Cloning, Characterization, and Expression of cDNAs Encoding Human (cid:68) 1 -Pyrroline-5-carboxylate Dehydrogenase*

(cid:68) 1 -Pyrroline-5-carboxylate dehydrogenase (P5CDh; EC 1.5.1.12), a mitochondrial matrix NAD (cid:49) -dependent dehydrogenase, catalyzes the second step of the proline degradation pathway. Deficiency of this enzyme is asso-ciated with type II hyperprolinemia (HPII), an autoso- mal recessive disorder characterized by accumulation of (cid:68) 1 -pyrroline-5-carboxylate (P5C) and proline. As an initial step in understanding the biochemistry of human P5CDh and molecular basis of HPII, we utilized published peptide sequence data and degenerate primer polymerase chain reaction to clone two full-length human P5CDh cDNAs, differing in length by 1 kilobase pair (kb). Both cDNAs have the identical 1689-base pair open reading frame encoding a protein of 563 residues with a predicted molecular mass of 62 kDa. The long cDNA contains an additional 1-kb insert in the 3 (cid:42) -untranslated region that appears to be an alternatively spliced in- tron. The conceptual translation of human P5CDh has 89% sequence identity with the published human P5CDh peptide sequences and 42 and 26% identity with Saccharomyces cerevisiae and Escherichia coli P5CDhs, respec- tively, as well as homology to several other aldehyde dehydrogenases. Both P5CDh cDNA clones detect a sin- gle 3.2-kb transcript on Northern blots of multiple human E. cells Inc.). Ten isolates the Sequenase 2.0 DNA sequencing pu- tative P5CDh PCR (cid:97) - 32 P]dCTP (3000 mmol) to probe 5 (cid:51) 5 from a human retinal cDNA library. Hybridization in 0.5 M 7% SDS and 1 m M EDTA 56 two times each 56 0.1 M 2 HPO 5% SDS and 1 m M EDTA and one time min 56 0.04 M HPO 7.2) 1% SDS and 1 m M EDTA. Two positive plaques were purified two additional cycles of screening. cDNAs were subcloned into the pBluescript ll KS( (cid:49) ) and transfected into DH5- (cid:97) competent cells. One (P5CDhR2, 1.8 kb) of the inserts sequenced.Toextend the 5 (cid:57) -end of the P5CDh cDNA, a 5 (cid:57) rapid amplification of the cDNA end (RACE) with 2 (cid:109) g of human HepG2 (cid:49) RNA, an antisense primer corresponding to the most 5 (cid:57) -region of the human retinal P5CDh cDNA (5 (cid:57) -CAGGGCAGCCTCAATGGC-3

⌬ 1 -Pyrroline-5-carboxylate dehydrogenase (P5CDh; EC 1.5.1.12), a mitochondrial matrix NAD ؉ -dependent dehydrogenase, catalyzes the second step of the proline degradation pathway. Deficiency of this enzyme is associated with type II hyperprolinemia (HPII), an autosomal recessive disorder characterized by accumulation of ⌬ 1 -pyrroline-5-carboxylate (P5C) and proline. As an initial step in understanding the biochemistry of human P5CDh and molecular basis of HPII, we utilized published peptide sequence data and degenerate primer polymerase chain reaction to clone two full-length human P5CDh cDNAs, differing in length by 1 kilobase pair (kb). Both cDNAs have the identical 1689-base pair open reading frame encoding a protein of 563 residues with a predicted molecular mass of 62 kDa. The long cDNA contains an additional 1-kb insert in the 3-untranslated region that appears to be an alternatively spliced intron. The conceptual translation of human P5CDh has 89% sequence identity with the published human P5CDh peptide sequences and 42 and 26% identity with Saccharomyces cerevisiae and Escherichia coli P5CDhs, respectively, as well as homology to several other aldehyde dehydrogenases. Both P5CDh cDNA clones detect a single 3.2-kb transcript on Northern blots of multiple human tissues, indicating the long cDNA containing the 3-untranslated intron represents the predominant transcript. The P5CDh structural gene appears to be single copy with a size of about 20 kb localized to chromosome 1. To confirm the identity of the putative P5CDh cDNAs, we expressed them in a P5CDh-deficient strain of S. cerevisiae. Both conferred measurable P5CDh activity and the ability to grow on proline as a sole nitrogen source. P5C 1 dehydrogenase (EC 1.5.1.12) is a mitochondrial matrix NAD ϩ -dependent enzyme catalyzing the irreversible conversion of P5C, derived either from proline or ornithine, to glutamate. This reaction is a necessary step in the pathway interconnecting the urea and tricarboxylic acid cycles (1). Human P5CDh is classified as a member of the aldehyde dehydrogenase (ALDh) superfamily (2) on the basis of substrate specificity and kinetic properties (3)(4)(5). The preferred substrate of P5CDh is glutamic ␥-semialdehyde, which is in spontaneous nonenzymatic equilibrium with P5C. Other substrates include succinic, glutaric, and adipic semialdehydes (3). Human P5CDh is a homodimer with a molecular mass of 142-175 kDa (3).
