INTRODUCTION

Type I collagen is the most abundant protein in vertebrates with levels reaching 5-65% of total protein synthesis in bone cells (1, 2). Two coordinately regulated genes, COL1A1 and COL1A2, give rise to the mRNA transcripts, each containing 50-51 exons, which are processed, translated, and post-transcriptionally modified. The resulting procollagen chains are assembled into a heterotrimeric structure and modified further before entering into the collagen fibril, which in turn is modified by its interaction with other extracellular molecules. The primary site of regulation for this complex series of events is thought to be at the level of transcription initiation. Most studies aimed at understanding the transcriptional regulation of COL1A1 have focused on cell-specific regulatory domains within the proximal promoter, the first intron, and more distal 5-upstream domains without attention to sequences accounting for the extremely high level of gene transcription (3-8).

We have been studying the regulation of the rat COL1A1 gene utilizing a family of transgenes, designated ColCAT, which are derived from the expression vector pSV₂CAT. These transgenes are expressed in a tissue-specific manner in bone, tendon, and skin of transgenic mice and have been used to elucidate a bone-specific regulatory element within the distal promoter (8). Another element, lacking in our constructs, accounts for high collagen expression in vascular smooth muscle (9). Although these constructs are expressed in a tissue-specific manner, they are expressed at only a fraction of the level of the endogenous gene. In contrast, a human COL1A1 minigene developed by Olsen et al. (10) achieves endogenous levels of expression in stably transfected NIH3T3 cells and in transgenic mice.

To identify domains that might achieve endogenous levels of expression, we chose to re-derive ColCAT in a new generation of expression vectors. The parental system that was chosen, pOBCAT (11), eliminates the problem of cryptic splice site selection which is inherent to SV40¹ tAg-containing constructs (e.g. pSV₂CAT). Instead, the pOBColCAT series of expression vectors relies on a splicing event upstream of the reporter gene between a collagen splice donor and an SV40 splice acceptor and includes most of the COL1A1 first intron. pOBColCAT contains 3.5 kb of upstream promoter, the first exon, and most of the first intron. The human COL1A1 minigene, in contrast, includes 2.3 kb of upstream promoter, the first five exons and introns fused with the terminal six introns and exons, 1.5 kb of 3-untranslated region (UTR), and 2.0 kb of 3-flanking sequence. Although pOBColCAT derived vectors are expressed to a much greater degree than their pSV₂CAT counterparts, they continue to be expressed below the level of the endogenous COL1A1 gene. Three potential explanations for the lower level of pOBColCAT expression include 1) the absence of essential regulatory elements that are present in the minigene; 2) decreased efficiency of mRNA processing; and 3) altered stability of the mRNA transcript.

To identify the elements associated with high levels of expression, we chose to study the major regions of the human COL1A1 minigene by introducing them into pOBColCAT and to assess their activity in stably transfected osteoblastic Py1a cells, a model known to express high levels of type I collagen (12). Constructs were designed 1) to determine whether or not the first intron has a role as a transcriptional enhancer; 2) to test for transcriptional enhancing activity of the COL1A1 3-flanking sequences and to determine their effect on mRNA stability; 3) to examine the potential transcriptional enhancing activity within the first five exons and introns fused with the terminal six introns and five exons that compose the body of the COL1A1 minigene and to measure their effect on mRNA stability; and 4) to assess whether the presence of multiple splicing units and/or their identity contributes to higher levels of expression.

EXPERIMENTAL PROCEDURES

Stable Transfection of Cultured Cells

All cell culture work involved the immortalized rat osteoblast cell line Py1a (12) and the use of standard cell culture techniques in F-12 medium supplemented with 5% fetal bovine serum (Life Technologies, Inc.). Stable transfection of DNA constructs (20 µg) along with the neomycin^Res gene (1 µg) was performed using calcium-phosphate precipitation (13). Selection of stable transfectants began after 48 h by adding the neomycin analog G418 (100-200 µg/ml, Life Technologies, Inc.) to the medium. Selection routinely occurred within 7 days of culture with G418 and resulted in at least 100 colonies/10-cm plate; cells were expanded for 1-2 additional weeks before pooling colonies for analysis of transgene expression.

Analysis of CAT Activity

Pooled populations of stably transfected cells were split and grown to confluence in 35-mm wells. Total protein in cell lysates was assayed in a fluor diffusion CAT assay as described previously (14). Briefly, 5 µl of cell extract was mixed with 200 µl of reaction mix (0.1 M Tris, pH 7.8, 0.313 M chloramphenicol, and 100,000 cpm of [³H]acetyl-CoA) in a 15 × 45-mm glass scintillation tube (Kimble Glass, Inc.) and overlaid with 4 ml of nonaqueous Econofluor-2 scintillation fluid (NEN Life Science Products). Samples were assayed in a Beckman scintillation counter, and the total protein content of the extract was determined by the BCA Protein Assay (Pierce). CAT activity was analyzed by linear regression and expressed as cpm/h/µg of protein. Each 35-mm well was assayed for CAT activity, and then the values from replicate wells were averaged. Therefore one value (n = 1) was obtained for each population of cells. The percent CAT activity is given as the mean ± S.E. of the independent populations (n) of pooled stable transfections. All transfection experiments included a standard parental pOBColCAT construct (B15) in addition to one to five test constructs. CAT activity from each transfection was expressed relative to the parental construct, which was set to 100%, to allow for interassay comparison. Statistical analysis involved a one way analysis of variance followed by post hoc analysis using the Student-Newman-Keuls test. The level of significance for the analysis of variance was set at p < 0.001 and for the post-hoc test at p < 0.05.

