Different cis-Regulatory DNA Elements Mediate Developmental Stage- and Tissue-specific Expression of the Human COL2A1 Gene in Transgenic Mice

Transgenes

Cosmid clone cosHcol2 contained the complete human COL2A1 structural gene with a 9.1-kbp 5′ flanking and a 1.7-kbp 3′ flanking DNA (gift of E. Solomon, Imperial Cancer Research Fund, London, UK).

Reporter gene constructs were created in pPOLYIII I vector by fusing various lengths of 5′-flanking promoter sequences to a 125-bp 5′ untranslated region of human COL2A1 gene and an E. coli β-galactosidase (lacZ) reporter gene cassette with an initiation translation codon and polyadenylation signal, followed by various lengths of the first intron of the human COL2A1 gene. The COL2A1 content in the different constructs were as shown in Fig. 1. For ease of reference, the intron 1 sequences analyzed in this study have been subdivided into three regions, designated X, Y and Z, the coordinates of which are +701 to +2,387, +2,388 to +2,696, and +2,697 to +2,855, respectively (see Figs. 1 and 6).

Production of Transgenic Mice

Transgenic mice harboring the various COL2A1 constructs were generated by pronuclear injection of one-cell zygotes of F1 (CBA × C57BL6) hybrid mice as described (Hogan et al., 1986). Transgenic fetuses were collected at different stages of development and tested for integration of the transgene by PCR. PCR was performed on genomic DNA using oligonucleotide primers for the lacZ gene 5′-CGTAATAGCGAAGAGGCCCG-3′ (nucleotides [nt] 112–130; translation start = 1); and 5′-TCCAACGCAGCACCATCACC-3′ (complementary to nt 558–577). PCR reactions, containing 2.5 U Taq polymerase (Life Technologies, Inc., Gaithersburg, MD), were denatured at 95°C for 3 min, followed by 30 cycles of 95°C for 1 min, denaturing, 55°C for 1 min, annealing, and then 72°C for 1 min of elongation.

Section and Whole Mount In Situ Hybridization

In situ hybridization to sections was carried out as described previously (Cheah et al., 1991; Nonchev et al., 1996). Two hybridization probes used were human COL2A1-specific (KH14) and mouse Col2a1-specific (KH15). KH14 and KH15 probes each contained 70 bp of relatively less conserved 3′ untranslated sequence from equivalent regions of the human and mouse type II collagen genes corresponding to +31,102 to +31,171 (COL2A1) and +28,624 to +28,693 (Col2a1), i.e., KH14 is the human homologue of KH15. These probes were so designed because they contained short stretches of nonconserved bases and were species-specific when sections were treated with RNase posthybridization, and then washed at high stringency. Autoradiography and photography of sections were carried out as described previously (Cheah et al., 1991).

Whole mount in situ hybridization on 8.0–9.0 dpc CBA mouse embryos was carried out as described by Wilkinson (1992) using single-stranded, digoxgenin-UTP (Boehringer Mannheim, Mannheim, Germany)–labeled antisense riboprobes generated from pNJ61 (Ng et al., 1993), which contains exon 1–5B of Col2a1.

X-Gal Staining and Histochemistry

Mouse embryos or fetuses of various stages were harvested in PBS, pH 7.4, and then fixed for 15–60 min, according to their size, at 4°C in 4% paraformaldehyde in PBS. They were assayed for expression of lacZ by X-gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) staining essentially as described in Yee and Rigby (1993). After staining, fetuses were embedded in paraffin. 6-μm sections were counterstained with eosin.

Human COL2A1 Transgene Is Expressed Like the Endogenous Col2a1 Gene

The expression of 45.8-kbp transgene, cosHcol2, containing the whole COL2A1 gene, 9.1-kbp 5′ and 1.7-kbp 3′ flanking sequence was examined by in situ hybridization using riboprobes specific for the human transgene (KH14) and the mouse gene (KH15). The COL2A1 transgene and the mouse Col2a1 gene were coexpressed in the prechondrogenic mesenchyme and chondrocytes (Fig. 2, a, c, and e; Table I, COL2A1; Fig. 2, b, d, and f; Table I, Col2a1). COL2A1 was also expressed in the nonchondrogenic tissues that express the Col2a1 gene, such as the brain (Fig. 2 a) and otic vesicle (Fig. 2 c). However, COL2A1 transgene was not expressed in the mouse heart (Fig. 2 c) which normally expresses the Col2a1 gene strongly. No ectopic expression of the transgene was found. COL2A1 riboprobe did not display any nonspecific hybridization in nontransgenic mouse embryo (Fig. 2 g).

