Identification and Characterization of a Third Human, Rat, and Mouse Collagen Prolyl 4-Hydroxylase Isoenzyme*

Collagen prolyl 4-hydroxylases (C-P4Hs) catalyze the formation of 4-hydroxyproline by the hydroxylation of -X-Pro-Gly-triplets. The vertebrate enzymes are α2β2 tetramers, the β-subunit being identical to protein-disulfide isomerase (PDI). Two isoforms of the catalytic α-subunit, which combine with PDI to form [α(I)]2β2 and [α(II)]2β2 tetramers, have been known up to now. We report here on the cloning and characterization of a third vertebrate C-P4H α-subunit isoform, α(III). The processed human, rat and mouse α(III) polypeptides consist of 520–525 residues, all three having signal peptides of 19–22 additional residues. The sequence of the processed human α(III) polypeptide is 35–37% identical to those of human α(I) and α(II), the highest identity being found within the catalytically important C-terminal region and all five critical residues at the cosubstrate binding sites being conserved. The sequence within a region corresponding to the peptide-substrate binding domain is less conserved, but all five α helices constituting this domain can be predicted to be located in identical positions in α(I), α(II), and α(III) and to have essentially identical lengths. The α(III) mRNA is expressed in many human tissues, but at much lower levels than the α(I) and α(II) mRNAs. In contrast to α(I) and α(II), no evidence was found for alternative splicing of the α(III) transcripts. Coexpression of a recombinant human α(III) polypeptide with PDI in human embryonic kidney cells led to the formation of an active enzyme that hydroxylated collagen chains and a collagen-like peptide and appeared to be an [α(III)]2β2 tetramer. The catalytic properties of the recombinant enzyme were very similar to those of the type I and II C-P4Hs, with the exception that its peptide binding properties were intermediate between those of the type I and type II enzymes.


EXPERIMENTAL PROCEDURES
Isolation of cDNA Clones-PCR primers 1-␣(III) (5Ј-GTTAGGAATG-CAGCACTGTTT-3Ј) and 2-␣(III) (5-GCTGGAGCTGCAGGGTCT-3Ј) were used to obtain a 162-bp product from a Human Fetal Marathon-Ready cDNA (BD Biosciences), which was then used to screen a human umbilical vein endothelial cell (HUVEC) lambda cDNA library (Stratagene), 13 positive clones being obtained. The human ␣(III) cDNA was also analyzed by consecutive overlapping 5Ј-and 3Ј-RACE reactions in a Marathon-Ready human total fetal cDNA pool (BD Biosciences), and its 5Ј-end was obtained by 5Ј-RACE of HUVEC cDNA using Advantage-GC 2 polymerase (BD Biosciences). HUVEC cDNA was generated using the SMART RACE cDNA amplification kit (BD Biosciences), and total RNA isolated from cultured HUVEC cells with the RNeasy Midi kit (Qiagen). The full-length human ␣(III) cDNA was cloned into pUC18 in two steps. An ␣(III) fragment extending from nucleotide 184 to the poly(A) tail was first cloned into pUC18 using the SureClone ligation kit (Amersham Biosciences), after which this construct was digested with BamHI-ApaI and a similarly digested ␣(III) fragment covering bp 1-418 was ligated to it. The rat and mouse ␣(III) cDNAs were assembled by consecutive overlapping 5Ј-and 3Ј-RACE reactions using various rat and mouse Marathon-Ready cDNA pools (BD Biosciences) as templates and initial 5Ј-and 3Ј-RACE primers based on a 169 bp rat sequence (GenBank™ accession number AW921407). DNA sequencing was performed on ABI Prism 377 (Applied Biosystems). Nucleotide and amino acid sequence homology comparisons were made using the ClustalW service at the European Bioinformatics Institute.
Analysis of the Expression of Human ␣(III) Subunit mRNA-Human MTN Blot (BD Clontech) and Real TM Human Fetal mRNA Blot I and II (Invitrogen) containing 2 g poly(A)ϩ RNA per lane were hybridized using ULTRAhyb TM solution under the conditions specified by the manufacturer, with 0.5 g of a 2250-bp ␣(III) or a 1440-bp ␣(I) cDNA fragment as the probe, the blots being exposed for 72 and 4 h, respectively. PCR analysis of the Human MTC Panel I and Human Fetal MTC Panel (BD Clontech) was performed according to the manufacturer's protocol with the primer pairs ␣(I)panel-5Ј (GAAGGCGAGATTTCTAC-CATAGATAAAGTC) and ␣(I)panel-3Ј (CCTTGTACGTTGTCAGAATT-GGAATGAC), and ␣(III)panel-5Ј (GACATGGGGGATTATTACCATGC-CATTC) and ␣(III)panel-3Ј (ACCCAAAACTGGTGACCCTCAACCAC). The amounts of the ␣(I) and ␣(III) primers were optimized prior to the analyses to produce similar amounts of DNA from equal amounts of the ␣(I) and ␣(III) plasmids under the same PCR conditions. The preincubation was at 95°C for 1 min, followed by 26,30,34,38, and 42 cycles at 95°C for 30 s and 68°C for 1 min, 5-l aliquots being taken at each point and analyzed on 1% agarose gels.
