Characterization of human type III collagen expressed in a baculovirus system. Production of a protein with a stable triple helix requires coexpression with the two types of recombinant prolyl 4-hydroxylase subunit.

An efficient expression system for recombinant collagens would have numerous scientific and practical applications. Nevertheless, most recombinant systems are not suitable for this purpose, as they do not have sufficient amounts of prolyl 4-hydroxylase activity. Pro-α1 chains of human type III collagen expressed in insect cells by a baculovirus vector are reported here to contain significant amounts of 4-hydroxyproline and to form triple-helical molecules, although the T of the triple helices was only about 32-34°C. Coexpression of the pro-α1(III) chains with the α and β subunits of human prolyl 4-hydroxylase increased the T to about 40°C, provided that ascorbate was added to the culture medium. The level of expression of type III procollagen was also increased in the presence of the recombinant prolyl 4-hydroxylase, and the pro-α1(III) chains and α1(III) chains were found to be present in disulfide-bonded molecules. Most of the triple-helical collagen produced was retained within the insect cells and could be extracted from the cell pellet. The highest expression levels were obtained in High Five cells, which produced up to about 80 μg of cellular type III collagen (120 μg of procollagen) per 5 × 106 cells in monolayer culture and up to 40 mg/liter of cellular type III collagen (60 mg/liter procollagen) in suspension. The 4-hydroxyproline content and T of the purified recombinant type III collagen were very similar to those of the nonrecombinant protein, but the hydroxylysine content was slightly lower, being about 3 residues/1000 in the former and 5/1000 in the latter.

The collagens are a family of closely related but distinct extracellular matrix proteins. At least 19 proteins representing more than 30 gene products have now been defined as collagens, and at least another 10 proteins contain collagen-like domains. All collagen molecules consist of three polypeptide chains, called ␣ chains, that are coiled around one another into a triple-helical conformation. In some collagen types all three ␣ chains of the molecule are identical, while in others the mole-cule contains two or three different ␣ chains. The most abundant collagens form extracellular fibrils and are hence known as fibril-forming collagens, while others form supramolecular aggregates of other kinds (for recent reviews on collagens, see Refs. [1][2][3][4][5][6]. Nevertheless, many of the recently discovered collagens are present in tissues in such small quantities that it has not been possible to isolate them for characterization at the protein level, and thus many important questions concerning their structure and organization are still open (1)(2)(3)(4)(5)(6)(7). Some of the fibril-forming collagens are now in medical use, in applications ranging from biomaterials and drug delivery systems to trials for the potential of type II collagen as an oral toleranceinducing agent for the treatment of rheumatoid arthritis (8 -10). The collagens used in these applications have been isolated from animal tissues, and it is thus obvious that an efficient large scale recombinant system for expressing collagens would have numerous scientific and practical applications.
Recombinant expression of collagens has proven difficult to achieve, as their biosynthesis requires processing by up to eight specific posttranslational enzymes (1,6,11). Nevertheless, only one of these enzymes, prolyl 4-hydroxylase, which hydroxylates about 100 proline residues in each of the ␣ chains in the case of the fibril-forming collagens, is an absolute requirement, as 4-hydroxyproline-deficient polypeptide chains cannot form triple helices that are stable at 37°C (6,12). No attempts have previously been reported regarding the production of triplehelical collagens in insect cells, bacteria, or yeast, as insect cells are likely to have insufficient levels of prolyl 4-hydroxylase activity, and bacteria and yeast do not contain this enzyme at all. Recombinant collagens have recently been produced in mammalian cells, however (7,(13)(14)(15), as such cells do have adequate levels of prolyl 4-hydroxylase, at least for experiments in which the aim has not been to obtain a very high expression level.