The P5CDh genes of Escherichia coli (6), Salmonella typhimurium (7), and Saccharomyces cerevisiae (8,9) have been cloned and sequenced. The E. coli putA gene is bifunctional encoding both P5CDh and proline oxidase. The S. cerevisiae PUT2 gene encodes P5CDh only. Mutant strains of S. cerevisiae lacking P5CDh activity (8) are unable to use proline as a sole nitrogen source and provide a system for expression and analysis of any putative human P5CDh. No molecular information regarding mammalian, or specifically human, P5CDh is available. In humans, deficiency of P5CDh causes HPII, an autosomal recessive inborn error of metabolism (10, 11) characterized by a 10 -15 times accumulation of plasma proline (normal range 100 -350 M) and a 10 -40 times accumulation of plasma P5C (normal range 0.2-2 M) (1,10,11). Although some adults with HPII appear to be normal, this disorder may be causally related to neurologic manifestations, including seizures and mental retardation (1,12).
Utilizing published human P5CDh peptide sequences and degenerate primer PCR, we cloned two full-length P5CDh cDNAs of 2139 and 3150 bp. Both clones encode an identical 563-amino acid protein; the longer contains an additional 1-kb insert in the 3Ј-untranslated region. Functional complementation of a S. cerevisiae put2 mutant confirms the identity of these cDNAs as human P5CDh cDNAs.

EXPERIMENTAL PROCEDURES
cDNA Libraries and Plasmids-Human retina, kidney, and HepG2 cell libraries were gifts from Drs. J. Nathans, G. Bell, and A. Brake, respectively. A human smooth muscle 5Ј-STRETCH cDNA library was purchased from Clontech. Unless otherwise indicated pBluescript II KS(ϩ) or SK(ϩ) were used for all cloning manipulations.
Degenerate Primer PCR Cloning Strategy-Two degenerate oligonucleotide primers, one 5Ј-sense and one 3Ј-antisense, were designed from the published partial amino acid sequence of human P5CDh (4) (see Fig.  1). The 5Ј-sense primer (primer 1) was 5Ј-CGGAATTCGCNAT(A/C/ T)GA(A/G)GCNGCN(C/T)TNGC, the 3Ј-antisense primers was 5Ј-CGG-GATCCATNGG(T/C)TG(T/C)TGNCC(T/C)TC (primer 2). Restriction sites (primer 1, EcoRI; primer 2, BamHI) were included at the 5Ј-end of each primer. Thirty cycles of PCR were performed using primers 1 and 2 and human HepG2 cDNA library DNA (13) as template. Annealing * This work is supported in part by National Eye Institute Grant EY02948. Portions of this work have been presented in abstract form. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM 1 The abbreviations used are: P5C, ⌬ 1 -pyrroline-5-carboxylate; P5CDh, ⌬ 1 -pyrroline-5-carboxylate dehydrogenase; ALDh, aldehyde dehydrogenase; HPII, type II hyperprolinemia; RACE, rapid amplification of the cDNA end; MGP-ura Ϫ , minimal glucose medium containing pro-line without uracil; MGA-ura Ϫ , minimal glucose medium containing ammonium sulfate without uracil; UTR, untranslated region; PCR, polymerase chain reaction; bp, base pair(s); kb, kilobase pair(s). was for 1 min at 37°C, extension for 2 min at 72°C, and denaturing for 1 min at 94°C. A PCR product of appropriate