INTRODUCTION

The small leucine-rich pericellular proteoglycan (PG)¹ biglycan is composed of a protein core of 331 amino acids and two chondroitin/dermatan sulfate chains attached to its N-terminal region (1). Biglycan is closely related in primary structure to decorin (2) both being coded for by genes that consist of eight exons and seven introns and that probably arose from a common ancestor (3). The mature biglycan protein cores are highly conserved among species (4). As predicted from the presence of the respective functional domains, biglycan and decorin interact with fibronectin (5, 6) and type V collagen (7), implicating them in matrix assembly and cell-matrix interactions. Moreover, both biglycan and decorin core proteins bind TGF- with high affinity (8, 9), and since their synthesis is in turn controlled by this growth factor, they are thought to form a negative feedback loop regulating TGF- activity (10). This peculiar relationship to TGF-, a potent mitogen, might explain at least in part their involvement in the regulation of cell growth and proliferation (11, 12, 13). The recent discovery that decorin can suppress the transformed phenotype of human WiDr colon carcinoma cells, a cell type refractory to growth effects of TGF-, furthermore suggests a novel mechanism of tumor suppression that is independent of growth factor modulation (13). The ubiquitous distribution of both proteoglycans is consistent with their role in the development and maintenance of the structural and functional integrity of all organs. However, they may also fulfill tissue-specific functions; the recent identification of biglycan mRNA and protein in the peritubular cells of the human testis (14), which are actively involved in the paracrine dialogue between the seminiferous epithelium and the testicular interstitium, would suggest an active participation of biglycan in processes regulating spermatogenesis and Leydig cell function.

Biglycan displays a spatial and temporal regulatory pattern that is quite different from that of decorin; whereas the latter is mainly confined to connective tissue cells (chondrocytes, fibroblasts), biglycan is also synthesized by endothelial cells (15, 16, 17), by cells that possess a myogenic phenotype, such as vascular smooth muscle cells (14, 16, 18), myocytes (15), renal mesangial cells (19), and peritubular myofibroblasts of the human testis (14), and even by epithelial cells, e.g. keratinocytes (15) and glomerular cells of the kidney (20). During development, biglycan synthesis is decreased, while that of decorin is increased (21, 22, 23). In response to TGF- biglycan is up-regulated, whereas decorin is down-regulated (24, 25, 26). The reverse is true for glucocorticoid action (27). The largely inverse mode of regulation is in contrast to the great homology that biglycan and decorin share with respect to general gene structure and, together with the observation that changes in core protein and mRNA levels often parallel each other (22, 28, 29, 30), emphasize the importance of regulatory events at the gene and/or mRNA level in the control of their expression. Indeed, while regulation by TGF- appears, at least in human decorin, to involve a post-transcriptional mechanism(s) rather than changes in transcription (12), the density-dependent down-regulation of this gene by TNF- in human fibroblasts has clearly been shown to be mediated via a decrease in promoter activity (12).

Whereas the human decorin promoter has been studied in some detail (31, 12), little information is as yet available for the human biglycan gene. Although part of its 5-flanking sequence has been published along with the biglycan exon/intron gene sequence (3), no functional studies have been carried out yet. Shortly before submission of this manuscript, the homologous sequences in the murine biglycan gene were shown to possess promoter activity in HeLa cells (4). To elucidate the molecular mechanisms governing the regulation of human biglycan expression, we have characterized the transcriptional regulation of the endogenous gene and its 5-flanking region in various human tissues of reproductive and nonreproductive origin. In this report we show for the first time that the human 5-flanking region possesses promoter activity that can be modulated by IL-6 and TNF-. Furthermore, evidence is presented that the powerful up-regulation of biglycan mRNA by TGF-1 involves, at least in osteoblasts, neither alterations in transcriptional activity nor changes in the stability of cytoplasmic mRNA.

EXPERIMENTAL PROCEDURES

Human kidney tissue was obtained from postmortem specimens. Human testis tissue was collected from elderly men undergoing orchidectomy because of prostatic carcinoma. Seminiferous tubules and large testicular blood vessels were kindly provided by Dr. S. Ergün (Hamburg, Germany).

MG-63 human osteosarcoma cells (ATCC CRL-1427) were purchased from the American Type Culture Collection, T47D human breast cell carcinoma (ATCC HTB-133) and SK-UT-1 human leiomyosarcoma cells (ATCC HTB-114) were a gift from Dr. B. Gellersen (Institute of Hormone and Fertility Research, Hamburg). WI-38 human embryonal lung fibroblasts were kindly provided by Dr. D. O. Schachtschabel (University of Marburg). MG-63 and T47D cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 2 mM L-glutamine and either nonessential amino acids (MG-63) or 10 ng/ml insulin (T47D). SK-UT-1 cells were cultured in 1:1 Dulbecco's modified Eagle's medium:Ham's F-12 nutrient mixture with 10% fetal bovine serum and 2 mM glutamine. All media contained 50 units/ml penicillin and 50 µg/ml streptomycin.

