Expression of Wild-Type and Modified Proα Chains of Human Type I Procollagen in Insect Cells Leads to the Formation of Stable [α1(I)]2α2(I) Collagen Heterotrimers and [α1(I)]3 Homotrimers but Not [α2(I)]3 Homotrimers*

Insect cells coinfected with a baculovirus coding for the proα1(I) chain of human type I procollagen and a double promoter virus coding for the α and β subunits of human prolyl 4-hydroxylase produced homotrimeric [proα1(I)]3procollagen molecules. The use of an additional virus coding for the proα2(I) chain led to the formation of a heterotrimeric molecule with the correct 2:1 ratio of proα1 to proα2 chains of type I procollagen (proα1(I) and proα2(I) chains, respectively), unless the proα1(I) chain was expressed in a relatively large excess. Replacement of the sequences coding for the signal peptide and the N propeptide of the proα1(I) chain with those of the proα1(III) chain increased level of expression of the proα1(I) chain, whereas no similar effect was found when the corresponding modification was made to the virus coding for the proα2(I) chain. Molecules containing such modified N propeptides were found to be processed at their N terminus more rapidly than those containing the wild-type propeptides. TheT m of the type I collagen homotrimer was similar to that of the heterotrimer, both values being about 42–43 °C when determined by circular dichroism. The wild-type proα2(I) chain formed no homotrimers. Replacement of the C propeptide of the proα2(I) chain with that of the proα1(I) chain or proα1 chain of type III procollagen (proα1(III) chain) led to the formation of homotrimers, but the α2(I) chains in such molecules were completely digested by pepsin in 1 h at 22 °C. The data thus suggest that, in addition to control at the level of the C propeptide, other restrictions may exist at the level of the collagen domain that prevent the formation of stable homotrimeric [proα2(I)]3 molecules in insect cells.

Insect cells coinfected with a baculovirus coding for the pro␣1(I) chain of human type I procollagen and a double promoter virus coding for the ␣ and ␤ subunits of human prolyl 4-hydroxylase produced homotrimeric [pro␣1(I)] 3 procollagen molecules. The use of an additional virus coding for the pro␣2(I) chain led to the formation of a heterotrimeric molecule with the correct 2:1 ratio of pro␣1 to pro␣2 chains of type I procollagen (pro␣1(I) and pro␣2(I) chains, respectively), unless the pro␣1(I) chain was expressed in a relatively large excess. Replacement of the sequences coding for the signal peptide and the N propeptide of the pro␣1(I) chain with those of the pro␣1(III) chain increased level of expression of the pro␣1(I) chain, whereas no similar effect was found when the corresponding modification was made to the virus coding for the pro␣2(I) chain. Molecules containing such modified N propeptides were found to be processed at their N terminus more rapidly than those containing the wild-type propeptides. The T m of the type I collagen homotrimer was similar to that of the heterotrimer, both values being about 42-43°C when determined by circular dichroism. The wild-type pro␣2(I) chain formed no homotrimers. Replacement of the C propeptide of the pro␣2(I) chain with that of the pro␣1(I) chain or pro␣1 chain of type III procollagen (pro␣1(III) chain) led to the formation of homotrimers, but the ␣2(I) chains in such molecules were completely digested by pepsin in 1 h at 22°C. The data thus suggest that, in addition to control at the level of the C propeptide, other restrictions may exist at the level of the collagen domain that prevent the formation of stable homotrimeric [pro␣2(I)] 3

molecules in insect cells.
The collagen superfamily now includes at least 19 proteins formally defined as collagens and more than 10 additional proteins with collagen-like domains (for reviews, see Refs. [1][2][3][4][5][6]. Type I collagen is the most abundant member of this family and was the first to be characterized. Its molecule is a heterotrimer consisting of two identical ␣1(I) chains and a slightly different ␣2(I) chain that are coiled around one another into a triple-helical conformation. The molecule is synthesized in the form of a precursor in which the pro␣1(I) 1 and pro␣2(I) chains have propeptide extensions at both their N-and Cterminal ends. In addition to this heterotrimer, several tissues contain small amounts of a molecule with a chain composition of [␣1(I)] 3 known as the type I collagen homotrimer (1). In fact, early renaturation experiments with individual ␣ chains of type I collagen indicated that they are able to form both [␣1(I)] 2 ␣2(I) heterotrimers and [␣1(I)] 3 homotrimers although the former were favored, and the T m of the latter was slightly lower than that of the former (7). Homotrimers with the structure of [␣2(I)] 3 were also obtained at low temperatures, but their yield was much lower and the T m was only about 20 -24°C (7).