size was gel-purified using the Gene Clean kit (Bio 101), ligated into the pBluescript ll SK(ϩ) using the primer EcoRI and BamHI restriction sites, and transfected into E. coli DH5-␣ cells (Life Technologies, Inc.). Ten isolates were sequenced with the Sequenase version 2.0 DNA sequencing kit (U. S. Biochemical Corp.) according to the manufacturer's protocols. The putative P5CDh PCR product was labeled with [␣-32 P]dCTP (3000 Ci/ mmol) as described elsewhere (14) and used to probe Ͼ5 ϫ 10 5 plaques from a human retinal cDNA library. Hybridization was done in 0.5 M Na 2 HPO 4 (pH 7.2) with 7% SDS and 1 mM EDTA at 56°C (15). Filters were washed two times for 10 min each at 56°C in 0.1 M Na 2 HPO 4 (pH 7.2) with 5% SDS and 1 mM EDTA and one time for 10 min at 56°C in 0.04 M Na 2 HPO 4 (pH 7.2) with 1% SDS and 1 mM EDTA. Two positive plaques were purified through two additional cycles of screening. These cDNAs were subcloned into the pBluescript ll KS(ϩ) and transfected into DH5-␣ competent cells. One (P5CDhR2, 1.8 kb) of the inserts was sequenced.
To extend the 5Ј-end of the P5CDh cDNA, we performed a 5Ј rapid amplification of the cDNA end (RACE) with 2 g of human HepG2 poly(A) ϩ RNA, an antisense primer corresponding to the most 5Ј-region of the human retinal P5CDh cDNA (5Ј-CAGGGCAGCCTCAATGGC-3Ј, primer 3) and a Clontech 5Ј-AmpliFinder RACE kit. Hybridization, clone isolation and sequencing were performed as described for the human retinal P5CDh cDNA clone.
PCR Amplification of Genomic DNA and cDNA-The sequences of the primers used to amplified the 3Ј-UTR of genomic DNA and cDNAs are listed below. Primer 4 (5Ј-AGCTACGCGTACATGCAGTG) primer 5 (5Ј-GGAATGGCCACACCAGCTC) correspond to sequences flanking the insert. Primer 6 (5Ј-TGTGGGGCACAGGGGGC) and primer 7 (5Ј-GTGTCACCATGGCACAGT) are at the 5Ј-and 3Ј-ends of the insert, respectively. We performed PCR amplification on genomic DNA (100 ng) and both the long and short forms of human P5CDh cDNAs (1 pg) as follows. An initial denaturation step of 4 min at 96°C, followed by 30 cycles of 1 min at 96°C, 1 min at 61°C, and 2 min at 72°C. The amplified fragments were then electrophoresed in 1% agarose. For sequencing, the fragments were excised, gel purified, and sequenced directly with their PCR primers.
DNA and Deduced Amino Acid Sequence Analysis-We sequenced all cDNA clones in both directions. DNA sequences were aligned and compared using the MacVector (Kodak) and the BLAST programs (16). We compared the deduced amino acid sequence of the putative fulllength human P5CDh cDNA clone to that of S. cerevisiae and E. coli P5CDhs and those of five members of the human aldehyde dehydrogenase superfamily with the MegAlign program (DNASTAR).
Nucleic Acid Analysis and Chromosome Localization-Total and poly(A) ϩ RNA from human tissue culture cells were isolated as described elsewhere (17). A multiple tissue Northern blot filter (Clontech) was hybridized under high stringency conditions as suggested in manufacturer's manual. Human ␤-actin was used to verify RNA quality and quantity.
