Two Silencers Regulate the Tissue-specific Expression of the Collagen II Gene*

Collagen II, the major component of cartilage, is synthesized primarily by chondrocytes and by certain cells in the eye. Previously, we have studied the regulatory regions of the collagen II gene by DNA transfection assays (Horton, W., Miyashita, T., Kohno, K., and Yamada, Y. (1987) Proc. Natl. Acad. Sci. U.S.A. 84, 8864-8868). These studies show that both the promoter and an enhancer sequence in the first intron are required for high transcriptional activity in chondrocytes. These elements do not show significant activity in cells which do not synthesize collagen II, such as in muscle cells and fibroblasts. In this report, we have constructed plasmids containing various deletions of the promoter of the collagen II gene, fused to a reporter gene for chloramphenicol acetyltransferase (CAT) and transfected them into both chick embryonic fibroblasts and HeLa cells. We have found that silencer elements in the collagen II promoter region reduce CAT activity 11-fold in fibroblasts, while not affecting the enhancer-mediated transcription in chondrocytes. Deletions in the promoter showed that most of the silencing activity was localized in two sites, between -360 and -460 base pairs and between -620 and -700 base pairs. Furthermore, a fragment containing these two sequences in a thymidine kinase promoter CAT construct reduced the activity of the promoter in an orientation independent fashion. Sequence analysis revealed that the two silencer regions are homologous and contain consensus motifs for silencer elements found in other genes. Gel retardation experiments showed that nuclear factors from HeLa cells bind specifically to a DNA fragment containing the silencer, whereas chondrocyte nuclear extracts did not show any activity. Thus, our study indicates that the expression of the collagen II gene is controlled by both negative and positive elements to ensure that the gene is only expressed in suitable cells.

Two Silencers Regulate the Tissue-specific Expression of the Collagen II Gene* (Received for publication, July 27, 1989) Pierre Savagner, Tomoyuki Miyashita$, and Yoshihiko Yamada From  Collagen II, the major component of cartilage, is synthesized primarly by chondrocytes and by certain cells in the eye. Previously, we have studied the regulatory regions of the collagen II gene by DNA transfection assays (Horton, W., Miyashita, T., Kohno, K., and Yamada, Y. (1987) Proc. Natl. Acad. Sci. U. S. A. 84, 8864-8868).
These studies show that both the promoter and an enhancer sequence in the first intron are required for high transcriptional activity in chondrocytes. These elements do not show significant activity in cells which do not synthesize collagen II, such as in muscle cells and fibroblasts. In this report, we have constructed plasmids containing various deletions of the promoter of the collagen II gene, fused to a reporter gene for chloramphenicol acetyltransferase (CAT) and transfected them into both chick embryonic fibroblasts and HeLa cells. We have found that silencer elements in the collagen II promoter region reduce CAT activity ll-fold in fibroblasts, while not affecting the enhancer-mediated transcription in chondrocytes. Deletions in the promoter showed that most of the silencing activity was localized in two sites, between -360 and -460 base pairs and between -620 and -700 base pairs. Furthermore, a fragment containing these two sequences in a thymidine kinase promoter CAT construct reduced the activity of the promoter in an orientation independent fashion. Sequence analysis revealed that the two silencer regions are homologous and contain consensus motifs for silencer elements found in other genes. Gel retardation experiments showed that nuclear factors from HeLa cells bind specifically to a DNA fragment containing the silencer, whereas chondrocyte nuclear extracts did not show any activity. Thus, our study indicates that the expression of the collagen II gene is controlled by both negative and positive elements to ensure that the gene is only expressed in suitable cells.

Gene transcription
is regulated by multiple cis-acting elements, such as promoters, enhancers, and silencers. Promoter elements are needed in all cases for transcription initiation. The transcription level is also regulated by both enhancer and sequences are recognized by trons-acting factors, which bind to specific sites with variable affinity. This DNA-binding complex results in the enhancement, or the inhibition of the transcriptional complex initiating in the promoter region (l-3).
Cartilage is composed of specific components including collagen II, cartilage proteoglycan, and link protein (4)(5)(6), In cartilage, these components appear to be synthesized in a coordinate fashion. The coordinate expression of these elements could be due to some common regulatory elements shared by their genes. We have used the al( II) collagen gene as a model for the study of gene regulation, because its expression is restricted mainly to chondrocytes in cartilage, although it is also present in some other tissues, including the vitreous and nucleus pulposus (7).
We have earlier identified and characterized regulatory regions of the rat collagen II gene, which include a promoter and an enhancer (8,9). These studies show that a promoterenhancer construct is needed in order to have a high level of transcription in chondrocytes (9). This enhancer, located in the first intron, is active only in chondrocytes, Thereafter, we have examined if the promoter could also control the tissue specificity of collagen II gene expression. In this report, we have identified silencing elements in the 5'-region which suppress activity of the collagen II promoter in various cell types, but not in chondrocytes.

