Conserved cis-regulatory regions in a large genomic landscape control SHH and BMP-regulated Gremlin1expression in mouse limb buds
- 6k Downloads
Mouse limb bud is a prime model to study the regulatory interactions that control vertebrate organogenesis. Major aspects of limb bud development are controlled by feedback loops that define a self-regulatory signalling system. The SHH/GREM1/AER-FGF feedback loop forms the core of this signalling system that operates between the posterior mesenchymal organiser and the ectodermal signalling centre. The BMP antagonist Gremlin1 (GREM1) is a critical node in this system, whose dynamic expression is controlled by BMP, SHH, and FGF signalling and key to normal progression of limb bud development. Previous analysis identified a distant cis-regulatory landscape within the neighbouring Formin1 (Fmn1) locus that is required for Grem1 expression, reminiscent of the genomic landscapes controlling HoxD and Shh expression in limb buds.
Three highly conserved regions (HMCO1-3) were identified within the previously defined critical genomic region and tested for their ability to regulate Grem1 expression in mouse limb buds. Using a combination of BAC and conventional transgenic approaches, a 9 kb region located ~70 kb downstream of the Grem1 transcription unit was identified. This region, termed Grem1 Regulatory Sequence 1 (GRS1), is able to recapitulate major aspects of Grem1 expression, as it drives expression of a LacZ reporter into the posterior and, to a lesser extent, in the distal-anterior mesenchyme. Crossing the GRS1 transgene into embryos with alterations in the SHH and BMP pathways established that GRS1 depends on SHH and is modulated by BMP signalling, i.e. integrates inputs from these pathways. Chromatin immunoprecipitation revealed interaction of endogenous GLI3 proteins with the core cis-regulatory elements in the GRS1 region. As GLI3 is a mediator of SHH signal transduction, these results indicated that SHH directly controls Grem1 expression through the GRS1 region. Finally, all cis-regulatory regions within the Grem1 genomic landscape locate to the DNAse I hypersensitive sites identified in this genomic region by the ENCODE consortium.
This study establishes that distant cis-regulatory regions scattered through a larger genomic landscape control the highly dynamic expression of Grem1, which is key to normal progression of mouse limb bud development.
KeywordsLacZ Expression Grem1 Expression Evolutionary Conserve Region Posterior Mesenchyme Fibroblast Growth Factor Signal Transduction
During embryonic development, the spatio-temporal expression of transcriptional regulators and morphogenetic signals is tightly controlled. In fact, different types of congenital malformations are caused by mutations affecting specific cis-regulatory and non-coding genomic regions that alter the expression of genes or gene clusters in specific tissues (reviewed in ref. [1, 2, 3]). Genome-wide functional mapping revealed many of the large genomic landscapes that control the expression of developmental regulator genes in a global and tissue-specific manner . This study and others reveal that the cis-regulatory regions controlling gene expression in a particular tissue are often located several hundred kilobases (kb) or even megabases (Mb) up- or downstream of the transcriptional start sites in other loci, and may control the expression of several genes [5, 6, 7]. Recently, it has been shown how multiple cis-regulatory regions interact to control the expression of the 5’HoxD gene complex in the presumptive digit territory. The transcriptionally active part of the 5’HoxD gene cluster forms a so-called regulatory archipelago in which dispersed cis-regulatory elements co-operate to control gene expression in the distal limb bud by interacting with proximal promoters over large distances . In fact, regulatory landscapes with distant and dispersed cis-regulatory regions seem to be a recurring theme in the dynamic spatio-temporal regulation of genes that are essential for limb bud development (reviewed in ref. [2, 9, 10], see also below).