Baculoviruses have proven to be very efficient expression vectors for the large scale production of various recombinant proteins in insect cells. The proteins produced in this expression system are usually correctly processed, properly folded, and disulfide-bonded (16,17), but as described in this paper, our initial experiments indicated that recombinant collagen polypeptide chains produced in a baculovirus system do not form triple helices that are stable at 37°C. A recent success in producing a fully active human prolyl 4-hydroxylase ␣ 2 ␤ 2 tetramer in insect cells by coinfection with two recombinant baculoviruses, one coding for the ␣ subunit and the other for the ␤ subunit (18), suggested that it might be possible to produce collagens with stable triple helices by using three baculoviruses, one of them coding for the recombinant collagen polypeptide chains and two coding for the ␣ and ␤ subunits of prolyl 4-hydroxylase. The present paper reports on the special features of such a system with human type III collagen as the test collagen and on the characterization of the recombinant collagen produced.

Construction of the Baculovirus Transfer Vector and Generation of the Recombinant Virus-A BglII
site was created 16 base pairs upstream of the translation initiation codon to a full-length cDNA for the pro-␣1 chain of human type III procollagen (19) by polymerase chain reaction, and the cDNA was digested with the BglII and XbaI restriction enzymes. The BglII-XbaI fragment was then ligated to pVL1392 (Invitrogen). The recombinant pVL construct was cotransfected into Spodoptera frugiperda Sf9 insect cells with a modified Autographa californica nuclear polyhedrosis virus DNA using the BaculoGold transfection kit (Pharmingen), and the resultant viral pool was collected, amplified, and plaque-purified (17). The recombinant virus, termed rhproCIII, was checked by a polymerase chain reaction-based method (20).
Analysis of Recombinant Proteins in Insect Cell Cultures-Insect cells (Sf9 or High Five; Invitrogen) were cultured in TNM-FH medium (Sigma) supplemented with 10% fetal bovine serum (BioClear) or in a serum-free HyQ CCM3 medium (HyClone), either as monolayers or in suspension in spinner or shaker flasks at 27°C. To produce recombinant proteins, insect cells seeded at a density of 5-6 ϫ 10 5 /ml were infected with the rhproCIII virus and with the viruses for the ␣ subunit (virus ␣59) and ␤ subunit of human prolyl 4-hydroxylase (18), the rhproCIII virus being used in a 5-10-fold excess over the other two. Ascorbate (80 g/ml) was added to the culture medium daily. The cells were harvested 48 -120 h after infection, washed with a solution of 0.15 M NaCl and 0.02 M phosphate, pH 7.4, homogenized in a 0.3 M NaCl, 0.2% Triton X-100, and 0.07 M Tris buffer, pH 7.4, and centrifuged at 10,000 ϫ g for 20 min. The remaining cell pellet that was insoluble in the homogenization buffer was further solubilized in 1% SDS and analyzed by SDS-PAGE. 1 The cell culture medium was concentrated 10 times in an ultrafiltration cell (Amicon) with a PM-100 membrane. Aliquots of the supernatants of the cell homogenates and the concentrated cell culture medium were analyzed by denaturing SDS-PAGE, followed by staining with Coomassie Brilliant Blue or Western blotting with an antibody to the N-propeptide of human type III procollagen. Other aliquots were studied by a radioimmunoassay for the trimeric N-propeptide of human type III procollagen (Farmos Diagnostica) and a colorimetric method for 4-hydroxyproline (21). Still further aliquots were digested with pepsin for 1 h at 22°C (22), and the thermal stability of the pepsin-resistant recombinant type III collagen was measured by rapid digestion with a mixture of trypsin and chymotrypsin (22). Prolyl 4-hydroxylase activity was assayed by a method based on the hydroxylation-coupled decarboxylation of 2-oxo[1-14 C]glutarate (23).