Genomic Cloning and Sequencing of the Human Biglycan 5-Flanking Region

A human male placenta genomic library in the Lambda Fix II-bacteriophage containing DNA fragments partially digested with Sau3AI was purchased from Stratagene (Heidelberg, Germany). Approximately 1 million plaques of the amplified library were screened with a 672-bp human biglycan cDNA (3 probe, nt +438 to +1109), including exons 4-7 as well as a part of exons 3 and 8 (3). After three rounds of screening, 11 plaques remained positive, of which phage DNA was isolated according to the Quiagen protocol (Diagen, Düsseldorf, Germany). Southern blots of SacI-digested phage DNA were then re-hybridized to a second, 179-bp biglycan probe which was generated by reverse transcription-polymerase chain reaction from human testicular RNA using primers P-46 and P+133 (sequences given below) in order to obtain clones which contain sequences upstream from exon 1. Of the nine clones selected, all indicated the presence of a ~1.8-kb SacI fragment hybridizing to the 5 probe. After subcloning the DNA insert of one clone designated hBGN 18 I into the SacI site of the Bluescribe plasmid (Stratagene) nested deletions were generated from both directions using the ExoIII-mung bean kit (Stratagene). These were further subjected to restriction analysis. Sequencing of both strands according the method of Sanger et al. (32) revealed the presence of 504 bp of the first intron, the entire exon 1, and 1218 bp of 5-flanking sequences of the human biglycan gene.

Isolation of total RNA from tissue samples was performed according to the protocol of Chirgwin et al. (33), whereas tumor cells and nuclei were lysed directly in the wells using RNA-Clean (AGS, Heidelberg) and further processed as detailed in the manufacturer's instructions. The RNA was formaldehyde-denatured, fractionated on 1.2% agarose gels, transferred by capillary blotting onto Hybond N membranes (Amersham Buchler, Braunschweig, Germany), and fixed by UV-cross-linking. The Northern blots were probed with a 1144-bp biglycan cDNA (nt +124 to +1267) as described previously (14).

Ribonuclease protection assays (RPA) were performed using the RPA II kit (Ambion, Austin, TX). All DNA templates for the in vitro transcription reactions were cloned such as to generate antisense probe RNA by using T7 RNA polymerase. For determination of the transcriptional start site(s), a 402-bp NciI-fragment of the human biglycan gene (nt 148 to +254) was excised from the genomic SacI fragment, filled in with dNTPs using Klenow enzyme, and subcloned into the SmaI site of the Bluescribe plasmid. The resulting construct was linearized with BamHI yielding a cRNA probe of 432 bp (including 30 bp of the polylinker). A plasmid containing a 107-bp cDNA insert from the coding region of human glyceraldehyde-3-phosphate dehydrogenase (nt +452 to +558) (34) cloned into the SmaI site of the Bluescribe vector was digested with HindIII prior to transcription of a 167-bp cRNA probe. At the end of the in vitro transcription reaction containing [-³²P]CTP (800 Ci/mmol; Amersham-Buchler) as the labeled nucleotide, the DNA template was removed by digestion with DNase I and the full-length cRNA enriched by gel purification. All subsequent steps were done according to the kit instructions, hybridizing 2-10 µg of total RNA with 1-2 × 10⁵ cpm (or a 5-fold molar excess) of the labeled antisense RNA probe (specific activity, 3 × 10⁸ cpm/µg) and digesting unbound probe with a 1:100 dilution of the RNase A and T1 mixture. Protected fragments were resolved by 6% polyacrylamide gel electrophoresis and visualized by autoradiography (Kodak X-Omat AR, Eastman Kodak Co.).

Five 5 oligonucleotide primers designated P-110, P-78, P-46, P-1, and P+124 were synthesized. The first three were designed to be complementary to genomic sequences upstream of the main transcriptional initiation site (previously determined by RPA and primer extension), whereas oligonucleotide P+124 spanned the exon 1/exon 2 border of the cDNA. The sequences were as follows: P-110 (5 CCCACCAGCCCCCTCCCTCC 3, nt 110 to 91), P-78 (5 CCGCCCTCTCCCCGCTGTCC 3, nt 78 to 59), P-46 (5 CCCGCCTGCCCAGCCTTTAGC 3, nt 46 to 26), P-1 (5 CTCTCTCCACAAACTGCCCAGG 3, nt 1 to +21), and P+124 (5 CCCTCTCCAGGTCCATCCGC 3, nt +124 to +143). Each of these primers was used in combination with primer P+623 (5 GAGCTGGGTAGGTTGGGCGG 3, the reverse and complement of nt +604 to +623). The choice of P+623 as 3 primer served as a precaution against obtaining false positives from amplification of contaminating genomic sequences. As a template, first strand cDNA was used which had been reverse-transcribed from total RNA of various tissues and cell lines (see Fig. 3) using standard procedures (35). Integrity of the cDNA preparations was checked by amplifying a 107-bp fragment from the cDNA for glyceraldehyde-3-phosphate dehydrogenase (nt +452 to +558). The PCR products were electrophoresed on a 1.5% agarose gel, alkali-denatured, transferred by Southern blotting to Hybond N, and hybridized to an internal 390-bp BglI-StuI fragment (encompassing exons 2, 3, and part of exon 4 of the human biglycan cDNA), labeled with [-³²P]dCTP to a specific activity of 5 × 10⁸ cpm/µg. The blot was washed at high stringency (0.1 × SSPE, 0.1% SDS at 65 °C) and exposed to x-ray film.