Analysis of CAT mRNA

Total RNA was isolated from cells that were grown to confluence in 35-mm plates according to a modification of the Peppel/Baglioni method (15). Analysis of RNA levels and structure was performed by Northern blot analysis, RNase protection, and reverse transcriptase-polymerase chain reaction, as appropriate. Northern blot analysis was performed using ³²P-labeled DNA hybridization probes generated by random oligonucleotide labeling (16). The following DNA fragments were used for hybridization: CAT (XbaI to XhoI fragment, mucleotides 543-1342, from pOB₂₅CAT), COL1A1 (HindIII to EcoRI fragment, nucleotides 1455-4758, from the human COL1A1 minigene), and 18 S ribosomal RNA (1.5-kb EcoRI fragment of HHCSA65, obtained from ATCC). Membranes were exposed to x-ray film and later digitized using a PhosphorImager (Molecular Dynamics) for signal quantitation.

Transgene structure and its level of expression relative to endogenous COL1A1 mRNA were assessed using RNase protection analysis (17). A riboprobe template was designed for use with the pOBColCAT family of vectors which reflects the correctly spliced fusion between the COL1A1 first exon and CAT. Reverse transcriptase-polymerase chain reaction was performed on total RNA derived from the tail tendon of transgenic mice harboring B15 (line T-254). In brief, 4-10 µg of total RNA was reverse transcribed using 2.5 µg of CAT-specific oligonucleotide (5-GCCACTCATCGCAGTACTGT-3). Subsequent polymerase chain reaction amplification employed the same CAT-specific oligonucleotide and a rat COL1A1 exon 1-specific oligonucleotide (5-TGAGGCCACGCATGAGCC-3). The resulting 959-bp polymerase chain reaction product was restriction digested with NotI and PvuII to generate the 300-bp fragment that includes the splice junction between exon 1 and CAT. This fragment was then ligated into the ColCAT-Riboprobe (17), which was first restriction digested with SstI, blunted with T4 DNA polymerase, and then restriction digested with NotI, generating the pOB-CAT Riboprobe B21. This construct was linearized with BglII (226 of the COL1A1 promoter) before T3 RNA polymerase-mediated transcription. Upon RNase protection this probe gives rise to a 417-bp band corresponding to the pOBColCAT transcript as well as two bands corresponding to the endogenous COL1A1 transcript 116 bp and 70 bp (for diagram, see Fig. 5). The presence of two protected bands for the endogenous gene results from prior mutagenesis of the rat COL1A1 first exon performed to destroy the translation start site. This mutagenesis results in an 8-bp mismatch upon RNA hybridization.

[View Larger Version of this Image (53K GIF file)]

The half-lives of CAT mRNA and the endogenous COL1A1 mRNA were determined using Northern blot analysis. Cells were grown to confluence in 35-mm wells and treated with the transcriptional inhibitor actinomycin D (Sigma, 6 µg/ml) at various time points before mRNA harvest. COL1A1 half-life determinations utilized 0-, 1-, 2.5-, 9-, 18-, and 24-h time points. CAT half-life determinations utilized 0-, 1-, 3-, and 6-h time points. Time point 0 represents cells not treated with actinomycin D. Ethidium bromide staining was used to assure equal loading between samples. Northern blot signals were digitized to allow for quantitation on a PhosphorImager. The half-life for each transcript was calculated using the equation t_1/2 = 0.693/u, where u = log(A₀/A) 2.303/t, A₀ = image value at time 0, and A = image value at time t (18).

Generation of DNA Constructs

The generation of most constructs required multiple cloning steps, the details of which will be provided upon request. Construct identity was confirmed by a combination of multiple restriction enzyme digests and DNA sequence analysis when appropriate. The standard nomenclature for this new family of constructs (e.g. pOB₂₅ColCAT 3.6/1.6/0) includes three basic elements: the parent construct (e.g. pOB₂₅) (11), the promoter and reporter gene (e.g. ColCAT), and a series of three numbers (e.g. 3.6/1.6/0). The first number refers to the size (in kb) of the promoter, the second number indicates the size (in kb) of the first intron and may indicate specific nucleotide deletions (e.g. 1.6786-1594), and the third number indicates the size (in kb) of the fragment introduced into the enhancer polylinker present downstream of the SV40 pA signal (0 signifies no fragment). Most constructs have a plasmid name, e.g. pOB₂₅ColCAT 3.6/1.6/0 and for simplicity a cloning number, e.g. B15.