Type I collagen is now used in many medical applications as a biomaterial and as a delivery system for certain drugs (8 -10). The collagen used in these applications has been isolated from animal tissues and is liable to cause allergic reactions in up to 3% of human subjects (11). It is obvious, therefore, that an efficient large-scale recombinant system for the production of type I collagen would have many practical applications.
We recently reported that insect cells provide an excellent system for the large-scale expression of native triple-helical human type III collagen, a homotrimer consisting of three identical ␣1(III) chains (12). Nevertheless, coexpression with the ␣ and ␤ subunits of human prolyl 4-hydroxylase, an ␣ 2 ␤ 2 tetramer (6,13,14), was required for the production of molecules with stable triple helices (12). The properties of the recombinant type III collagen were very similar to those of the corresponding nonrecombinant protein, and the highest expression levels obtained in suspension cultures were up to 40 mg/liter type III collagen, corresponding to 60 mg/liter type III procollagen.
The purpose of the present work was to study whether coexpression of the pro␣1(I) and pro␣2(I) chains of human type I procollagen in insect cells that also express human prolyl 4-hydroxylase will lead to the formation of heterotrimeric molecules with the correct 2:1 chain ratio. The association of type I procollagen chains begins with interactions among the C * This work was supported by grants from the Health Sciences Council of the Academy of Finland and from FibroGen Inc. (South San Francisco, CA) and by European Commission Grant B104-CT96-0537. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  propeptides, and it has been regarded as likely that their structural features favor the formation of heterotrimers rather than [pro␣1(I)] 3 homotrimers and prevent the formation of [pro␣2(I)] 3 homotrimers (1,6,15). We therefore also studied whether pro␣2(I) chains in which the C propeptide has been replaced with that of the pro␣1(I) or pro␣1(III) chain will form homotrimeric molecules or whether additional restrictions exist at the level of the collagen domain, as suggested by the early renaturation experiments with isolated ␣2(I) chains (7).

Construction of Baculovirus Transfer Vectors and Generation of Recombinant
Viruses-A full-length cDNA for the human pro␣1(I) chain (16) was digested with XbaI and ligated to pVL1392 (Invitrogen). A full-length cDNA Hp2010 (17) for the human pro␣2(I) chain was cloned as a blunt-ended fragment into the EcoRV site of pSp72 (Promega), generating pSp72-C1A2. A BglII site was created 9 bp upstream of the translation initiation codon by PCR, and the full-length cDNA was digested with BglII and BamHI and ligated to pVL1392. The pVL constructs were cotransfected into Spodoptera frugiperda Sf9 cells with a modified Autographa californica nuclear polyhedrosis virus DNA using the BaculoGold transfection kit (Pharmingen), and the resultant viral pools were collected, amplified, and plaque-purified (18). The recombinant viruses were termed C1A1 and C1A2, respectively.
The sequences coding for the signal peptides and N propeptides of the human pro␣1(I) and pro␣2(I) chains were replaced with those coding for the corresponding regions of the human pro␣1(III) chain by PCR. A construct pSp72-C1A1 was created by cloning the full-length cDNA for the pro␣1(I) chain into the XbaI site of pSp72. Two fragments, the first including a 5Ј BglII site 16 bp upstream of the translation initiation codon and the sequences coding for the pro␣1(III) chain up to the N propeptide cleavage site, and the second starting from the codon for the first amino acid in the N telopeptide of the pro␣1(I) chain and continuing up to an internal DraIII site, were generated by PCR. These fragments were ligated into BglII-DraIII digested pSp72-C1A1, and the resultant construct was termed pSp72-C1A1NproIII. The full-length C1A1NproIII was digested with BglII and XbaI and ligated to pVL1392. For the generation of a corresponding pSp72-C1A2NproIII construct, two fragments, the first as above and the second starting from the codon for the first amino acid in the N telopeptide of the pro␣2(I) chain and continuing to an internal SacII site, were generated by PCR. These fragments were ligated into BglII-SacII digested pSp72-C1A2, and the resultant construct was termed pSp72-C1A2NproIII. The full-length C1A2NproIII was digested with BglII and SmaI and ligated to pVL1392. The pVL constructs were cotransfected into Sf9 cells as above, and the recombinant viruses were termed C1A1NproIII and C1A2 NproIII, respectively.