Vertebrate limb bud development is controlled by interactions between two main signalling centres, the apical ectodermal ridge (AER) and the zone of polarizing activity (ZPA) located in the posterior limb mesenchyme. These two signalling centres interact as part of a self-regulatory feedback signalling system involving several signalling pathways . In particular, ZPA cells produce the Sonic Hedgehog (SHH) signal, which together with GREMLIN1-mediated antagonism of Bone Morphogenetic Proteins (BMPs) in the posterior-distal limb bud mesenchyme propagates Fibroblast Growth Factor (FGF) signalling in the AER [11, 12, 13, 14]. This SHH/GREM1/AER-FGF feedback-signalling loop promotes distal progression of limb bud outgrowth and formation of the autopod which gives rise to carpals and digit rays (reviewed in ref. ). SHH expression in the posterior limb bud mesenchyme is controlled by a cis-regulatory region located about 800 kb upstream of the transcriptional start site. Deletion of this ZPA regulatory region (ZRS) results in loss of distal limb structures , while point mutations within the ZRS result in anterior ectopic Shh expression and duplications of the thumb and/or anterior digits in different species [17, 18]. In cells actively transcribing Shh, the ZRS loops out of its chromosomal territory to the Shh transcription unit, which reveals how distant cis-regulatory elements control gene expression . Recently, several transcriptional regulators have been identified that control Shh expression by directly interacting with the distant ZRS region. These include HOX proteins, the bHLH transcription factor HAND2, and ETS transcription factors, providing a glimpse of the complex transcriptional regulation of Shh in the posterior limb bud mesenchyme [20, 21, 22].
SHH signalling is transduced by the GLI family of transcriptional regulators and inhibits the constitutive processing of the full-length GLI3 activator to its repressor form GLI3R (see e.g. ref. ). Of the three GLI transcription factors expressed during limb bud development, only Gli3 is essential on its own (see e.g. ref. ). The inactivation of Gli3 alters morphogenetic feedback signalling and results in formation of additional anterior digits [23, 25, 26, 27]. In particular, GLI3 restrains the proliferation of mesenchymal progenitors in the anterior limb bud mesenchyme and promotes initiation of digit ray chondrogenesis by directly repressing Grem1 expression during handplate formation .
We have previously shown that the expression of Grem1 in the limb bud mesenchyme is controlled by a large genomic landscape downstream of its transcription unit . In fact, several of the classical limb deformity mutations that disrupt distal limb bud development and formation of metanephric kidneys in the mouse are caused by deletions affecting this cis-regulatory landscape rather than directly altering the Grem1 gene products [7, 13, 28]. Molecular and genetic analysis in the mouse identified a 70 kb genomic region located downstream of the Grem1 transcription unit within the neighbouring Formin1 (Fmn1) gene that is required for Grem1 expression in the limb bud mesenchyme [7, 29]. Detailed analysis of this Grem1 genomic landscape revealed similarities with the global control region (GCR) that controls the expression of 5’HoxD genes in the limb bud mesenchyme , but did not reveal the structural nature and transacting factors and/or signalling pathways controlling these cis-regulatory regions.
To gain further insights into the Grem1 landscape, the potential cis-regulatory activities of the three highest evolutionarily conserved genomic regions within the 70 kb Grem1 critical region were analysed. In addition to its expression in the posterior mesenchyme, emphasis was also given to the coordinated distal-anterior expansion of Grem1 expression, which is key to orderly progression of limb bud development [14, 30]. Combining BAC with conventional transgenic approaches, we identified a 9 kb genomic region that is able to recapitulate major but not all aspects of the dynamic spatio-temporal Grem1 expression in the limb bud mesenchyme. This 9 kb region was termed Grem1 Regulatory Sequence 1 (GRS1), and contains a core sequence that is essential to express a LacZ transgene under control of a ß-globin minimal promoter in a Grem1-like pattern. The GRS1 region drives gene expression into the posterior and subsequently distal-anterior mesenchyme, i.e. reproduces aspects of the distal-anterior expansion of Grem1 expression during progression of limb bud development. By crossing GRS1 transgenic mice into different mutant contexts, we establish that the GRS1 region is controlled by inputs from both the SHH and BMP signalling pathways in limb buds. In addition, chromatin immunoprecipitation (ChIP) shows that GLI3 interacts with the conserved HMCO1 sequences in the GRS1 region. The functionally relevant cis-regulatory regions identified by the present and two previous studies [31, 32] map to the endogenous DNAse I hypersensitive sites within the Grem1 genomic landscape recently identified by the ENCODE consortium [33, 34]. This indicates that the Grem1 cis-regulatory landscape is composed of at least five active cis-regulatory regions that control the spatio-temporal expression of Grem1 in mouse embryos and limb buds.