ColCAT 3.6 has been described previously (19). It contains sequences between 3518 and +115 bp of the rat COL1A1 gene, which includes upstream promoter and part of the first exon, fused to the CAT gene. This sequence was introduced into pSV₂CAT, which contains the tAg splicing unit and pA signal from the SV40 virus.

B15 (pOB₂₅ColCAT 3.6/1.6/0) contains sequences between 3518 and +1594 bp of the rat COL1A1 gene, which includes the upstream promoter, first exon, and most of the first intron. This sequence was introduced into the pOB₂₅CAT expression vector, which includes the 16 S splice acceptor site and pA signal from the SV40 virus (11). The pOBCAT family of vectors differs from pSV₂CAT in that the SV40 tAg splicing unit has been removed. (This construct has been described previously; see Refs. 9 and 20.)

Human COL1A1 minigene was provided by Dr. D. Prockop and has been described previously (10). In brief, this construct contains approximately 2.3 kb of upstream promoter, exons and introns 1-5, exons and introns 47-52, as well as the 3-UTR and about 2 kb of 3-flanking region. The sequence between exons 5 and 47 includes the proximal 80 nucleotides of intron 5, an SalI linker and the distal 301 nucleotides of intron 46.

B30 (pOB₂₅ColCAT 3.6/1.6786-1594/0) is analogous to the parent construct (B15) but has the intron sequences from AflII to BamHI (nucleotides 786-1594) deleted.

B31 (pOB₂₅ColCAT 3.6/1.6516-1594/0) excludes sequences from AsuII to BamHI (nucleotides 516-1594).

B33 (pOB₂₅ColCAT 3.6/1.6342-1594/0) excludes sequences from BstEII to BamHI (nucleotides 342-1594).

B46 (pOB₂₅ColCAT 3.6/1.6291-1594/0) was generated, in part, using nuclease Bal-31 and excludes sequences from nucleotide 291 to 1594.

B60 (pOB₂₅ColCAT 3.6/1.6281-1594/0) excludes sequences between HinPI and BamHI (nucleotides 281-1594).

B36 (pOB₂₅ColCAT 3.6/1.6516-786/0) excludes sequences between AsuII and AflII (nucleotides 516-786).

B34 (pOB₂₅ColCAT 3.6/1.6342-516/0) excludes sequences between BstEII and AsuII (nucleotides 342-516).

B47 (pOB₂₅ColCAT 3.6/1.6119-1594/0) has the first intron and the distal third of the first exon replaced with the 16 S splice donor signal from the SV40 virus. The 16 S signal donor was obtained as an XhoI to BamHI fragment from pOB₄CAT (22) and introduced into B15 in place of the NotI to BamHI fragment (nucleotides 119-1594).

3-End Constructs

B18 (pOB₂₅ColCAT 3.6/1.6/3.5) was generated by introducing and orienting a 3.5-kb ClaI-ClaI fragment from the 3-end of the human COL1A1 minigene into the enhancer polylinker of B15. This fragment includes sequences from 74 bp upstream of the translation termination codon, the 3-UTR, and approximately 2 kb of 3-flanking sequence. Because this construct contains the SV40 pA signal upstream of the polylinker (and therefore upstream of the COL1A1 3-end fragment) it was designed to assess the transcriptional enhancing activity of this region.

B9 (pOB₂₈ColCAT 3.6/1.6/3.5) is analogous to the B18 construct but differs in that the SV40 pA sequences have been omitted from the pOB₂₈CAT (11) version which allows for pA to occur at the endogenous COL1A1 sites included in the 3-end minigene fragment. The presence of the endogenous 3-UTR in the transcripts derived from this construct adds the potential for COL1A1-specific post-translational regulation.

B40-2 (pOB₂₅ColCAT 3.6/1.6/(1-546-51)) was generated by introducing a 3.1-kb fragment from HindIII to PmlI (nucleotides 1455-4610) from the body of the human COL1A1 minigene (10) into the downstream polylinker of B15, analogous to the 3-end construct (B18). The fragment includes the distal 232 bp of the first intron through the proximal 89 bp of intron 51. In addition, the SalI site (nucleotide 2466) present in the original minigene construct has been replaced by an XhoI site. This fragment was introduced in both orientations, which resulted in the second construct B40-14 (pOB₂₅ColCAT 3.6/1.6/(51-465-1). Because these constructs contain the SV40 pA signal upstream of the polylinker (and therefore upstream of the minigene body) they were designed to assess the transcriptional enhancing activity of this region.