Various lengths of the 5-flanking region from the biglycan gene were cloned upstream from the luciferase gene into the polylinker region of the promoterless vectors pGL2-Basic and pGL2-Enhancer (Promega, Heidelberg). The original 1.8-kb genomic SacI fragment was digested with RsrII (+39) and subsequently blunted by filling-in with Klenow polymerase. Having thus generated a common 3 end, the 1.26-kb SacI-RsrII/blunt-ended fragment (1218 to +42) was either ligated directly into the luciferase plasmids previously digested with SacI and HindIII/blunt (to give BGNSac-Luc) or was further shortened at the 5 end utilizing conveniently located restriction endonuclease sites: BglII (at 985), BamHI (at 686), or BglI (at 160). The BglII- and BamHI-digested fragments were ligated into the pGL2 plasmids restricted with BglII and HindIII/blunt (to generate BGNBglII-Luc and BGNBam-Luc), while the 3 overhang of the BglI site (nt 156 to 154) was removed by incubation with Mung bean nuclease prior to inserting the resulting double blunt-ended fragment into the HindIII-digested, blunt-ended pGL2 plasmid (to give BGNBglI-Luc).

Starting from BGNBglI-Luc further deletions were prepared by PCR using the above mentioned 5 primers (110, 78, 46, and 1) and the 3 primer P+133 (5 CTGGAGAGGGGAGGCGCCAG 3, the reverse and complement of nt +114 to +133). PCR products were gel-purified, digested with RsrII, and blunted with Klenow polymerase. Ligation into the filled-in HindIII site of the pGL2 plasmids generated constructs BGN-110-Luc, BGN-78-Luc, BGN-46-Luc, and BGN-1-Luc. The identity of the PCR-amplified fragments and their correct orientation with respect to the 5 and 3 boundaries of the insert was verified by sequencing.

Cells (cultured in six-well plates for 24 h) were transiently transfected with various human biglycan promoter-luciferase chimeric plasmids by lipofection using LipofectAMINE Reagent (Life Technologies, Inc., Eggenstein, Germany). Plasmid DNA for transfections was purified by affinity chromatography on Jet-Star DNA purification columns (GENOMED Inc., Bad Oeynhausen, Germany), according to the kit instructions. For transfection, a constant amount of 2.5 µg of DNA/well was used. The amounts of LipofectAMINE added and the number of transfected cells had been optimized in preliminary experiments for each cell type as follows: MG-63, 2 µl of LipofectAMINE and 2 × 10⁵ cells; T47D, 8 µl of LipofectAMINE and 4 × 10⁵ cells; SK-UT-1, 10 µl and 5 × 10⁵ cells. The cotransfected -galactosidase-coding plasmid pCH110 (Pharmacia Biotech Inc., 0.5 µg/well) served as an internal standard for transfection efficiency. After a 6-h incubation the transfection mixture was removed and replaced with normal growth medium. Following an additional incubation period of 18 h (SK-UT-1), 24 h (MG-63), or 42 h (T47D), respectively, cells were lysed and the luciferase activity measured using the Luciferase Assay system (Promega) in a Luminometer 1250 (Bio-Orbit Oy, Turku, Finland). -Galactosidase activity was determined with the Galacto-Light kit (Tropix/Serva, Heidelberg). In another set of experiments, cells were incubated after transfection with the indicated concentrations of IL-6, TNF- (both human, recombinant, Boehringer, Mannheim, Germany), or TGF-1 (human, recombinant, Biomol, Hamburg), in regular growth medium for various periods of time up to 40 h. In all transfection experiments that included a stimulation with one of the aforementioned agents, the SV40-enhancerless constructs were used.

TGF- Stimulation and DRB Treatment of MG-63 Cells

Initial experiments indicated that the time course and final extent of TGF--induced up-regulation of biglycan mRNA was essentially identical irrespective of serum deprivation of cells prior to TGF- addition. For analysis of the temporal response, RNA was isolated from almost confluent cultures stimulated with 5 ng/ml TGF-1 in normal growth medium for 0, 10, and 24 h, respectively. To analyze whether TGF-1 affects biglycan cytoplasmic mRNA stability, cultures were pretreated with TGF-1 (5 ng/ml), while control cells only received vehicle. Further transcriptional activity was then inhibited by adding DRB (Sigma, Deisenhofen, Germany), a selective inhibitor of RNA polymerase II, to the medium at a final concentration of 60 µM. After a further incubation period of 0, 6, 12, 24, 48, or 72 h, respectively, cells were lysed, and total RNA was isolated as described above.