Total RNA from human fetal epiphyseal cartilage was a gift from Dr. E. Vuorio, University of Turku, Finland, while total RNA from human fibroblasts (N-09) was isolated using the RNeasy Midi kit (Qiagen). RT-PCR was performed using a SMART RACE cDNA amplification kit (BD Clontech), and PCR analysis as for the MTC panels, using the ␣(I) and ␣(III) primers described above and the ␣(II) primers ␣(II)panel-5Ј (GAGGAGGCCACCACAACCAAGTCA) and ␣(II)panel-3Ј (GACCTTG-TGGATCAACAGAAGTTGACTG).
Expression and Analysis of a Recombinant ␣(III) Polypeptide in Insect and Mammalian Cells-The full-length human ␣(III) cDNA was cut into two pieces from pUC18 with BamHI-EcoRI digestion and cloned into a similarly digested baculovirus vector pVL1393 (BD Pharmingen). The ␣(III) cDNA lacking sequences coding for the signal peptide was generated by PCR and cloned into pACGP67A (BD Pharmingen) in-frame with the GP67 signal sequence. The recombinant baculoviruses were generated and the Sf9 and High Five insect cells were coinfected with viruses encoding the ␣(III) and PDI polypeptides and analyzed as described previously (17,18).
Soluble and insoluble fractions of the insect and mammalian cell homogenates were analyzed by 8% SDS-PAGE under reducing or nonreducing conditions and 8% nondenaturing PAGE followed by Coomassie Blue staining or Western blotting with a PDI antibody 5B5 (Dako), a polyclonal ␣(I) antibody K17 (19) or a monoclonal antibody VTT1081 against the ␣(III) subunit generated at the Technical Research Centre of Finland by immunizing mice with a recombinant ␣(III) polypeptide (amino acids 177-501) expressed in Escherichia coli using the pET expression system (Novagen) and purified from the inclusion bodies (20). P4H activity was assayed by methods based on the formation of 4-hydroxy[ 14 C]proline in a [ 14 C]proline-labeled substrate consisting of nonhydroxylated procollagen polypeptide chains or the hydroxylationcoupled decarboxylation of 2-oxo[1-14 C]glutarate (21). K m and K i values were determined as described (17). Native immunoprecipitation was performed with Protein G Sepharose 4 Fast Flow according to the instructions provided by the manufacturer (Amersham Biosciences). The cell lysates were precleared, immunoprecipitated with 1-5 g of the antibodies VTT1081 against the ␣(III) subunit or anti-FLAG (Sigma) as a negative control, and the Sepharose and the bound antibodyprotein complexes were washed three times with a Triton X-100 buffer (17,18). The samples were then boiled in SDS-PAGE sample buffer and analyzed by SDS-PAGE followed by Western blotting. N-glycosidase F treatment was carried out according to the instructions provided by the manufacturer (Roche Applied Science). Gel filtration was performed in a calibrated Superdex 200 column (Amersham Biosciences). Recombinant human type I and II C-P4Hs were expressed in insect cells, and the type I enzyme was purified as described (6,18).

RESULTS AND DISCUSSION
Cloning of the Human, Rat, and Mouse ␣(III) Subunits-A sequence homology search identified several human ESTs representing a gene product with similarity to the conserved Cterminal regions of the C-P4H ␣(I) and ␣(II) subunits. EST AA116081 was used to design primers for the initial 5Ј-and 3Ј-RACE reactions and amplification of a 162-bp probe, which was then used to screen a human umbilical vein endothelial cell cDNA library, the 5Ј-end of the cDNA being obtained by 5Ј-RACE reactions. The cDNA clones cover 43 bp of the 5Јuntranslated sequence, a 1635-bp open reading frame, and 591 bp of the 3Ј-untranslated sequence, with a canonical polyadenylation signal ATTAAA followed 15-bp downstream by a poly(A) tail.