B41 (pOB₂₅ColCAT 3.6/1.6 +(1-546-51)/0) was generated by introducing the 3.1-kb genomic fragment described for B40-2 into the distal portion (*) of the first intron of B15 (e.g. pOB₂₅ColCAT 3.6/1.6 */0). This construct was designed in a configuration that would allow for proper mRNA splicing given that both ends of the fragment were derived from intronic sequences (as described above). Again, this construct was expected to provide insight into the role of this region as a transcriptional enhancer as well as its role in mRNA processing and stability.

B45 (pOB₂₅ColCAT 3.6/1.6+50-51-51/0) was generated by introducing a genomic fragment from the COL1A1 minigene containing the distal 249 bp of intron 50, exon 51, and the proximal 89 bp of intron 51 into the distal portion (*) of the first intron of B15 (e.g. pOB₂₅ColCAT 3.6/1.6 */0). This construct was designed to assess the role of the penultimate splicing event in mRNA processing of the COL1A1 gene.

B57.14 (pOB₂₅ColCAT 3.6/1.6(11-12-12)/0) was generated by introducing a genomic fragment from the COL1A2 gene containing the distal 110 bp of intron 11, exon 12, and the proximal 280 bp of intron 12 into the distal portion (*) of the first intron of B15 (e.g. pOB₂₅ColCAT 3.6/1.6 */0). A second construct was also generated, B57.10 (pOB₂₅ColCAT 3.6/1.6(12-12-11)/0) with the fragment in the opposite orientation. The first construct (B57.14) was designed to assess the role of a heterologous splicing event on the mRNA processing efficiency of B15. The second construct (B57.10) was intended to serve as a control for potential transcriptional enhancing activity of this region.

RESULTS

B15 had 50-100-fold greater expression than ColCAT 3.6 in stably transfected Py1a cells (see Fig. 1 and Table I). Northern blot analysis of total RNA derived from the same stable transfectants revealed a prominent CAT transcript, of the predicted size, from B15 compared with at least three less intense bands from ColCAT 3.6 (Fig. 1). The larger size of the full-length ColCAT 3.6 transcript compared with B15 was consistent with the fact that B15 has 450 fewer bp than ColCAT 3.6 as a result of replacing the tAg splicing unit.

[View Larger Version of this Image (43K GIF file)]

Table I. Summary of CAT activity and mRNA half-life data This data is a summary of the CAT activity and mRNA half for the B15 construct and the intron, 3 end and body/processing constructs. The percent CAT activity is given as the mean ± S.E. of the independent populations (n) of pooled stable transfections. All constructs were normalized to B15, which was set to 100% to allow for interassay comparisons. mRNA half data is given in hours and was determined using Northern blot analysis. Statistical significance of CAT activity data and mRNA half-life is indicated by (*) as performed using analysis of variance (<0.001) followed by post-hoc testing with Student-Newman-Keuls (P < 0.05). Name Detailed name n Relative CAT activity P mRNA half-life % h Parent constructs ColCAT 3.6   ColCAT 3.6 4 2  ± 1* ND B15   pOB25ColCAT 3.6/1.6/0 15 100  ± 6 1.9  ± 0.6 Intron deletion constructs B30   pOB25ColCAT 3.6/1.6  786-1594/0 6 113  ± 6 2.4  ± 0.6 B31   pOB25ColCAT 3.6/1.6  516-1594/0 6 14  ± 3* ND B33   pOB25ColCAT 3.6/1.6  342-1594/0 3 85  ± 7 1.2  ± 0.2 B46   pOB25ColCAT 3.6/1.6  291-1594/0 6 58  ± 8* ND B60   pOB25ColCAT 3.6/1.6  281-1594/0 6 39  ± 9* ND B36   pOB25ColCAT 3.6/1.6  516-786/0 6 162  ± 10* ND B34   pOB25ColCAT 3.6/1.6  342-516/0 6 210  ± 31* ND B47   pOB25ColCAT 3.6/1.6  119-1594/0 6 94  ± 8 1.0  ± 0.7 3-end constructs B18   pOB25ColCAT 3.6/1.6/3.5 10 161  ± 17* 2.7  ± 1.1 B9   pOB28ColCAT 3.6/1.6/3.5 6 80  ± 6 2.9  ± 0.9 Body/processing constructs B40-2; B40-14   pOB25ColCAT 3.6/1.6/(1-546-51) 10 492  ± 37* 1.1  ± 0.4 B41   pOB25ColCAT 3.6/1.6 + (1-546-51)/0 6 0.6  ± 0.1* 3.0  ± 0.9 B45   pOB25ColCAT 3.6/1.6 + (50-51-51)/0 6 27  ± 3* ND B57.14   pOB25ColCAT 3.6/1.6 + (11-12-12)/0 6 40  ± 6* ND B57.10   pOB25ColCAT 3.6/1.6 + (12-12-11)/0 6 62  ± 6* ND

A series of internal deletions of the COL1A1 first intron was generated and tested in stably transfected cells. The specific deletions were guided, in part, by previously identified "regulatory" domains. For example, the construct B31 excludes the highly conserved AP1 binding site located between nucleotides 557 and 563 of the rat COL1A1 first intron (5, 6). In total, seven internal deletion constructs were generated using a combination of restriction enzyme sites and in one construct nuclease, Bal 31 digestion. Figs. 1 and 2 and Table I represent a summary of the CAT activity data for ColCAT 3.6, B15, and the intron deletion constructs in stably transfected Py1a cells.