Results and Discussion
Identification of conserved limb bud regulatory regions within the Grem1genomic landscape
Spatio-temporal activity of the GRS1region during mouse limb bud development
To analyze comparatively the activity of the GRS1 region with respect to the spatio-temporal regulation of Grem1 expression, transgenic mouse strains expressing the GRS1-ßglob-LacZ reporter construct were established (Figure 4). Seven transgenic founders were obtained. In three independent strains, LacZ was expressed in the posterior fore- and hindlimb bud mesenchyme in a pattern comparable to the transgenic founder embryos (Figure 3B and data not shown). Two of these three transgenic strains were analyzed in detail and used to study the spatio-temporal LacZ distribution (Figure 4B, C and Additional file 4). Initial analysis revealed that ß-galactosidase activity of the LacZ transgene was detected earlier in one strain and is significantly higher than in the other strain (Figure 4B, compare to Figure 4C). As levels did not change during subsequent generations (data not shown), this is likely due to differences in transgene copy number and/or integration site. Therefore, the copy number of both strains was determined by real-time qPCR analysis (Additional file 4). This analysis revealed that ten copies of the GRS1-ßglob-LacZ transgene were integrated into the genome of the strain expressing higher LacZ levels (GRS1 tg10 ), while only two copies were detected in the GRS1 tg2 strain (Additional file 4). In GRS1 tg10/+ embryos, ß-galactosidase activity is first detected at ~ E10.25 in the posterior forelimb bud mesenchyme and continuously increases until ~ E11.75 (Figure 4B). From ~ E11.0 onwards, scattered positive cells were detected in the anterior mesenchyme, and this anterior expression increased during distal limb bud outgrowth to form a crescent in the distal-anterior autopod (right panels, Figure 4B). By E11.75, ß-galactosidase activity was rather variable, such that the anterior crescent separated from the posterior domain in some forelimb buds. In GRS1 tg2/+ embryos, the spatio-temporal pattern is similar, but expression levels are significantly reduced due to the lower transgene copy number, resulting in detection of ß-galactosidase activity from only ~ E10.5 onwards (Figure 4C, arrowhead). In both strains, GRS1 drove expression of the ßglob-LacZ reporter specifically in the limb bud mesenchyme, despite some low ß-galactosidase activity detected in the developing eyes of GRS1 tg10/+ transgenic embryos (Additional file 4). This analysis reveals the robust nature of the GRS1 region, which functions in positive regulation of Grem1 expression in the limb bud mesenchyme. The posterior domain of ß-galactosidase activity at E10.5 (Figure 4B) is comparable to the Grem1 transcript distribution in wild-type limb buds at earlier stages (E9.5-E10.25, Figure 4D). This temporal delay is likely due to postponed transcriptional activation and/or up-regulation of the GRS1 transgene as the establishment of a posterior LacZ expression domain is also only apparent at E10.5 in GRS1 tg10/+ transgenic limb buds (Additional file 5). Therefore, additional cis-regulatory regions likely control the temporally correct early onset of Grem1 expression. Similar delays in the onset of LacZ reporter gene expression have been previously observed by analyzing the ZRS cis-regulatory region that controls Shh expression in mouse limb buds . During distal progression of limb bud development, endogenous Grem1 expression expands distal-anterior within the developing handplate and begins to fade by E11.75 (Figure 4D), due to termination of the SHH/GREM1/AER-FGF feedback loop and GLI3-mediated repression in the anterior limb bud mesenchyme [11, 26, 39, 40]. In contrast, ß-galactosidase activity remains high in the posterior limb bud mesenchyme of GRS1 transgenic embryos and the distal-anterior expansion of its expression is significantly delayed and only occurs as transcription of the endogenous Grem1 locus is starting to terminate (E11.5 onwards Figure 4B, C, compare to Figure 4D).
