Assembly of stable human type I and III collagen molecules from hydroxylated recombinant chains in the yeast Pichia pastoris. Effect of an engineered C-terminal oligomerization domain foldon.

The C-propeptides of the pro alpha chains of type I and type III procollagens are believed to be essential for correct chain recognition and chain assembly in these molecules. We studied here whether the 30-kDa C-propeptides of the human pC alpha 1(I), pC alpha 2(I), and pC alpha 1(III) chains, i.e. pro alpha chains lacking their N-propeptides, can be replaced by foldon, a 29-amino acid sequence normally located at the C terminus of the polypeptide chains in the bacteriophage T4 fibritin. The alpha foldon chains were expressed in Pichia pastoris cells that also expressed the two types of subunit of human prolyl 4-hydroxylase; the foldon domain was subsequently removed by pepsin treatment, which also digests non-triple helical collagen chains, whereas triple helical collagen molecules are resistant to it. The foldon domain was found to be very effective in chain assembly, as expression of the alpha 1(I)foldon or alpha 1(III)foldon chains gave about 2.5-3-fold the amount of pepsin-resistant type I or type III collagen homotrimers relative to those obtained using the authentic C-propeptides. In contrast, expression of chains with no oligomerization domain led to very low levels of pepsin-resistant molecules. Expression of alpha 2(I)foldon chains gave no pepsin-resistant molecules at all, indicating that in addition to control at the level of the C-propeptide other restrictions at the level of the collagen domain exist that prevent the formation of stable [alpha 2(I)]3 molecules. Co-expression of alpha 1(I)foldon and alpha 2(I)foldon chains led to an efficient assembly of heterotrimeric molecules, their amounts being about 2-fold those obtained with the authentic C-propeptides and the alpha 1(I) to alpha 2(I) ratio being 1.91 +/- 0.31 (S.D.). As the foldon sequence contains no information for chain recognition, our data indicate that chain assembly is influenced not only by the C-terminal oligomerization domain but also by determinants present in the alpha chain domains.

The collagen superfamily of proteins includes more than 20 types of collagen and more than 15 additional proteins that have collagen-like domains. All collagen molecules consist of three polypeptide chains, called ␣ chains, 1 that are coiled around each other into a triple helix and contain the triplet sequence -Gly-X-Y-, in which the Y position amino acid is often 4-hydroxyproline. The most abundant collagens form fibrils and are therefore known as fibril-forming collagens, whereas others form other kinds of supramolecular structures. The molecules of the most abundant fibril-forming collagen, type I, consist of two ␣1(I) chains and one ␣2(I) chain, whereas the molecules of type III collagen are [␣1(III)] 3 homotrimers. In addition to the type I collagen heterotrimer, most tissues also contain a small amount of type I collagen with a chain composition of [␣1(I)] 3 , known as the type I collagen homotrimer (for reviews, see Refs. [1][2][3][4]. The fibril-forming collagens are synthesized as procollagen molecules with N-and C-terminal propeptides. Chain assembly begins with association of the three C-propeptides through a process directed by their structures (2,5). Renaturation experiments with individual type I collagen ␣ chains that lack any propeptides have indicated, however, that the chains form both [␣(I)] 2 ␣2(I) heterotrimers and [␣1(I)] 3 homotrimers, although the process is very slow and the T m values of the molecules formed are lower than of those present in vivo (6,7). Based on these findings and numerous subsequent studies carried out in a large variety of biological systems over a period of more than 25 years, the C-propeptides are now believed to be essential for correct chain recognition and to play a crucial role in chain assembly in vivo (2,5,8).