A sequence homology search with the human ␣(III) cDNA identified a 169-bp rat sequence that is 89% identical to the coding nucleotides at the 3Ј-end of the human sequence. The rat sequence information was used to design primers for 5Ј-RACE reactions in rat cDNA pools, and 5Ј-and 3Ј-RACE reactions in mouse cDNA pools. The cDNA fragments obtained cover the full-length coding sequences of the rat and mouse ␣(III) subunit cDNAs, including 25 and 13 bp of the 5Ј-untranslated region, respectively.
The human, rat, and mouse ␣(III) cDNA sequences have been deposited in the GenBank™ with accession numbers AY313448, AY313450, and AY313449, respectively.
Amino Acid Sequences of the Human, Rat, and Mouse ␣(III) Subunits-The human and rat ␣(III) cDNAs encode polypeptides of 544 amino acids, the mouse polypeptide being two residues shorter. Putative signal peptides (22) of 19, 19, and 22 residues are present in the N termini of the human, rat, and mouse ␣(III) polypeptides, respectively, the lengths of the processed human and rat ␣(III) subunits thus being 525 amino acids and that of the mouse ␣(III) 520 amino acids (Fig. 1). The processed human ␣(III) subunit is slightly longer than the processed human ␣(I) and ␣(II), which consist of 517 and 514 residues, respectively (4, 6). The human ␣(III) sequence shows 91 and 94% identity to the rat and mouse sequences, respectively, the latter two being 95% identical.
The overall sequence identity between the processed human ␣(III) and ␣(I) subunits is 35%, and that between ␣(III) and ␣(II) 37% ( Fig. 1), while the identity between ␣(I) and ␣(II) is higher, 65%. The identity is highest within the catalytically important C-terminal region (1)(2)(3)17), the 120 C-terminal residues of the human ␣(III) subunit being 56 -57% identical to those of human ␣(I) and ␣(II) (Fig. 1), while the identity between ␣(I) and ␣(II) in this region is 80%. All four critical residues at the catalytic site, the two histidines and one aspartate that bind the Fe 2ϩ atom and the lysine that binds the C-5 carboxyl group of the 2-oxoglutarate in position ϩ10 with respect to the second iron-binding histidine (1-3, 17), are conserved in all these ␣-subunits (Fig. 1). The residue that binds the C-5 carboxyl group of 2-oxoglutarate in the HIF asparaginyl hydroxylase (FIH) is also a lysine, but it is present in position ϩ15 with respect to the first iron-binding histidine (23,24). The other 2-oxoglutarate-dependent dioxygenases, including the HIF-P4Hs and lysyl hydroxylases, differ from the C-P4Hs and FIH in that their 2-oxoglutarate binding residue is an arginine (3, 9 -11, 25, 26). Furthermore, a fifth critical residue, a histidine that is probably involved in the binding of the C-1 carboxyl group of the 2-oxoglutarate to the Fe 2ϩ atom and the decarboxylation of this cosubstrate (17), is likewise conserved in all the C-P4H ␣-subunits (Fig. 1).
The peptide-substrate binding domain of the C-P4Hs is distinct from the catalytic domain and is located between Phe-144 and Ser-244 in the human ␣(I) subunit (27,28). Recent NMR studies have shown that this domain is composed of five ␣ helices that can also be accurately predicted (29) based on amino acid sequence (28). The sequence identity between the human ␣(I) peptide-binding domain and the corresponding regions of the human ␣(II) and ␣(III) subunits is 57 and 35%, respectively, while that between the ␣(II) and ␣(III) polypeptides is 34%. However, all five ␣ helices can be predicted (29) to be located in identical positions in all three polypeptides and to be identical in length, the only exception being the slightly shorter fifth helix in the ␣(III) polypeptide ( Fig. 1). It thus seems probable that the structures of the peptide-binding domains of all three ␣-subunits are highly similar despite their relatively low overall amino acid sequence identity. Detailed studies to verify this aspect would nevertheless require comparison of the crystal-based structures of complexes of the three domains with a peptide substrate, which is clearly beyond the scope of this study.
Expression of the ␣(III) Subunit mRNA in Various Human Tissues-Expression of the ␣(III) and ␣(I) subunit mRNAs was studied by Northern hybridization, the blots being exposed for 72 or 4 h, respectively. The ␣(III) mRNA was found to be about 2.7 kb in size, whereas the ␣(I) mRNA is about 3.0 kb ( Fig. 2A). The highest ␣(III) mRNA expression levels were found in the placenta, adult liver, and fetal skin, low levels being detected in the fetal liver, lung and muscle ( Fig. 2A). However, as exposure times shorter than 72 h produced no visible bands with the ␣(III) probe, the relative expression levels of this mRNA must be much lower than those of the ␣(I) mRNA in all the tissues studied.