To replace the sequence coding for the C propeptide of the pro␣2(I) chain with that coding for the C propeptide of the pro␣1(III) chain, two fragments were generated by PCR, the first extending from an internal AvrII site of the cDNA coding for the pro␣2(I) chain to the codon for the last amino acid of the C telopeptide and the second from the codon for the first amino acid of the C propeptide of the pro␣1(III) chain to a BamHI site created 51 bp downstream of the translation stop codon. These fragments were ligated into AvrII-BamHI digested pSp72-C1A2, and the resultant construct was termed pSp72-C1A2CproIII. Another modified C1A2 construct coding for both the C telopeptide and the C propeptide of type III procollagen was also generated. Two fragments were made by PCR, the first extending from the internal AvrII site of the cDNA for the pro␣2(I) chain to the codon for the last amino acid of the triple-helical region, and the second from the codon for the first amino acid of the C telopeptide of the pro␣1(III) chain to a BamHI site created 51 bp downstream of the translation stop codon. These fragments were ligated into pSp72-C1A2 as above, and the resultant construct was termed pSp72-C1A2Ctelo-proIII. The C telopeptide and C propeptide of the pro␣2(I) chain were also replaced by those of the pro␣1(I) chain. Two fragments were generated by PCR, the first one being the same as used for the construction of pSp72-C1A2Ctelo-proIII and the second extending from the codon for the first amino acid of the C telopeptide of the pro␣1(I) chain to a BamHI site created 20 bp downstream of the translation stop codon. These fragments were ligated into pSp72-C1A2 as above, and the resultant construct was termed pSp72-C1A2Ctelo-proI. The full-length C1A2CproIII, C1A2Ctelo-proIII, and C1A2Ctelo-proI were digested with BglII and BamHI and ligated to pVL1392, the pVL constructs being cotransfected into Sf9 cells as above. The recombinant viruses were termed C1A2CproIII, C1A2Ctelo-proIII, and C1A2Ctelo-proI, respectively.
Analysis of Recombinant Proteins-High Five cells (H5, Invitrogen) were cultured either as monolayers or in suspension in TNM-FH medium (Sigma) or SF900IISFM medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum (BioClear) at 27°C. To produce recombinant proteins, cells seeded at a density of 5-6 ϫ 10 5 cells/ml in monolayers or 1-1.5 ϫ 10 6 cells/ml in suspension were infected with different combinations of viruses coding for the native or modified type I procollagen chains together with a double-promoter virus 4PH␣␤ 2,3 coding for the ␣ (19) and ␤ (20) subunits of human prolyl 4-hydroxylase (21). The collagen-coding viruses were used in a 5-10-fold excess over the 4PH␣␤ virus (12). L-ascorbic acid phosphate (80 g/ml) (Wako) was added to the culture medium daily. The cells were harvested 72 h after infection, washed with a solution of 0.15 M NaCl and 0.02 M phosphate, pH 7.4, homogenized in a solution of 0.3 M NaCl, 0.2% Triton X-100, and 0.07 M Tris, pH 7.4, and centrifuged at 10,000 ϫ g for 20 min. The Triton X-100 soluble proteins were analyzed by SDS-PAGE, followed by staining with Coomassie Brilliant Blue or Western blotting with a polyclonal antibody against the N propeptide of type I (PINP) or type III procollagen (PIIINP) (Farmos Diagnostica) or a monoclonal antibody 95D1A recognizing the collagenous regions of various collagen chains. 4 Aliquots of the Triton X-100 supernatants were incubated with pepsin (0.2 mg/ml) for 1 h at 22°C (22), and some samples were subsequently digested with a combination of trypsin (0.1 mg/ml) and chymotrypsin (0.25 mg/ml) for 2 min at various temperatures (22).
