Molecular Enzymology of Mammalian Δ1-Pyrroline-5-carboxylate Synthase

Δ1-Pyrroline-5-carboxylate synthase (P5CS; EC not assigned), a mitochondrial inner membrane, ATP- and NADPH-dependent, bifunctional enzyme, catalyzes the reduction of glutamate to Δ1-pyrroline-5-carboxylate, a critical step in the de novo biosynthesis of proline and ornithine. We utilized published plant P5CS sequence to search the expressed sequence tag data base and cloned two full-length human P5CS cDNAs differing in length by 6 base pairs (bp) in the open reading frame. The short cDNA has a 2379-bp open reading frame encoding a protein of 793 residues; the long cDNA, generated by “exon sliding,” a form of alternative splicing, contains an additional 6-bp insert following bp +711 of the short form resulting in inclusion of two additional amino acids in the region predicted to be the γ-glutamyl kinase active site of P5CS. The long form predominates in all tissues examined except gut. We also isolated the corresponding long and short murine P5CS transcripts. To confirm the identity of the putative P5CS cDNAs, we expressed both human forms in γ-glutamyl kinase- and γ-glutamyl phosphate reductase-deficient strains ofSaccharomyces cerevisiae and showed that they conferred the proline prototrophy. Additionally, we found expression of the murine putative P5CS cDNAs conferred proline prototrophy to P5CS-deficient Chinese hamster ovary cells (CHO-K1). We utilized stable CHO-K1 cell transformants to compare the biochemical characteristics of the long and short murine P5CS isoforms. We found that both confer P5CS activity and that the short isoform is inhibited by l-ornithine with a K i of ∼0.25 mm. Surprisingly, the long isoform is insensitive to ornithine inhibition. Thus, the two amino acid insert in the long isoform abolishes feedback inhibition of P5CS activity by l-ornithine.

⌬ 1 -Pyrroline-5-carboxylate synthase (P5CS; EC not assigned), a mitochondrial inner membrane, ATP-and NADPH-dependent, bifunctional enzyme, catalyzes the reduction of glutamate to ⌬ 1 -pyrroline-5-carboxylate, a critical step in the de novo biosynthesis of proline and ornithine. We utilized published plant P5CS sequence to search the expressed sequence tag data base and cloned two full-length human P5CS cDNAs differing in length by 6 base pairs (bp) in the open reading frame. The short cDNA has a 2379-bp open reading frame encoding a protein of 793 residues; the long cDNA, generated by "exon sliding," a form of alternative splicing, contains an additional 6-bp insert following bp ؉711 of the short form resulting in inclusion of two additional amino acids in the region predicted to be the ␥-glutamyl kinase active site of P5CS. The long form predominates in all tissues examined except gut. We also isolated the corresponding long and short murine P5CS transcripts. To confirm the identity of the putative P5CS cDNAs, we expressed both human forms in ␥-glutamyl kinase-and ␥-glutamyl phosphate reductase-deficient strains of Saccharomyces cerevisiae and showed that they conferred the proline prototrophy. Additionally, we found expression of the murine putative P5CS cDNAs conferred proline prototrophy to P5CS-deficient Chinese hamster ovary cells (CHO-K1). We utilized stable CHO-K1 cell transformants to compare the biochemical characteristics of the long and short murine P5CS isoforms. We found that both confer P5CS activity and that the short isoform is inhibited by L-ornithine with a K i of ϳ0.25 mM. Surprisingly, the long isoform is insensitive to ornithine inhibition. Thus, the two amino acid insert in the long isoform abolishes feedback inhibition of P5CS activity by L-ornithine.
Proline can be synthesized either from glutamate or ornithine and is nonessential for full-term human infants and adults (1, 2) but conditionally indispensable for premature neonates (3,4). ⌬ 1 -Pyrroline-5-carboxylate (P5C), 1 which is in nonenzymatic equilibrium with glutamic ␥-semialdehyde, is a common intermediate in these pathways (5). Formation of P5C from glutamate in humans has been proposed to be catalyzed either by P5C synthase, a bifunctional enzyme with both ␥-glutamyl kinase (␥-GK) and ␥-glutamyl phosphate reductase (␥-GPR) activities or a complex of separate ␥-GK and ␥-GPR enzymes (6). Hu et al. (7) showed that plant P5CS is bifunctional, catalyzing the conversion of glutamate to P5C. By contrast, P5CS activity in prokaryotes and lower eukaryotes like Saccharomyces cerevisiae requires the combined functions of separate ␥-GK and ␥-GPR enzymes. In rats, P5CS activity is highest in small intestine with measurable activity in colon, pancreas, and thymus (8). Since proline may be a neurotransmitter and because the concentration of proline in the extracellular fluid of the central nervous system is low (Ͻ5 M) (9 -14), P5CS may be of particular importance in certain regions of the central nervous system. Wakabayashi et al. (8) showed that detectable P5CS activity is present in rat cerebrum and cerebellum; however, the regional distribution of this enzyme in the central nervous system has not been well characterized. P5C can also be synthesized from ornithine in a reaction catalyzed by ornithine ␦-aminotransferase (OAT) (15).
The P5CS genes of plants (7, 16 -18) and the ␥-GK and ␥-GPR genes of prokaryotes (e.g. in Escherichia coli; see Ref. 19) and S. cerevisiae (20,21) have been cloned and sequenced. Mutant strains of S. cerevisiae auxotrophic for proline because of mutations in ␥-GK or ␥-GPR genes are known and provide a potential system for expression and analysis of human P5CS. Additionally, CHO-K1 cells are proline auxotrophs and lack both P5CS and OAT activities (22)(23)(24).
Early studies of Chinese hamster-human somatic cell hybrids (25) and chromosome localization of glutamate oxaloacetate transaminase (26), a marker linked to P5CS, suggested that the human P5CS gene is on chromosome 10. During the course of our work, Aral et al. (27) reported the sequence of a human cDNA that, based on sequence similarity, appeared to encode P5CS. Direct evidence showing that this cDNA encodes * This work was supported in part by National Eye Institute Grant (5RO1EY02948). 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.
To better understand the molecular enzymology and regulation of mammalian P5CS, we utilized homology to Vigna P5CS to clone two full-length human P5CS cDNAs and the corresponding murine P5CS cDNAs. Expressing these cDNAs in mutant yeast strains and in CHO-K1 cells, we showed both forms of mammalian P5CS encode bifunctional enzymes and complement growth phenotype of mutant cells. More interestingly, we found that the short but not the long P5CS isoform is sensitive to feedback regulation of ornithine. Our results suggest a mechanism for differential regulation of P5CS activity.

