| Multiple enhancer regions govern the transcription of CCN2 during embryonic development
Stephanie Frost1, Ke Liu1, Ian M H Li1, Blandine Poulet1, Eithne Comerford1,2, Sarah De Val3 and George Bou-Gharios1
Institute of Ageing and Chronic disease, William Henry Duncan Building, University of Liverpool, West Derby Street, Liverpool L7 8TX, UK
Institute of Veterinary Science, Leahurst Campus, University of Liverpool, Chester High Road, Neston, CH64 7TE, UK
3. Ludwig Institute for Cancer Research, Oxford University, Old Road Campus Research Building, Headington. Oxford OX3 7DQ
CCN2 is a critical matricellular protein that is expressed in several cells with major implications in physiology and different pathologies. However, the transcriptional regulation of this gene remains obscure. We used the Encyclopaedia of DNA Elements browser (ENCODE) to visualise the region spanning from 300kb upstream to the CCN2 start site in silico in order to identify enhancer regions that regulate transcription of this gene. Selection was based on three criteria associated with enhancer regions: 1) H3K4me1 and H3K27ac histone modifications, 2) DNase I hypersensitivity of chromatin and 3) inter-species conservation. Reporter constructs were created with sequences spanning each of the regions of interest placed upstream of an Hsp68 silent proximal promoter sequence in order to drive the expression of β-galactosidase transgene. Each of these constructs was subsequently used to create transgenic mice in which reporter gene production was assessed at the E15.5 developmental stage.
Four functional enhancers were identified, with each driving distinct, tissue-specific patterns of transgene expression. An enhancer located -100kb from the CCN2 transcription start site facilitated expression within vascular tissue. An enhancer -135kb upstream of CCN2 drove expression within the articular chondrocytes of synovial joints. The other two enhancers, located at -198kb and -229kb, mediated transgene expression within dermal fibroblasts, however the most prevalent activity was found within hypertrophic chondrocytes and periosteal tissue, respectively.
These findings suggest that the global expression of CCN2 during development results from the activity of several tissue-specific enhancer regions in addition to proximal regulatory elements that have previously been demonstrated to drive transcription of the gene during development.
CCN2- CCN family member 2
ENCODE- Encyclopaedia of DNA Elements
H3K4me1- monomethylation of 4th lysine of histone 3
H3K4me3- trimethylation of 4th lysine of histone 3
H3K27ac- acetylation of 27th lysine of histone 3
TGF-β – Transforming growth factor beta
TSS- transcription start site
X-gal - 5-Bromo-4-chloro-3-indolyl-β-D-galactopyranoside
CCN2, also known as connective tissue growth factor (CTGF), is an important matricellular protein. As with the other CCN protein family members, CCN2 has several domains that mediate signalling between cells and the extracellular matrix within a broad range of tissues. Aberrant expression of CCN2 has been associated with several fibrotic diseases such as??, in addition to being associated with osteoarthritis and cancer (Planque and Perbal 2003; Omoto et al. 2004; Leask et al. 2009).
In vivo manipulation of CCN2 expression has been a fundamental strategy in understanding the localisation and function of the protein. Perhaps the most profound insight into the role of CCN2 has come as a result of global knockout of gene expression (Ivkovic et al. 2003). Whilst this is not embryonically lethal, CCN2-/- mice nonetheless exhibit premature mortality caused by respiratory failure as a consequence of rib malformation. This reflects the severe chondrodysplasia demonstrated by these mice. An important aspect of the endochondral-related phenotype exhibited with CCN2 knockout is defective angiogenesis which reinforces abnormality in the development and maintenance of endochondral growth plate (Ivkovic et al. 2003). This tallies with the presence and function of CCN2 in endothelial cells and vascular remodelling, which has been well established in several other studies (Shimo et al. 1999; Friedrichsen et al. 2003; Hall-Glenn et al. 2012). Skeletal dysmorphism displayed in CCN2-/- mice is compounded by disruption of intramembranous ossification which results in abnormal craniofacial development (Kawaki et al. 2008). As with deletion of the gene, murine skeletal development is also perturbed by overexpression of CCN2. When driven by a collagen type XI promoter this causes premature chondrocyte hypertrophy leading to incomplete cartilage development and accelerated angiogenesis; culminating in reduced bone length and density (Nakanishi et al. 2001). This contrasts overexpression directed by a collagen type II promoter which causes increased postnatal bone length and density (Tomita et al. 2013). Further examination of this model has revealed a chondro-protective function of CCN2 with increased chondrocyte proliferation, supporting findings from the knockout model where chondrocyte proliferation and stress response were impaired (Hall-Glenn et al. 2013; Itoh et al. 2013). These studies therefore indicate that CCN2 is of imperative importance in the developmental delineation of both cartilaginous and osseous tissues.