Quantification and Purification of the Collagen-Two assays were used to measure the level of expression of the wild-type or modified type I collagen. The first was based on measurement of the 4-hydroxyproline content, assuming that all the hydroxylatable proline residues in the ␣1(I) and ␣2(I) chains had been converted to 4-hydroxyproline. Aliquots of the Triton X-100 supernatants were hydrolyzed at 110°C for 16 h and studied by a colorimetric method for 4-hydroxyproline (23). The second assay was based on densitometry of the Coomassie-stained collagen ␣ chain bands in SDS-PAGE using known amounts of type I collagen (Chemicon) as a standard. The amounts of the ␣1(I) and ␣2(I) chains were estimated by densitometry of the Coomassie-stained bands using a Bioimage instrument (Millipore Corp.).
The recombinant type I collagen was purified as described previously (12), with the exception that the collagen was precipitated with 4 M NaCl.
Other Assays-Amino acid analyses of the homotrimeric and heterotrimeric recombinant type I collagens were performed in an Applied Biosystems 421 or Beckman system 6300 amino acid analyzer. The melting curves were determined in a Jasco J-500 spectropolarimeter with a temperature-controlled quartz cell of path length of 1 cm (Gilford) (12).

Expression of a Recombinant Human Type I Procollagen
Homotrimer in High Five Cells-A recombinant virus coding for the pro␣1(I) chains was generated and used to infect High Five cells together with a double-promoter virus coding for the ␣ and ␤ subunits of human prolyl 4-hydroxylase. The cells were cultured either as monolayers or in suspension, harvested 48 and 72 h after infection, homogenized in a buffer containing 0.2% Triton X-100, and centrifuged. The Triton X-100 soluble proteins of the cell homogenates were then digested with pepsin at 22°C for 1-4 h. Samples were analyzed by SDS-PAGE under reducing conditions followed by Coomassie staining or Western blotting. Two bands corresponding to the pro␣1(I) and pN␣1(I) chains were seen in the Coomassie-stained gel of nonpepsinized samples 48 h after infection (Fig. 1A, lane 2) while only pN␣1(I) chains were seen at 72 h (Fig. 1A, lane 3). The presence of the pro␣1(I) and pN␣1(I) chains was confirmed by Western blotting using the 95D1A antibody against the collagenous regions of various collagens (Fig. 1B, lane 1) and the PINP antibody against the N propeptide of human type I procollagen (data not shown). In the case of the pepsin-digested samples, a major band corresponding to the ␣1(I) chains was seen in the Coomassie-stained gel (Fig. 1A, lane 4), and the same band was identified in the Western blot using the 95D1A antibody (Fig. 1B, lane 2). The level of expression of the human type I collagen homotrimer was about 10 -20 mg/liter, which is lower than the figure of up to 40 mg/liter obtained for type III collagen in the same cells (12). As in the case of type III collagen expression (12), only a minor fraction of the total type I collagen homotrimer produced in High Five cells was found in the culture medium (data not shown).
Sequences at the 5Ј ends of DNA constructs influence the expression level of many polypeptides in the baculovirus system (18). Because the expression level obtained for type I homotrimer was less than that obtained for type III collagen, we decided to study whether the level of expression of the former can be increased by replacing the sequences coding for the signal peptide and the N propeptide of the pro␣1(I) chain with those of the pro␣1(III) chain. A new virus, C1A1NproIII, was generated and High Five cells were infected and analyzed as above. A faint band corresponding to the hybrid pro␣1(I) chains and a major band corresponding to the hybrid pN␣1(I) chains were seen in Coomassie-stained SDS-PAGE 48 h after infection (Fig. 1A, lane 5), and these bands were also stained in Western blotting using the 95D1A (Fig. 1B, lane 3) and PIIINP (data not shown) antibodies. A major band corresponding to the hybrid pN␣1(I) chains and a minor band with the mobility of fully processed ␣ chains were seen 72 h after infection (Fig. 1A,  lane 6). The latter band could not be stained by the PIIINP antibody, suggesting that it indeed represented fully processed ␣ chains. In the case of the pepsin-digested samples, a major band corresponding to the ␣1(I) chain was seen both in the Coomassie-stained SDS-PAGE (Fig. 1A, lane 7) and in the Western blot stained by the 95D1A antibody (Fig. 1B, lane 4). About a 3-fold increase was obtained in the level of expression of the type I collagen homotrimer using the C1A1NproIII virus, the level ranging up to about 60 mg of collagen/liter, which corresponds to about 90 mg of procollagen.
