, pancreas and kidney, which normally express Sox9 but not Sox10 (Figure 6C). They also lost the ability to express TgSOM in the lung, likely because this tissue no longer expresses Sox10 in the absence of Sox9 (34). In contrast, they maintained TgSOM activity in dorsal root ganglia, a tissue that strongly expresses Sox10. By demonstrating that Sox9 and Sox10 are sufficient to activate a SOM-driven reporter and that Sox9 is necessary for this activation in multiple tissues, these data suggest that the proteins may similarly mediate the activity of the endogenous SOM enhancer.

Figure 6.

Sox9 is sufficient and necessary to activate TgSOM. (A) 10T1/2 cells were transfected with empty (grey bars) or SOX9 (red bars) expression plasmid along with Sox9 reporters harboring no enhancer (−), 1 or 4 copies of SOM or 4 copies of Col2a1 enhancer. Reporter activities are presented relative to the activity of the Sox9 promoter-only reporter. (B) 10T1/2 cells were transfected with the 4-copy SOM reporter and 0–120 ng of FLAG-Sox expression plasmid. Top, relative reporter activities. Bottom, western blot of cell extracts hybridized with FLAG antibody. The migration level and Mr of protein standards are indicated. Arrowheads, Sox proteins. (C) X-gal-stained sections from E11.5 Sox9 and Sox9 littermates.

Sox9 directly activates SOM

Sox9 binds its known genomic targets at sites matching the Sox consensus sequence CA/TTTGA/TA/T or presenting 1 or 2 nucleotide variations in this sequence (26,35,36). It binds single sites as a monomer in Sertoli cells, and binds site pairs as a homodimer in somatic tissues, provided that these pairs are oriented head to head and separated by 3–5 bp. Applying these parameters, we found 32 evolutionarily conserved putative binding sites for Sox9, including seven pairs, in the SOM sequence (Figure 7A and Supplementary Figure S5). ChIP revealed that Sox9 specifically binds to SOM in RCS cells (Figure 7B), and EMSAs showed that SOX9 is able to bind most predicted sites in vitro, but prefers pairs (Figure 7C and Supplementary Figure S6A). The mutation of any Sox9-dimer site reduced SOM activity by 33–92% in RCS cells and SOX9-expressing 10T1/2 cells (Figure 7D and Supplementary Figure S6B). The amplitude of these effects was larger than expected for additive contributions of the sites, strongly suggesting site cooperativity. Finally, the combined mutation of multiple pairs abrogated enhancer activity in vitro as well as in transgenic embryos (Figure 7D and E). We thus concluded that Sox9 dimers directly bind to multiple sites in the SOM sequence and thereby cooperatively activate the enhancer.

Figure 7.

Sox9 directly activates SOM. (A) Identification of putative Sox binding sites in SOM. Underneath the UCSC vertebrate conservation plot of SOM, pink and purple arrowheads represent predicted sites for Sox9 monomers and dimers, respectively; green double arrows, EMSA probes and blue double arrows, SOM regions amplified by PCR in ChIP assay. (B) ChIP assay of Sox9 binding to SOM in RCS cells. The SOM a and b regions, and regions located 2-kb upstream or downstream of SOM were amplified by PCR using input material, chromatin precipitated with non-immune (n.i.) IgG or chromatin precipitated with Sox9 antibodies. PCR products are visualized following resolution in electrophoresis gels. Sox9 antibodies precipitated chromatin containing the SOM a and b regions, but not sequences flanking SOM, demonstrating specific binding of Sox9 to the enhancer. DNA markers and their length (in bp) are shown on either side of the gel. (C) EMSA of SOX9-expressing COS1 cell extracts incubated with probes 1–14. The migration levels of free probe and probe bound by SOX9 dimers (dim), SOX9 monomers (mon) and non-specific protein (ns) are shown. (D) Activity of 4-copy SOM reporters mutated (M) in the Sox9 paired sites 4–13 or in all these sites. (E) Activity of TgSOM mutated in the Sox9 binding pairs 4–13 in an E14.5 transgenic founder embryo. Ectopic X-gal staining was seen in serous glands in one of five transgenic embryos, but no staining was seen anywhere else in any embryo.

DISCUSSION

This study newly identified an enhancer of the Sox9 gene and demonstrated its underlying mechanisms and roles in vivo. The enhancer is located 70 kb 5′ of Sox9 and is highly conserved in vertebrates. It is active in multiple somatic tissues and we therefore call it SOM. It is directly targeted by the Sox9 protein, which thereby amplifies its own gene expression to benefit the development and adult regeneration of at least one organ, the pancreas.

SOX9 was proposed to be critically controlled by remote cis-elements almost 20 years ago when alterations in its surrounding 2-Mb gene desert were found to cause CD and associated malformations (1–3). Subsequent transgenic studies lent support to this concept by identifying putative tissue-specific enhancers across the gene desert, but the actual roles of these enhancers in vivo remain untested. The present study thus represents a departure from previous studies by newly identifying SOM and by demonstrating that this enhancer quantitatively contributes to Sox9 expression in multiple somatic tissues during development and adult tissue regeneration. The fact that SOM is not absolutely required for Sox9 expression supports the notion that other enhancers also participate in Sox9 transcription. Sox9 is not unique in this regard, as many other genes also are controlled by multiple, often remote, enhancers (37). The −1-Mb limb-specific enhancer of the Sonic Hedgehog (Shh) gene works solo (38), but multiple enhancers in the >800-kb-wide transcriptional archipelago of the HoxD cluster work in concert (39). Some enhancers have qualitative roles, conferring gene-specific spatiotemporal transcription, while others have quantitative roles, conferring robustness to transcription. Further studies are thus necessary to reveal the roles of putative Sox9 enhancers previously reported, unearth more enhancers and demonstrate how these enhancers functionally interact with one another and with SOM.