The GRS1 cis-regulatory region integrates inputs from the SHH and BMP signalling pathways
BMPs control Grem1 activation in the posterior limb bud mesenchyme and directly modulate its expression as part of the self-regulatory SHH/GREM1/AER-FGF feedback signalling system. Therefore, a conditional Smad4 loss-of-function allele  in combination with the Prx1-Cre recombinase strain  was used to inactivate Smad4 (Smad4 Δ/ΔM ) and thereby canonical BMP signal transduction in the limb bud mesenchyme. In Smad4 Δ/ΔM mutant limb buds, endogenous Grem1 expression is down-regulated but not lost, while Shh expression remains (Benazet et al., submitted). In Smad4 Δ/ΔM GRS1 tg10/+ forelimb buds, ß-galactosidase activity appears normal in the posterior mesenchyme, while distal-anterior expansion fails to occur (Figure 6C, compare to Figure 4B). This indicates that GRS1 activity is modulated by SMAD4-mediated signal transduction during the progression of limb bud development. These results indicate that also the distal-anterior expansion of Grem1 expression depends on BMP activity and agree with the observation that genetic lowering of Bmp4 results in a global reduction of Grem1 expression in the limb bud mesenchyme .
Finally, in Grem1-deficient embryos, BMP activity is increased due to reduced BMP antagonism, which results in up-regulation of non-functional Grem1 transcript . As BMP signal transduction modulates GRS1 activity during advanced limb bud development (Figure 6C), we also determined the potential influence of Grem1 deficiency on GRS1 activity. In Grem1 Δ/Δ GRS1 tg10/+ E10.5 limb buds ß-galactosidase activity is initially similar to GRS1 tg10/+ limb buds (Figure 6D, compare to Figure 4B). Subsequently, ß-galactosidase activity is increased and by E11.75 the entire Grem1 Δ/Δ GRS1 tg10/+ limb bud is positive (Figure 6D). These alterations of ß-galactosidase activity in Smad4 Δ/ΔM and Grem1 Δ/Δ forelimb buds reveal that GRS1 activity is extensively modulated by changes in BMP activity, in particular also the anterior expansion of its expression. Taken together, this genetic analysis (Figure 6) shows that GRS1 activity critically depends on SHH and is modulated extensively by BMP signal transduction. This analysis identifies the GRS1, located ~70 kb downstream of the Grem1 transcription unit as a cis-regulatory region that integrates inputs by both SHH and BMP signal transduction. However, as previous analysis provided good evidence that Grem1 is activated by mesenchymal BMP signalling upstream of establishing the SHH/FGF feedback loop , these studies point to the existence of additional unknown BMP response regions that activate Grem1 expression in the posterior limb bud mesenchyme.