The thermal stability of the type I collagen homotrimer was studied by digestion with a mixture of trypsin and chymotrypsin at various temperatures. These experiments indicated that the T m of the type I collagen homotrimer was about 40°C, a value which remained unaffected by replacement of the type I N propeptide (data not shown).
Purification of the Recombinant Type I Collagen Homotrimer-The type I procollagen homotrimer was expressed in High Five cells in suspension in shaker flasks using either the C1A1 or the C1A1NproIII virus together with the 4PH␣␤ virus. The recombinant type I collagen was purified as described previously for the recombinant type III collagen (12), with minor modifications. The purified type I collagen homotrimer was studied by amino acid and CD spectrum analyses. The amino acid composition agreed with that reported for the human ␣1(I) chains (24), except that the hydroxylysine content was about 80% of that in the nonrecombinant protein (Table I).
The T m of the recombinant type I collagen homotrimer was 42.8 Ϯ 1.2°C (determined from four individual samples) (Fig.  2).
Expression of Recombinant Human Type I Procollagen Heterotrimer in High Five Cells-To study whether collagens consisting of more than one type of ␣ chain can be assembled in insect cells, a recombinant virus C1A2 coding for the pro␣2(I) chain was generated and used to infect High Five cells either with or without the C1A1 virus but in the presence of the 4PH␣␤ virus. Equal MOI amounts of the C1A1 and C1A2 viruses were used in a 5-10-fold excess over the 4PH␣␤ virus. The cells were cultured and homogenized as above, and the samples were analyzed by SDS-PAGE under reducing conditions followed by Coomassie staining. When the pro␣1(I) and pro␣2(I) chains were coexpressed, bands corresponding to the ␣1(I) and ␣2(I) chains with an approximate ratio of 2 to 1 (1.89 Ϯ 0.20 to 1, determined from the integrated densitometry values of four individual samples) were seen in Coomassiestained SDS-PAGE of pepsinized samples (Fig. 3, lane 2). When the pro␣2(I) chains were expressed alone, no pepsin-resistant band was seen (Fig. 3, lane 3). Thus, the pepsin-resistant ␣2(I) chains seen in lane 2 must have been present in heterotrimeric molecules. As the ratio of the ␣1(I) to ␣2(I) chains was about 2:1, essentially all the ␣1(I) chains must likewise have been present in heterotrimers. The highest expression levels obtained for the type I collagen heterotrimer were found to be about 20 mg/liter.
To study whether the level of expression of the type I procollagen heterotrimer can be improved by replacing the signal sequence and N propeptide of the pro␣2(I) chain with those of the pro␣1(III) chain, a new virus C1A2NproIII was generated. Several experiments were performed to express the type I procollagen heterotrimer using different combinations of equal MOI amounts of the viruses C1A1, C1A2, C1A1NproIII, and C1A2NproIII, all in the presence of the 4PH␣␤ virus. Pepsinresistant ␣2(I) chains were detected in all the samples by Coomassie staining of SDS-PAGE, but significant differences were found in the ␣1(I) to ␣2(I) chain ratios upon densitometry (Fig. 3, lanes 4 -7). When insect cells were coinfected with the C1A1 and C1A2 viruses coding for the wild-type pro␣1(I) and pro␣2(I) chains, the ratio of pepsin-resistant ␣1(I) to ␣2(I) chains was consistently 2 to 1 (Fig. 3, lane 2). However, when the C1A2 virus was used together with the C1A1NproIII virus, an excess of the pepsin-resistant ␣1(I) chains was found in several experiments and the ␣1(I) to ␣2(I) chain ratio varied, being about 2-5 to 1 (3.11 Ϯ 1.09 to 1, determined from the integrated densitometry values of four individual samples) (Fig. 3, lanes 4 -5). The level of expression of type I collagen obtained by coinfection with the C1A1NproIII and C1A2 viruses was about 40 mg/liter. When insect cells were coinfected with the C1A1 or C1A1NproIII and C1A2NproIII viruses, the formation of heterotrimeric type I collagen appeared to be very inefficient, the ratio of the pepsin-resistant ␣1(I) to ␣2(I) chains being about 5-10 to 1 (Fig. 3, lanes 6 -7).