Expression of the ␣(III) mRNA was studied further by PCR analysis of human multitissue cDNA panels under conditions that produced similar amounts of DNA from equal amounts of ␣(III) and ␣(I) plasmids. The highest ␣(III) mRNA expression levels were seen in the placenta and fetal kidney, liver and lung FIG. 1. Alignment of the amino acid residues of the processed human, rat, and mouse C-P4H ␣(III), and human ␣(I) and ␣(II) subunits. Gaps (Ϫ) are introduced for maximal alignment and dots (q) are shown above every tenth residue. The two histidines and one aspartate that bind the Fe 2ϩ atom, the lysine that binds the C-5 carboxyl group of 2-oxoglutarate, and a further histidine, 10 residues after the lysine, that is probably involved in the binding of the C-1 carboxyl group of 2-oxoglutarate to the Fe 2ϩ atom and the decarboxylation of this cosubstrate (17) are indicated by asterisks. The positions of cysteine residues () and potential N-glycosylation sites (Ϫ) in ␣(III) are indicated above the human ␣(III) sequence. Residues that are identical between an ␣(III) subunit and ␣(I) or ␣(II) are shown by white letters on a black background. The five ␣ helices of the ␣(I) peptide-substrate binding domain (28) are indicated below the ␣(I) amino acids 144 -244 with black cylinders and the predicted (29) ␣ helices of the corresponding ␣(III) and ␣(II) regions with white and gray cylinders, respectively. (Fig. 2B). The expression levels of the ␣(III) mRNA were again much lower than those of the ␣(I) mRNA in all the tissues studied, since no ␣(III) mRNA was detected in these samples after 34 cycles, whereas the ␣(I) mRNA was already clearly seen in many tissues after 30 cycles (Fig. 2B). Furthermore, PCR analysis of human fibroblast and fetal epiphyseal cartilage cDNA pools indicated that the ␣(III) mRNA was expressed at much lower levels than the ␣(I) and ␣(II) mRNAs in both samples (Fig. 2C). Our data thus demonstrate that the ␣(III) mRNA is expressed at low levels in many fetal and adult tissues, and presumably also in many cell types. Further studies will be needed to demonstrate whether there are cells that express this mRNA at levels higher than those of the other C-P4H ␣-subunit mRNAs.
Organization of the Human ␣(III) Subunit Gene-A search of High Throughput Genomic Sequences from the GenBank TM data base indicated that the human ␣(III) subunit gene is present on chromosome 11q12 (accession number AC006595), while the human ␣(I) and ␣(II) genes are located on chromosomes 10q21.3-23.1 (1) and 5q31 (30), respectively. The exonintron organization of the human ␣(III) gene is very similar to those of the ␣(I) and ␣(II) genes (30,31). However, a number of 5Ј-RACE reactions carried out with various human, rat, and mouse cDNAs (details not shown) indicated that the ␣(III) gene lacks an exon corresponding to the first exon of the ␣(I) and ␣(II) genes (30,31) and that only one ␣(III) exon, number 1, corresponds to exons 2 and 3 in the ␣(I) and ␣(II) genes (Fig.  3A). The subsequent exon-intron boundaries are mainly con-served, except that the ␣(III) exon 3, corresponding to exon 5 in ␣(I) and ␣(II), is 86 bp longer and that the region corresponding to exon 6 in ␣(I) and ␣(II) is split into two exons, 4 and 5, in the ␣(III) gene, while the ␣(III) sequence corresponding to exons 11 and 12 in ␣(I) and exons 10 and 11 in ␣(II) is fused into one exon, number 9 (Fig. 3A). Small differences were also found in the lengths of some of the exons (Fig. 3A) Two forms of ␣(I) and ␣(II) mRNA resulting from mutually exclusive alternative splicing of the homologous exons 9 and 10 in the human ␣(I) gene and exons 12a and 12b in the human ␣(II) gene have previously been identified (30,31). Careful sequence homology analysis of all introns in the human ␣(III) GenBank TM clone AC006595 indicated that none of them contained any sequences with similarity to those in the preceding or following exon, thus excluding the presence of consecutive homologous exons that might be subject to mutually exclusive alternative splicing in the ␣(III) gene (details not shown).
