Transcriptional modulation may be mediated by cis-regulatory elements distant from their target genes. Mutations in a conserved locus about 1 Mb upstream of the Shh coding region often affect Shh expression and are associated with preaxial polydactyly (PPD) defects. To understand the molecular mechanism, we analyzed a novel mouse PPD model with a T-to-A point mutation in this distant locus. A core element of mutation (CEM) with putative enhancer activity was identified by promoter activity assay and shown to contain a matrix attachment region. HnRNP U preferentially bound to the mutant but not the wild-type CEM. Interestingly, HnRNP U also bound to the 5′-UTR of the Shh gene, which was not located in the nuclear matrix in wild-type embryonic cells, as indicated by chromatin immunoprecipitation. We propose that the 5′-UTR of Shh was pulled into the nuclear matrix by HnRNP U when the CEM was mutated, and consequently affected Shh expression. Therefore, distant cis-elements may modulate gene expression by altering HnRNP U's affinity for certain mediator proteins and nuclear relocation.
Gene regulatory elements can engage in transcriptional modulation of distant target genes. Mutations in such regulatory elements may cause various diseases (1–3). Recently, several models have been proposed to explain the molecular mechanisms underlying such long-distance gene expression regulation (4,5). By analyzing the physical proximity of chromosomal fragments, the distant regulatory elements are often found to directly interact with the target genes (6). However, the intermediary proteins and other related molecules are largely unknown.
The SHH signaling secreted from zone of polarizing activity (ZPA) in the posterior mesenchymal region is crucial for determining the anterior–posterior axis in limb development. Interestingly, a highly conserved intronic sequence (termed mammals-fishes-conserved-sequence 1, MFCS1) residing 1 Mb upstream of the Shh coding sequence has been identified in many preaxial polydactyly (PPD) mutants in both human and mouse (7–11). Several studies indicated that MFCS1 may contain cis-acting elements for Shh expression regulation because no ectopic digit is observed when the MFCS1 mutant allele and inactive Shh allele are on the same chromosome (8,10). Most of these MFCS1 mutations resulted in ectopic Shh expression at the anterior mesenchyme of mouse limb buds. Elimination of this regulatory element abolished Shh expression in limb buds without affecting other Shh-expressing tissues, suggesting MFCS1 acts in a limb-specific manner (12). Not only the most distant cis-regulatory module, located 1000 kb away from the promoter proximal to Shh, MFCS1 adds more complexity by lying into the intron of an unrelated gene, Lmbr1. Unfortunately, few studies have been carried out to provide clear explanations of the molecular mechanism for such a long-range gene expression regulation (4).
Here, we reported a new mouse PPD mutant (DZ) with a novel point mutation in MFCS1 induced by N-ethyl-N-nitrosourea (ENU). Detailed analysis indicated that the nuclear factor HnRNP U displayed different affinities to the wild-type and mutant core element of mutation (CEM). The stronger binding affinity to mutant CEM may bring the enhancer to the immediate vicinity of 5′-UTR of Shh and then facilitate ectopic expression in the anterior mesenchyme of developing limb buds.
A novel PPD model with ectopic Shh expression
A novel DZ mouse strain with full (100%) penetration of the PPD phenotype was generated on C57BL/6J background by ENU mutagenesis (13). Although only hind limbs were affected in heterozygous mutants (DZm/+), both forelimbs and hindlimbs displayed defects in homozygous mutants (DZm/m). DZm/m mice also displayed hemimelia in hind limbs, with shortened and curved tibia and fibula or total loss of fibula (Supplementary Material, Fig. S1A and B). In general, the phenotypes were more severe in hind limbs than in forelimbs both in zeugopods and in autopods, although stylopods showed no obvious abnormality in DZ mutants.
Interestingly, when DZm/+ mice were crossed with CAST/Ei strain mice, the limb abnormality completely disappeared in the mutant F1 offspring. The PPD phenotype returned in some progeny when backcrossing the F1 offspring to C57BL/6J background, indicating the existence of modifier loci suppressing the PPD phenotype in the CAST/Ei background. At least two genomic regions, located on chromosomes 3 (18.5–52.5 cM) and 16 (9.61–57.8 cM), were mapped tightly to the modifiers using MIT microsatellite markers throughout the whole genome in the backcrossing mice (Supplementary Material, Table S2).