[View Larger Version of this Image (35K GIF file)]

Deletion of the distal half of the first intron (nucleotides 786-1594) in construct B30 did not change CAT activity. Deletion to nucleotide 516 in construct B31 decreased CAT activity by 10-fold. Further deletion, however, to nucleotide 342 in construct B33 restored CAT activity to near full levels. Deletion to nucleotide 291 in construct B46 and to nucleotide 281 in construct B60 also resulted in significant drops in CAT activity. Two additional internal deletion constructs (B36 and B34) designed to characterize further the changes in activity observed between nucleotides 342 and 786 both revealed increases in CAT activity. Analysis of mRNA half-life of selected constructs revealed no change in mRNA stability (Table I). Northern blot analysis of selected constructs revealed CAT transcripts of the predicted size with the notable exception of B31 (Fig. 2, lane 3). This construct expressed less CAT mRNA that was larger in size than predicted, raising the possibility of alternative mRNA processing (e.g. cryptic splice site selection) and decreased cytoplasmic mRNA accumulation. This larger band may also be present in Fig. 2, lanes 1 and 2, but at a much smaller proportion of the total. Note the absence of this larger transcript when the COL1A1 first intron is replaced, B47 (Fig. 2, lane 5).

An alternative approach to assess the role of the COL1A1 first intron involved removing it entirely and replacing it with an SV40 splicing unit. Analysis of Py1a cells stably transfected with B47 revealed no change in CAT activity compared with B15 (Fig. 2 and Table I). Furthermore, Northern blot analysis revealed a single CAT transcript confirming the integrity of this splicing event (Fig. 2, lane 5) as well as no change in mRNA half-life (Table I). The slight increase in mRNA mobility observed can be explained, in part, by the deletion of 80 bp from exon 1 in the cloning of this construct, plus slightly faster mobility of the entire lane as judged by the position of the 18 S ribosomal RNA band.

Further evidence of alternative intron 1 splicing was provided by RNase protection of RNA from selected constructs. The full-length RNase protection probe is designed to protect transcripts that include exon 1 fused with CAT sequences (417-bp band, see Fig. 5). Cleavage of the probe between these two regions would give rise to the two smaller bands corresponding to exon 1 and CAT alone (221 bp and 196 bp; see Fig. 5). The most likely event that would give rise to a cleavage between these regions would by the inclusion of intron sequences resulting from alternative slice site selection. RNase protection of B15 revealed three bands corresponding in size to the full-length riboprobe as well as the two smaller region of exon 1 and CAT (see Fig. 5, lane 2). Addition evidence for this alternative mRNA processing can also be seen with B18, B40 and B41 (see Fig. 5, lanes 3, 4, and 5, respectively).

Contribution of the 3-Flanking Sequences to Transgene Activity

The transcriptional enhancing activity of the 3-end of the COL1A1 minigene was assessed by introducing it into the downstream enhancer polylinker of B15 generating B18. Analysis of this construct in stably transfected Py1a cells revealed a modest 61% increase in CAT activity compared with B15 (Fig. 3 and Table I). This result represents the pooled data from two independent experiments. The first experiment revealed no increase in CAT activity for B18 compared with B15; in contrast, the second experiment revealed a 2-fold increase in CAT activity for B18. The second experiment may be confounded, however, by an absolute decrease in the overall CAT activity resulting from suboptimal sample preparation. Northern blot analysis of mRNA derived from the second experiment revealed no increase in CAT mRNA from B18 compared with B15 (Fig. 3, compare lanes 1 and 6). Because mRNA was harvested from representative samples, however, quantitative comparisons between lanes cannot be made. Furthermore, Northern blot analysis of B18 revealed that even in the presence of the two endogenous COL1A1 pA signals, this construct utilizes the SV40 pA nearly exclusively (Fig. 3, lane 6). In addition, mRNA stability was unchanged when compared with B15 (Table I). RNase protection of mRNA derived from B18 revealed a minimal increase in CAT mRNA compared with B15 when normalized to COL1A1 levels (see Fig. 5, compare lanes 2 and 3).

CAT activity and Northern blot analysis data for B15 and the 3-end constructs B18 and B9.