Identifying the cis-Acting Elements for COL2A1-specific Expression

To test the activity of the cis-acting elements that may direct lineage-specific expression of the COL2A1 gene, a COL2A1-lacZ transgene (Fig. 1, pKL9), containing the 6.1-kbp (−1 to −6,100) 5′ flanking sequence, 125-bp 5′ untranslated region and 2.1-kbp intron 1 (+701 to +2,855, regions X, Y, and Z, refer to Materials and Methods) was made. The lacZ gene was expressed in the first and second branchial arches, otic vesicle, frontonasal mesenchyme, and the notochord in the 9.5-dpc embryo (Fig. 3, c–e; Table I). At 13.5 dpc, lacZ activity was detected in the craniofacial chondrogenic tissues (e.g., the frontonasal, occipital, parietal, and facial mesenchyme), the axial (prevertebrae and ribs) and appendicular skeleton (Fig. 4, a–e, limb cartilages). Expression of the pKL9 transgene at both ages was very similar to that of cosHcol2 (Fig. 2, a, c and e) and Col2a1 (Figs. 2 b, d, f and 3 b; Table I). These observations suggest that the regulatory elements of pKL9 are sufficient for driving the COL2A1-specific expression in transgenic mice. Some necessary regulatory elements may, however, be missing in pKL9 since at 9.5 dpc lacZ transgene expression was absent from the dorsal half of the otic vesicle and the brain (Fig. 3, c–e). A construct (pKL8) containing only the 6.1-kbp 5′ flanking region but without the intron 1 sequence linked to the reporter gene did not express in all the transgenic embryos (Table I). Some critical tissue-specific regulatory elements are therefore present within the first intron of the gene.

DNA Modules for COL2A1 Expression during Early Organogenesis

In the mouse, Col2a1 is expressed early at 8.0–8.5 dpc. A distinct craniocaudal gradient of Col2a1 mRNA expression was observed in the paraxial mesoderm (Fig. 3 a). In addition, the midline notochord displayed a uniform expression throughout its length (Fig. 3 a). A 309-bp fragment (region Y) of the first intron of the COL2A1 gene (84.1% homologous to the minimal enhancer of the rat type II collagen gene [Krebsbach et al., 1996]) when linked in normal orientation to the lacZ gene and 6.1-kbp 5′ flanking COL2A1 sequence (pKL80.3), could direct transgene expression like that of endogenous Col2a1 in early organogenesis stage mouse embryos. β-galactosidase activity was found in the paraxial mesoderm and was expressed strongly in the node and notochord at 8.5 dpc (Fig. 3, f–i; Table I). At 9.5 dpc, pKL80.3 transgene was expressed in the first and second branchial arch and the paraxial mesoderm (Fig. 3, j and k), like that of the endogenous Col2a1 (Fig. 3 b). pKL80.3 expression was stronger in the newly formed somites than in the more mature somites (Fig. 3 j). These findings indicate that region Y, together with 6.1-kbp 5′ flanking DNA, was sufficient to drive lacZ expression in tissues that normally express mouse Col2a1 during early organogenesis (Table I).

Identification of Minimal 5′ Flanking Sequence for COL2A1 Expression

The 6.1-kbp 5′ flanking sequence was systematically examined in conjunction with 2.1 kbp of the intron 1 sequence (regions X, Y, and Z) to determine the minimal upstream DNA required to regulate appropriate lacZ expression. The 5′ flanking sequence was reduced to 1.5 kbp (−1 to −1,500) (pKL10, and pKL10 which differs by the orientation of the intron 1 sequence, refer to Fig. 1) and also to 90 bp (−1 to −90, pKL12, refer to Fig. 1). None of these constructs showed any significantly different expression patterns from one another (Fig. 3, l–n; Table I) or that of pKL9 (Fig. 3, c–e). Constructs pKL9 and pKL80.3 were expressed in chondrogenic tissues (Fig. 4, a–d and e–j; Table I) and were appropriately downregulated in the hypertrophic chondrocytes like the endogenous Col2a1 (Fig. 4, i and j; Ng et al., 1993; Lui et al., 1995). pKL10, pKL10R, and pKL12 transgenes were expressed in the same prechondrogenic and chondrogenic tissue types as pKL9 and pKL80.3 (Table I). Reduction of the 5′ flanking sequence of pKL9 to 1500 (−1 to −1,500, pKL10 and pKL10R) or 90 bp (−1 to −90, pKL12) did not seem to affect the expression of the transgene (Fig. 4, k–m at 13.5 dpc). Transgene expression was found in the notochord when the transgene was regulated by 6.1-kbp, 1.5-kbp, or 90-bp upstream sequence in combination with 2.1-kbp first intron (Fig. 3, c and e, pKL9; Fig. 3 l, pKL10; Fig. 3 m, pKL10R; Fig. 3 n, pKL12). Similar notochord expression was also achieved with region Y and 6.1-kbp 5′ flanking DNA (Fig. 3, f–k, pKL80.3). Therefore, the 5′ flanking sequence −1 to −90 with 2.1-kbp intron 1 sequence is sufficient for COL2A1 expression, and 6.1 kbp is required if only region Y of intron 1 is present.