In further experiments, the ratio of viruses coding for the pro␣1(I) and pro␣2(I) chains was varied by keeping the amount of C1A2 virus constant but using different amounts of either the C1A1 virus (Fig. 4, lanes 2-5) or C1A1NproIII virus (Fig. 4,  lanes 6 -9) so that the MOI ratio of the C1A1 or C1A1NproIII virus to the C1A2 virus was 0.25 (Fig. 4, lanes 2 and 6), 0.5 (Fig.  4, lanes 3 and 7), 0.75 (Fig. 4, lanes 4 and 8), or 1.0 (Fig. 4, lanes  5 and 9). Pepsin-digested samples were then studied by SDS-PAGE followed by Coomassie staining and densitometry of the bands. The amount of pepsin-resistant ␣1(I) chains increased with increasing amounts of the C1A1 or C1A1NproIII virus as could be expected. In addition, the amount of pepsin-resistant ␣2(I) chains increased in a similar manner even though the amount of C1A2 virus was kept constant (Fig. 4). The ratio of pepsin-resistant ␣1(I) to ␣2(I) chains varied with the original C1A1 and C1A2 viruses, from about 3:1 at a virus ratio of 0.25 ( Fig. 4 lane 2) to about 2:1 with ratios of 0.75 and 1.0 (Fig. 4,    5). When the C1A1NproIII virus was used instead of the C1A1 virus, the 2:1 ratio was obtained with the MOI ratio 0.75, while a 3:1 ratio was obtained with a MOI ratio of 1.0 (Fig. 4, lanes 8 and 9).
The T m of the type I collagen heterotrimer when studied by digestion with a mixture of trypsin and chymotrypsin at various temperatures was about 40°C (Fig. 5).
Purification of the Recombinant Type I Collagen Heterotrimer-The type I procollagen heterotrimer was expressed in High Five cells in suspension using the C1A1 and C1A2 viruses together with the 4PH␣␤ virus, and the recombinant protein was purified as above. The amino acid composition of the purified recombinant collagen corresponded well to that reported for the nonrecombinant human protein (24) (Table I), with the exception that the hydroxylysine content was only about 30%. The T m for the purified collagen when determined by CD analyses was 41.5 Ϯ 0.8°C (determined from three individual samples) (Fig. 2).
Expression of Pro␣2(I) Chains with Modified C Propeptides and C Telopeptides-To study whether pro␣2(I) chains with a modified C propeptide or C propeptide and C telopeptide are able to form homotrimers, High Five cells were infected with any of the viruses C1A2CproIII, C1A2Ctelo-proIII, or C1A2Ctelo-proI together with the 4PH␣␤ virus. In the virus C1A2CproIII, the sequences coding for the C propeptide of the pro␣2(I) chain had been replaced by those coding for the C propeptide of the pro␣1(III) chain, whereas in C1A2Ctelo-proIII and C1A2 Ctelo-proI, the sequences coding for both the C telopeptide and C propeptide of the pro␣2(I) chain had been replaced by those of the pro␣1(III) or pro␣1(I) chain, respectively. When any of these viruses was used, a faint band corresponding to the pro␣2(I) chain with the modified C propeptide was seen in SDS-PAGE after both Coomassie staining and Western blotting using the 95D1A antibody (as shown for C1A2CproIII in Fig. 6A, lanes 2 and 4). A considerable portion of the modified pro␣2(I) chains was found in a Triton X-100 insoluble, 1% SDS soluble fraction (Fig. 6A, lanes 3 and 5), a result that differs distinctly from those obtained in expression experiments involving the pro␣1(I) chain (data not shown).