Given that transcription is the first step in the expression of a gene, understanding the regulatory sequences through which gene transcription is controlled is therefore critical. A region of 160kb encompassing -57kb upstream to +100kb downstream of the CCN2 coding sequence (Tg(Ctgf-EGFP)FX156Gsat) has been used to drive the expression of an eGFP reporter gene as a surrogate of endogenous CCN2 expression (Gong et al. 2003). Studies using this model have described transgene expression within several tissues including retinal vasculature and cartilage (Pi et al. 2011; Hall-Glenn and Lyons 2011). The presence of regulatory sequence within this 160kb region is reinforced by other studies which have demonstrated the ability of a region spanning from -4kb to the CCN2 coding sequence to drive expression of a β-galactosidase reporter gene within the dermal vasculature and intervertebral disc (Huang et al. 2010; Hall-Glenn et al. 2012). The promoter is a non-coding element found within the vicinity of the transcription start site (TSS) for a gene, and serves as the primary point from which basal gene expression is directed and modulated (Roy and Singer 2015). In vitro studies have identified several transcription factors binding motifs within the proximal promoter region for CCN2, enabling identification of some signalling pathways that control expression of the gene. Transforming growth factor-beta (TGF-β) was one of the first cell-signalling molecules to be implicated as a regulator of CCN2 with the discovery of direct interaction through a binding motif within 200bp from the human sequence start site (Grotendorst et al. 1996). Subsequent studies have demonstrated multiple indirect signalling interactions through which TGF-β can modulate CCN2 expression including via Smad, MEK/ERK and Ets-1 binding motifs within the promoter region (Holmes et al. 2001; Leask et al. 2003; Geisinger et al. 2012). Several other signalling pathways have been associated with transcriptional regulation of CCN2 including hypoxia inducible factor-1 (HIF1), thrombin and Sox9 (Chambers et al. 2000; Higgins et al. 2004; Oh et al. 2016). Moreover, there may be competition between transcription factors in binding to shared regulatory sequence motifs in the vicinity of CCN2, such as occurs with Sox9 and TCF-LEF (Huang et al. 2010). Thus far, attempts to elucidate the regulatory elements that dictate the transcription of CCN2 have failed to fully recapitulate the expression pattern of the gene (Friedrichsen et al. 2003; Ivkovic et al. 2003). This therefore suggests that additional non-coding elements located further away from the CCN2 coding sequence are responsible for wider, cell-lineage specific transcription of the gene.
In addition to the promoter, there are several other forms of non-coding regulatory regions; such as enhancers, which confer greater refinement of gene transcription. The purpose of an enhancer is to mediate robust, cell-specific regulation of gene transcription (Heintzman et al. 2009; Kieffer-Kwon et al. 2013). There may be numerous enhancers that regulate the transcription of a single gene, however each may function within a discrete cellular context (Bonn et al. 2012). In addition, the activity of these regions may be seemingly redundant, such as in the case of shadow enhancers where several enhancers drive analogous patterns of target gene expression within a single cell or tissue type (Cannavò et al. 2016). A defining feature of an enhancer is the ability to drive expression independent of its positioning relative to location of the target gene (Nobrega et al. 2003). In terms of mechanistic function, enhancers form dynamic platforms that bring about topological organisation of chromatin that results in the initiation of gene transcription (Spitz and Furlong 2012). Consensus motifs within enhancers are bound by cell lineage-specific transcription factors, triggering a cascade of protein-protein interactions that are prerequisite in gene transcription such as mediator, p300 and cohesin (Petrenko et al. 2016). As this occurs, the chromatin becomes looped and the promoter and enhancer regions are brought within close proximity to one another, whilst the protein machinery required for transcription including RNA polymerase II is assembled at the start site (Kagey et al. 2010; Roy and Singer 2015).