[View Larger Version of this Image (42K GIF file)]

Transcriptional enhancing activity as well as effects on mRNA stability were assessed using the construct B9, which is analogous to B18 but excludes the SV40 pA signal. This construct is dependent, therefore, on the endogenous COL1A1 pA signals for critical mRNA processing steps that include terminal exon definition (21), splicing (22, 23), and pA (21, 24). Analysis of this construct in stably transfected Py1a cells revealed no change in CAT activity compared with B15, however, when compared with B18 there was a 2-fold decrease in activity (Fig. 3 and Table I). As predicted, Northern blot analysis of B9 revealed two transcripts of approximately 1-kb difference in size analogous to that seen for COL1A1 mRNA (Fig. 3, lane 7). Analysis of mRNA stability revealed no difference when compared with either B15 or B18 (Table I).

The transcriptional enhancing activity of the body of the minigene, designated (1-547-51), was assessed by introducing it into the downstream enhancer polylinker of B15 generating B40-2. Analysis of this construct in stably transfected Py1a cells revealed a 5-fold increase in CAT activity compared with B15 (Fig. 4 and Table I). Furthermore, the analysis of the same region introduced in the opposite orientation B40-14 gave rise to the same increase in activity. Northern blot analysis of B40-2 revealed increased CAT mRNA as compared with B15 (Fig. 4, compare lanes 1 and 8). RNase protection revealed an increase in CAT mRNA from B40-2 when normalized to endogenous COL1A1 mRNA (Fig. 5, compare lanes 2 and 4).

[View Larger Version of this Image (41K GIF file)]

As an additional test of the role of this region in transcriptional regulation, it was introduced into the distal segment of the first intron of B15 in a configuration that would allow for proper mRNA processing, B41. Analysis of this construct in stably transfected Py1a cells revealed that CAT expression was less than 1% the level of B15. Northern blot analysis of B41 revealed the presence of a larger transcript corresponding in size to that predicted by the inclusion of the minigene coding region, which adds 1244 bp to the CAT transcript (Fig. 4, lane 9). Northern blot analysis also revealed an increase in expression similar to that seen for B40, suggesting that translational efficiency of B41 was impaired (Fig. 4, compare lanes 1, 8, and 9).

The full-length RNase protection probe is designed to protect a transcript that includes exon 1 fused with CAT sequences (see "Experimental Procedures"). The transcript resulting from B41, which includes the minigene body introduced into the distal portion of the first intron, should not protect the full-length riboprobe. Upon RNase protection of this transcript, however, a full-length transcript was evident in addition to the two expected individual bands corresponding to exon 1 and CAT (see Fig. 5, lane 5). This "fully protected" band probably arises from failure of the RNase T1/T2 to cleave completely the probe opposite the predicted 1.2-kb loop structure containing the exons from the human minigene. Based on the size of the transcript on Northern blot analysis (Fig. 4, lane 9) and the greater intensity of the two lower bands compared with the full-length upper band on RNase protection (Fig. 5, lane 5), it would appear that this transcript was processed normally, unlike the previous constructs containing only the first intron.

Regulation of mRNA processing events may also contribute to the overall level of COL1A1 expression. We were interested in addressing several general hypotheses related to the role of the penultimate COL1A1 splicing unit and the minimum number of splicing units required for efficient mRNA processing. To examine the role of the COL1A1 penultimate splicing unit a fragment containing the distal 249 bp of intron 50, exon 51, and the proximal 89 bp of intron 51 was introduced into the distal segment of the first intron of B15 in a configuration that would allow for proper mRNA processing, B45. Analysis of this construct in stably transfected Py1a cells revealed that CAT expression levels were 73% less than B15 (Fig. 6 and Table I). As an additional test of the minimum number of splicing units required for efficient processing, a heterologous splicing unit from the COL1A2 gene was inserted into the first intron generating B57.14. The analysis of this construct revealed a 60% decrease in CAT activity compared with B15 (see Fig. 6 and Table I). Furthermore, the analysis of the same region inserted in the opposite orientation (B57.10), designed as a control for transcriptional enhancing activity, revealed a decrease of 38%.

[View Larger Version of this Image (18K GIF file)]

DISCUSSION

The level of gene expression at any given time must involve the regulation of tissue specificity and the amplitude of expression. A number of sites have been implicated in the tissue-specific expression of the COL1A1 gene (9, 25-27), but little attention has been paid to understanding the regulation of the amplitude of expression in a given tissue. We and others have demonstrated that tissue specificity of COL1A1 expression in bone is localized to the 5-upstream flanking sequences (14, 17, 28, 29). Despite contributing to tissue specificity, these sequence elements do not confer levels of expression comparable to endogenous COL1A1. In contrast, the human COL1A1 minigene is thought to be expressed at endogenous levels (10), suggesting that the elements required for endogenous activity are absent from our constructs and present within the human COL1A1 minigene. To localize these putative elements further, we chose to introduce those sequences unique to the human COL1A1 minigene, 3-flanking sequences, the minigene body, and the first intron, into a new family of collagen expression vectors.