Double Immunostaining of Sf9 Cells-Sf9 cells were grown on glass slides and fixed in 100% ethanol at Ϫ20°C. Alternatively, cells in monolayer culture were detached, washed twice with a solution of 0.15 M NaCl and 0.02 M phosphate, pH 7.4 (washing solution), suspended in cold ethanol, and spread on silanated (24) glass slides. They were then incubated with 1% bovine serum albumin in 0.15 M NaCl and 0.02 M phosphate, pH 7.4, for 15 min, followed by incubation for 30 min in a 1:50 dilution of a mouse monoclonal antibody to the ␤ subunit (5B5, Dako) and a rabbit polyclonal antibody to the ␣ subunit of human prolyl 4-hydroxylase in the above solution containing bovine serum albumin. The cells were subsequently washed 4 times for 20 min with the washing solution and incubated in a 1:10 dilution of a sheep anti-mouse Ig-rhodamine F(ab)Ј2 fragment (Boehringer Mannheim) and a sheep anti-rabbit IgG fluorescein F(ab)Ј2 fragment (Boehringer Mannheim) in the bovine serum albumin solution for 30 min, washed with the washing solution, rinsed with distilled water, and mounted in Glycergel (Dako). The samples were photographed using a Leitz Aristoplan microscope equipped with epi-illuminator and filters for fluorescein isothiocyanate and tetramethyl rhodamine B isothiocyanate fluorescence.
Purification of Recombinant Type III Collagen-Insect cells expressing the recombinant type III procollagen were washed with a solution of 0.15 M NaCl and 0.02 M phosphate, pH 7.4, homogenized in a cold 0.2 M NaCl, 0.1% Triton X-100, and 0.05 M Tris buffer, pH 7.4 (20 ϫ 10 6 cells/ml), incubated on ice for 30 min, and centrifuged at 16,000 ϫ g for 30 min. Unless otherwise mentioned, all the following steps were performed at 4°C. The supernatant was chromatographed on a DEAE cellulose column (DE-52, Whatman) equilibrated and eluted with a 0.2 M NaCl and 0.05 M Tris buffer, pH 7.4, the void volume being collected. The pH of the sample was lowered to 2.0 -2.5, and the sample was digested with a final concentration of 150 g/ml of pepsin for 1 h at 22°C. The pepsin was irreversibly inactivated by neutralization of the sample followed by overnight incubation on ice. The recombinant type III collagen was precipitated by adding solid NaCl to a final concentration of 2 M and centrifuging at 16,000 ϫ g for 1 h. The pellet was dissolved in a 0.5 M NaCl, 0.5 M urea, and 0.05 M Tris buffer, pH 7.4, for 1 day, and the sample was digested with pepsin as above a second time. The sample was then chromatographed on a Sephacryl HR-500 gel filtration column (Pharmacia Biotech Inc.), eluted with a solution of 0.2 M NaCl and 0.05 M Tris, pH 7.4, dialyzed against 0.1 M acetic acid, and lyophilized.
Other Assays-The amount of purified type III collagen obtained was determined by Sircol collagen assay (Biocolor). Amino acid analysis of the purified type III collagen was performed in an Applied Biosystems 421 amino acid analyzer. The melting curve was determined in a Jasco J-500 spectropolarimeter, equipped with a temperature-controlled quartz cell of 1-cm path length (Gilford). The collagen concentration for the sample shown was 104 g/ml in 0.05% acetic acid. Thermal transition curves were recorded at a fixed wavelength (221 nm) by raising the temperature linearly at a rate of 30°C/h using a Gilford temperature programmer.

Expression of Recombinant Human Type III Procollagen in
Sf9 and High Five Cells-In order to ascertain whether it is possible to produce full-length pro-␣1 chains of human type III procollagen in insect cells and whether the pro-␣1(III) chains produced in these cells can form triple-helical molecules, a recombinant baculovirus coding for the pro-␣1(III) chains was generated and used to infect Sf9 and High Five cells. The cells were cultured in TNM-FH medium supplemented with 10% fetal bovine serum or in a serum-free medium if the cell culture medium was to be analyzed, harvested 72 h after infection, homogenized in a buffer containing 0.2% Triton X-100, and centrifuged. Samples of the Triton X-100 soluble protein fraction and the concentrated cell culture medium were then analyzed by SDS-PAGE under reducing conditions, followed either by Coomassie staining (Fig. 1A) or Western blotting with an antibody to the N-propeptide of type III procollagen (Fig. 1B). Some of the samples were digested with pepsin for 1 h at 22°C before the SDS-PAGE (Fig. 1, A and B, lanes with odd numbers). The triple helix of collagens is resistant to proteolytic enzymes, whereas nontriple-helical pro-␣1(III) chains and the propeptides of triple-helical procollagen molecules are digested.