Most studies of collagen synthesis have used vertebrate cells, which usually possess sufficient levels of all the specific cotranslational and posttranslational enzymes needed for collagen processing. In recent years experiments have also been performed using recombinant expression in cultured insect cells (9 -11), yeasts (12)(13)(14) and plants (15). These cells have been shown to assemble partially (13,15) or fully (9 -12, 14) hydroxylated recombinant collagen chains into molecules with partially (13,15) or fully (9 -12, 14) stable triple helices, provided that they co-express the two types of subunit of a recombinant prolyl 4-hydroxylase. This key enzyme of collagen synthesis is an ␣ 2 ␤ 2 tetramer in vertebrates (16 -18). Most of the triple helical collagen molecules assembled in insect cells, yeasts, and plants are not secreted (9 -15) but accumulate within the endoplasmic reticulum (19). Studies of these systems have confirmed that the N-propeptides are not required for the assembly of triple helical molecules, as they can be swapped between collagen types (10,11) or omitted (14,15,20). A recent study reported, highly surprisingly, that assembly of triple helical recombinant type I collagen molecules from par-tially hydroxylated ␣ chains in the yeast Saccharomyces cerevisiae does not even require the C-propeptides (20). The triple helical molecules had an ␣1(I) to ␣2(I) chain ratio of 5:1, however, suggesting that they consist of mixtures of [␣1(I)] 2 ␣2(I) heterotrimers and [␣1(I)] 3 homotrimers (20).
In the present work we have studied whether fully hydroxylated recombinant human type I collagen ␣ chains lacking C-propeptides can assemble effectively into triple helical molecules in the yeast Pichia pastoris when it co-expresses the two types of subunits of a recombinant human prolyl 4-hydroxylase (12,14). As the collagen domains of the ␣ chains of type III collagen, unlike those of type I, contain two cysteine residues at their C-terminal ends, which are involved in the formation of interchain disulfide bonds (see Ref. 21), we also determined whether type III collagen ␣ chains lacking their C-propeptides would be assembled more effectively than those of type I. One major aspect studied here was whether the 30-kDa C-propeptides of the pro␣1(I), pro␣2(I), or pro␣1(III) chains could be replaced by foldon, a 29-amino acid peptide that is normally located at the C terminus of the polypeptide chains in the bacteriophage T4 fibritin, a three-stranded ␣ helical coiled-coil protein, and appears to be essential for the assembly of the fibritin molecule (22)(23)(24). The addition of foldon to the C terminus of the peptide (Pro-Pro-Gly) 10 has been shown to markedly increase the T m of the triple helical 4-hydroxyproline-free molecules that are formed from these peptides even in the absence of foldon (25,26). No data are available to indicate whether foldon could be used to replace the C-propeptides of pro␣ chains in the assembly of stable triple helical collagen molecules.
We were unable to confirm the effective assembly of triple helical collagen molecules from ␣ chains lacking the C-propeptides in yeast cells (20), at least in P. pastoris, but it was clear that the C-propeptides could be replaced by the foldon sequence. Expression of ␣1(I)foldon or ␣1(III)foldon chains led to an even more efficient assembly of homotrimeric molecules with stable triple helices than in the cases of chains expressed with their own C-propeptides. Furthermore, co-expression of the ␣1(I)foldon and ␣2(I)foldon chains led to the effective formation of type I collagen heterotrimers with the correct 2:1 chain ratio, indicating that the chain composition is determined not only by the C-terminal oligomerization domain but also by determinants present in the collagen domain of the polypeptide chains, at least in the case of very small oligomerization domains such as foldon.

P. pastoris Expression Vectors and Generation of Recombinant
Strains-The P. pastoris host strain yJC300 (his4, arg4, ade1) and the expression vectors pBLADESX and pBLARGIX (27) were gifts from Dr. James Cregg, (Keck Graduate Institute of Applied Life Sciences), and the vectors pPIC3K, pPICZB, and pPICZ␣A were from Invitrogen. The recombinant strains were generated by the electroporation method according to the manufacturer's instructions (Invitrogen (32)). The recombinant strains were of the methanol utilization-plus phenotype. PC␣1(I), ␣1(I)foldon, and ␣1(I) Strains-A cDNA for the ␤ subunit of human prolyl 4-hydroxylase lacking the signal sequence and flanked by EcoRI restriction sites (12) was cloned into pPICZ␣A in frame with the S. cerevisiae ␣ mating factor (␣MF) pre-pro sequence. The expression cassette encoding the ␣MF-␤ polypeptide was digested from pPICZ␣A␤ with BamHI-BglII and cloned into pBLADESX. To generate a recombinant P. pastoris strain expressing human prolyl 4-hydroxylase ␣ 2 ␤ 2 tetramers, the pBLARGIX␣ (14) and pBLADESX␤ were linearized with HincII and SpeI, respectively, and cotransformed into the yJC300 strain. The resulting strain was named Arg␣Ade␤,hisϪ.