Human genomic DNA was isolated from human lymphocytes according to the method of Kunkel (18). Genomic DNA (10 g) was digested with EcoRI, BamHI or HindIII separated on a 0.8% agarose gel and transferred to GeneScreen plus membrane (DuPont NEN). DNA was cross-linked to the membrane using a UV-autocross linker (Stratagene). Hybridization and autoradiography were performed with the same conditions used for library screening. Chromosome mapping was done by using a PstI-digested monochromosomal somatic cell hybrid mapping panel (Oncor) as described in manufacturer's manual. All hybridizations were with human P5CDhS as the probe.
Construction of Human P5CDh Yeast Expression Vector-We used the 2-m based pSM703 E. coli/yeast shuttle vector plasmid containing URA3 and amp R selectable markers and a multiple cloning site downstream of an S. cerevisiae phosphoglycerate kinase promoter and upstream of LacZ sequences (kindly provided by S. Michaelis). We cloned P5CDhS and P5CDhL into the EcoRI site in the sense orientation relative to the phosphoglycerate kinase promoter in pSM703 and designated the two derivative plasmids, pHsP5CDhS1 and pHsP5CDhL1, respectively (Fig. 2).
To replace the phosphoglycerate kinase promoter and 5Ј-UTR of the human P5CDh cDNAs with S. cerevisiae PUT2 5Ј sequence, we used a PCR-based cloning strategy to construct pHsP5CDhS2 and pHsP5CDhL2. Briefly, we made use of a NarI site 7 bp downstream of the predicted start ATG to delete the 5Ј-UTR of the human P5CDh cDNAs. A 3Ј-antisense primer, primer 8 (5Ј-CGCGGGCGCCGGCAG-CAGCATAATTCCTGTGAATTTG), contains complementary sequences of human P5CDh cDNA at position ϩ1 to ϩ18 (including the NarI site at ϩ10) and yeast PUT2 5Ј-UTR position Ϫ1 to Ϫ16. A 5Ј-primer, primer 9 (5Ј-CCCAAGCTTGATCCATTAAACTGGAAACAC), introduces a 5Ј-HindIII cloning site and corresponds to yeast PUT2 promoter sequence bp Ϫ435 to Ϫ414 relative to the translational start site. After a 30-cycle amplification, we gel-purified the 450-bp product fragment and substituted it for the HindIII-NarI fragment of the plasmids pHsP5CDhS1, pHsP5CDhL1 (see Fig. 2). The nucleotide sequences at the ligation junctions in plasmids pHsP5CDhS2, and pHsP5CDhL2 were confirmed by sequencing. pKB13, a plasmid containing S. cerevisiae PUT2 gene was a gift from M. Brandriss (19). We subcloned the yeast PUT2 gene from pKB13 to pSM703 to form pScPUT2.
Protein and Enzyme Assays-Protein concentrations were determined using the bicinchoninic acid reagent (Pierce) with bovine serum albumin serving as a standard. P5CDh activity was assayed radioisotopically (11) in soluble extracts of saturated overnight cultures of MB1472 grown in medium MGA-ura ϩ , HV1 grown in MGA-ura Ϫ medium, and HV2, HV3, and HV4 grown in MGP-ura Ϫ medium, were harvested, flash frozen, and disrupted in liquid N 2 .