The stability of the modified pro␣2(I) chains was studied by pepsin treatment of the Triton X-100 soluble fraction. No pepsin-resistant ␣2(I) chains were seen in Coomassie-stained SDS-PAGE (as shown for C1A2CproIII in Fig. 6A, lane 6) when the digestion was performed for 1 h at 22°C (as in the cases of the [␣1(I)] 3 homotrimer and [␣1(I)] 2 ␣2(I) heterotrimer, above). In contrast, pepsin-resistant ␣2(I) chains were seen in the cases of all three modified pro␣2(I) viruses when the digestion was performed for 1 h at 4°C (as shown for C1A2CproIII in Fig. 6B, lane 1). Nevertheless, the pepsin-resistant ␣2(I) chains dissappeared when the digestion at 4°C was prolonged (Fig. 6B, lanes  2-4). Further experiments indicated that no pepsin-resistant ␣2(I) chains were present when the cells were expressing unmodified pro␣2(I) chains, even when the digestion was performed at 4°C (Fig. 6C). DISCUSSION The data reported here indicate that coexpression of the pro␣1(I) and pro␣2(I) chains of human type I procollagen in insect cells leads to the formation of heterotrimeric molecules with the correct 2:1 chain ratio. The data further indicate that expression of pro␣1(I) chains without pro␣2(I) chains effectively leads to the formation of homotrimeric molecules. The formation of the heterotrimers was nevertheless clearly favored, as the ratio of pepsin-resistant ␣1(I) to ␣2(I) chains remained at 2:1 in the coexpression experiments unless the pro␣1(I) chains were expressed in a relatively large excess. Replacement of the sequences coding for the signal peptide and the N propeptide of the pro␣1(I) chain with those of the pro␣1(III) chain increased the level of expression of the pro␣1(I) chain about 3-fold, whereas no corresponding effect was seen when a similar modification was made to the pro␣2(I) chain. The highest expression level obtained for the type I collagen homotrimer with the modified construct was 60 mg/ liter, thus being slightly higher than that previously obtained for type III collagen in High Five cells (12).
The heterotrimeric and homotrimeric procollagen molecules produced in High Five cells were found to be processed with time, as partial conversion of the wild-type pro␣1(I) and pro␣2(I) chains to pN␣1(I) and pN␣2(I) chains (i.e. cleavage of the C propeptides) was seen in all the expression experiments (although not shown for the pro␣2(I) chain in the coexpression experiments). Although the N propeptides of the wild-type pro␣1(I) and pro␣2(I) chains appeared to be stable, the N propeptide of the pro␣1(III) chain artificially transferred into the pro␣1(I) chain was less so, as some of the modified pro␣1(I) chains were processed to ␣(I) chains even in the case of homotrimeric molecules in which the three type III N propeptides should be able to form a correctly folded trimer. Similar processing of the N propeptide of the pro␣1(III) chain in insect cells was also found when this propeptide was transferred to the pro␣1(II) chain, 3 whereas the wild-type N propeptide of the pro␣1(II) chain 3 5. Analysis of the thermal stability of the recombinant human type I collagen heterotrimer by trypsin/chymotrypsin digestion. Cells were infected with the C1A1 and C1A2 viruses together with the 4PH␣␤ virus and harvested 72 h after infection, and the cell extract was digested with pepsin for 1 h at 22°C. The samples were subsequently treated with a mixture of trypsin and chymotrypsin at temperatures between 32 and 42°C as described (22). The samples were analyzed by 8% SDS-PAGE under reducing conditions followed by Coomassie staining. Lane P shows a sample after pepsin digestion without trypsin/chymotrypsin digestion; lanes 32-42 show samples treated with the trypsin/chymotrypsin mixture at the temperatures indicated at the bottom of the lanes. The arrows show the positions of the ␣1(I) and ␣2(I) chains.