Separate Regulatory Elements Direct Expression in Different Components of the Cartilaginous Skeleton

The pKL10 construct was modified by deleting region Y+Z from the intron 1 fragment (refer to Fig. 1, pKL10X). This transgene was expressed in sites that normally would express Col2a1 (Fig. 5, a–e; e.g., craniofacial mesenchyme, paraxial mesoderm, and prevertebrae), except the notochord, the second branchial arch (Fig. 5, a and b, 9.5 dpc), and the digits of the limbs (Fig. 5, c and f, 14.5 dpc). Transgene expression was generally patchy (Fig. 5 a) and reduced in limbs, digits, the upper trunk, vertebral cartilage, and rib cage (Fig. 5 c). A similar expression pattern was observed (Fig. 5, g–l) for construct pKL12X containing the minimal 90-bp (−1 to −90) 5′ flanking sequence and region X (refer to Fig. 1). Together, these results suggest that at least two separable modules of DNA are present in intron 1. Region X may contain regulatory elements for the first branchial arch, facial skeleton, and lower trunk vertebrae. Indeed, deletion of region X from pKL10R (Fig. 1, pKL10RX) led to loss of expression in most tissues except the limbs and digits (Fig. 5, m–q; Table I). Region Y+Z may contain regulatory elements for regulating COL2A1 expression in the second branchial arch, notochord, upper trunk (axial), and limb (appendicular) cartilage.

Positive and Negative Regulatory Elements for Expression in the Ventral Midbrain

The lacZ reporter was expressed in the neuroepithelium of the ventral midbrain for constructs containing 1.5-kbp (−1 to −1,500) 5′ fragment and regions X+Y+Z (Fig. 3, l and m; Table I, pKL10 and pKL10R). In contrast, midbrain expression was absent for constructs containing additional sequences from −1,500 to −6,100 (Fig. 3, c and d, pKL9) or only 90-bp 5′ fragment (Fig. 3 n, −1 to −90 pKL12). This suggests that negative elements that suppress midbrain expression may be localized in the sequence between −1,500 and −6,100, and positive elements may be found between −90 and −1,500.

Although pKL12 was not expressed in the midbrain of the transgenic embryo, the removal of region Y+Z from pKL12 (i.e., pKL12X) has resulted in the reexpression of the lacZ reporter in the midbrain (Fig. 5, g and h; Table I d). Exchanging region X for Y+Z (pKL10RX) resulted in loss of midbrain expression (Table I). These findings therefore suggest that negative regulatory elements controlling midbrain expression may be localized in region Y+Z. These elements may act in an orientation-independent manner, since midbrain expression was similar for pKL10 and pKL10R (refer to Fig. 3, l and m). Deletion of the same region (Y+Z) from pKL10 (i.e., pKL10X), which was expressed in the midbrain, has no effect on the expression of the reporter (compare Fig. 3 l with Fig. 5, a and b). Therefore, it suggests that the positive elements found between −90 and −1,500 could overcome the negative effect of region Y+Z.

Tissue-specific Expression Requires Both COL2A1 5′ and Intron 1 Sequences

To investigate whether region Y needs to interact specifically with the COL2A1 promoter, we replaced the 6.1-kbp 5′ flanking sequence of pKL80.3 with heterologous promoter fragments from the human β-globin and mouse type X collagen (Col10a1) genes (refer to Fig. 1, pKL21 and pWP16 constructs). The human β-globin promoter fragment has been shown to cooperate with tissue-specific elements from heterologous genes, such as myogenin (Yee and Rigby, 1993) and Hoxa2 (Nonchev et al., 1996), to drive tissue-specific lacZ expression in 9.5-dpc transgenic mice embryos. The Col10a1 promoter fragment can direct lacZ expression in developing bone of 14.5-dpc transgenic mouse fetuses (Poon, W.S., and K.S.E. Cheah, unpublished data).