Confluent MG-63 cells were stimulated for 8 h with 5 ng/ml TGF-1 in normal growth medium, while control cells received only vehicle. All subsequent steps were performed according to a protocol that had previously been optimized for MG-63 cells (36). Cells were rinsed in ice-cold phosphate-buffered saline, transferred to a microcentrifuge tube and lysed with cold lysis buffer (10 mM Tris-HCl, pH 7.6, 10 mM NaCl, 3 mM MgCl₂, and 0.5% Nonidet P-40). The isolated nuclei (5 × 10⁶) were resuspended in 150 µl of 1 × reaction buffer (10 mM Tris-HCl, pH 7.5, 5 mM MgCl₂, 300 mM KCl, 180 units/ml RNasin). The transcription reaction was started by the addition of another 150 µl of 1 × reaction buffer containing twice the final amount of unlabeled rNTPs (4 mM ATP, 4 mM UTP, 6 mM GTP, 40 µM CTP) and 450 µCi of [-³²P]CTP (3000 Ci/mmol, Amersham-Buchler). After an incubation period of 45 min at 32 °C, in vitro labeled transcripts were extracted using RNA Clean. Equal amounts of radiolabeled transcripts (4 × 10⁶ cpm/ml) were hybridized over a 4-day period to denatured plasmid DNA immobilized on Hybond N using a slot blot apparatus (Schleicher & Schüll, Dassel, Germany). Each filter contained 10 µg each of the Bluescribe plasmid alone and the Bluescribe plasmid carrying either a 1144-bp (nt +100 to +1243) human biglycan cDNA or a 939-bp (nt +5 to +943) (37) human fibronectin cDNA, respectively. Hybridization was carried out at 60 °C for 3 days. Membranes were washed twice in 2 × SSPE, 0.1% SDS and twice in 1 × SSPE, 0.1% SDS all at 60 °C and subsequently exposed to Kodak X-Omat x-ray film at 80 °C with two intensifying screens for 4 weeks.

Densitometric scanning of autoradiographs was performed with the program Java (Jandel Scientific Software, Erkrath, Germany). Exposure times were chosen as to stay within the linear range of the x-ray film used. Intensities of ribosomal RNA bands were determined from ethidium bromide-stained gels after scanning of photographs and computeral integration using the program OPTIMAS (Optimas Corp.).

RESULTS

Cloning and Sequence Analysis of Human Biglycan 5-Flanking Region

Full sequencing of the ~1.8-kb insert of the hBGN 18 I clone displayed identity of the sequence with that published by Fisher et al. (3) and an additional extension 714 bp further upstream. Computer-aided analysis (program FACTOR/HUSAR) (38) revealed the presence of several putative cis-acting regulatory elements. In addition to those already identified within the GC-rich first 504 bp (3), seven further Sp1 binding sites, partly overlapping each other, and five sequence motifs conforming to consensus sequences for the transcription factor AP-2 were found (Fig. 1). Throughout the 5-flanking sequence, five IL-6 response elements (IL-6RE) type II are spread (39, 40, 41). One, possessing the consensus core sequence CCGGGAA, is found at 150; the others occur either in reverse complement form at 533 (TTCCAG) and 1182 (TCCCAG) or as a tandem repeat (CTGGGA) at 1210 and 1190 with the latter one spaced by only two nucleotides from the copy at 1182. The discovery of two glucocorticoid response elements at 312 and 695 is of interest in the light of recent reports showing that dexamethasone can decrease steady state levels of biglycan mRNA in osteoblasts and bone marrow stromal cells (27) and that it can inhibit the TGF--induced increase of biglycan production and mRNA levels in cultured human skin fibroblasts (42).

Nucleotide sequence and putative regulatory elements of the 5-flanking region of the human biglycan gene.

One potential binding site for the transcription factor AP-3 can be found in reverse complement form at 519 and another one for nuclear factor-B (NF-B) is present at 647. Further putative binding sites include three AP-1 sites at 672 and 695 and +24, two TREs (one at 696 overlapping the AP-1 site and the other one in reverse complement form at 666). The identification of a binding site for the liver-specific transcription factor C/EBP at 709 is interesting in view of the particularly strong biglycan expression in the liver during the transformation of fat-storing (Ito) into myofibroblast-like cells (43). A TGF--negative element (TGF-NE) (44) extends from 819 to 810; sequence elements recognized by another transcription factor involved in TGF- regulation, namely NF1, are found at 364 and 874. The presence of an E-box-like sequence (CATTTG) at 845, which is considered to be a target for muscle-specific transcription factors (45), is worth mentioning, since biglycan is expressed in a variety of cells with a myogenic phenotype. Several sequences bound by members of the Ets family of oncogenes (Ets-1, PEA3, and PU-boxes) were present; these as well as some other sequences identified by the FACTOR program have not been indicated in Fig. 1 for reasons of clarity.