Endogenous GLI3 proteins are part of the chromatin complexes interacting with specific parts of the GRS1region in mouse limb buds
Spatio-temporal Grem1 expression is regulated by the interaction of multiple cis-regulatory regions far downstream of the transcription unit
Previous genetic analysis showed that Grem1 transcription in limb buds is initiated by BMP signalling and up-regulated under the influence of SHH and AER-FGF signaling as limb bud development progresses . Finally, Grem1 expression is terminated concomitantly with the initiation of digit formation by high levels of FGF signal transduction and GLI3-mediated repression in the anterior mouse limb bud mesenchyme [26, 40]. This dynamic spatio-temporal regulation of Grem1 expression indicates that the activity of different cis-regulatory elements may change over time. The spatio-temporal activity of the GRS1 transgene (Figure 4B) indicates that it is primarily regulated by the SHH/GREM1/AER-FGF feedback signaling system as limb bud development progresses. In particular, its activity is first apparent when the SHH/GREM1/AER-FGF feedback is already established, and the expected termination does not occur, as ß-galactosidase activity remains in the posterior mesenchyme after the endogenous Grem1 transcripts have been down-regulated (Figure 4B, D). These results indicate that FGF-mediated termination of Grem1 expression and the underlying self-regulatory feedback signalling system  does not occur by FGF signal transduction impacting the GRS1 region. In addition, neither the 70 kb critical region of the Grem1 landscape (Grem1-LacZ, Figure 2A) nor transgenes derived from this region (Figures. 234 and ref. ) are able to accurately recapitulate the entire spatio-temporal distribution of Grem1 transcript in the limb bud mesenchyme. Therefore, either the interactions among these elements or additional as yet unknown cis-regulatory regions located outside the critical region must also participate in the regulation of Grem1 expression. Indeed, there is circumstantial evidence for the latter, as deletion of the GC-rich Fmn1 coding exon 9 (located ~200 kb downstream of the Grem1 transcription unit) results in non-complementation with a Grem1 null allele, causing a hypomorphic Grem1 limb skeletal phenotype, and significantly reduced Grem1 expression (ref.  and A.Z and R.Z., unpublished data).
In this study, we identify GRS1 as a distant cis-regulatory region that encompasses the highly conserved HMCO1 region and show that it controls important aspects of Grem1 expression in the posterior and distal-anterior limb bud mesenchyme. GRS1 activity depends critically on SHH signalling and endogenous GLI3 protein complexes interact with specific regions within HMCO1 in wild-type limb buds. In addition, the anterior expansion of GRS1 activity is also regulated by BMP signal transduction. These results together with previous studies reveal the large genomic architecture composed of several distant cis-regulatory regions that control the highly dynamic Grem1 expression in response to signalling inputs from the BMP, SHH and FGF pathways. It is likely that the interactions among these dispersed cis-regulatory regions provide the dynamic Grem1 expression in limb buds with the necessary robustness, but at the same time allow for evolutionary plasticity of its expression .
All the experiments were conducted in strict accordance with Swiss law following the 3R principles and the conduct defined by the Basel Declaration. All studies involving mice were classified as grade zero by the Animal Welfare and Ethics Commission of both cantons of Basel and Argovia, which implies minimal suffering.
The Shh loss-of-function allele , the Gli3 extra-toes-J allele , the Grem1 null allele  and the GRS1 tg/+ alleles (this study) were maintained in an NMRI background, while the Smad4 conditional allele (Smad4 flox , ) and the Prx1-Cre transgene  were maintained in a C57BL/6 J background.
BAC constructs were engineered using ET cloning as previously described . The deletions ΔHMCO1, ΔHMCO2 and ΔHMCO3 were induced in the Grem1-LacZ BAC clone using a Kanamycin resistance selection cassette. All primer sequences are available upon request. See Additional file 8 for the genomic coordinates of HMCO1-3.
An expression cassette was constructed in pKS-Bluescript using LacZeocin reporter under control of the minimal human ß-globin promoter sequences . Various combinations of HMCO sequences were inserted downstream into this cassette. The HMCO1, -2 and -3 core sequences were amplified by PCR directly from mouse genomic DNA (primers available on request). The 9 kb GRS1 region was initially subcloned as a 9.6 kb Bgl2 fragment from BAC RP23-113 H17 (chr2: 113,611,499-113,847,299 (mm10). BacPac Resources, Children’s Hospital, Oakland, USA). All constructs were linearized with Ksp1 before microinjection.