The ability of an enhancer to form these interactions is reliant on permissive nucleosomal structure and epigenetic profile (Thurman et al. 2012). Chromatin state is highly dynamic and specific to cell type and differentiation state, which is reflected in temporospatial precision in enhancer utilisation (Kieffer-Kwon et al. 2013; Nord et al. 2013). Characteristics of chromatin are therefore useful in the identification of enhancer regions and the context in which they may function (Bonn et al. 2012). Firstly, open chromatin is signified by hypersensitivity to DNase I treatment (Thurman et al. 2012). Epigenetic modifications of the histone proteins such as acetylation of lysine 27 of histone 3 (H3K27ac) in conjunction with monomethylation of lysine 4 of histone 3 (H3K4me1) are associated with active enhancers. H3K4me1 alone indicates poised enhancers, whereas trimethylation of this residue denotes promoter regions (H3K4me3) (Heintzman et al. 2007; Creyghton et al. 2010; Rada-Iglesias et al. 2011). A further feature of an enhancer is a high degree of sequence conservation between evolutionarily disparate species, denoting selection on the basis of sequence function (Visel et al. 2008). However, sequence divergence does not necessarily deplete an enhancer’s capacity to function (Taher et al. 2011).
The paramount need for rigorous control of CCN2 expression across many cell and tissue types during development and beyond suggests that it is highly likely that multiple non-coding regulatory elements are utilised in ensuring robust and intricate patterns of physiological gene expression. Identification of these enhancer regions is a fundamental aspect of further understanding the context in which CCN2 is expressed and the factors that mediate this process. In order to assess this principle, we examined the non-coding region upstream of the CCN2 gene, identifying several enhancers that were capable of driving discrete patterns of reporter gene activity in vivo.
MATERIALS AND METHODS
Identification of candidate enhancers of CCN2
The Encyclopaedia of DNA Elements (ENCODE) browser (genome.ucsc.edu) build NCBI37/mm9 was used to visualise the region 300kb upstream of the CCN2 coding sequence (Waterston et al. 2002; Kent et al. 2002). Given the aforementioned imperative role of CCN2 in cartilaginous tissue, information pertaining to limb tissue was prioritised in the prediction of enhancer regions. ENCODE assimilated tracks regarding E14.5 limb-specific histone modifications of H3K4me1, H3K3me3, H3K27ac were procured from ENCODE/Ludwig Institute for Cancer Research. DNase I hypersensitivity information was gathered from ENCODE/University of Washington regarding thoracic and pelvic limb bud at E11.5, in addition to lung derived fibroblast at 8 weeks. Information regarding conservation of sequences was gathered from Multiz Alignment (Blanchette et al. 2004). Five candidate enhancers were identified based on convergence of peak locations for each of these tracks. Genomic regions were chosen to encompass the entirety of the widest track peak with additional sequence of +/-100bp, with each having a final size of approximately 2kb. This ensured that all potential regulatory sequences were contained within the region, and that insertion into the genome would not perturb function
Creation of β-galactosidase reporter constructs
Each of the enhancer regions was amplified from purified from murine genomic DNA (Promega) and cloned into the Hsp68-lacZ-Gateway vector (Pennacchio et al. 2006).
For the -100kb, -135kb, -198kb and -254kb constructs, regions amplified using PCR with primers; -100kb F 5’-ACCAGATCAGACACCGAGCAATA, R 5’- TGGTTAATGGCTCACGTGGATTC; -135kb F 5’- GAAGCGCAAGAAGGAAGACCAAAG, R 5’- CAGCTCCTTTGCCTTTGCACTGTA; 198kb F 5’- GGTCTTAGGCAAGCAAATCTCTG, R 5’- CATTGAAGAGTCCAAGAAGCAGG. PCR products were cloned into pCR8/GW TOPO entry vector (ThermoFisher). LR recombination reaction was then used to sub-clone enhancer fragment into Hsp68-lacZ-Gateway destination vector. For the -229kb reporter plasmid, the region was amplified with primers containing 5’ sites for ApaI and HindIII restriction enzymes on the forward and reverse primers respectively; F 5’- GTGGTACCGGGCAATTTTAACAAGGCTGAGTA, R 5’- GACGCTAGCTCTCAGGTTCTCAGTCAGTTCTTT. PCR product was agarose gel electrophoresis purified using QIAquick Gel Extraction Kit (Qiagen) before restriction digest, further purification and ligation into Hsp68-lacZ-Gateway vector using T4 DNA ligase (New England Biolabs), in accordance with manufacturer’s guidelines. Presence, specificity and orientation of each construct insert was verified using PCR and restriction digest.
All animal work was carried out in compliance with UK Home Office regulation and subject to review from the local institutional ethical committee.
Each of the Hsp68-LacZ constructs containing the sequences of interest was linearised via restriction enzyme digest with ApaI and NotI in order to remove bacterial maintenance and selection sequence with agarose gel extraction before and column purification (Merck) and elution in water for embryo transfer (Sigma Aldrich). These constructs were then microinjected into the pro-nuclei of 200 fertilised C57BL6xCBAF1 eggs, and transferred into eight foster mothers as described previously (Chiang et al. 2017). At E15.5, embryos were genotyped using construct specific primers and whole mount embryos were stained as indicated below.