A P. pastoris expression cassette encoding human type I pC␣1 chains (procollagen chains lacking their N-propeptides) was digested from pPICZBpC␣1(I) (14) with PmeI-NotI and cloned into pPIC3K to generate pPIC3KpC␣1(I). A cDNA coding for the 29-amino acid foldon domain of fibritin (22-24) (GenBank™ accession no. AAD42679) was generated by annealing the oligonucleotide Foldon-5Ј1 (5Ј-AGCTTTA-TATTCCTGAAGCTCCAAGAGATGGGCAAGCTTACGTTCGTAA-3Ј) with Foldon-5Ј2 (5Ј-CCATCTTTACGAACGTAAGCTTGCCCATCTCTT-GGAGCTTCAGGAATATAA-3Ј), and Foldon-3Ј1 (5Ј-AGATGGCGAAT-GGGTATTCCTTTCTACCTTTTTATCACCAGCATAAGC-3Ј) with Foldon-3Ј2 (5Ј-GGCCGCTTATGCTGGTGATAAAAAGGTAGAAAGGAATA-CCCATTCG-3Ј). The oligonucleotides (Invitrogen) were designed so that HindIII and NotI overhangs (underlined) are created at the 5Ј and 3Ј ends of the annealed Foldon-5Ј and Foldon-3Ј fragments, respectively, and cohesive overhangs at the 3Ј and 5Ј ends. The foldon fragments were co-ligated into HindIII-NotI digested pBluescript (Stratagene) to generate pBSfoldon. To replace the sequence coding for the C-propeptide of the pC␣1(I) chain with that coding for foldon, two fragments were generated by PCR, the first extending from an internal BamHI site in the pC␣1(I) cDNA to the codon for the last amino acid of the C-telopeptide and the second from the first codon of the foldon cDNA to the NotI site following the stop codon. These fragments were co-ligated into BamHI-NotI-digested pPIC3KpC␣1(I) to generate pPIC3K␣1(I)foldon. To delete the C-propeptide, a fragment extending from the internal BamHI site of pC␣1(I) cDNA to the codon for the last amino acid of the C-telopeptide followed by a stop codon, and a NotI site was created by PCR and ligated into BamHI-NotI-digested pPIC3KpC␣1(I) to generate pPIC3K␣1(I). The pPIC3KpC␣1(I), pPIC-3K␣1(I)foldon, and pPIC3K␣1(I) constructs were linearized with SalI and transformed into the Arg␣Ade␤,hisϪ strain expressing recombinant human prolyl 4-hydroxylase tetramers. Schematic representations of the pC␣1(I), ␣1(I)foldon and ␣1(I) chains are shown in Fig. 1.