The anterior–posterior axis in the limb was controlled by the activity of the SHH protein in posterior mesenchyme of the developing limb (14). Ectopic expression of Shh in the anterior limb was frequently observed in PPD mutants (15–19). Shh expression was normal in both forelimbs and hind limbs in mutant embryos at E10.5 (data not shown). However, ectopic expression of Shh was clearly detected at the anterior margin of the mesenchyme in hindlimbs of DZm/m embryos at E12 (Fig. 1A). In forelimbs, although ectopic Shh expression was not detected, we observed ectopic PTCH1, a trans-membrane SHH receptor, expressed in the anterior mesenchyme, suggesting the existence of ectopic SHH signal (Fig. 1B). Many signals such as BMPs, FGFs and SHH signals regulated the delicate processes of limb development in a complex network (20,21). To investigate whether the ectopic SHH-secreting cells established a new functional signal center, we examined the expression of Fgf8 and Dkk1. Fgf8 was known to be expressed throughout the apical ectodermal ridge (AER) and can induce and maintain normal Shh expression (22). In DZm/m embryos, Fgf8 expression was enhanced at the most anterior region of the AER in both forelimbs and hindlimbs, compared with wild-type embryos (Fig. 1C). Dkk1, expressed in a stripe of the anterior necrotic zone (ANZ) and its posterior counterpart the PNZ, as well as in the AER, was downstream of BMP signals and was associated with the programmed cell death cascade (21). In DZ mutants, Dkk1 expression was remarkably reduced in these areas, especially in the ANZ (Fig. 2D), indicating reduced apoptosis at the most anterior region. Therefore, ectopic SHH signal secreted in the anterior regions in DZ mutants could interact with FGF and BMP signals and cause the disruption of anterior–posterior patterning and eventual limb malformation.
Enhancer within cis-acting element around mutation
Lineage analysis and mapping of the DZ mutation using microsatellite markers narrowed the PPD candidate region near D5Mit255 (Entrez gene ID 61732) on chromosome 5 (no recombination was observed between them in 49 meioses). Sequencing data identified a novel single mutation (T8545A) located in intron 5 of Lmbr1, in a reported cis-regulatory element (MFCS1) of Shh, which is located 1 Mb upstream of the transcription start site (9,12) (Supplementary Material, Fig. S1C).
Wild-type MFCS1 may contain a regulatory element for limb-specific expression of Shh, since transgenic reporter genes controlled by wild-type MFCS1 were specifically expressed in the posterior mesenchyme of mouse limb buds (9,11,23). To confirm the transcriptional regulatory activity of the core sequence neighboring the DZ mutation site, a 150 bp fragment (CEM) containing four previously reported point mutations, the Belgian1 (A/T), Belgian2 (T/C) and Cuban (G/A) mutations in human, and the M100081 (A/G) mutation in mouse (9,10), was sub-cloned into a minimal pSV40 promoter-luciferase vector and used for a promoter assay in HEK293T cells. The wild-type CEM significantly increased luciferase activity compared with the control promoter only. Meanwhile, the CEM with DZ mutation dramatically enhanced luciferase activity even more than the wild-type CEM (Fig. 2A). This observation indicated that the wild-type CEM harbored a potential enhancer. Furthermore, the enhancer with a T-to-A transition in DZ mutation was more efficient than the wild-type one.