Staining for β-galactosidase activity
All E15.5 embryos (4-8 pups from each foster mother) were dissected in cold PBS before fixation (0.2% glutaraldehyde, 0.1M sodium, phosphate buffer pH 7.4, 5mM EGTA pH 8.0, 2mM MgCl2, and 2% formaldehyde) at room temperature for 45 minutes Three wash stages of 30 minutes at room temperature in rinse solution (0.1M sodium phosphate buffer, 2mM MgCl2, 0.1% sodium deoxycholate, 0.2% NP-40 substitute) were carried out before staining overnight at room temperature in 5-Bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal) stain solution (1mg/mL X-gal, 5mM potassium ferricyanate, 4mM potassium ferrocyanate in rinse solution). The rinse steps were then repeated before whole mount imaging using Olympus SFX10 microscope and overnight fixation in 10% neutral buffered formalin at 4ºC.
For clearing of embryos harbouring the -229kb enhancer lacZ construct, embryos were transferred into 1% potassium hydroxide (KOH) solution at room temperature for 6 days before substitution of solution to 0.8% KOH/20% glycerol and further incubation at room temperature for 6 days. This process was repeated with the glycerol ratio being increased by 20% each time the solution was changed. Upon reaching 100% glycerol, whole mount images were taken of the embryos as before.
Histology and imaging
Fixed embryos were processed for histology prior to paraffin embedding. Embryos were sectioned (6µm) in sagittal orientation up to half way into the embryo. Sections were counterstaining with eosin Y and imaged using Olympus BX60 and Carl Zeiss axiocam.
The non-coding region upstream of CCN2 contains several enhancers of gene transcription
Five potential enhancers were identified in silico within the murine genome of CCN2 gene (Figure 1A). These were located at -254kb, -229kb, -198kb, -135kb and -100kb from the TSS. Each of these regions contained track peaks for enhancer associated chromatin characteristics in samples originating from embryonic limb (E11.5 and E14.5). In addition, each demonstrated a high degree of sequence conservation when compared to evolutionarily distant species.
Whole mount imaging of X-gal stained E15.5 embryos allowed gross differences in β-galactosidase transgene activity to be assessed (Figures 1B-F). For each construct, at least three founders were required to stain in a similar fashion before we report the common features of this enhancer activity. Four of the five predicted enhancer regions; -100kb, -135kb, -198kb and -229kb, were able to drive transgene activity at this time point using our criteria. Negligible transgene expression was observed in embryos harbouring -254kb lacZ construct across several founders (Figure 1B). For the -229kb region superficial transgene expression was observed globally with the exception of the ear, abdomen, paws and tail; although these regions were not completely devoid of staining (Figure 1C). The strongest transgene expression was observed within skeletal elements, including the parietal and frontal regions of the cranium, ribs and limbs. The region at -198kb drove the strongest pattern of transgene expression with dermal X-gal staining observed across the whole embryo, with the exception of the digits (Figure 1D). These patterns contrasted with the staining observed for the -135kb and -100kb enhancers which both demonstrated expression within distinct anatomical locations. The -135kb region drove strong transgene expression in the articular regions of both the fore and hind limbs, with staining restricted to the elbow, wrist, knee and ankle (figure 1E). For the -100kb enhancer, staining was punctate across the dermis (figure 1F).
-100kb upstream enhancer of CCN2 functions within dermal microvasculature
Several founder embryos in which β-galactosidase staining was visualised within the vicinity of blood vessels (Figure 2A-C arrows) were embedded in wax, with histological sectioning revealing that transgene activity only occurred within the microvasculature, and more specifically was located within the superficial vascular plexus of the dermis (Figures 2 F, G and I). β-galactosidase activity was concomitant with the localisation of the developing dermal capillary blood vessels. Furthermore, erythrocytes were localised within areas of staining which also exhibited branching in a vessel like manner (Figures 2F and G). There was no staining in any other type of blood vessel, nor within cartilaginous, muscle or osseous tissue (Figure 2 A, E, G and H).