The level of pro-␣1(III) chain expression was too low for these to be detected in the Coomassie-stained SDS-PAGE (Fig. 1A, lanes 2, 4, 6, and 8), but they could be seen by Western blotting in samples of the Triton X-100 soluble proteins (Fig. 1B, lanes 2 and 6) and cell culture media (Fig. 1B, lanes 4 and 8) in the case of both the Sf9 and High Five cells. After the pepsin digestion, ␣1 chains of type III collagen were seen in the High Five cells in the Coomassie-stained gel (Fig. 1A, lane 7). The pepsin-resistant ␣1(III) chains were not detected in the Western blot (Fig. 1B, lanes 3, 5, 7, and 9), since the antibody used reacts only with the N-propeptides of pro-␣1(III) chains, which were apparently digested by the pepsin.
One possible explanation for the low level of expression of pepsin-resistant type III collagen could be that insect cells have insufficient amounts of prolyl 4-hydroxylase activity. In order to study this possibility, insect cells were coinfected with three recombinant baculoviruses, one of them coding for the pro-␣1(III) chain as above, and the other two coding for the ␣ and ␤ subunits of human prolyl 4-hydroxylase. The cells infected with the three viruses were then cultured, harvested, and analyzed as above, the Coomassie-stained SDS-PAGE being shown in Fig 6) and that of pepsin-resistant ␣1(III) chains ( Fig. 2A, lanes 3 and 7) in both the Sf9 and High Five cells. No pro-␣1(III) chains ( Fig. 2A,  lanes 4 and 8) or ␣1(III) chains ( Fig. 2A, lanes 5 and 9) could be detected in the medium samples in the Coomassie-stained gel, but a minor amount of pro-␣1(III) chains was seen in the Western blot (Fig. 2B, lanes 4 and 8).
The efficiency of multiple baculovirus infection was assessed by immunocytochemical staining of the insect cells. Sf9 cells were coinfected with two recombinant viruses coding for the ␣ and ␤ subunits of prolyl 4-hydroxylase and immunostained with antibodies to these two subunits (Fig. 3). When the analysis was performed 48 h after infection, 87% of the cells were found to express at least one of the two types of subunit, with 90% of the cells expressing one type of subunit also expressing the other.
Prolyl 4-Hydroxylase Activity in the Insect Cells-The 0.2% Triton X-100 extracts of the cell homogenates were analyzed for prolyl 4-hydroxylase activity with an assay based on the hydroxylation-coupled decarboxylation of 2-oxo[1-14 C]glutarate (23). As reported previously (25), a significant level of prolyl 4-hydroxylase activity was found in both the Sf9 and High Five cells, that in the High Five cells being distinctly greater than that in the Sf9 cells (Table I). Infection of the cells with a virus coding for the pro-␣1(III) chains had only minor effects on this activity, whereas the activity in the cells infected with the virus coding for the pro-␣1(III) chains together with viruses coding for the two types of subunit of human prolyl 4-hydroxylase was markedly higher (Table I).
Effect of Recombinant Prolyl 4-Hydroxylase on the Level of Type III Procollagen Expression-Sf9 and High Five cells were infected with the virus coding for the pro-␣1(III) chains either with or without viruses coding for the two types of subunit of prolyl 4-hydroxylase (Table II). The level of expression of total type III procollagen was measured with a radioimmunoassay for the trimeric N-propeptide, and the amount of 4-hydroxyproline formed in the cells was determined by a colorimetric assay. Both values were used to calculate the amount of type III collagen produced, assuming that all the pro-␣1(III) chains formed triple-helical molecules and that all the hydroxylatable proline residues in the pro-␣1(III) chains had been converted to 4-hydroxyproline. Starting out from the known structure of type III procollagen and the amount of 4-hydroxyproline in type III collagen, the amount of type III collagen in the samples was calculated by multiplying the N-propeptide values by 7 and the 4-hydroxyproline values by 8. If all the pro-␣1(III) chains are present in fully hydroxylated, triple-helical molecules, both assays should give identical values, whereas if some of the type III procollagen is converted to type III collagen and the propeptides are subsequently degraded, then the values calculated from the 4-hydroxyproline assay should be higher. All these measurements were made 72 h after infection.