Heterogeneous nuclear RNP U preferentially bound to the mutant CEM
Identification of the trans-acting factors for the CEM may shed light on the molecular mechanism for the long regulation of Shh expression. We speculated that some of these factors would display different binding affinities to the wild-type and mutant CEMs. To test this hypothesis, a 29 bp oligonucleotide from the CEM covering the DZ (T/A), Cuban (G/A) and M100081 (A/G) mutation sites (indicated in Supplementary Material, Fig. S1C) was labeled and incubated with nuclear extract from E12 limb buds in electrophoresis mobility shift assay (EMSA). For the wild-type probe, three specific binding bands were identified (Fig. 2B). When the DZ (T/A) mutant probe was used, bands B and C were dramatically enhanced, although band A disappeared. This revealed that the single-base substitution of the DZ (T/A) mutation had indeed resulted in the change of protein–DNA-binding status. All three bands were effectively competed with increasing concentrations (10-, 50- and 250-fold) of non-labeled cold oligonucleotides. The DZ mutant probe showed much stronger affinity to the binding factor than the wild-type probe, since even 50-fold mutant cold oligonucleotide completely competed with the radio-labeled wild-type probe, whereas >250-fold wild-type cold oligonucleotide was needed to achieve similar effect (Fig. 2B). Moreover, band B might represent a loose binding complex because it was outcompeted easily by cold oligonucleotide. Nuclear extracts from limbs of either wild-type or DZm/m E12 embryos showed no difference in EMSA results (data not shown).
To identify which trans-acting factor(s) could bind to the sequence within the CEM, we purified DNA-binding protein using streptavidin magnetic beads that tethered dimers of the 29 bp oligonucleotide used in the above EMSA assay. All samples displayed the same band patterns with different nuclear extracts (wild-type or DZm/m E12 embryo limbs) and with different oligonucleotides (wild-type or mutant oligonucleotide) (Fig. 2D and E). One explanation is that an excess of oligonucleotide was added, compared with the nuclear extract, whereas an unrelated 29 bp control oligonucleotide displayed a different pattern, suggesting the previous binding bands were specific for the CEM oligonucleotide (Fig. 2E). Finally, the mass spectrometry analysis identified several DNA-binding proteins by SDS–PAGE, including heterogeneous nuclear ribonucleoprotein U (HnRNP U, also called scaffold attachment factor A, SAF-A, 120 kDa) and Non O (non-POU domain-containing octamer-binding protein, 54 kDa) (Fig. 2D and E).
To define the identified proteins showing specific binding bands in EMSA with the mutant oligonucleotide (bands B and C), we performed supershift assay with anti-HnRNP U and anti-Non O antibodies. As a result, band B was shifted when anti-HnRNP U antibody was added (Fig. 2C), whereas anti-Non O antibody did not disturb the EMSA-binding bands (data not shown). This supershift result suggested HnRNP U as a candidate trans-acting factor that may bind to the CEM.
The binding of HnRNP U to the mutant CEM oligonucleotide was further tested by knocking down endogenous HnRNP U with specific siRNAs in HEK293T cells. As shown in Figure 3A, two siRNAs (siU-1 and siU-2) for HnRNP U downregulated HnRNP U expression by >70%. The treatment with siU-1 and siU-2 also reduced band B to ∼50% of the level of non-target control in EMSA (Fig. 3B). In contrast, overexpression of HnRNP U in HEK293T significantly enhanced band B (Fig. 3C). To further confirm the specificity of this interaction, an oligonucleotide from the osteopontin (OPN) promoter, which is known to bind HnRNP U (24), was added cold as a specific competitor in EMSA. It strongly disturbed the DNA–protein interaction with nuclear extracts from both HEK293T cells and embryo limb buds (Fig. 3B and D). All these results indicated that HnRNP U was one of the trans-acting factors binding to the CEM, with significantly enhanced affinity for the DZ (T/A) mutation.
HnRNP U interacted with both the mutant CEM and the Shh 5′-UTR
To verify the interaction between CEM and HnRNP U in vivo, we employed chromatin immunoprecipitation (ChIP) in E12 limbs of wild-type and DZm/m embryos. HnRNP U antibody precipitated the CEM from DZm/m limbs but not from wild-type limbs (Fig. 4A).This provides further support for enhanced DNA–protein binding between the mutant CEM and HnRNP U in vivo. Interestingly, the Shh 5′-UTR was also pulled down by the HnRNP U antibody in both wild-type and DZm/m mice (Fig. 4A). The transcriptional co-activator P300 has previously been described as a bridging factor binding to HnRNP U and regulatory DNA elements (25). But, in both wild-type and DZm/m limbs, the interactions between HnRNP U and the mutant CEM or Shh 5′-UTR were independent of P300.