The possible presence of alternatively spliced exons was studied further by PCR analysis of human multitissue cDNA panels and sequencing of the products. The primer pairs used in Fig. 3, panel B, were from the boundary between exons 1 and 2 and from exon 9, while those used in Fig. 3, panel C, were from exons 7 and 13. PCR products of the expected lengths of 1041 (Fig. 3B) and 667 bp (Fig. 3C) were obtained from all the tissues studied. An additional product of about 650 bp was obtained with the primers used in panel B, but sequencing indicated that it was a nonspecific product not related to any ␣(III) transcript. Our data thus exclude the presence of mRNA species lacking one or more exons or containing an additional exon, as has been reported for one of the isoenzymes (isoenzyme 2) of a related enzyme, lysyl hydroxylase (32). Furthermore, sequencing of the 1041 and 667-bp PCR products indicated that they all had identical sequences (details not shown). Our data thus appear to exclude the possibility that the ␣(III) mRNA may be subject to alternative splicing, at least in any of the tissues studied.
Expression of an Active Recombinant Type III C-P4H-Recombinant baculoviruses encoding the human ␣(III) subunit, either with its native signal peptide or with the baculoviral signal peptide GP67, were generated and used to infect insect cells together with a virus encoding human PDI (18). No P4H activity was generated, however, and no association between the ␣(III) and PDI polypeptides was detected, whereas high levels of P4H activity were obtained in control coinfection experiments with viruses encoding ␣(I) and PDI and a band corresponding to a type I C-P4H tetramer was readily seen in nondenaturing PAGE (data not shown). N-terminal sequencing of the recombinant ␣(III) polypeptides indicated that the native signal peptide was retained in the majority of the polypeptides, while the GP67 signal peptide was cleaved incorrectly (data not shown). Possible explanations for the lack of any type III C-P4H assembly in insect cells may therefore lie in inefficient transport of the ␣(III) polypeptide into the ER and its incorrect processing.
To study the assembly of type III C-P4H in mammalian cells, stably transfected human embryonic kidney HEK-293 cell lines coexpressing either the recombinant ␣(III) and PDI or ␣(I) and PDI polypeptides were generated. The cells were harvested at confluency, homogenized in a Triton X-100 buffer, centrifuged and the soluble fractions analyzed by SDS-PAGE followed by Western blotting with antibodies to the PDI, ␣(III), and ␣(I) polypeptides (Fig. 4). PDI was found in both stably transfected cell lines (Fig. 4A), whereas the ␣(III) and ␣(I) polypeptides were seen only in the cell lines expressing the recombinant ␣(III) or ␣(I) polypeptide, respectively (Fig. 4, B and C). No immunoreactive bands were detected in cells transfected with an expression construct encoding ␤-galactosidase (Fig. 4), indicating that the endogenous levels of the PDI, ␣(III), and ␣(I) polypeptides were below the detection limit of the Western blots.
P4H activity in the HEK cell lines was analyzed by a method based on the formation of 4-hydroxy[ 14 C]proline in a [ 14 C]proline-labeled substrate consisting of nonhydroxylated procollagen polypeptide chains (21). The P4H activity levels in the soluble fraction from cells coexpressing the recombinant ␣(III) and PDI or ␣(I) and PDI polypeptides were 3.5-and 6-fold, respectively, relative to that in the cells expressing recombinant ␤-galactosidase (Table I).
The Type III C-P4H Is Probably an ␣ 2 ␤ 2 Tetramer, in Which PDI Serves as the ␤-Subunit-To analyze whether the ␣(III) and PDI polypeptides are present in the same molecule, the Triton X-100 buffer-soluble fraction from cells coexpressing the two polypeptides was immunoprecipitated with an antibody to the ␣(III) polypeptide and the precipitate was analyzed by SDS-PAGE under reducing conditions followed by Western blotting. The antibody was found to precipitate both the ␣(III) and PDI polypeptides from the soluble fraction from cells coexpressing these two polypeptides (Fig. 5, lanes 1 and 4), but not the ␣(I) or PDI polypeptide from cells coexpressing these two (Fig. 5, lanes 2 and 5). An anti-FLAG antibody used as a negative control precipitated neither of the two polypeptides from cells coexpressing ␣(III) and PDI (Fig. 5, lanes 3 and 6). The data thus indicate that the ␣(III) and PDI polypeptides are very likely to be present in the same molecule.
The subunit composition of the type III enzyme was studied further by gel filtration of soluble fractions from cells coexpressing recombinant polypeptides in a calibrated Superdex 200 column. The elution position of the P4H activity was found to be identical for cell lines coexpressing either the ␣(III) and PDI or the ␣(I) and PDI polypeptides (Fig. 6A). This elution position also corresponded to that of a purified recombinant type I C-P4H tetramer, no activity being detected in elution positions corresponding to an enzyme dimer or the ␣-subunit monomer (Fig. 6A).