The β-globin promoter (pKL21) was unable to direct transgene expression in the Col2a1 pattern despite the presence of region Y of intron 1 (seven expressing embryos showing low and nonspecific expression, data not shown). Furthermore, type X collagen promoter was also unable to substitute for the COL2A1 5′ flanking sequence and no expression was seen in seven transgenic embryos. The specific requirement for intron 1 sequence is further demonstrated by the inappropriate expression of pKL28 construct, in which the intron 1 fragment was replaced by the SV-40 enhancer (refer to Fig. 1). Only one of six transgenic embryos expressed lacZ and was not in a pattern characteristic of the Col2a1 (pKL9/pKL80.3) pattern (data not shown).

Different Combinations of DNA Modules Confer Specific Expression of the COL2A1 Gene

In this study, we have identified DNA elements that regulate the expression of the COL2A1 gene by assaying the expression of COL2A1-lacZ transgenes in mouse embryos. We found that the 6.1-kbp upstream region and the first intron of COL2A1 contain regulatory elements required for transcription in tissues that normally express the endogenous mouse type II collagen gene. The first intron operates in an orientation-independent manner. These results agree with other studies using cell transfection assays and transgenic mice, which demonstrate an enhancer role of the first intron of type II collagen genes (Horton et al., 1987; Wang et al., 1991; Seghatoleslami et al., 1994; Metsäranta et al., 1995; Mukhopadhyay et al., 1995; Zhou et al., 1995; Krebsbach et al., 1996; Lefebvre et al., 1996).

In our experiments, neither the first intron nor the COL2A1 5′ flanking DNA could be substituted by enhancer or promoters of unrelated genes. These results are consistent with the report that the promoter and first intron of the rat type II collagen gene act synergistically (Savagner et al., 1995), but contrast with another report showing that mouse Col2a1 intron 1 sequences linked to the human β-globin or adenovirus major late promoters could direct transcription in chondrocytes (Mukhopadhyay et al., 1995; Zhou et al., 1995). These contradictory findings could be related to differences on the strength of the promoter–enhancer interaction and the proximity of the promoter to sequences in the first intron, which may affect the efficiency of transcription. In the constructs tested in the present study, the first intron sequences are separated from the promoter sequence by at least 3 kbp of lacZ DNA. This may hamper the formation of the correct promoter-enhancer loop, and thus diminishes the activity of the heterologous promoters. Furthermore, in other studies, multiple intron 1 sequences are placed immediately upstream of heterologous promoters. This arrangement may facilitate the stabilization of promoter–intron interactions allowing the expression of the transgene (Mukhopadhyay et al., 1995; Zhou et al., 1995).

The requirement for proper physical configuration and/ or differing combination of integrity of regulatory elements for transgene expression has also been shown in the present study. Appropriate expression of the lacZ transgene is achieved by a combination of either the 6.1-kbp 5′ flanking fragment and region Y, or a 90-bp 5′ flanking sequence and the 2.1-kbp intron 1 sequence containing X+Y+Z. Therefore, regions X and Z may contain some elements that interact with the short (90-bp) 5′ flanking sequence. The deletion of sequences −1,500 to −6,100 resulted in the loss of lacZ expression. Both combinations of regulatory elements were equally effective for correct tissue-specific expression, despite sequence differences between modules. The results are consistent with the notion that sufficient lengths of specific combinations of the 5′ flanking sequence and intron 1 of COL2A1 are essential to achieve the appropriate conformation for establishing the necessary heteromeric complex to activate transcription.