To study the expression of type I collagen heterotrimers, the linearized pPICZBpC␣2(I) and pPICZB␣2(I)foldon constructs were transformed into the above strains expressing pC␣1(I) and ␣1(I)foldon chains, respectively. PC␣1(III), ␣1(III)foldon, and ␣1(III) Strains-To delete the sequence encoding the N-propeptide of pro␣1(III) chains, a PCR fragment extending from the PmeI site of the alcohol oxidase 5Ј sequence of pPICZB to the end of the signal sequence of pro␣1(III) and followed directly by the sequence coding for the N-telopeptide of pro␣1(III) until an internal NdeI site was amplified using pPICZBpro␣1(III) (12) as a template and a 5Ј-GCCCATATGAATCATACTGTGCCAAAATAATAGTGGGATG-AAGCA-3Ј oligonucleotide (with the nucleotides corresponding to the signal sequence and N-telopeptide shown in bold and italics, respectively, and the NdeI site underlined) as the reverse primer. The fragment was cloned into PmeI-NdeI digested pPICZBpro␣1(III) to generate pPICZBpC␣1(III). To change the expression vectors, the pro␣1(III) and pC␣1(III) expression cassettes were digested from the pPICZB vectors with PmeI-NotI and cloned into pPIC3K to generate pPIC3Kpro␣1(III) and pPIC3KpC␣1(III). To replace the C-propeptide of pC␣1(III) with foldon, two fragments were generated by PCR, the first extending from an internal AvrII site in the pC␣1(III) cDNA to the 3Ј end of the C-telopeptide sequence and the second encoding foldon as described above. The fragments were co-ligated into AvrII-NotI-digested pPIC3KpC␣1(III) to generate pPIC3K␣1(III)foldon. To delete the Cpropeptide, a fragment extending from the internal AvrII site to the 3Ј end of the C-telopeptide sequence followed by a stop codon, and a NotI site was created by PCR and ligated into AvrII-NotI-digested pPIC3KpC␣1(III) to generate pPIC3K␣1(III). The pPIC3Kpro␣1(III), pPIC3KpC␣1(III), pPIC3K␣1(III)foldon, and pPIC3K␣1(III) constructs were linearized with StuI and transformed into the Arg␣Ade␤,hisϪ strain.
Culture and Induction of P. pastoris Strains-Cells were cultured in 25-ml shaker flasks in a buffered glycerol complex medium, pH 6.0, with 1 g/liter yeast extract and 2 g/liter peptone. Expression was induced in a buffered minimal methanol medium, pH 6.0, and methanol was added every 12 h to a final concentration of 0.5%. Amino acids were added up to 100 g/liter as required.
Analysis of the Recombinant Collagens-Cells were harvested after a 60-h methanol induction at 30°C, washed once, and suspended in cold (4°C) 5% glycerol, 1 mM Pefabloc SC, and 50 mM sodium phosphate buffer, pH 7.4. The cells were broken by vortexing with glass beads, and the lysate was centrifuged at 10,000 ϫ g for 30 min. Aliquots of the soluble fractions were analyzed by SDS-PAGE under reducing conditions followed by Western blotting with polyclonal type I or type III collagen antibodies (Rockland). Further aliquots were digested with pepsin for 2 h at 22°C or 16 h at 4°C, the thermal stability of the pepsin-resistant recombinant collagens was studied by digestion with a mixture of trypsin and chymotrypsin for 2 min at various temperatures (28), and the samples were analyzed by SDS-PAGE under reducing or nonreducing conditions followed by Coomassie Blue staining or Western blotting as described above. The amounts of the collagen chains were estimated by densitometry of the Coomassie Blue-stained bands using a GS-710 calibrated imaging densitometer (Bio-Rad). 3 Homotrimers-It has previously been shown that stable recombinant human type I and type III collagen homotrimers and type I collagen heterotrimers can be produced in P. pastoris by expressing full-length pro␣1(I) or pro␣1(III) chains alone or by co-expressing pro␣1(I) chains with pro␣2(I) chains in a recombinant strain also expressing human prolyl 4-hydroxylase ␣ 2 ␤ 2 tetramers (12,14). Deletion of the N-propeptides had no effect on the chain assembly, the pC␣1(I) and pC␣2(I) chains producing heterotrimeric pCcollagen molecules with the correct 2:1 chain ratio (14). In addition, the expression levels of the type I pCcollagen homotrimers and heterotrimers were 1.5-3-fold relative to those of the corresponding procollagen trimers (14).