Gel filtration fractions showing the highest P4H activity levels were analyzed further by SDS-PAGE and nondenaturing PAGE followed by Western blotting. Immunoreactive bands corresponding to the ␣(III) and PDI polypeptides were seen in SDS-PAGE of the fractions from cells coexpressing ␣(III) and PDI, whereas no bands were detected with the ␣(I) antibody (Fig. 6B). Likewise, immunoreactive bands corresponding to the ␣(I) and PDI polypeptides were seen in SDS-PAGE of the fractions from cells coexpressing ␣(I) and PDI (Fig. 6C), while no bands were found with the ␣(III) antibody (data not shown). Nondenaturing PAGE of the fractions from cells coexpressing ␣(III) and PDI showed a band that had a mobility similar to that of the purified recombinant type I C-P4H tetramer and could be stained with the ␣(III) and PDI antibodies (Fig. 6, D and F), but not with the ␣(I) antibody (Fig. 6D). Similarly, nondenaturing PAGE of the fractions from cells coexpressing ␣(I) and PDI showed a band reactive with the ␣(I) and PDI antibodies (Fig. 6E), but not with the ␣(III) antibody (data not shown). It thus seems very likely that the type III C-P4H has a similar subunit composition to the type I and type II C-P4Hs (1-3) and is an [␣(III)] 2 ␤ 2 tetramer in which PDI serves as the ␤-subunit.
Previous coexpression studies in insect cells argue against the presence of a mixed tetramer containing the ␣(I) and ␣(II) subunits in a single molecule (6). C-P4H ␣-subunits from nonvertebrate species have also been cloned and characterized, including one ␣-subunit from Drosophila melanogaster (33) and three, known as PHY-1 to PHY-3, from the nematode Caenorhabditis elegans (34 -36). The D. melanogaster ␣-subunit also forms an ␣ 2 ␤ 2 tetramer with PDI, whereas the C. elegans PHY-1 and PHY-2 form a mixed PHY-1/PHY-2/(PDI) 2 tetramer, neither of them forming any tetramer in the absence of the other (35). Both do form small amounts of PHY/PDI dimers, however, but PHY-2 with particularly low efficiency (35). PHY-3 differs from the others in that it does not interact with PDI, its molecular composition being currently unknown (36). Our data do not exclude the possibility that the ␣(III) subunit may also be able to form a mixed tetramer with some other ␣-subunit type, but they do exclude the possibility that it could resemble the C. elegans PHY-1 and PHY-2 polypeptides (35) in being unable to form any active tetramer in the absence of a second ␣-subunit type and in being able to form an active ␣␤ dimer with PDI. It may be noted that some C-P4H ␣-subunitlike polypeptides such as those characterized from Paramecium bursaria Chlorella virus 1 (37) and certain plants (1-3, 38) differ from those discussed above in being active P4H monomers.
The Additional Cysteine Residue Present in the Human ␣(III) Subunit Is Not Involved in Intrachain or Interchain Disulfide Bonding-The human ␣(III) subunit contains six cysteine residues that are also found in the rat and mouse polypeptides, while the human ␣(I) and ␣(II) subunits contain five conserved cysteines in positions identical to those of cysteines 1 and 3-6 in ␣(III), the ␣(II) subunit having an additional cysteine between the ␣(III) cysteines 5 and 6 ( Fig. 1). Site-directed mutagenesis studies of human ␣(I) have indicated that intrachain disulfide bonds essential for tetramer assembly are formed between cysteines corresponding to the third and fourth and the fifth and sixth cysteines in ␣(III), while the first cysteine is not involved in any disulfide bonding (39,40). The additional cysteine (residue 226) present in human ␣(III) is located in helix 4 of the peptide-substrate binding domain, while the first cysteine (residue 165) is located in helix 1 (Fig. 1). The ␣(I) peptide-binding domain has recently been crystallized (41) and its structure solved. 2 Inspection of the crystal structure 2 indicated that the ␣(I) residues corresponding to ␣(III) Cys-165 and Cys-226 are far from each other. As the structures of the ␣(I) and ␣(III) peptide-substrate binding domains are likely to be highly similar (see above), it seems evident that the ␣(III) Cys-165 and Cys-226 cannot form any intrachain disulfide bond.