Shh 5′-UTR was relocated to nuclear matrix with mutated CEM
As reported previously, HnRNP U could interact with DNA segments termed matrix attachment regions (MARs) and therefore interfere with chromosomal structure and gene transcription (26,27). Interestingly, MAR-associated DNA prediction (MarFinder, http://www.futuresoft.org/MAR-WIZ/) indicated that the genomic DNA surrounding the DZ (T/A) mutation site contains a MAR (Fig. 4B). To verify this in vivo, we isolated the MAR and loop fraction from chromatin of E12 embryo limbs by the LIS method (28). By comparing the ratio of loop fraction to matrix fraction, we found that both the mutant and wild-type CEMs from E12 embryo limbs were enriched in the MAR fraction (ratio <1), although the Shh 5′-UTR was mainly located in the loop fraction (ratio >1). However, the distribution of Shh 5′-UTR in the loop fraction was significantly decreased when the CEM was mutated (Fig. 4C). When the loop/matrix ratio in the anterior and posterior halves of limb buds was further assessed, the relocation of Shh 5′-UTR from loop to matrix fraction was obviously detected in the anterior halves of limb buds, where Shh was ectopically expressed (Fig. 4C). Meanwhile, the loop/matrix ratios of the Shh 5′-UTR in Shh-expressing halves, including the posterior halves of both wild-type and DZ limb buds and the anterior halves of DZ limb buds, were significantly lower than those of the anterior halves of wild-type limb buds where Shh was not expressed. This suggests a functional association between Shh expression and its 5′-UTR distribution.
A variety of experimental evidence clearly demonstrates that the CEM interacts with the Shh gene located 1 Mb away and moderates Shh transcription in cis. One likely scenario is that these two segments are pulled into proximity by a common protein complex in the nucleus. The docking of the regulatory elements, mostly within or near the CEM, and related trans-acting factors could influence the spatiotemporal expression of Shh. Following this hypothesis, mutation of the CEM might prevent or alter the formation of regulatory DNA–protein complexes and subsequently result in defects of limb development through aberrant transcriptional regulation of Shh. In this study, we provided strong evidence to support this hypothesis and suggested that HnRNP U, the multifunctional mediator protein, played a crucial role in this long-distance gene regulation.
HnRNP U has been reported to play a structural role in chromatin organization as one of the core components of the HnRNP complex, with other HnRNPs (29). One property of HnRNP U is that it forms large aggregates containing 100 or more molecules in vitro and induces DNA loops to gather around it (30). Furthermore, HnRNP U could specifically recognize MARs, which could interact with the underlying nuclear skeleton (26,27). All these features make HnRNP U a good candidate for a regulator of long-distance gene expression.
With DZ mutant mice, we confirmed that HnRNP U preferentially binds MFCS1 with the DZ (T/A) mutation rather wild-type both in vivo and in vitro. Results from chromatin IP also indicate that HnRNP U directly interacts with the Shh 5′-UTR in both wild-type and DZm/m limb cells. Moreover, we found that the fraction of Shh 5′-UTR in the nuclear matrix increased significantly when MFCS1 was mutated, accompanied by a high expression of Shh, especially in the anterior halves of limb buds, where Shh was ectopically expressed (Fig. 4C and D). Therefore, we propose that, in the anterior limbs of wild-type mouse, Shh locus is in an inactive status when mainly resided in the chromosome territory (CT). At that time, HnRNP U only binds to the Shh 5′-UTR. In the anterior limbs of DZ mouse, the enhanced interaction between HnRNP U and mutant MFCS1, as well as the homophilic adhesive interaction between HnRNP U proteins, might cause relocation of the Shh 5′-UTR into the nuclear matrix in the nuclei of mutant cells. The new location changes the surrounding environment of the Shh 5′-UTR, with different transcription factors, and therefore alters the spatiotemporal pattern of Shh expression (Fig. 5). The emergence of the ectopic SHH signal center in the anterior limb mesenchyme eventually leads to PPD defects.