Isolation and Characterization of Human
P5CDh cDNA-We selected two human P5CDh peptide sequences corresponding to conserved regions of human and yeast proteins to design degenerate PCR primers (Fig. 1). Amplifying with primer 1 and 2 we obtained the expected ϳ 240-bp product. The deduced amino acid sequence of the amplified fragment (P5CDhP-1) had 53% identity to the corresponding region of yeast P5CDh and contained perfect (15/15) matches for the human peptides (II and III) predicted to be included (Fig. 1). We used P5CDhP-1 to isolate putative human P5CDh cDNAs from human retinal, muscle and kidney cDNA libraries. The 5Ј end of one cDNA clone (P5CDhM19) extended into the 5Ј untranslated region to bp Ϫ6 (where ϩ1 is the A of the initiation methionine codon). 5Ј-RACE with HepG2 cDNA identified a 300-bp fragment (P5CDhR1) that had a 286-bp overlap with the 5Ј-end of P5CDhM19 and extended an additional 24 bp 5Ј-ward to Ϫ30. Of the cDNAs that extended to the poly(A) tail, two had a 3Ј-UTR of 1415 bp and one had a 3Ј-UTR of 404 bp. We constructed two composite P5CDh cDNA clones (Fig. 2) The open reading frame of these clones encodes a 563-amino acid protein with a predicted molecular mass of 62 kDa. The overall amino acid sequence of the putative human P5CDh has a 42 and 26% identity to those of S. cerevisiae and E. coli, respectively (Fig. 1). The predicted active site residues (Glu 314 and Cys 348 ) are completely conserved among these three proteins as well as five other members of the ALDh family (Fig.  3A) (2, 6). A region corresponding to a possible NAD ϩ -binding motif is also highly conserved (Fig. 3B). There are three potential N-glycosylation sites (consensus ϭ N-X-(S/T)) located at Asn 193 , Asn 211 , and Asn 283 of the unprocessed P5CDh protein (Fig. 1) (22-24).
As expected for a mitochondrial matrix protein, the N terminus of the putative P5CDh has no acidic residues and several arginine, leucine, and serine residues (25,26). Hendrick et al. (25) suggested that a fair predictor of the cleavage site is the sequence R-X-(⌽)-X-X-(S)-where the R is located at Ϫ10 relative to the cleavage site and ⌽ ϭ a hydrophobic residue, usually F, V, L, or I. Additionally, an R is often present at Ϫ2 relative to the cleavage site. With these considerations, there are two possible mitochondrial leader cleavage sites in the conceptual translation of P5CDh (indicated by the vertical arrowheads in Fig. 1). The more N-terminal site has an R at Ϫ10, a W at Ϫ8, an A at Ϫ5, and an R at Ϫ2; the more distal site has an R at Ϫ10, a K at Ϫ8, an S at Ϫ5, and a K at Ϫ2. We favor the more N-terminal site which would predict a leader sequence of 24 residues containing 4 R, 6 L, and 1 S residues and would yield a processed P5CDh of 539 amino acids.
Tissue Distribution and Size of Human P5CDh Transcripts-Northern blot analysis of mRNA isolated from multiple human tissues showed one predominant transcript of about 3.2 kb (Fig.  4A). Liver had the highest expression followed by skeletal muscle, kidney, heart, brain, placenta, lung, and pancreas.
Interestingly, we did not detect P5CDh transcripts corresponding to the P5CDhS (predicted length 2.2 kb), indicating that the predominant transcript in the tissues examined corresponds to the P5CDhL cDNA.
To determine if the structure of the genomic sequence encoding P5CDh matches the 3Ј-end of the P5CDhL cDNA, we compared the products of PCR amplification of genomic DNA and cDNA using primers flanking and/or at the ends of the insert. We found that when amplifying with either genomic DNA or P5CDhL cDNA as template and primer pairs 4 ϩ 5, 4 ϩ 7, 6 ϩ 5, and 6 ϩ 7, we detected 1,242-, 881-, 776-, and 415-bp product fragments, respectively, on both lanes (Fig. 5, lanes G and L). We detected only a 233-bp product fragment when using P5CDhS cDNA as template and primer pair 4 ϩ 5 (Fig. 5, lane  S). These results indicate that the 3Ј-UTR sequence of P5CDhL is colinear with genomic DNA and that the 1011-bp sequence is spliced out in P5CDhS.
Southern Analysis and Chromosome Mapping of P5CDh Gene-To estimate the size and complexity of the human P5CDh structural gene, we hybridized Southern blots of human genomic DNA with radiolabeled P5CDhS. The probe detects genomic DNA fragments totaling ϳ 20 kb in size suggesting the P5CDh structural gene is single copy and relatively simple (Fig. 4B).
Using a monochromosomal somatic cell hybrid mapping panel and radiolabeled P5CDhS as probe, we mapped the human P5CDh structural gene to human chromosome 1 (Fig. 4C).