COL2A1 Activity in the Midbrain Is Regulated by Positive and Negative Regulatory Elements

Transgene expression in the midbrain is found to be modulated by a combination of positive and negative elements. DNA sequences in the upstream region −1,500 to −6,100 and region Y+Z of intron 1 seem to exert a suppressive activity, whereas the sequences in the −1 to −1,500 region and region X are required for the expression of the COL2A1 transgene (Fig. 6). The data also suggest that expression of COL2A1 in different skeletal elements may be mediated by different combinations of regulatory elements. For example, 1.5-kbp with region Z+Y gave limb expression only (Fig. 5 m) but 1.5-kbp region X gave expression in the axial skeleton but not in the limbs (Fig. 5 c). Recently, cis-acting elements that control α2(XI) collagen gene expression in different cartilaginous components of the skeleton (Tsumaki and Kimura, 1995) and separable DNA modules for tissue and site-specific expression of the α1(I) (Sokolov et al., 1995; Rossert et al., 1995, 1996) and α1(VI) collagen genes (Braghetta et al., 1996) have been described. The differential impact of selective enhancer and repressor elements of the COL2A1 gene on lacZ reporter activity in the brain suggests lineage-specific or region-specific deployment of regulatory elements may be a common mechanism of gene regulation.

Some Regulatory Motifs Identified by Cell Transfection Assays Have No Function In Vivo

Region Y contains the analogous sequences to those in the rodent type II collagen genes shown to be important for enhancer activity in transfected chondrocytes (Wang et al., 1991; Krebsbach et al., 1996; Lefebvre et al., 1996) but not all the motifs identified were present in this region. An AT-rich element previously shown to be important for the enhancer activity and for transcriptional regulation of the Col2a1 gene (Krebsbach et al., 1996) was absent in region Y, suggesting that this sequence is not essential for tissue-specific expression of COL2A1 in vivo.

The 90-bp upstream region of COL2A1 is highly conserved among human, mouse, and rat and contains many potential transcription factor–binding motifs, including a potential Sp1-binding site (−75 to −80) (Vikkula et al., 1992). In the rat Col2a1 gene, a potential Sp1-binding site outside this 90-bp region (−114 to −119) interacts with the first intron enhancer via a complex consisting of Sp1 factor and a zinc finger protein, CIIZFP (Savagner et al., 1995). Our results showing appropriate transgene expression in the absence of this Sp1-binding site (construct pKL12) suggest either that the site is redundant or the other potential Sp1-binding site (−74 to −80) or yet undefined elements within the 90-bp sequence are sufficient for gene expression. The increased specific binding activity to Sp1 motifs found for nuclear extract prepared from non- Col2a1–expressing cells such as fibroblasts and dedifferentiated chondrocytes suggest that the Sp1-binding site may mediate negative regulation of the gene (Dharmavaram et al., 1997).

The rat type II collagen gene contains two sequence elements at positions −421 to −436 and −657 to −666 which behave as silencers in cell transfection assays (Savagner et al., 1990). This region is not well conserved between human and rat type II collagen gene (Vikkula et al., 1992) and the silencing activity has not been tested in vivo. Omitting the −90 to −1,500 region of COL2A1 did not change the tissue-specificity of transgene expression, suggesting that the putative silencer elements in the 5′ flanking DNA do not play any role in the regulation of COL2A1 activity in transgenic mice.

Molecular Mechanisms for Regulation of COL2A1 Expression: A Model of the Organization of COL2A1 Lineage-specific Regulatory Elements

The main structural and functional features of the DNA modules regulating lineage-specific expression of human COL2A1 (Fig. 6) are: (a) early organogenesis and chondrogenic expression are regulated by the interaction of elements in the upstream (−1 to −6,100) region and the 309-bp region Y of the first intron and of elements in −1 to −90 region and region X of the first intron; (b) expression in the midbrain tissue is mediated by interactions between positive elements in the −1 to −1,500 region and the negative elements in −1,500 to −6,100 and region Y+Z of the first intron; (c) critical regulatory elements for gene expression in the notochord are found in the region −1 to −1,500 and region Y of the first intron.

Such an interplay of positive and negative regulation to achieve defined expression patterns has been found for many developmentally regulated genes (Ip and Hemavathy, 1997). Transcription factors such as Bicoid, Dorsal, snail, and Elf-1 regulate patterning in the Drosophila embryo by restricting gene expression to specific sites (Novina and Roy, 1996; Ip and Hemavathy, 1997). We and others have recently shown that SOX9 directly regulates the mouse and human type II collagen genes via SRY/SOX binding motifs found in the Y region of the first intron (Bell et al., 1997; Lefebvre et al., 1997). However, the expression of SOX9 does not always transactivate the Col2a1 gene, especially in nonchondrogenic tissues, suggesting that other factors are also required (Bell et al., 1997). Future challenges lie in identifying those additional factors which bind to and interact with the various activating/repressing DNA modules in vivo, to direct appropriate tissue-specific expression of COL2A1 during key differentiation events in organogenesis and chondrogenesis.