Recently, Amano et al. (31) directly measured the physical distance between the MFCS1 and Shh transcripts by 3D-FISH and the chromosomal conformation capture assay. In accordance with our speculation, these two loci co-localized in a small percentage of cells (18%) among shh-expressing cells and separated when shh was turned off, suggesting a functional association between Shh expression and the Shh–MFCS1 interaction. They also found that the activation of Shh was accompanied by its looping out from its CT, and MFCS1 was required for this protrusion. But the relationship between shh's active state and looping out from its CT is still unclear.
Chromatin is arranged into repeating 50–200 kb loop domains anchored to components of the nuclear matrix or chromosome scaffold by MARs and occupies restricted CT in the nucleus (32,33). The nuclear matrix associations that we demonstrated in this study further suggest that the Shh-MFCS1 interaction is mediated by underlying dynamic nuclear matrix structures, and related binding proteins such as HnRNP U. This is consistent with Amano's results. The relocation into or nearby the nuclear matrix of Shh 5′-UTR from its CT causes the change of its surrounding transcriptional factors (Fig. 5). Since the nuclear matrix and related transcription factors are dynamically regulate throughout the nucleus among different cell types, they may play an important role in regulating gene expression.
Nevertheless, there are still some discrepancies to be addressed in future studies. First, our model does not rule out the possibility that the Shh 5′-UTR interacts with its upstream regulatory element around MFCS1 nearby region independently of HnRNP U. The fact that the co-localization of MFCS1 and Shh was not influenced by the deletion of MFCS1 in mice supports this notion (12,31). It would be much more interesting to examine whether other previously identified PPD mutations in MFCS1 also display enhanced affinity for HnRNP U. Indeed, the M100081 mutation showed a similar binding pattern to that of the DZ mutation in EMSA. Second, our model did not exclude the possibility of other factors or genomic DNA fragments participating in the interaction between the Shh 5′-UTR and MFCS1. In our backcross breeding experiment, at least two possible modifiers on chromosomes 3 and 16 (Supplementary Material, Table S2) suppress the PPD phenotype. Dissection of the molecular mechanism of this suppression may shed light on the inter-chromosomal regulation of gene expression.
In summary, our study proposes for the first time that the nuclear matrix and its associated proteins are important for the long-distance regulation of Shh expression during limb development. Identification of more long-range cis-acting regulatory elements and related binding factors will help us explore the mechanism of gene expression regulation in temporal and spatial patterns, as well as genetic information from non-coding genomic regions.
MATERIALS AND METHODS
Generation and mapping of DZ mice
C57BL/6J, C3H/HeJ and CAST/Ei mice were obtained from the Jackson Laboratory (Bar Harbor, ME, USA). The founder mouse of the DZ mutant line was generated in the C57BL/6J background by ENU mutagenesis (13). The mutation was mapped in N2 progeny backcrossed to C3H/HeJ by linkage analysis with microsatellite markers from the Whitehead/MIT database (http://www-genome.wi.mit.edu). The point mutation was determined by sequencing candidate genomic loci from DZm/m mice. This novel point mutation (T/A) formed a new restriction site for Tsp509I, which can be used for genotyping (primers in Supplementary Material, Table S3).
Skeletal staining and whole-mount in situ hybridization
Embryos were collected at different stages. Skeletal staining with Alcian Blue and Alizarin Red and whole-mount in situ hybridization were performed as described (34,35). The Shh, Dkk1 and Fgf8 in situ probes were kindly provided by Professor Rossant (Samuel Lunenfeld Research Institute, Toronto, Canada) and Professor E.M. De Robertis (UCLA, CA, USA), respectively. The Ptch1 in situ probe was synthesized from a plasmid containing Patched1 partial cDNA (926–1441 bp).
Briefly, target fragments were inserted into the pGL3-promoter plasmid (Promega). Luciferase activity was determined using a dual-luciferase reporter assay system following the manufacturer's protocol (Promega). Luciferase activity was measured using the GloMax™ 96 Microplate Luminometer with Dual Injectors (Promega).