Generation of transgenic founder embryos and transgenic strains
BAC and ßglob-LacZ transgenic constructs were injected into the pronucleus of fertilized mouse eggs. Several founder embryos for each construct were scored for ß-galactosidase activity (Additional file 3). For the GRS1 ßglob-LacZ transgene, seven independent transgenic mouse strains were established by crossing founders into the NMRI background. Transgene copy numbers were determined by real time qPCR  using the Bio-Rad CFX96 Real-Time PCR System in combination with the iQ SYBR Green Supermix (Bio-Rad). For each mouse, 20 ng of genomic DNA were analysed in triplicate. The primers to amplify the DBH genomic region were used for normalization and non-transgenic littermate DNA was used as control. The normalized control levels were set to 2 as the primers also amplified both alleles of the endogenous HMCO1 regions. The 2-ΔΔCt formula was used to calculate and normalize transgene copy numbers. In addition to three different F1 males, three embryos from the F2 and three embryos from the F3 generation were analysed for the two transgenic strains shown to ascertain stable transmission of the transgenes. Primers sequences are available on request.
Detection of ß-galactosidase activity
Embryos were isolated in ice-cold PBS and staged according to somites numbers, then fixed in 1% formaldehyde, 0.2% glutaraldehyde, 0.02% NP40, 0.01% sodium deoxycholate in 1x PBS for 20–30 minutes at 4°C. Subsequently, embryos were washed three times in 1x PBS for 5 minutes at room temperature and incubated overnight at 37°C and in the dark in 1 mg/mL X-Gal, 0.25 mM K3Fe(CN6), 0.25 mM K4Fe(CN6), 0.01% NP-40, 0.4 mM MgCl2 and 1% sodium deoxycholate to detect ß-galactosidase activity, which colours cells blue. To stop the reaction, embryos were washed three times in 1x PBS for 5 minutes each at room temperature.
Chromatin Immunoprecipitation (ChIP)
Forelimbs and hindlimbs of 10 wild-type or Gli3 Xt/Xt embryos at E11.5 were dissected and processed for ChIP as described  using a polyclonal anti-GLI3 antibody (#2676, Genentech, ). To compute the level of enrichment of a given region, the Ct values of both input and ChIP samples were compared with those of a negative control amplicon located in the mouse ß-actin locus . All results (mean ± SD) were obtained by analysing three completely independent experiments per genotype. The significance of all differences was assessed using the two-tailed, non-parametric Mann–Whitney test. The coordinates of the relevant amplification are shown in Additional file 7.
The authors are grateful to A. Offinger and her staff for excellent animal care. L. d'Amato and Christina Torres are thanked for technical assistance. S. Scales (Genentech) provided the GLI3 antibodies. We thank group members for critical input on the manuscript. The ENCODE Group at University of Washington (J. Stamatoyannopoulos and coworkers) are acknowledged for the genome-wide DNase I hypersensitive site mapping data made available freely through the ENCODE Data Coordination Center at UCSC (release date on March 29, 2012). NM received a EU Marie Curie postdoctoral fellowship. This research is supported by grants of the Swiss National Science Foundation (grants no. 31003A_130803 to RZ and 310000_122558 to AZ), a EU reintegration grant (PERG-GA-2009-246576 to JLR) and the University of Basel. The authors declare no conflict of interest.
- 3.Zuniga A, Probst S, Zeller R: The molecular basis of human congenital limb malformations. 2012, Wiley Interdisciplinary Reviews: Developmental BiologyGoogle Scholar
- 18.Maas SA, Suzuki T, Fallon JF: Identification of spontaneous mutations within the long-range limb-specific Sonic hedgehog enhancer (ZRS) that alter Sonic hedgehog expression in the chicken limb mutants oligozeugodactyly and silkie breed. Dev Dyn. 2011, 240: 1212-1222. 10.1002/dvdy.22634.PubMedCentralCrossRefPubMedGoogle Scholar
- 52.Morgan BA, Conlon FL, Manzanares M, Millar JB, Kanuga N, et al: Transposon tools for recombinant DNA manipulation: characterization of transcriptional regulators from yeast, Xenopus, and mouse. Proc Natl Acad Sci U S A. 1996, 93: 2801-2806. 10.1073/pnas.93.7.2801.PubMedCentralCrossRefPubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.