The possibility was not excluded, however, that Cys-226 might form an interchain disulfide bond, which is not found in the type I or type II C-P4H (1). This possibility was studied by analyzing the soluble fraction from cells coexpressing the ␣(III) and PDI polypeptides by SDS-PAGE under nonreducing condi-  5. Analysis of immunoprecipitates obtained with an antibody against the ␣(III) polypeptide. Stably transfected HEK cells coexpressing recombinant human ␣(III) and PDI or ␣(I) and PDI polypeptides were harvested and homogenized as in Fig. 4. Soluble fractions of the cell lysates were immunoprecipitated with an ␣(III) antibody and the samples were analyzed by SDS-PAGE under reducing conditions followed by Western blotting with antibodies against ␣(III) (lanes 1-3) and PDI (lanes 4 -6). Immunoprecipitation of the soluble fraction from cells coexpressing the ␣(III) and PDI polypeptides with an anti-FLAG antibody was used as a negative control (lanes 3 and 6). The polypeptides expressed are indicated on top of the lanes, and the antibodies used in the immunoprecipitation (IP), and immunoblotting are indicated below the lanes. The anti-␣(III) antibody was found to precipitate trace amounts of protein also from cells coexpressing the ␣(I) and PDI polypeptides, possibly because of a presence of trace amounts of endogenous nonrecombinant type III C-P4H in the HEK cells.
tions followed by Western blotting. The mobility of ␣(III) was found to be identical under nonreducing and reducing conditions and to correspond to that of the polypeptide monomer (Fig. 7A, lanes 1 and 3). It is thus evident that the additional cysteine present in the human ␣(III) polypeptide is not involved in either intrachain or interchain disulfide bonding. This cysteine is not likely to be involved in the catalytic mechanism, either, as the mechanism of 2-oxoglutarate dioxygenases is currently well understood and does not involve any cysteine residue (1)(2)(3), and as the type I C-P4H, which has very similar catalytic properties (see below), contains only one cysteine not involved in disulfide bonding (corresponding to Cys-165 in ␣(III)) and this can be mutated to serine with no loss of catalytic activity (40). The rat and mouse ␣(III) polypeptides contain one additional cysteine at position Ϫ8 with respect to Cys-226 in the human polypeptide (Fig. 1), but its role was not studied further.
The Two N-Glycosylation Sites Have No Role in the Activity of the Type III C-P4H-The human, rat, and mouse ␣(III) polypeptides contain two N-glycosylation sites that are conserved between species but not in ␣(I) and ␣(II) (Fig. 1), which also have two N-glycosylation sites but in different positions (4 -6). Site-directed mutagenesis studies of the human ␣(I) subunit have shown that glycosylation has no role in the assembly of the type I C-P4H tetramer or its catalytic activity (40). To study whether N-glycosylation has any role in the catalytic activity of the human type III C-P4H, the soluble fraction from cells coexpressing the ␣(III) and PDI polypeptides was treated with N-glycosidase F and analyzed by SDS-PAGE under reducing conditions followed by Western blotting. A distinct increase was found in the mobility of the ␣(III) polypeptide after N-glycosidase F digestion (Fig. 7B, lane 2) indicating that it had been synthesized in the HEK-293 cells in an Nglycosylated form (Fig. 7B, lane 1). The soluble fraction from cells treated with N-glycosidase F was subsequently assayed for P4H activity using a [ 14 C]proline-labeled substrate consisting of nonhydroxylated procollagen polypeptide chains (21). No difference in enzyme activity was found between the N-glycosidase F-treated (78,000 dpm/mg) and nontreated (80,000 dpm/ mg) soluble fractions, the activity levels in this experiment being slightly higher than those shown in Table I. Our data thus indicate that N-glycosylation of the ␣(III) polypeptide, like that of the ␣(I) polypeptide, plays no role in the catalytic activity of the corresponding C-P4H.