PLasmids
Cons&c&-Collagen II promoter-CAT' constructs were derived from pCI1 1 (9)  5'upstream sequences from the collagen II gene were transfected into CEF and HeLa cells, using the calcium phosphate method. CAT activity of the construct (pCII-977) containing the longest 5'upstream region (-977 to +llO bp) was decreased by 85% compared to the activity of the construct (pCII-52) containing the shortest promoter region (-52 to +110 bp) (Fig. 1). Successive constructs elongating from -52 to 5'-direction showed decreased activity. Two main regions were involved in this decrease, which each inhibited the CAT activity by 20%. The first region was located between -700 and -620 bp (CIISl) and the second between -460 and -360 bp (CIIS2). Similar results were obtained when these constructs were transfected into HeLa cells (Fig, 1).
Plasmids including internal deletions of the promoter region were also prepared, and their activity was examined in CEF and HeLa cells. (Fig. 1). Deletion from -110 to -52 (pCI1 IDl) did not modify significantly the silencer activity, compared with the control construct pCII-977. However, deletion from -400 to -312 (pCI1 ID2) increased more than 4fold in CEF the CAT activity, compared with pCII-977. This increase was even higher when deleting the -4OO-to -52-bp region (pCI1 ID3), both in CEF and in HeLa cells. The results suggested that these deletions were removing a silencer element.
To confirm that the silencer activity detected in these experiments was not due to different transfection efficiency, plasmid pCHll0 containing 1acZ gene for ,&galactosidase, under the control of the SV40 early promoter was cotransfected with collagen II/CAT constructs in CEF (Fig. 2). The three CAT constructs, pCII-312, pCII-880, and pCII-312E, which include the collagen II enhancer, gave similar @-galactosidase activity, whereas CAT activity of pCII-880, but not pCII-312, was strongly decreased as shown in Fig. 1. The inclusion of the enhancer to pCII-312 did not change CAT activity in CEF. These results indicate that transfection efficiency was similar for these constructs. The activity of the constructs was also checked by ribonuclease protection experiments. RNAs from transfected cells were hybridized with a RNA probe encoding the collagen II 5'untranslated region Promoter-To study the promoter specificity of these negative regulatory elements, a 482-bp sequence (-783 to -301 bp) including the two silencing regions was coupled to the thymidine kinase promoter, in both orientations. When these constructs were transfected into CEF, CAT activity was reduced to a tenth of the expression observed with the thymidine kinase promoter alone (Fig. 4). However, all of them had similar activity in chick embryonic chondrocytes (data not shown). These results show that the collagen II negative regulatory elements also function with a heterologous promoter and are independent of the orientation. They also confirm the silencers are inactive in chick embryonic chondrocytes.

Gel Retardation
Anulysis-A '*P-labeled lOO-bp fragment containing the silencer CIISl was incubated with nuclear extracts from different sources (Fig. 5). The DNA fragment was shifted with HeLa cells nuclear extracts indicating the formation of complexes. The retarded band was competed by the unlabeled lOO-bp fragment, but not by a 176-bp fragment from the first intron, containing the collagen II enhancer, indicating specific DNA-protein interactions (Fig. 5, lanes 2 -4). Chick chondrocytes or rat chondrosarcoma nuclear extracts did not show any detectable activity (Fig. 5 and the beginning of the CAT gene and treated with RNAse (12). The levels of the protected RNA were in agreement with results deduced from the CAT assays (data not shown).
Collagen II Negative Regulatory Elements Have Weak Activity in Chondrocytes-Previously, an 800-bp enhancer located in the first intron of collagen II was shown to be required with the promoter to give a high level of CAT activity in chondrocytes (9). To test if the negative regulatory elements affect the enhancer-mediated transcription of the gene, plasmids containing the enhancer either with or without the negative regulatory regions (respectively, pCII-977E and pCII-312E) were transfected into chick chondrocytes and fibroblasts. The transcription level of pCII-977, which included the silencer elements, was decreased by about 30% in chondrocytes compared to pCII-312E, which did not include them (Fig. 3). However, the two constructs gave an activity more than 13 times higher than pCII-312 or pCII-977, the constructs missing the enhancer element. In fibroblasts, the enhancer did not have any activity in either cases. These results indicate that the negative regulatory elements are weakly functional in chondrocytes, where the enhancer appears to be essential for the transcription.