The basal activity of the pOBColCAT family of constructs was 50-100-fold higher than the original ColCAT constructs, which were based on the pSV2CAT design now known to be expressed poorly because of inefficient splicing of the tAg splicing unit. Although this was a possible explanation for the high activity seen with B15, it remained possible that the first intron contained a regulatory domain. To test this possibility a series of internal deletion constructs was examined. The region deleted in B30 includes the previously defined orientation-dependent, negative cis-regulatory element (7, 30) and revealed no change in CAT activity compared with B15 (Fig. 2 and Table I).

The next deletion, B31, removed the highly conserved AP1 binding site (5). It resulted in an nearly 10-fold drop in CAT activity (Fig. 2 and Table I), suggesting the importance of this regulatory element. Northern blot analysis and RNase protection of this construct, however, suggest an alternative explanation for the observed decrease in expression. The presence of a slightly larger mRNA transcript resulting from this construct (Fig. 2, lane 3) is consistent with alternative splice site selection and consistent with cleavage of mRNA transcripts between exon 1 and CAT sequences as suggested by RNase protection analysis (Fig. 5). It is possible, for example, that in the generation of this internal deletion a splice donor may have been created or unmasked which is able to compete successfully with the upstream COL1A1 splice donor. This would give rise to a segment of retained first intron and a larger transcript. Furthermore, the decrease in CAT mRNA levels might be explained by competing splicing events that decrease mRNA processing efficiency. In support of this possibility, Baker demonstrated that the elimination of the 19 S splice donor from pOB₂₅CAT, to eliminate competing splicing events, resulted in a 4-fold increase in CAT activity (11). A second construct with a smaller internal deletion that also removed this AP1 site (B36) showed no evidence of cryptic splicing and no loss of CAT activity, suggesting that this AP1 site may not be crucial for basal activity.

The next deletion, B33, essentially restored full CAT activity and resulted in a normally spliced transcript (Fig. 2 and Table I). Two additional constructs were designed to assess whether or not this region contained additional regulatory elements. Analysis of B34 revealed a 2-fold increase in activity, consistent with the deletion of a negative element or the unmasking of a positive element (Fig. 2 and Table I). It could also result from the removal of the cryptic donor site. The analysis of B36 revealed a more modest 62% increase in activity (Fig. 2 and Table I). Further deletion of the first intron to nucleotide 291 (B46) and to nucleotide 281 (B60) showed a progressive drop in CAT activity of 42 and 61%, respectively (Fig. 2 and Table I).

Although each of the above changes in activity may represent endogenous regulatory elements, it seems prudent not to assign too much weight to these subtle changes in light of the structural mRNA changes observed for B31. In addition to a potential role of the first intron in transcriptional regulation, this region is also involved in mRNA processing, which makes an internal deletion analysis inherently complex and potentially misleading. Alternative methods should be considered to assess the transcriptional enhancing activity of this region. These results suggest that the use of internal deletion constructs to assess transcriptional enhancer activity should be complemented with an analysis of mRNA structure (for related discussion, see Ref. 31).

A total intron replacement construct, B47, revealed no change in CAT activity compared with B15 (Fig. 2 and Table I). These results suggest that the endogenous COL1A1 first intron, in the context of 3.6 kb of upstream promoter and in stably transfected Py1a cells, is not required for high levels of B15 expression. Furthermore, these results suggest that the replacement of the SV40 tAg splicing unit explains the increase in expression seen for B15 compared with ColCAT 3.6. However elements within the first intron may exist which are preferentially used is a subset of collagen producing cells that would only be revealed by construct analysis in a transgenic mouse.

Contribution of the COL1A1 Minigene 3-Flanking Sequence

Several groups have examined the role of the COL1A1 3-flanking and untranslated (3-UTR) regions in transcriptional regulation as well as mRNA stabilization (32, 33). Recently, Sokolov et al. (27) reported that 90% of the 3-UTR, in the context of the human COL1A1 minigene, was not essential for tissue-specific expression. The level of expression from this construct, however, was markedly reduced overall (27). Our analysis of the transcriptional enhancing activity of the human COL1A1 minigene 3-flanking sequences involved the comparison of two constructs: B15 and B18, a modified version containing the 3-end fragment cloned into the "enhancer" polylinker downstream of the SV40 pA signal. Use of this polylinker allows for the study of enhancer elements without their being incorporated into the transcript (11). The comparison of transcriptional activity between these two constructs was possible because both constructs are processed in a similar manner which eliminates the possibility of differential mRNA stability/processing. CAT activity from B18 revealed only a modest increase compared with B15. These data support the conclusion that this region does not contribute significantly to increased transcriptional activity. These results are consistent with recent observations by Määttä et al. (34) who concluded that the COL1A1 3-end does not influence transcriptional regulation in the context of transiently transfected Rat-1 cells.