Catalytic Properties of the Recombinant Type III C-P4H-Catalytic properties were studied using the Triton X-100 solu-FIG. 6. Analysis of the assembly of recombinant human type III and I C-P4Hs in HEK cells. A, stably transfected HEK cells coexpressing recombinant human ␣(III) and PDI or ␣(I) and PDI polypeptides were harvested and homogenized as in Fig. 4. Soluble fractions were applied to a calibrated Superdex 200 column, and the eluted fractions were assayed for P4H activity (dpm). The void volume (V) and the elution positions of recombinant type I C-P4H (known to correspond to a molecular weight of 350,000 rather than the true value of 240,000, Ref. 1), aldolase (158 kDa), PDI dimer (120 kDa), PDI (60 kDa), ovalbumin (43 kDa), and chymotrypsinogen A (25 kDa) are shown. The gel filtration fractions from cells coexpressing recombinant ␣(III) and PDI and ␣(I) and PDI were analyzed by SDS-PAGE under reducing conditions (B and C) and by nondenaturing PAGE (D and E) followed by Western blotting with the antibodies indicated on top of the lanes, the expressed polypeptides being indicated below the panels. Panel F shows Coomassie Blue-stained nondenaturing PAGE of purified recombinant type I C-P4H (18). ble extracts from the HEK cells as sources of the recombinant type III and type I C-P4Hs, while those of the recombinant human type II C-P4H were studied with a corresponding extract from insect cells (6). The activity assay used is based on the hydroxylation-coupled decarboxylation of 2-oxo[1-14 C]glutarate (21). The K m values of the type III enzyme for the cosubstrates Fe 2ϩ , 2-oxoglutarate and ascorbate were essentially identical to those of the type I and type II C-P4Hs (Table II). The minor differences in the K m for Fe 2ϩ may be due to the presence of trace amounts of iron in the nonpurified enzyme preparations. The IC 50 for pyridine-2,4-dicarboxylate, an effective competitive inhibitor of the type I and type II C-P4Hs with respect to 2-oxoglutarate, was likewise very similar to those of the type I and type II C-P4Hs, the IC 50 for pyridine-2,5-dicarboxylate being 4 -7-fold, however (Table II). These data agree with the finding that all five critical residues identified at the cosubstrate binding sites (17) are fully conserved in all three types of ␣ subunit (see above), the difference in the IC 50 for pyridine-2,5-dicarboxylate probably being due to effects of some of the nonconserved residues.
The K m of the type III C-P4H for the peptide substrate (Pro-Pro-Gly) 10 was slightly higher than that of the type I C-P4H, but distinctly lower than that of the type II C-P4H (Table II). Poly(L-proline) is an effective competitive inhibitor of the type I C-P4H, whereas the type II C-P4H is inhibited by poly(L-proline) only at very high concentrations (1)(2)(3)(4)(5)(6). The IC 50 of the type III C-P4H for poly(L-proline), M r 5000, was 30 M (Table II), being thus five times that of the type I C-P4H but only one-tenth of that of the type II C-P4H. Site-directed mutagenesis studies have shown that most, although not all, of the difference in the peptide binding between the type I and type II C-P4Hs can be explained by the presence of a glutamate and glutamine in the peptide-binding domain of the ␣(II) subunit in positions corresponding to Ile-182 and Tyr-233 in ␣(I) (27). The ␣(III) polypeptide contains Trp-199 and Arg-253 in the corresponding positions (Fig. 1). It seems possible that the trypto-phan residue may be more favorable for peptide binding than the glutamate present in the ␣(II) subunit, and this might at least in part explain why the peptide binding properties of the type III C-P4H are intermediate between those of the type I and type II enzymes with respect to the K m for (Pro-Pro-Gly) 10 , and especially with respect to inhibition by poly(L-proline). CONCLUSIONS C-P4H was long believed to be of one type only, until an isoform of the ␣-subunit, ␣(II), was cloned and characterized from mouse and human sources (5,6). The present study reports on the identification of a further vertebrate ␣-subunit isoform, the ␣(III) subunit, the human polypeptide showing 35-37% sequence identity to the human C-P4H ␣(I) and ␣(II) subunits (4, 6) but no distinct sequence similarity to the HIF-P4Hs (9,10). This degree of identity is considerably lower than the 65% observed between the human ␣(I) and ␣(II) subunits (4,6). The ␣(III) polypeptide also resembles the C-P4H ␣-subunits rather than the HIF-P4Hs in that its residue that binds the C-5 carboxyl group of the 2-oxoglutarate is a lysine (17) rather than an arginine (9 -11) and that the IC 50 of the type III enzyme for pyridine 2,5-dicarboxylate is about 7 M, whereas it is much more than 300 M for all three human HIF-P4Hs (42). The ␣(III) polypeptide, like the ␣(I) and ␣(II) polypeptides, appears to form with PDI an ␣ 2 ␤ 2 tetramer that hydroxylates collagen chains and collagen-like peptides, its elution position in gel filtration and mobility in nondenaturing PAGE excluding the presence of any active dimer or monomer. The ␣(III) mRNA was found to be expressed in many human tissues, but at much lower levels than the ␣(I) and ␣(II) mRNAs, no general trend being found for any distinct difference between embryonic and adult tissues. Our data do not, however, exclude the possibility that some cell types may express the ␣(III) mRNA at levels higher than those of the other C-P4H ␣-subunit mRNAs.