EMSA and supershift assay
Nuclear extracts were collected from limb bud tissue of wild-type and DZm/m E12 embryos with the nuclear extract kit (Active Motif). EMSA was performed using the gel shift assay system (Promega). Probes are shown in Supplementary Material, Table S3. For the supershift assay, binding mixtures included rabbit anti-HnRNP U (N250) serum prepared in our laboratory, rabbit anti-Non O serum (kindly provided by Professor Xiao Han, Nanjing Medical University, P.R. China) or normal rabbit serum as control (Upstate).
DNA-binding protein purification
According to the kit manufacturer's protocol (Roche Applied Science), double-stranded oligonucleotides (58mers) with double repeats, phosphorylated at the 5′ end, were used for self-primed PCR. Nuclear extracts from wild-type control and DZm/m E12 embryo limbs were purified on streptavidin magnetic particles with tethered oligonucleotide. After eluate dialysis, SDS–PAGE was performed. After staining with Coomassie Brilliant Blue G350 solution, the mass spectrum was used to identify the different DNA-binding proteins.
Cell culture, siRNA transfection and western blot
Cell culture reagents were purchased from Life Technologies/Gibco BRL (Invitrogen). The siRNAs (siU-1 and siU-2) were designed to target mouse HnRNP U mRNA (RIBOBIO Co., P.R. China). Seventy-two hours after transfection into HEK293T cells, nuclear proteins were extracted by the nuclear extraction kit (Active Motif). Western blotting was performed with anti-HnRNP U (N250) serum to check the efficiency of siRNA with a standard protocol.
ChIP was carried out following the EZ-ChIP kit manufacturer's protocol (Upstate). In brief, wild-type and DZm/m E12 embryo limbs were homogenized and cross-linked with 1% formaldehyde. The following antibodies were used: anti-HnRNP U (N250), anti-p300 (C20) (SC-585X, Santa Cruz), anti-RNA polymerase II and normal rabbit IgG, the latter two supplied in the kit as a positive and a negative control, respectively. Products were evaluated for specificity by PCR with the primers listed in Supplementary Material, Table S3 for the Gapdh promoter region, DZ mutation site and Shh promoter and 5′-UTR. All the bands were quantified and normalized by their corresponding inputs.
Cells were harvested from wild-type and DZm/m E12 embryo limbs. The preparation of nuclear matrices was adapted from the procedure of Mirkovitch et al. (27). Specifically, whole limbs or the anterior and posterior halves of limbs from 30 embryos were pooled together and lysed. Nuclei were collected and incubated at 37°C to stabilize the nuclear scaffolds. Lithium 3,5-diiodosalicylic acid (LIS) was added to extract histones. Nuclear halos were collected and digested by EcoRI (New England Biolabs). After proteinase K treatment, DNA was extracted from both the soluble fraction and the insoluble pellet. Specific regions of DNA were detected by real-time PCR with SybrGreen staining on the ABI PRISM 7700 (Applied Biosystems, USA). Mouse immunoglobulin µ MAR (Ig µ) was detected as a positive control (28). All primers are listed in Supplementary Material, Table S3.
Data are presented as mean ± SEM. The two-tailed Student's t-test was used for comparison between two groups. For multiple comparisons, results were analyzed by factorial ANOVA using the GraphPad Prism 4 software (*P < 0.05; **P < 0.01). A value of P < 0.05 was considered statistically significant.
This work was supported by the National Science Foundation (30825024) and the Ministry of Science and Technology of China (2006BAI23B00, 2005CB522501 and 2006CB943500).
We thank Professor Janet Rossant (Samuel Lunenfeld Research Institute, Toronto, Canada) and Professor E.M. De Robertis (UCLA, USA) for probe plasmids; Professor Xiao Han (Nanjing Medical University, P.R. China) for control cell lines and antibodies; and Professor Qingshun Zhao for helpful discussion.
Conflict of Interest statement. None declared.