A considerable variation was found in the values obtained in different experiments, as shown in Table II, but a number of conclusions can still be made from the results. First, the amount of 4-hydroxyproline formed was distinctly higher in the cells infected with the prolyl 4-hydroxylase-coding viruses than in their absence in all the experiments. Second, the level of expression obtained in the High Five cells was consistently higher than that obtained in the Sf9 cells. Third, the level of type III collagen produced in the cells coinfected with the prolyl 4-hydroxylase-coding viruses was always higher when calculated from the 4-hydroxyproline values than from the radioimmunoassay values, suggesting either that some of the N-propeptides of type III procollagen had been degraded or that some of the fully 4-hydroxylated pro-␣1(III) chains had remained nontriple-helical. The highest type III collagen expression values were seen in the High Five cells that also expressed prolyl 4-hydroxylase, the amount of cellular type III collagen in these cells being about 41-81 g/5 ϫ 10 6 cells (Table II). The amount of type III collagen found in the culture medium, as measured with the radioimmunoassay, was about 25-50% of the total in the Sf9 cells and about 10 -30% that in the High Five cells.
Experiments were also performed in which High Five cells were grown in suspension in spinner or shaker flasks. A similar effect of the prolyl 4-hydroxylase-coding viruses was seen as above. The highest expression levels found in such experiments were approximately 40 mg of cellular type III collagen/liter of culture in 72 h, about 80 -90% of the total collagen produced being found in the cell pellet and 10 -20% in the medium (details not shown).
Time Course for the Synthesis and Secretion of Type III Procollagen in High Five Cells-Since the level of expression of type III collagen in the High Five cells was found to be about 3-10-fold relative to that in the Sf9 cells, High Five cells were selected for the subsequent experiments. The time course for the expression of hydroxylated type III procollagen was studied in High Five cells infected with the three recombinant viruses. The cells were harvested at 48, 72, 96, and 120 h after infection, and the amount of type III collagen produced was calculated from the radioimmunoassay and 4-hydroxyproline values as above. In three complete experiments the amount of cellular type III collagen was found to be highest at about 72 h, the value decreasing distinctly at 96 and 120 h (details not shown). The amount of prolyl 4-hydroxylase activity in the cell homogenate was likewise highest 72 h after infection (details not shown). The radioimmunoassay showed the amount of type III procollagen in the medium to increase continuously, being about 30 -45% of the total at 120 h. The amount of 4-hydroxyproline in the culture medium increased even more, but evidently most of that was present in collagen degradation products, as no band corresponding to the ␣1 chains of type III collagen was seen at 96 or 120 h by SDS-PAGE analysis (details not shown). Since the amount of cellular type III collagen appeared to be highest 72 h after infection, this time point was used in the subsequent experiments.

Conformational Integrity of the Recombinant Type III Collagen-Association of the pro-␣1(III) chains into trimers was studied by SDS-PAGE analysis under nonreducing conditions.
Essentially all the pro-␣1(III) chains synthesized were found as disulfide-bonded trimers, judging from the disappearance of the band corresponding to monomeric pro-␣1(III) chains and the appearance of a protein band of high molecular weight (Fig.  4, lane 2). After pepsin digestion, the band corresponding to the recombinant type III procollagen was converted to one corresponding to type III collagen, and the protein remained in the form of the trimer, thus indicating the existence of disulfide bonds between the ␣1(III) chains (Fig. 4, lane 3). Virtually all the type III procollagen expressed was soluble in the homogenization buffer containing Triton X-100, as no band corresponding to type III procollagen was seen in the Triton X-100-insoluble, SDS-soluble fraction (Fig. 4, lane 4).