Functional Complementation of P5CDh-deficient Yeast-To confirm the predicted function of our P5CDh cDNAs, we constructed several yeast/human P5CDh hybrid minigenes (Fig. 2) for expression in S. cerevisiae strain MB1472, a proline utili-zation (PUT) mutant with a partially deleted P5CDh gene (put2). MB1472 is unable to utilize proline as a sole nitrogen source and does not revert to PUTϩ phenotype. Transformants were selected on MGP-ura Ϫ or MGA-ura Ϫ and further tested for the coordinate appearance of URAϩ and PUTϩ phenotypes. For each transformation, three plasmids were isolated and shuttled into E. coli 294 cells. The plasmids were isolated from the 294 cell transformants, characterized by restriction enzyme mapping and sequencing and used to retransform MB1472 selecting on MGP-ura Ϫ . In all cases, these plasmid-dependent, PUTϩ transformants were also found to be URAϩ indicating that the cDNA carried by the plasmid was responsible for rescue of the proline utilization phenotype. Initially, in these experiments we used the pHsP5CDhS1 and pHsP5CDhL1 plasmids which contain the entire human P5CDh cDNAs including its 30-bp 5Ј-UTR downstream from the phosphoglycerate kinase promoter. However, no growth was detected when we transferred transformants to the selective MGP-ura Ϫ medium (data not shown). We reasoned that this failure to complement might be the result of reduced translational efficiency due to the presence of the human 5Ј-UTR (28). Accordingly, we reengineered the human P5CDh cDNA clones, replacing the human

FIG. 3. Comparison of the amino acid sequence of motifs in human
P5CDh to that of P5CDh of S. cerevisiae and E. coli and five members of the ALDh family. A, the sequence around the predicted active site residues E314 and C348; B, the sequence around the possible NAD ϩ binding motif. Residues are numbered to the left. Residues identical in five or more of the sequences are shown in black. HsP5CDh, human P5CDh; HsALDh1, human ALDh type I; HsALDh2, human ALDh type II; HsALDh3, human ALDh type III; RnALDh4, rat ALDh type IV; HsSSADh, human succinate semialdehyde dehydrogenase; and RnSSADh, rat succinate semialdehyde dehydrogenase. The asterisks indicate completely conserved residues thought to be involved in the active site and corresponding to E314 and C348 of HsP5CDh.
P5CDh 5Ј-UTR with that of S. cerevisiae PUT2 under the control of the PUT2 promoter and leaving the entire human P5CDh reading frame unchanged (plasmids pHsP5CDhS2 and pHsP5CDhL2). These recombinant cDNAs did restore the PUTϩ growth phenotype in MB1472 transformants. The yeast P5CDh gene (PUT2) did complement more efficiently than the human: MB1472 cells transformed with the PUT2 gene (strain HV2), grew on MGP-ura Ϫ after 3-5 days; while MB1472 harboring pHsP5CDhS2 or pHsP5CDhL2 (strain HV3 or HV4, respectively), required 7-10 days of incubation on MGP-ura Ϫ for equivalent growth (Fig. 6).
To confirm that the complementing activity in strains HV2, HV3, and HV4 was due to the expression of a functional P5CDh, we assayed the activity of this enzyme in extracts of these cells. Substantial P5CDh activity was detectable in HV2, 4.4 nmol of product/h/mg (4.2-4.6) (mean (range)); HV3, 4.2 (4.0-4.4); and HV4, 2.9 (2.6 -3.5). No activity was detected in yeast transformed with vector alone (strain HV1). We concluded that the protein encoded by pHsP5CDhS2 or pHsP5CDhL2 is human P5CDh and that the human enzyme is able to function in the heterologous yeast system. DISCUSSION Using published peptide sequences and degenerate primer PCR, we isolated two types of cDNAs encoding human P5CDh. Both encode the identical 563 amino acid protein but differ in their 3Ј-UTR with one containing a 1011 bp insert. The protein predicted by conceptual translation of these cDNAs has excellent agreement with the P5CDh peptide sequences published by Hempel et al. (4); all the peptides are recognizable with an overall identity of 89% (93/105 residues) with 7 of the 12 mismatched residues located at the C termini of three peptides. As expected for a protein found in the mitochondrial matrix, the N-terminal region of human P5CDh has features characteristic of a mitochondrial targeting sequence. The most likely cleavage site yields a leader of 24 amino acids and a mature P5CDh of 539 amino acids with a molecular mass of 59 kDa which is in excellent agreement with published values for human and rat P5CDh (3,5,29). Expression studies in yeast mutants lacking P5CDh activity confirm that the encoded protein is human P5CDh.