An alignment of the human biglycan 5-flanking sequence with 609 bp of the murine biglycan promoter (4) revealed a high degree of conservation within the first ~220 bp upstream of exon 1 (+1 in the human corresponds to +62 in the mouse sequence); apart from one short insertion and two short deletions in the human sequence, there is an overall homology of 90% in this region. In particular, two Sp1 sites (at 46 and 78 in the human sequence) and two AP-2 motifs (at 60 and 101 in the human sequence) are perfectly conserved. A comparison of the entire human biglycan 5-flanking region with that of the human decorin gene revealed little similarity over the ~1000 bp of available promoter sequence, except that both possess a TGF-NE and potential NF-B and AP-1 binding sites.

In order to establish whether the same promoter region is utilized by cells of different tissue origin, we performed a RPA, which employed a 402-nt genomic cRNA probe extending 148 bp upstream of exon 1. A cluster of five protected fragments, consisting of one predominant band of 135 nt surrounded by four minor ones, was present in most samples (Fig. 2). MG-63 osteosarcoma cells served as a positive control (46). To locate more precisely the nucleotide corresponding to the major protected band in the RPA, primer extension was performed. One extended product was visible, that corresponded precisely with the major RPA band and mapped to the same site that had previously been shown to be the principal transcriptional initiation site in human skin fibroblasts (3) (data not shown).

Upon closer inspection, one minor protected fragment of considerably larger size than that of the major initiation site was evident in most RNA sources albeit at different intensities (Fig. 2, small arrowhead). To determine whether this represents another less frequently used transcriptional initiation site or only probe molecules which escaped degradation by formation of intramolecular hybrids, we performed reverse transcription-polymerase chain reaction analysis. A series of 5 oligonucleotide primers (P-110, P-78, P-1, and P+124), and a common 3 primer (P+623), were tested for their ability to amplify the corresponding sequences in reverse transcribed RNA extracted from seminiferous tubules, testicular blood vessels, T47D, HeLa, SK-UT-1, and MG-63 cells (Fig. 3). All samples, except HeLa cells, were positive when 5 primer P+124 was used (data not shown). However, in SK-UT-1, no amplification product was obtained with P-1 (5 base identical with nt 1 of the 5-flanking region), indicating that exon 1 is smaller than in the other tissues investigated. In T47D cells, the most upstream located mRNA cap site corresponded to a nucleotide located somewhere between primers P-110 and P-78, since P-78 but not P-110 was still able to prime. Only RNA from testicular tissues and MG-63 cells contained biglycan gene transcripts that initiated even upstream from P-110. Thus, although the results indicate a remarkable homogeneity among tissues in the use of the major transcriptional initiation site(s), there is an evident tissue-specific difference in the use of minor upstream start sites.

Inclusion of a glyceraldehyde-3-phosphate dehydrogenase probe allowed for semiquantification of signals between tissues and cells in the RPA (Fig. 2). Moderate amounts of specific biglycan mRNA were present in total RNA from kidney (lane 1), whole testis (with apparent interpatient variations, lanes 2-4), and seminiferous tubules (including the peritubular cells, lane 5), and were highly enriched in testicular blood vessels (arteries and veins, lane 6), confirming results from in situ hybridization experiments (14). Among the cell lines tested, MG-63 (lanes 10-12) and WI-38 cells (lane 14) displayed the strongest signal, while those of T47D (lanes 8 and 9) were extremely weak. As expected from the PCR data (see above), the initiation cluster was absent from HeLa (lane 7) and SK-UT-1 RNA (lane 13).

TGF-1 is known as a potent positive regulator of biglycan expression. To contribute to the elucidation of the mechanism of its action, we initially analyzed the temporal response of biglycan mRNA levels to TGF-1 in MG-63 cells. TGF-1 caused an elevation of up to 3-fold the constitutive level with a saturation being reached after a stimulation period of 24 h (Fig. 2, lanes 10-12). The relative signal intensities of the protected bands corresponding to the major start sites were increased proportionally indicating that TGF- does not change the initiation pattern qualitatively.

Promoter Activity of Human Biglycan 5-Flanking Region in T47D, SK-UT-1, and MG-63 Cells

Being primarily interested in biglycan expression in the human testis we originally intended to use human peritubular myofibroblasts for functional studies. Unfortunately, this was hampered by two problems: first, the only available testicular tissue from elderly men did not permit extraction of peritubular myofibroblasts of sufficient purity (Dr. M. Davidoff, personal communication), second, to our knowledge no transformed counterparts of testicular myofibroblasts of human origin have yet been established. Thus, we decided to use primarily cell lines established from other reproduction-associated human tissues, namely T47D and SK-UT-1 cells. This enabled us to carry out a comparative analysis of promoter activity/regulation in cell lines both of epithelial (T47D) and of mesenchymal (SK-UT-1, MG-63) origin.