DISCUSSION
We describe here two negative regulatory elements in the 5'-upstream region of the collagen II gene which reduced the transcriptional activity of the collagen II promoter when transfected in either fibroblasts or HeLa cells. These sequences were also active in either orientation on a thymidine kinase promoter, and at least one of them binds to nuclear factors in HeLa cells.
A variety of silencers have been characterized (16,(19)(20)(21)(22)(23)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37)(38)(39)(40). Some appear to be promoter-specific, like the 5'upstream element regulating the gamma crystallin gene (30). This specific activity for the homologous promoter may be due to the silencer's flanking regions. However, the collagen II silencers do not appear to be promoter-specific, since the thymidine kinase promoter was also inhibited by a fragment containing both CIISl and CIIS2. Several other reports have described consensus negative regulatory sequences which were active on heterologous promoters including a sequence found in the genes for lysozyme (19), insulin (16), b-interferon (21), immunoglobulin heavy chain (20), and growth hormone (23). This sequence, ACCCTCTCT, is also included in a long interspaced rat repetitive element (16) and in the repetitive CR1 element in the avian genome (17). Interestingly, more than 50,000 copies of this repetitive element are found in the rat genome. Although their physiological function is still unclear, these repetitive elements are likely to be involved in gene regulation, possibly by preventing the transcription of different portions of the genome. We also found homologous motifs repeated in both orientation in the c-myc gene promoter region, close to a distinct sequence described as a silencer (26, 27). Moreover, the sequence TACTCACAGG, included in the c-myc silencer, appears highly homologous to an inverted motif found in CIISl. Since these two consensus sequences are located close one to each other in CIISl, they could interact by participating in the formation of a silencing complex with nuclear factors. The formation of this complex could also involve the two inverted repeats found in CIISl, which concern both of the silencer consensus sequences. A distinct sequence, CACCTCC, is shared by CIISl and CIIS2 and is located one helix turn from the consensus sequence ACCCTCTCT in CIISl. As this region does not share homology with any described consensus sequences, it could have a cartilage-specific activity, although this has not been shown directly.
Silencers are likely to be binding sites for nuclear factors. Our data suggest such soluble binding factors are present in nuclei from nonproducing HeLa cells, but are missing in nuclei from chondrocytes or chondrosarcoma cells that synthesize collagen II. Such factors could be specific for silencers or be similar to factors binding to activation sites like enhancers. In yeast, a transcriptional factor RAP 1 has been shown to bind to the upstream activation sites of different sets of genes, as well as to a silencer (31). In these cases, silencers could be considered as nonoperative enhancers which mimic the first steps of enhancer-mediated transcription and then block it at a further stage. A similar situation has been described for the c-myc gene. The c-fos protein has been found to be part of a complex binding to a c-myc negative regulatory element which includes partially the region showing homology with CIISl (27). This complex also involves the transcriptional factor APl, encoded by c-@z (41, 42). However, its DNA-binding site appears to include also a sequence recognized by the well described octamer-binding protein involved in the transcriptional control of immunoglobulin genes (26). These very different factors could mediate activation or repression of the gene by modifying their relative binding affinities. Such a "push and pull" mechanism has also been proposed for the regulation of low density lipoprotein receptor gene, where the sterol-dependent binding of a protein to a regulatory sequence could inhibit the activity of an adjacent Spl binding site (32). This is more likely to occur when negative regulatory regions are located close to enhancer sites. Such a localization has been reported in several other cases including the murine IgH (20), the T cell receptor a/6 locus (43), and the mouse al(I) collagen gene. The al(I) and ~2(1) collagen gene shows some homologies with the cxl(II) collagen gene. The 5'-upstream region and the first intron include negative regulatory elements (33, 34), as well as enhancer elements (33,44,45). Furthermore, we also found a negative regulatory region, located upstream from the 800-bp enhancer complex in the first intron, which represses the activity of the collagen II promoter, in a position-dependent manner.' The function of silencers in uiuo is still unclear. They could be responsible at least partially for tissue-specific protein expression. Collagen II is expressed mainly by chondrocytes. Our data suggest that collagen II silencers express their activity in non-chondrocytic cells where the collagen II promoter would be active otherwise, as these cells do not require the collagen II enhancer to start transcription from the promoter. Therefore, silencers are not functional in chondrocytes, likely because these cells do not provide functional nuclear factors binding to the silencer, as suggested by our gel retardation experiments.
Conversely, chondrocytes and other cells producing collagen II could be the only cells that provide nuclear factors able to bind specifically to the collagen II enhancer and initiate a transcription complex as suggested by gel retar-

C-MYC
GTCCCCCGTCTACCTAGAT r-------t CCCCATCCARCCTGGCY-I-EAACACCTCCAGGGATGGACGGC L m--m-mm J ATCC4AACCTCTCkjGGCAGCCTGnCCAGCACC CTG?XCGCflGCGATGAllTATwACAAGGATGCGGl7TGTCAAA dation experiments3 Such compensation mechanisms have been described previously in the case of a heterologous enhancer (19). However, this may be the first report of a tissuespecific compensation mechanism, between regulatory elements derived from the same gene. In U~VO, the modulation of silencer activity may also be directed by an extracellular stimulus, such as observed for the chicken ovalbumin gene, where the promoter repression by a negative regulatory element is relieved by steroid hormones (46). The balance between silencer or enhancer accessible binding sites and available transcriptional factors could be a key element in the understanding of tissue-specific expression of collagen II.