The thermal stability of the type III collagen expressed under different cell culture conditions was studied using digestion with a mixture of trypsin and chymotrypsin after heating to various temperatures (22), ascorbate being either added to the cell culture medium daily as usual or omitted during infection. The Triton X-100-soluble proteins were first digested with pepsin for 1 h at 22°C to convert the type III procollagen to type III collagen (22), and the trypsin/chymotrypsin digestion was then performed for aliquots of the pepsin-treated samples. When the pro-␣1(III) chains were expressed in the absence of prolyl 4-hydroxylase and ascorbate, the T m of the type III collagen was found to be about 32-34°C (Fig. 5A). The presence of either ascorbate or prolyl 4-hydroxylase without the other caused virtually no increase in thermal stability (Fig. 5, B and C), but   when the pro-␣1(III) chains were produced in the presence of both, the T m was increased considerably, being about 40°C (Fig. 5D).
Purification and Characterization of Recombinant Type III Collagen-Type III procollagen was expressed in High Five cells cultured either as monolayers or in suspension in shaker flasks. The cells were harvested 72 h after infection, homogenized in a buffer containing 0.1% Triton X-100, and centrifuged, and the supernatant of the cell homogenate was passed through a DEAE cellulose column to remove nucleic acids. The flow-through fractions containing the type III procollagen were pooled and digested with pepsin. This converted the type III procollagen to type III collagen and digested most of the noncollagenous proteins. The type III collagen was then concentrated by salt precipitation, solubilized, and treated with pepsin as above. The type III collagen was finally separated from pepsin and other remaining contaminants by gel filtration on a Sephacryl S500-HR column. The fractions containing the type III collagen were pooled, dialyzed, and lyophilized.
The purified type III collagen was analyzed by 5% SDS-PAGE under reducing (Fig. 6, lane 2) and nonreducing (Fig. 6,  lane 3) conditions. No contaminants were seen in the Coomassie-stained gel, and the ␣1(III) chains were disulfide-bonded. Amino acid and CD spectrum analyses were performed on the purified type III collagen. The amino acid composition corresponded well with that reported for the human protein (Table  III), although the 4-hydroxyproline content was slightly lower. A distinct difference was found in the amount of hydroxylysine, which was about 3 residues/1000 amino acids in the recombinant type III collagen rather than 5/1000 amino acids in the authentic human type III collagen. The T m of the recombinant type III collagen was 40.8 Ϯ 1.3°C (Ϯ S.D., n ϭ 4), and virtually all of the recombinant protein was stable at 37°C (Fig. 7). DISCUSSION The data reported here indicate that it is possible to achieve large scale expression of native-type triple-helical human collagens in insect cells. The High Five cells gave consistently higher production rates than the Sf9 cells, the highest rates seen in High Five cells when cultured in monolayers being about 80 g of cellular recombinant human type III collagen/5 ϫ 10 6 cells, which corresponds to about 120 g of type III procollagen. The largest amount of cellular type III collagen produced when the High Five cells were cultured in suspension in spinner or shaker flasks was about 40 mg/liter, corresponding to about 60 mg/liter of type III procollagen.
Prolyl 4-hydroxylase plays a central role in the biosynthesis of all collagens, as 4-hydroxyproline residues are essential for the folding of the newly synthesized polypeptide chains into triple-helical molecules (6,12,26). When the pro-␣1 chains of type III procollagen were expressed in insect cells alone, without recombinant prolyl 4-hydroxylase, considerable amounts of 4-hydroxyproline were generated in the cells and the pro-␣1 chains formed triple-helical molecules, as indicated by the resistance of the collagenous domains of these chains to pepsin digestion at 22°C. However, the T m of the triple helices of such The cells that expressed only the pro-␣1 (III) chains or these chains together with the ␣ and ␤ subunits of prolyl 4-hydroxylase were analyzed 72 h after infection.