Multiple mechanisms must be considered when assessing the potential role in gene expression for the body of a gene. This region may influence gene expression via transcriptional regulatory elements, mRNA stability events, or through the regulation of mRNA processing steps (see below). Although little attention as been focused on the COL1A1 body as a potential transcriptional regulator, the presence of DNase I-hypersensitive sites within this region is suggestive. Although not correlated with transcriptional activation, at least two hypersensitive sites have been identified within the body of the COL1A1 gene (35). The first site is located within intron 5,² and the second site exists 1 kb upstream of the EcoRI site present within exon 52 (36). Both sites are theoretically present, at least in part, within the body constructs we evaluated.

Our analysis of the transcriptional enhancing activity of the minigene body involved the comparison of four constructs: B15, B40-2, B40-14, and B41. In B40-2 and B40-14 the minigene body was introduced into the downstream enhancer polylinker in a configuration that prevents incorporation of the fragment into the mRNA transcript. Both constructs were expressed at levels 5-fold higher than the parent construct, B15. These data suggest that an orientation-independent transcription enhancer is present within this region. Furthermore, the presence of an enhancer in B41 is suggested by the high level of CAT mRNA on Northern blot analysis (Fig. 4, lane 9). The most likely explanation for the 500-fold drop in CAT activity between B40 and B41 is a decrease in translational efficiency possibly related to the long 5-UTR produced in the construct.

The mRNA half-life for COL1A1 in Py1a cells (8 h) is severalfold longer than that for CAT mRNA (1-3 h). It was postulated that the minigene 3-UTR or body might contribute to this increase in stability. The role of the COL1A1 3-UTR in mRNA stability was assessed by introducing this fragment into the downstream enhancer polylinker in the absence of the SV40 pA signal (B9). Transcription termination in this construct must occur at either of the endogenous COL1A1 pA signals. As predicted, Northern blot analysis of B9 (Fig. 3, lane 7) revealed two transcripts of approximately 1-kb difference in size analogous to that seen for COL1A1 mRNA. The analysis of mRNA half-life for B9 (2.9 h) suggests that the minigene 3-UTR does not contribute to higher levels of mRNA stability in this system (Table I).

The effect of the COL1A1 minigene body on mRNA stability was assessed using B41. mRNA half-life for this construct (3.0 h) suggests that the minigene body also does not contribute to higher levels of mRNA stability (Table I). Although it is possible that neither the 3-UTR nor the minigene body contributes to the increase in mRNA stability seen for the endogenous gene, it is also possible that CAT-specific mRNA (in)stability elements may have a dominant negative effect on reporter gene half-life.

The regulation of mRNA processing events, e.g. splicing, is increasingly being recognized as an important component of gene expression. A recent analysis of specific splicing events implicated the penultimate splicing event in mediating overall splicing efficiency (37). More recently, Neel et al. (38) have presented evidence for cooperative interactions between introns during pre-mRNA processing, suggesting that the splicing rate of a downstream intron increases with the addition of upstream introns. Nesic and Maquat (37) also suggest that a "network of interactions" forms across the terminal two exons which is essential for efficient mRNA processing. In light of these observations, experimental systems that routinely contain only one intron may not adequately model endogenous genes that commonly utilize two or more splicing units.

To determine if the penultimate COL1A1 splicing unit itself or the addition of a second, heterologous splicing unit was a critical determinant of efficient gene expression, we generated three constructs: B45, B57.14, and B57.10. B45 includes the COL1A1 penultimate splicing unit introduced upstream of the SV40 pA signal, whereas B57.14 was generated by introducing an heterologous splicing unit from COL1A2. B57.10 represents a control for transcriptional activity with the heterologous splicing unit introduced in reverse orientation. A 73% decrease in CAT activity for B45 and a 60% decrease for B57.14 suggests that the introduction of these sequences into the first intron may have interfered with an already efficient processing event (see Fig. 6 and Table I). Analysis of B57.10 revealed a similar decrease of 38%. Although it is possible that this decrease could be attributed to a decrease in transcriptional activity, which may partially explain the decrease observed for B57.14, it is more likely that all three constructs interfere with efficient mRNA processing or translational efficiency. Although a positive role for the penultimate splicing unit was not borne out here, it remains possible that technical problems with construct design may limit specific conclusions.

The elements responsible for high levels of COL1A1 expression have yet to be identified but are thought to be contained within the context of the human COL1A1 minigene (10). To localize these putative regulatory domains further, we chose to subdivide and introduce fragments of this minigene construct into a new family of efficiently expressed collagen reporter genes (pOBColCAT). We conclude, in the context of stably transfected Py1a cells, that the COL1A1 first intron and 3-flanking sequences have minimal roles, if any, in the basal regulation of this gene. In contrast, the body of the COL1A1 minigene confers a 5-fold, orientation-independent increase in transcriptional enhancing activity over the basal activity of B15. Furthermore, these data correlate with the presence of at least two previously described DNase I-hypersensitive sites (35). Future studies will be designed to characterize this increase in transcriptional activity further. Finally, we believe that the study of transcriptional enhancer activity should be accompanied by an analysis of mRNA structure and that internal deletion analyses of regions undergoing mRNA processing events should be done with caution.