Assembly of Stable Type I [␣1(I)] 3 and Type III [␣1(III)]
To study the effect of replacement of the C-propeptide by the foldon sequence and deletion of the C-propeptide on the assembly of type I and III collagen homotrimers, expression constructs encoding pC␣1(I), ␣1(I)foldon, and ␣1(I) chains or pro␣1(III), pC␣1(III), ␣1(III)foldon, and ␣1(III) chains were transformed by electroporation into a P. pastoris strain expressing active human prolyl 4-hydroxylase. The strains were cultured in buffered glycerol complex medium, and expression was induced in buffered minimal methanol medium, methanol being added every 12 h. The cells were harvested 60 h after induction, broken by vortexing with glass beads, and aliquots of the soluble fraction of the cell extracts were analyzed by SDS-PAGE under reducing conditions followed by Western blotting with a polyclonal antibody to type I or type III collagen. Bands corresponding to the pC␣1(I), ␣1(I)foldon, and ␣1(I) chains ( Fig. 2A, lanes 1-3) and pro␣1(III), pC␣1(III), ␣1(III)foldon, and ␣1(III) chains (Fig. 3A, lanes 1-4) were seen in the immunoblots.
The foldon domain was found to be very efficient for chain assembly, as increased amounts of pepsin-resistant ␣1(I) and ␣1(III) chains were seen in the samples from cells expressing the ␣1(I)foldon and ␣1(III)foldon chains, respectively (Figs. 2, B and C, and 3B). Densitometry of the bands in 11 individual samples indicated that the increase in the expression level of pepsin-resistant type I collagen homotrimers in the strains expressing ␣1(I)foldon chains was up to 3.0-fold that in strains expressing pC␣1(I) chains at the highest levels. Similar experiments indicated a 2.4-fold increase in the expression level of pepsin-resistant type III collagen homotrimers in strains expressing ␣1(III)foldon chains. The ␣1(I) and ␣1(III) chains lacking any oligomerization domain also became assembled into triple helical molecules (Figs. 2, B and C, lanes 3, and 3B, lane  4), but their assembly levels were distinctly lower than those of the pC␣1(I), pC␣1(I)foldon, pC␣1(III), and ␣1(III)foldon chains (Figs. 2, B and C, and 3B), and the chains lacking any oligomerization domain were susceptible to degradation (Figs. 2, B and  C, lanes 3, and 3B, lane 4).
The ␣1(III) chains contain at the C-terminal end of their collagen domain two cysteine residues that form interchain disulfide bonds (see Ref. 21). To study whether the ␣1(III) chains expressed with the foldon domain are disulfide-bonded and thus correctly aligned, pepsin-digested samples from strains expressing pC␣1(III) and ␣1(III)foldon chains were analyzed by SDS-PAGE under nonreducing conditions followed by Coomassie staining. The vast majority of the chains expressed with the foldon sequence (Fig. 4B, lane 2) was found as disulfide-bonded trimers, the extent of disulfide bonding being at least as high, if not higher, than in the case of chains expressed with their own C-propeptides (Fig. 4A, lane 2). It thus seems evident that the chains expressed with the foldon domain were correctly aligned. The thermal stability of the type I collagen homotrimers assembled from the pC␣1(I), ␣1(I)foldon, and ␣1(I) chains was analyzed by digestion with a mixture of trypsin and chymotrypsin after heating to various temperatures (28). The T m of the recombinant type I collagen homotrimers expressed in all three strains was between 38 and 40°C ( Fig. 5A-C), as reported previously for human type III collagen produced in P. pastoris in shaker flasks (12). 3 Homotrimers-To study whether pC␣2(I) chains with the C-propeptide replaced with foldon are able to form homotrimers, strains were generated that expressed pC␣2(I) and ␣2(I)foldon chains in the presence of human prolyl 4-hydroxylase. When these strains were cultured, induced, and harvested as described above and analyzed by SDS-PAGE under reducing conditions followed by Western blotting, full-length pC␣2(I) and ␣2(I)foldon chains were seen in the immunoblots of the soluble extracts (Fig. 6A,  lanes 1 and 2). Several additional immunoreactive bands, corresponding to degradation products, were also present (Fig. 6A,  lanes 1 and 2), however, and in the case of samples digested with pepsin at either 22 or 4°C, no pepsin-resistant ␣2(I) chains were seen in either strain (Fig. 6, B and C, lanes 1  and 2).