b The amount of type III collagen was calculated by multiplying the amount of the trimeric N-propeptide of type III procollagen determined by radioimmunoassay by 7, or the amount of 4-hydroxyproline by 8, as described under "Results." c ND, not determined. molecules was about 6 -8°C lower than of those produced in the presence of the recombinant enzyme. Also, the level of expression of type III collagen was lower in the absence of recombinant prolyl 4-hydroxylase than in its presence, probably because many of the partially hydroxylated polypeptide chains failed to form triple-helical molecules even at 27°C and were rapidly degraded. The insect cell system was found to resemble human fibroblasts (27,28) in that the presence of ascorbate in the culture medium was necessary to produce collagen molecules with stable triple helices.
Previous experiments had demonstrated that a fully active human prolyl 4-hydroxylase tetramer can be produced in insect cells by infecting them with two recombinant baculoviruses, one of them coding for the ␣ subunit of human prolyl 4-hydroxylase and the other the ␤ subunit (18). Nevertheless, the recombinant enzyme had on all previous occasions been extracted from the insect cell homogenates, and the assays had been performed in the presence of the polypeptide substrate and the various cosubstrates in vitro (18,25,29). Although the findings suggested that the enzyme may also be active in insect cells in vivo, the present data constitute the first demonstration that this is indeed the case. Double-immunostaining experiments demonstrated that about 90% of the insect cells expressing one of the two types of subunit of human prolyl 4-hydroxylase also expressed the other.
The ␤ subunit of prolyl 4-hydroxylase is a highly unusual multifunctional polypeptide (6,12,26,30), being identical to the enzyme protein-disulfide isomerase (31,32), which is regarded as the in vivo catalyst of disulfide bond formation in the biosynthesis of various secretory and cell surface proteins, including collagens (6,12,30). Although insect cells have a small amount of endogenous protein-disulfide isomerase activity, this activity is markedly increased when the cells are infected with a recombinant baculovirus coding for the human protein-disulfide isomerase/␤ subunit polypeptide (33). In agreement with this, the recombinant pro-␣1(III) chains produced in insect cells expressing the two types of subunit of recombinant prolyl 4-hydroxylase were found to be properly disulfide-bonded when studied by SDS-PAGE under nonreducing conditions. A major difference between the insect cell system studied here and human fibroblasts is that most of the triple-helical collagens produced by the insect cells were found to be retained within the cells, whereas collagenous molecules are rapidly secreted from human fibroblasts after formation of their triple helices (27). The low rate of secretion of triple-helical recombinant collagens from insect cells may be related to the finding  that such cells often secrete secretory proteins poorly in comparison with many vertebrate cells (33)(34)(35). The properties of the purified human type III collagen pro-duced in insect cells were found to be very similar to those of type III collagen extracted from various tissues (1)(2)(3)(4)(5)(6). In particular, the 4-hydroxyproline content and the T m of the triple helices, when determined by CD analysis, were found to be very similar to those of nonrecombinant type III collagen. Interestingly, the hydroxylysine content of the recombinant collagen was found to be about 60% of that of type III collagen extracted from various mammalian tissues even though no recombinant lysyl hydroxylase was coexpressed. This indicates that insect cells must have a considerable level of lysyl hydroxylase activity. We did not study the level of glycosylation of the 3 hydroxylysine residues formed per ␣ chain, as even the ␣ chains of type III collagen extracted from various tissues contain only about 0.1 residue of galactosylhydroxylysine and 0.8 of glycosylgalactosylhydroxylysine (36). However, as insect cells appear to have relatively high levels of prolyl 4-hydroxylase and lysyl hydroxylase activity, they may well also have relatively high levels of collagen glycosyltransferase activities so that some of the hydroxylysine present in the recombinant ␣1(III) chains may be glycosylated. The insect cell system studied here should allow various recombinant human collagens to be produced for use in medical applications. Furthermore, the insect cell system should make it possible to produce large quantities of various collagens that are present in tissues in amounts too small to be characterized at the protein level.