In transient transfection assays, a 5 deletion construct containing as little as 78 bp upstream of exon 1 displayed full promoter activity. This sequence, which contains five GC-boxes and one putative AP-2 site, is apparently sufficient to drive biglycan expression in a non-cell-specific manner (Fig. 4B). A construct further truncated to 46 merely displayed residual promoter activity, whereas, when the adjacent 114 bp relative to P-78 were included (192 to +42, BGNBglI-Luc), promoter activity was considerably increased, pointing to the presence of a strong non-tissue-specific positive element(s). However, a significant decrease in luciferase activity was detected for a construct of intermediate size (BGN-110-Luc) and for chimeras containing more than 192 bp of 5-flanking sequence. This finding suggests that further element(s) with negative activity reside between nt 78 and 110 as well as upstream of nt 192. Upon transfection, the cell-specific regulation of the various constructs was more markedly pronounced in T47D and MG-63 than in SK-UT-1 cells.

Structure and differential promoter activity of the biglycan 5-flanking region-luciferase constructs in T47D, SK-UT-1, and MG-63 cells.

Responsiveness of the Biglycan Promoter to IL-6, TNF-, and TGF-1 Stimulation

The discovery of multiple putative IL-6 response elements suggested that the human biglycan promoter could be subject to regulation by IL-6. To test this possibility, we transiently transfected T47D cells using the BGNSac-Luc construct with subsequent IL-6 stimulation. As shown in Fig. 5, a dose-dependent increase in luciferase activity was detected when IL-6 was added. Both the plasmid vector itself (pGL2-Basic) and the plasmid used for normalization of transfection efficiency (pCH110) did not respond to IL-6 stimulation (data not shown).

The addition of TNF-, another potent mediator of inflammatory processes, to confluent human dermal fibroblasts (12) and endothelial cells (17) was reported to cause a reduction of biglycan mRNA steady state levels. Since a similar effect has been shown to be mediated via TNF- response elements in the decorin gene (12), it was of interest to investigate whether the biglycan promoter, too, was responsive to TNF-. TNF- caused a dose- and time-dependent decrease in promoter activity in T47D cells. At concentrations between 10 and 25 ng/ml the maximal inhibition down to 38% of the original activity was achieved with 25 ng/ml over a 40-h stimulation period in cells transfected with the BGNSac-Luc construct but only to 56% in cells transfected with the BGN-78-Luc construct, respectively (Fig. 6). At the time of lysis, total cell numbers of untreated cells were not significantly different from those treated with TNF-, indicating that the decrease in luciferase activity was not the result of lower cell numbers. Other growth factors screened (platelet-derived growth factor AA, epidermal growth factor) failed to affect biglycan promoter activity in T47D cells.

Inhibition of human biglycan promoter activity by TNF- in T47D cells.

Furthermore, we sought to determine whether TGF-1 increases steady state biglycan mRNA levels via an increase in promoter activity. MG-63 cells were transfected with the BGNSac-Luc and BGNBam-Luc constructs and stimulated with TGF-1. Surprisingly, this resulted in a statistically significant decrease in luciferase activity (Fig. 7). Since this decrease was rapid (observable 9 h after TGF- addition) it was not the consequence of the antiproliferative effect this growth factor has on MG-63 cells (47). Although it is an attractive possibility, the down-regulation could not be explained by the presence of the TGF-1 negative element, since BGNSac-Luc, containing the TGF-NE was not significantly inhibited to a greater extent than the TGF-NE-lacking BGNBam-Luc. Thus, the promoter sequences contained in our genomic clone failed to account for the increase in biglycan mRNA levels after TGF-1 stimulation.

Effect of TGF-1 on BGN

Further Investigation of the Mechanism(s) by Which TGF-1 Modifies Biglycan Steady State mRNA Levels in MG-63 Cells

Biglycan promoter activity could not be influenced positively by TGF-1 (Fig. 7). Although most functional TGF- regulatory elements characterized so far in other gene promoters are located within the first 1000 bp upstream of the respective transcription start site (44, 48, 49, 50), the possibility remained that potential positive TGF- response elements reside elsewhere in the biglycan gene and that our constructs were missing such element(s). To overcome this potential limitation of the reporter gene assay, we performed nuclear run-on assays (Fig. 8). The faint signals repeatedly obtained indicated a low basal level of biglycan transcription. Densitometric scanning of the autoradiographic signals along with their statistical analysis showed that relative transcription rates did not differ significantly between TGF-1-stimulated and control cells 8 h after TGF- addition. Under the same conditions, fibronectin transcription was markedly elevated, while the plasmid vector alone displayed no hybridization above the background level. Since 8 h corresponds to the mid point of the 2.5-3-fold increase in biglycan mRNA levels, we consider it unlikely that transcription was increased at some earlier or later time. Thus, it appears that the TGF-1 effect on biglycan mRNA levels does not occur at the transcriptional level. We therefore investigated the possibility of a post-transcriptional mechanism(s) of regulation, the most common of which is the stabilization of cytoplasmic mRNA. After stimulating MG-63 cells for 24 h with TGF-1, the transcriptional inhibitor DRB was added, and the decay of already transcribed mRNA was monitored over the following 3 days by Northern hybridization (Fig. 9). Since we observed significant cell death beyond 48 h of DRB treatment, the densitometric data from the 72-h time point were not included in the statistical calculation of cytoplasmic biglycan mRNA half-life. This was estimated to be 2.5 days, irrespective of whether the cells were treated with TGF-1 or not. To exclude the possibility of incomplete transcriptional inhibition, the same Northern blot was rehybridized with a platelet-derived growth factor- receptor probe. The half-life for this mRNA in control cells was determined to be 3-4 h (data not shown), which is in close agreement with earlier studies in this cell line (3.5 h) (36). From these results it can be concluded that TGF-1 modulates biglycan mRNA levels, at least in human osteoblastic cells, via a nuclear post-transcriptional mechanism.