Lack of Assembly of Any Stable [␣2(I)]
Assembly of Stable Type I Collagen Heterotrimers-The Cpropeptides are believed to have an essential role in the selective association of procollagen chains in a type-specific manner (2,5,8). The effect of replacement of the C-propeptides of both the pC␣1(I) and pC␣2(I) chains with foldon on the assembly of type I collagen heterotrimers was studied by transforming expression constructs encoding pC␣2(I) and ␣2(I)foldon chains into the above described recombinant strains expressing pC␣1(I) and ␣1(I)foldon strains, respectively. The strains were cultured and harvested as described above. Bands corresponding to the pC␣1(I) and pC␣2(I) chains and ␣1(I)foldon and ␣2(I)foldon chains were seen in the immunoblots of the recombinant strains (details not shown). Two pepsin-resistant polypeptides corresponding to the ␣1(I) and ␣2(I) chains were seen in samples from the two strains co-expressing pC␣1(I) and  Fig. 2. Soluble fractions of cell lysates from strains expressing pC␣2(I) (lanes 1) and ␣2(I)foldon (lanes 2) chains were analyzed without pepsin treatment (A) and after digestion with pepsin at 22°C for 2 h (B) or 4°C for 16 h (C) by 8% SDS-PAGE under reducing conditions followed by Western blotting using a polyclonal antibody against type I collagen.
It has previously been reported that the pC␣1(I) and pC␣2(I) chains are assembled into type I collagen heterotrimers with the correct 2:1 chain ratio, 1.92 Ϯ 0.13 (S.D.) (14). Densitometry of the pepsin-resistant ␣1(I) and ␣2(I) bands from 10 individual strains co-expressing ␣1(I)foldon and ␣2(I)foldon chains indicated that the ␣1(I) to ␣2(I) chain ratio was 1.91 Ϯ 0.31 (S.D.) (Fig. 7, lane 2), and an increased quantity of pepsinresistant ␣1(I) and ␣2(I) chains was also seen in these samples (Fig. 7). Densitometry of the bands in seven individual samples indicated that the increase in the expression level of type I collagen heterotrimers in these strains was 2.1-fold relative to that obtained with the strains co-expressing pC␣1(I) and pC␣2(I) chains. The T m of the recombinant type I collagen heterotrimers produced in the strains co-expressing pC␣1(I) and pC␣2(I) chains or ␣1(I)foldon and ␣2(I)foldon chains was between 38 and 40°C (data not shown).

DISCUSSION
Numerous studies carried out in a large variety of biological systems have demonstrated that the C-propeptides are essential for correct chain recognition and chain assembly in the procollagen molecules in vivo (2,5,8). These events are then followed by nucleation and alignment of the collagen domains of the pro␣ chains, driven mainly by sequences present at the C-terminal end of these domains and propagation of the triple helix from the C terminus toward the N terminus (2,5,8,29). Where the C-propeptides of the pro␣ chains in the precursor forms of all three main fibril-forming collagens, types I-III, consist of about 245 amino acid residues (1,30), the critical region in the C-propeptides required for correct chain recognition consists only of a discontinuous sequence of 15 amino acids (8). No data are available, however, on any effective replacement of the C-propeptides by a smaller domain. One study performed in a cell-free translation system in the presence of semipermeabilized cells indicated that the C-propeptides can be replaced with a transmembrane domain, albeit at a much lower efficiency (31).
The data reported here indicate that the short, 29-amino acid foldon sequence can effectively replace the large C-propeptides of the human pC␣1(I), pC␣2(I), and pC␣1(III) chains in chain assembly in P. pastoris cells that also express the two types of subunit of human prolyl 4-hydroxylase. The foldon domains could subsequently be removed by pepsin treatment, leading to collagen molecules with stable triple helices. Expression of the ␣1(I)foldon and ␣1(III)foldon chains gave homotrimeric [␣1(I)] 3 and [␣1(III)] 3 collagen molecules in pepsin treatment, whereas co-expression of the ␣1(I)foldon and ␣2(I)foldon chains gave heterotrimeric collagen molecules with the correct 2:1 chain ratio. Replacement of the C-propeptides by the foldon sequence also increased the levels of the pepsin-resistant homotrimeric and heterotrimeric molecules by more than 2-fold. It therefore appears that the chains containing foldon also became assembled even more efficiently than those containing the Cpropeptides, whereas the nonassembled chains were rapidly degraded.