(12) appeared to be quite stable when present in their wild-type chains. The N propeptide is likely to fold back on the triple helical collagen domain (25) and may thus interact with sequences in it. The data obtained with the modified constructs suggest that such an interaction may be lost or reduced when the N propeptide is replaced with that of another pro␣ chain. As procollagen proteinases are extracellular enzymes (1,6,15), the processing of procollagen molecules found here was probably due to some nonspecific intracellular proteinases.
The properties of the purified type I collagen heterotrimer were very similar to those of the corresponding nonrecombinant protein. Although early studies on the type I collagen homotrimer produced by renaturation of individual ␣ chains suggested that the T m of the homotrimer may be slightly lower than that of the heterotrimer (7), no lower T m was found here for the homotrimer whether studied by digestion of crude cell extracts with a mixture of trypsin and chymotrypsin at various temperatures or by CD spectrum analysis. The slightly lower T m found for the homotrimer produced by the renaturation experiments (7) may thus be a property limited to that special case.
The wild-type pro␣2(I) chain did not form any homotrimeric molecules, as no pepsin-resistant ␣2(I) chains were seen in the absence of the pro␣1(I) chains. In agreement with this conclusion, the amount of pepsin-resistant ␣2(I) chains obtained in the coexpression experiments increased with increasing expression levels of the pro␣1(I) chains. Attempts were also made to produce homotrimeric molecules of pro␣2(I) chains by replacing the wild-type C propeptide with that of the pro␣1(III) or pro␣1(I) chain. Previous data on the formation of type III procollagen [pro␣1(III)] 3 homotrimers (12) and the present data on the formation of type I procollagen [pro␣1(I)] 3 homotrimers clearly demonstrate that the C propeptides used in the modified pro␣2(I) constructs become effectively associated in insect cells. Pepsin-resistant ␣2(I) chains were found in these experiments only when the digestion was performed at 4°C, while the chains were completely digested at 22°C. These data agree with those obtained in renaturation experiments with individual ␣2(I) chains in which the T m of the [␣2(I)] 3 homotrimer was only 22-24°C (7). Evaluation of the potential interchain interactions in type I collagen has likewise suggested that the [␣2(I)] 3 homotrimer should have fewer stabilizing interactions than the heterotrimer or [␣1(I)] 3 homotrimer (26). The present data differ from recent results obtained in translation experiments in a rabbit reticulocyte lysate system in vitro, in which truncated pro␣2(I) chains with an internal de-letion of 830 amino acids in the collagen domain did form triple helices with a T m of 35°C provided that the C propeptide of the pro␣2(I) chain had been replaced by that of the pro␣1(III) chain (27). This difference may be due to the large deletion used in the pro␣2(I) chain in the cell-free assembly experiments (27). Our data suggest that in addition to control at the level of the C propeptide, additional restrictions may exist at the level of the collagen domain of the full-length pro␣2(I) chain, which prevent the formation of stable homotrimeric molecules in cells.
Hsp-47 is regarded as a chaperone that may be specifically involved in the assembly and/or secretion of collagens (28,29). The present data indicate, however, that formation of the type I procollagen heterotrimers and homotrimers with stable triple helices did not require the presence of any recombinant Hsp-47. The possibility is not excluded that the cells may have markedly up-regulated expression of an insect Hsp-47, and that this protein assisted folding of the human procollagen molecules. This possibility does not seem very likely, however, as baculovirus infection interferes with the synthesis of cellular proteins (30). It is thus more likely that assembly of the type I procollagen heterotrimers and homotrimers did not require Hsp-47. The failure to obtain [pro␣2(I)] 3 homotrimers with stable triple helices from pro␣2(I) chains having modified C propeptides in insect cells and in renaturation experiments starting from individual ␣2(I) chains in vitro (7) might be due to the lack of Hsp-47 if this chaperone were especially important for folding of the pro␣2(I) chains. The current data on Hsp-47 (28, 31) does not suggest any such specificity, however. The failure of insect cells to secrete most of the type I and type III (12) procollagen molecules might likewise be due to lack of Hsp-47, but it has been reported previously that insect cells are also poor at secreting many other recombinant secretory proteins (32)(33)(34). It thus seems more likely that synthesis and assembly of recombinant heterotrimeric and homotrimeric type I procollagen molecules in insect cells does not require Hsp-47.