SOM features multiple evolutionarily conserved sequences that mediate its cooperative activation by Sox9 dimers. This cooperative mechanism fits with the observation that SOM makes a difference only when Sox9 expression is high, in development and following injury, thus presumably when enough Sox9 molecules are present to cooperatively activate SOM. It also fits with the dimeric mode of action of Sox9 that has been demonstrated in several somatic tissues, and the fact that mutations in the SOX9 dimerization domain cause CD (36,40). In contrast, these mutations do not cause XY sex reversal, and Sox9 works as a monomer in the gonad, including in TES activation (21). Thus, although Sox9 may control its own expression in the gonad, as it does in somatic tissues, it may use a distinct enhancer and mode of action to achieve this function in each tissue type. By proving that Sox9 regulates its own gene expression, our study allows to more definitively identify Sox9 as a master transcription factor, i.e. a factor that crucially contributes to determining cell fate and that ensures this function at least in part by sustaining its own expression. Other factors of this caliber include the stem cell factor Sox2 and myogenic factor MyoD (41,42). It may also include Sox10, as this factor was recently proposed to orchestrate the activity of a neural crest-specific enhancer in the vicinity of its gene (43). Interestingly, Sox9 can activate this Sox10 enhancer, and Sox10 can activate SOM. Because Sox9 and Sox10 are co-expressed in several cell types (9), it is thus possible that these proteins upregulate each other’s gene in addition to their own gene.

While TgSOM activity matched Sox9 expression and the consequences of SOM deletion in many somatic tissues, some discrepancies were observed: TgSOM was not active and SOM deletion had no effect in heart valves, despite expression of Sox9; TgSOM was active at all ages in brain and cartilage, but SOM deletion affected Sox9 expression only in fetal brain and adult cartilage; TgSOM was active in tissues that express Sox10 but not Sox9, such as dorsal root ganglia. One reason could be that Sox9 molecules are rare or kept inactive in heart valves, precluding any SOM activity. An explanation for cartilage, brain and dorsal root ganglia could be that the threshold of Sox9 or Sox10 protein needed to activate SOM is lower in TgSOM than in the Sox9 locus because SOM is present as 4 copies directly linked to the Sox9 promoter in TgSOM, but present as a single copy and 70 kb away from the Sox9 promoter in the endogenous locus. Moreover, the number of copies and the site of integration of TgSOM in the mouse genome may also facilitate transgene activity. Finally, the lack of consequences of SOM deletion in cartilage and brain is likely due to other enhancers exerting compensatory effects.

Although SOM deletion had no major impact on mice under standard conditions, RNA assays revealed decreases from 18 to 37% in Sox9 expression in several tissues during development. While such decreases would likely be inconsequential for many genes, we reasoned that they could be consequential for Sox9, as SOX9 haploinsufficiency causes CD and associated malformations. Accordingly, pancreas development was as affected in SOM mice as it is in mice lacking one Sox9 allele in pancreas progenitors (32,44). Moreover, SOM mice showed a 2-fold impairment in Sox9 upregulation when subjected to acute pancreatitis, and recovered about twice as slowly as control mice. Based on these data, the importance of Sox9 dosage and the evolutionary conservation of SOM, we predict that SOM may have additional roles in development, adult physiology and such disease settings as tissue repair, fibrosis and cancer, but that these roles may become apparent only upon co-deletion of co-acting enhancers. In addition, our data raise the attracting possibility that SOM be used as a new genetic tool for various types of scientific and clinical applications. It could be used for instance to amplify SOX9 expression in CD and other diseases due to SOX9 deficiency or to drive expression of tumor suppressors in SOX9-positive cancers.

In conclusion, the findings from this study and previous studies suggest a model whereby Sox9 expression is controlled by multiple enhancers spread over a large genomic region (Figure 8). SOM, located at −70 kb, enhances Sox9 expression in developing and regenerating somatic tissues, and thereby allows Sox9 to properly achieve its functions. TES, located at −14 kb, may drive Sox9 expression in the undifferentiated gonad and testis. Both SOM and TES mediate a positive feedback loop of Sox9 auto-regulation. TES binds Sox9 as a monomer together with its partner Sf1, whereas SOM binds multiple, cooperatively acting, Sox9 dimers. Additional enhancers, yet to identified and characterized, must contribute along with TES and SOM to achieve qualitative and quantitative expression of Sox9 in multiple cell types.

Figure 8.

Model of Sox9 transcriptional regulation. The 2-Mb Sox9 locus is shown as a loop; the three Sox9 exons as dark grey blocks and transcription as an angled arrow. SOM is bound by Sox9 dimers (paired ovals), which upregulate Sox9 expression in a positive feedback loop (large, curved arrows) in somatic tissues. TES is bound by a Sox9 monomer (single oval), which cooperates with Sf1 (hexagon) to maintain Sox9 expression in testis (small, curved arrows). Other putative enhancers are bound by transcription factors (diamonds) that interact with the basal transcriptional machinery (straight arrows) to contribute to Sox9 expression in various tissues.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online: Supplementary Table 1 and Supplementary Figures 1–6.

FUNDING

National Institutes of Health/National Institute of Arthritis and Musculoskeletal and Skin Diseases [AR46249 and AR60016 to V.L.]; Morgenthaler (to T.J.M.); Arthritis Foundation postdoctoral fellowships (to T.J.M. and P.B.). Funding for open access charge: Cleveland Clinic.

Conflict of interest statement. None declared.