Effect of TGF-1 on gene transcription rate in human osteosarcoma cells.

T

TGF-1 does not stabilize biglycan mRNA in MG-63 cells.

DISCUSSION

The present study provides the first comprehensive structural and functional analysis of the human biglycan gene promoter. Location of transcription start sites and the presence of several features typical of mammalian gene promoters in a genomic fragment suggested that sequences possessing promoter activity reside in the immediate 5-flanking region. This was confirmed by transient transfections of luciferase reporter constructs into appropriate cell lines, which revealed sequences important both for basal expression and for mediating modulation of promoter activity by IL-6 and TNF-. By contrast, TGF-1, the most potent stimulator of biglycan expression, did not act at the promoter level nor did it stimulate biglycan mRNA by increasing cytoplasmic mRNA half-life.

Using RPA and PCR, we demonstrated that biglycan transcription is initiated from several closely spaced major sites as well as from minor ones located further upstream (and possibly also downstream), the choice of which is regulated in a tissue-specific fashion. Multiple start sites of transcription have also been found in the murine biglycan gene as determined by primer extension (4), and, interestingly, the most 3 located one of these corresponds precisely to the major initiation site in the human gene. The human (and murine) biglycans apparently belong to a class of eukaryotic genes whose promoters are characterized by a high GC content, multiple potential binding sites for the transcription factor Sp1, the absence of a consensus TATA and CAAT box, and by the feature that they initiate transcription from multiple sites spread over a fairly large region. These genes are either constitutively expressed at a low level (e.g. housekeeping genes) or, interestingly, are related to growth (51). Growth dependence is most evident in the skeletal system where down-regulation of biglycan expression with age parallels cessation of interstitial growth of the cartilage matrix (22). Furthermore, osteoblasts display the highest expression of biglycan in fetal and pubescent age and in culture during the preosteoblastic stage which is characterized by proliferation (21) prior to differentiation into mature osteoblasts.

In transient transfections, a minimum of 78 bp of 5-flanking sequence was required for full basal promoter activity in all cell lines tested and promoter activity remained high when additional sequences up to 200 bp were included in the constructs. This is in contrast to findings with the corresponding murine sequences transfected as CAT constructs into HeLa cells (4); despite the high homology in this region, a construct with 73 bp of 5-flanking sequence (corresponding to 128 bp in the human) showed only little transcriptional activity. These divergent results can possibly be explained by species-specific differences, or, alternatively, by the fact that the murine experiment used a heterologous transfection system.

In T47D cells transiently transfected with BGNSacLuc, IL-6 caused a strong increase in luciferase activity. Notably, analysis of the biglycan 5-flanking sequence revealed the presence of five sequence motifs that match known IL-6 response elements. The hexanucleotide CTGGGA had originally been shown to confer IL-6 responsiveness to the promoters of several acute-phase genes (39) and has been suggested to mediate IL-6 activation in the mouse hepatocyte growth factor gene (52). It is conceivable that the regulation of the human biglycan gene has a similar underlying mechanism. Besides regulating synthesis of acute phase proteins, IL-6 stimulates production of other extracellular matrix proteins in the rat liver, namely collagen (53) and TIMP-1 (54). Interestingly, IL-6 enhances biglycan promoter activity in the T47D cell line, though it simultaneously inhibits growth of these cells (55). This resembles the situation in MG-63 cells, which, although their growth is inhibited by TGF-1, respond to this growth factor with an increase in biglycan mRNA levels.

In contrast to the stimulatory effect of IL-6, TNF- caused a silencing of the human biglycan promoter. This is in line with two recent reports showing that TNF- decreases biglycan mRNA levels in endothelial cells (17) and in confluent human dermal fibroblasts (12). The transcription factor NF-B, for which a putative binding site has been identified further upstream, is involved in TNF- signaling (56). However, since a promoter construct truncated to 78 (which is devoid of NF-B sites) was still responsive to TNF-, albeit to a lesser extent, it is assumed that additional TNF--responsive sequence(s) reside downstream of nt 78, possibly in exon 1. Notably, in human decorin, a similar effect was shown to be mediated via two TNF- responsive elements, with one being located in the 5-untranslated region of this gene (12). Although no TNF- responsive element conforming to the consensus sequence published by Li and Stashenko (57) could be found, a palindromic sequence resembling this is present within the first 78 bp contained in our luciferase fusion constructs (CCGTCGG, 53 to 47). The exact nature of the TNF--responsive sequences and their cognate binding factor(s) remains to be elucidated.