An enhancer located -135kb upstream of CCN2 facilitates gene expression in chondrocytes
For the enhancer located -135kb upstream of the CCN2 TSS, only cartilaginous tissue exhibited transgene activity (Figure 3). Moreover, staining was restricted to chondrocyte near the articular joints of the elbow, wrist, knee, ankle and some digits (Figures 3 C, D, F, G and H). There was no transgene activity within other synovial joints including the shoulder and hip. The most striking example of β-galactosidase expression occurred within the elbow where throughout the sagittal plane of the joint, strongly stained chondrocytes were concentrated in the vicinity of the articular surface of the joint, with dissipation of staining towards the growth plate (Figure 3C). However, the entire population was not positive for transgene activity. There was no X-gal staining within other regions composed of hyaline cartilage; for example within the ribs or cartilaginous anlagen subject to endochondral ossification such as within the forelimb (Figures 3 B and C respectively). In addition, there was no transgene expression within fibrocartilaginous tissue such as within the intervertebral disc. These results therefore indicated that the -135kb enhancer region was able to drive transgene activity in a highly specific sub-population of articular chondrocytes.
The -198kb upstream enhancer of CCN2 demonstrated strong activity in several tissues of mesenchymal origin.
Histological analysis of β-galactosidase expression driven by the -198kb enhancer showed expression in several cell lineages. Firstly, fibroblastic cells within the dermis; more specifically within the loosely organised connective tissue of the reticular layer (r), exhibited strong transgene expression (Figures 4A, C, E). This contrasted with the epidermis, in which there was no transgene expression (Figure 4A arrowheads). Further transgene expression of fibroblastic origin was also observed within tendinous tissue of the hind paw (Figure 4B) and the sheath enveloping the vibrissae follicle (Figure 4C, v). Strong expression of the transgene was also observed in multiple cartilaginous tissues. However, this was not demonstrated in all chondrocytes. Hyaline cartilage displayed prominent X-gal staining, particularly within the ribs (Figure 4D), basioccipital bone and atlas (Figure 4E), limbs (Figures 4H and K) and articular surface of the hip joint (Figure 4J). However, there was stratification of the intensity of staining within tissue undergoing endochondral ossification which tallied with cellular differentiation state. This was clearly visible in both the scapula and humerus (Figures 4 G and H respectively). Hypertrophic chondrocytes (hc) demonstrated strong transgene activity, contrasting the neighbouring zone of proliferative chondrocytes (pc). The specificity in cartilage sub-type stained was reinforced by a lack of staining of the fibrocartilage within the intervertebral disc (Figure 4L). We also observed staining of osteoblastic cells within the primary ossification centres of the iliac bone (Figure 4I), scapula (Figure G os) and long bones including the humerus (Figure 4H os). There was negligible transgene expression within tissue that undergoes to intramembranous ossification such as within the cranium. For example, the orbital plate of the frontal bone was negative for any staining, whereas the nearby hypochiasmatic cartilage demonstrated punctate X-gal staining (arrowheads Figure 4 C).
The -229kb upstream enhancer of CCN2 drives expression primarily within osseous tissue.
We conducted soft tissue clearing of embryos exhibiting β-galactosidase expression driven by the enhancer -229kb upstream of CCN2 which allowed further refinement of the whole mount imaging (Figures 5A and 1B respectively). Staining was observed in several tissues derived from mesenchymal origin. As with the -198kb region, there was staining of the connective tissue within the reticular layer of the dermis (Figure 5J). However, the most prevalent transgene expression driven by this enhancer occurred within primitive osseous tissue. The prevalence of enhancer activity within this tissue type was exemplified upon examination of the ribcages of cleared whole mount embryos, with the observation that the dorsal areas (which are ossified) stained strongly, whereas the frontal portion that remains cartilaginous did not stain. The process of clearing also enabled the identification of X-gal staining confined to the mid-shaft regions of both fore and hind limbs (Figure 5A).
Upon sectioning of these tissues it was clear that this staining was specifically within the periosteum and early bone collar. Strong staining consistently occurred around tissue subject to both endochondral and intramembranous ossification, such as in the orbital plate of the frontal bone (Figure 5B), scapula (Figure 5E), ribs (Figure 5K and L) and limbs (Figures 5H, M and N). However, this pattern of periosteal expression did not extend to all tissues that undergoes ossification. For example, the primitive tissue of the small bones within the paw did not exhibit any transgene activity at this stage of development (Figure 5H). Transgene expression within cartilaginous tissue was sparse and occurred within the nasal region and fibrocartilage of the intervertebral disc (Figures 5G, O and P). The small degree of galactosidase activity within cartilaginous tissue was exemplified within the mandible, with strong staining of the osseous tissue and no blue staining of the chondrocytes within the Meckel’s cartilage (Figure 5D ma and mc respectively). Within endochondral tissue, osteoblastic cells within the primary ossification centres of the scapula and proximal portion of the femur exhibited staining, but there was no staining of the chondrocytic cells throughout the tissue, from the elongated hypertrophic cells proximal to the ossification centre to the densely packed proliferating cells.