Expression of ␣ chains without any C-terminal oligomerization domain led to inefficient assembly of triple helical molecules, as the amounts of pepsin-resistant [␣1(I)] 3 and [␣1(III)] 3 homotrimers recovered were distinctly smaller than those obtained when the chains were expressed with an oligomerization domain. This result disagrees with a recent report of efficient assembly of type I collagen heterotrimers from partially hydroxylated ␣ chains in the yeast S. cerevisiae (20). The reason for this difference is unknown, as the data reported for S. cerevisiae disagree not only with our data obtained with another yeast strain but also with those reported in numerous previous studies in a variety of biological systems (2,5,8).
Although the pro␣1(I) and pro␣2(I) chains can form both [pro␣1(I)] 2 pro␣2(I) heterotrimers and [pro␣1(I)] 3 homotrimers, the former are preferred (2,5). In agreement with this preference, co-expression of pro␣1(I) and pro␣2(I) chains in insect cells (10) and P. pastoris (14) led essentially to the formation of heterotrimers only, unless the pro␣1(I) chain was expressed in great excess (10). As the foldon sequence contains no information for chain recognition, it seemed possible that co-expression of ␣1 ( 3 trimers are thermally unstable (6, 7) and would thus be degraded, whereas the [␣1(I)] 3 homotrimer is stable, as shown here and previously (10,14). In this case the samples would contain about 75% pepsinresistant [␣1(I)] 2 ␣2(I) heterotrimers and 25% [␣1(I)] 3 homotrimers, and the ratio of ␣1(I) to ␣2(I) chains would be 3:1 rather than the observed 2:1. Our data thus indicate that even in the presence of the foldon sequence, the chain composition is influenced by determinants present in the ␣ chain domains. The chain ratio obtained here differs distinctly from the ␣1(I) to ␣2(I) chain ratio of 5:1 obtained for pepsin-resistant molecules when ␣ chains lacking any propeptides were expressed in S. cerevisiae (20). This difference may be due in part to a preferential degradation of the ␣2(I) chains in the absence of efficient trimer formation.
Renaturation experiments with ␣2(I) chains lacking any propeptides have demonstrated that they do form [␣2(I)] 3 homotrimers but that the T m of such homotrimers is only 22-24°C (6,7). Because the foldon domain has been reported to promote marked stabilization of the triple helix formed by (Pro-Pro-Gly) 10 -foldon, we also studied whether fully hydroxylated ␣2(I)foldon chains would form homotrimers that are converted to stable [␣2(I)] 3 molecules upon pepsin treatment. This possibility was supported by a report on in vitro translation experiments in a rabbit reticulocyte lysate system indicating that truncated pro␣2(I) chains with an internal deletion formed homotrimers with stable triple helices provided that the Cpropeptide of the pro␣2(I) chains had been replaced by that of the pro␣1(III) chains (8). On the other hand, full-length pro␣2(I) chains in which the C-propeptide had been replaced FIG. 7. Analysis of the assembly of type I collagen heterotrimers from modified pro␣1(I) and pro␣2(I) chains. The recombinant P. pastoris strains were induced and harvested and the cells broken as described in the legend for Fig. 2. Soluble fractions of cell lysates from strains co-expressing pC␣1(I) and pC␣2(I) chains (lane 1) or ␣1(I)foldon and ␣2(I)foldon chains (lane 2) were analyzed after digestion with pepsin by 8% SDS-PAGE under reducing conditions followed by Coomassie Blue staining. The arrows indicate ␣1(I) and ␣2(I) chains, respectively. with that of the pro␣1(III) chains formed no homotrimeric molecules with stable triple helices in recombinant expression in insect cells (10). The present data clearly indicate that the pro␣2(I)foldon chains formed no homotrimers with stable triple helices, a finding that supports the previous suggestion (10) 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 pro␣2(I) chains that prevent formation of stable [␣2(I)] 3 triple helices.