Human P5CDh has significant sequence similarity to the orthologous protein of lower eukaryotes and prokaryotes (26 -42% identity) as well as to other aldehyde dehydrogenases confirming that it is a member of the aldehyde dehydrogenase superfamily. The functional significance of the highly conserved residues common to these proteins, in particular the putative active site residues and NAD ϩ -binding motifs, will require additional studies with site-directed mutagenesis and functional assays. The human P5CDh structural gene is located on chromosome 1, at least 20 kb in size and appears to be a single copy with a relatively simple structure as indicated by the low number of hybridizing fragments detected in genomic Southern blot analysis.
Several lines of evidence indicate that the 1-kb insert in P5CDh is an alternatively spliced intron. First, the insert sequence has consensus sequences for donor, branch point and acceptor splice sites at 5Ј-and 3Ј-termini, respectively. Second, we found that human P5CDh cDNA is colinear with genomic DNA over the area containing the insert (Fig. 5). We found cDNAs encoding both the long and short forms in kidney and retinal cDNA libraries and we were able to detect both the long and short forms by reverse transcription and PCR amplification of HepG2 cell RNA (data not shown). However, our Northern blot data show that the transcript containing the insert is the predominant P5CDh mRNA (Ͼ95%) in all tissues examined. These results suggest that either this 3Ј-UTR intron is not removed from most transcripts or that its presence enhances mRNA stability so that the long transcripts preferentially accumulate. The 5Ј-splice site (AG gtggcc . . .) is not ideal particularly at the Ϫ4 to Ϫ6 positions with a "consensus value" of 0.66 where the value for typical introns ranges from 0.7-1.0 with a mean of 0.83 (27,30,31). This suggests that this intron is a poor substrate for splicing. Nevertheless, additional studies of transcript splicing and stability will be required to discriminate between these two possibilities. Other human cDNAs with alternatively spliced introns in the 3Ј-UTR have been described (32).
Yeast systems have proven valuable for the study of human genes, particularly those encoding enzymes (33). For example, human ornithine aminotransferase (34), P5C reductase (17), galactose-1-phosphate uridyltransferase (35), and cystathionine ␤ synthase (36) have all been shown to complement the growth phenotype of the relevant mutant strain. Our results show that expression of chimeric P5CDh cDNAs comprised of the 5Ј-UTR of the yeast PUT2 gene and the human open reading frame of P5CDh clearly complement the growth phenotype of put2 yeast and express P5CDh activity in the range observed for human fibroblasts (11). The failure to obtain this result with the complete human cDNAs may be explained by differences in the 5Ј-UTR of the two species. The PUT2 transcript has a typical S. cerevisiae 5Ј-UTR with an abundance of A nucleotides (47%), a scarcity of G nucleotides (11%) and an A in the Ϫ3 position relative to the ATG codon (28). By contrast, the 30-nucleotide 5Ј-UTR encoded by our human P5CDh cDNAs has 23% A, 20% G, and a G at the Ϫ3 position. A similar benefit of substituting a yeast 5Ј-UTR to obtain maximal expression of a human cDNA has been observed by others (37).
Biochemical studies suggest that P5CDh mutations are responsible for HP II, an inborn error of proline catabolism (1). Availability of the human P5CDh cDNAs described in this work will allow us to determine the molecular basis of this disorder. Furthermore, the ability to complement yeast strain MB1472 (put2) lacking P5CDh activity with the human P5CDh cDNA clones will provide an assay system to examine the functional consequences of mutations in human P5CDh.