In cells of mesenchymal origin, including MG-63 osteosarcoma cells, TGF-1 acts as a potent positive regulator of biglycan expression (58, 59, this study). However, in MG-63 cells we could neither find evidence for an effect on the transcriptional level as measured by RPA, transient transfection of promoter reporter constructs and nuclear run-on analysis, nor were we able to demonstrate a (positive) effect of TGF-1 on the stability of cytoplasmic biglycan mRNA. Therefore, a nuclear post-transcriptional mechanism appears likely, involving either hnRNA stability, processing of nascent biglycan transcripts or nuclear export of processed biglycan mRNA. Alternative splicing or polyadenylation are important control points in the expression of several genes. However, from Northern blot data there is no evidence that TGF- regulates biglycan expression by one of these mechanisms. In human embryonal lung fibroblasts, the TGF--mediated up-regulation of biglycan mRNA, which follows a similar time course as in MG-63 cells, was considered to occur at the transcriptional level (25), since it could be blocked by actinomycin D. This could be interpreted to indicate that the mechanisms that TGF- utilizes to increase biglycan mRNA differ in fibroblasts and osteosarcoma cells. However, divergent results from nuclear run-on assays and experiments with general transcription inhibitors are not mutually exclusive; in colon carcinoma cells, where TGF- mediates the up-regulation of heparan sulfate-proteoglycan core protein mRNA levels, nuclear run-on assays showed no changes in either heparan sulfate-PG-specific or general transcriptional activity, although experiments with actinomycin D indicated that DNA transcription was required (60). Therefore, if the enhancement of biglycan mRNA by TGF-1 in fibroblastic and osteoblastic cells is to involve a similar mechanism(s), then actinomycin D is assumed to prevent transcription of proteins that themselves act post-transcriptionally in their nuclei to regulate biglycan mRNA. It is noteworthy that a post-transcriptional event at the nuclear level has also been implicated in the Ha-ras oncogene-controlled down-regulation of fibronectin mRNA levels in the osteosarcoma cell line TE-85 (61) as well as in the glucocorticoid-induced up-regulation of bone sialoprotein in rat marrow cells and in the osteoblastic cell lines UMR 106-6 and ROS 17/2.8 (62). In the case of SPARC (osteonectin), another major protein of bone, a regulatory event of this kind has even been considered responsible for the up-regulation of SPARC mRNA by TGF- in human fibroblasts, since no changes in transcription and cytoplasmic mRNA stability after TGF- challenge could be detected (63). Intriguingly, the human and bovine SPARC genes share several other features with the human biglycan gene with respect to promoter structure (lack of TATA and CAAT boxes, multiple overlapping Sp1 sites, long pyrimidine stretches, untranslated exon 1) (64), the long half-life of its mRNA (>2 days) (63), and the high expression in tissues undergoing rapid proliferation and remodeling (e.g. bone). Hence, it is conceivable that the expression of both genes is modulated by TGF- through a common nuclear post-transcriptional mechanism, the exact nature of which is currently under investigation in our laboratory. Finally, a mechanistically related process might be operating in the TGF--induced down-regulation of decorin mRNA in human fibroblasts, since a reporter construct comprising ~1 kb of decorin promoter sequences failed to respond to TGF- upon transfection (12).

In contrast to decorin, where TNF- and TGF- exhibit an additive inhibitory effect on its mRNA levels, TNF- acts as a natural antagonist of TGF- and, possibly, of IL-6 action in controlling biglycan gene expression. Through differential release of TGF-1, IL-6, and TNF-, the cells of a given organ/tissue might be capable of finely tuning biglycan synthesis. If this balance is perturbed, e.g. during an inflammatory response or an immune reaction, and not properly restored, it could result in chronically increased biglycan expression, as is observed in glomerulonephritis (65), and glomerulosclerosis (20), pulmonary (29, 66) and hepatic fibrosis (67), and atherosclerosis (68). Recent data from our laboratory further indicate that this scenario might also apply for the human testis: biglycan mRNA and protein levels are increased in the thickened lamina propria,² a pathological alteration which occurs in degenerating seminiferous tubules and is caused by an accumulation of extracellular matrix material between the peritubular cell layers. TGF-, IL-6, and TNF- are each produced by specific testicular cell types (69, 70, 71) and thus may participate in regulating biglycan synthesis in peritubular myofibroblasts. The advent of stable cell lines of peritubular origin is eagerly awaited and will enable researchers to study biglycan metabolism directly in this cell type as well as its possible role in testicular function.

Our present study has demonstrated that the biglycan 5-flanking region possesses strong basal and inducible promoter activity and is thus assumed to represent the promoter as it operates in vivo. However, additional regulatory influences might even come from another gene(s), as has recently been proposed to explain the unique observation that the X chromosomal biglycan gene does not show the conventional correlation between gene dosage and expression rate (72). If so, it will be exciting to learn about the nature of this gene(s) and how it affects the activity of the proximal biglycan promoter to control the expression of this important proteoglycan.