EXPERIMENTAL PROCEDURES
cDNA Libraries and Plasmids-We obtained a human small intestine 5Ј-STRETCH cDNA library from CLONTECH and a human kidney cDNA from Dr. G. Bell. pBluescript II KS(ϩ) was used for all cloning manipulations.
Cloning by Homology Probing and cDNA Library Screening-We utilized the deduced amino acid sequence of Vigna aconitifolia P5CS (7) as a probe to screen the GenBank TM (EST) data base (National Center for Biotechnology Information, Bethesda, MD) with the BLAST algorithm (29). We obtained a candidate human EST clone (GenBank TM accession no. R13516) from the National Genome Center, Lawrence Livermore National Laboratory (Livermore, CA) and sequenced the cDNA insert. Using this putative P5CS EST clone as a probe, we screened Ͼ5 ϫ 10 5 plaques from both the human kidney and small intestine cDNA library as described (30,31). We purified two positive plaques, one (HsP5CS.K5) from kidney cDNA library and the other (HsP5CS.G1) from the small intestine cDNA library, through two additional cycles of screening, subcloned the inserts (HsP5CS.K5, 2.7 kb; HsP5CS.G1, 2.8 kb) into pBluescript II KS(ϩ), and sequenced them. Using the full-length human P5CS cDNA sequence as a probe, we also identified a full-length murine P5CS long form cDNA (GenBank TM accession no. AA1042889) and confirmed its identity by sequencing the insert (MmP5CS.long.1; 3.3 kb).
Genomic and Reverse Transcription PCR-We performed PCR amplification on either human or murine genomic DNA (100 ng) with primers DV2375 (5Ј-GGATTTCCATGATGAGCAGAAGC) and DV2374 (5Ј-CAGTCGGGCAGCCAGGCTATCATTATC), corresponding to sequences flanking the 6-bp insert in HsP5CS.long with an initial denaturation step of 6 min at 96°C, followed by 30 cycles of 1 min at 94°C, 2 min at 65°C, and 2 min at 72°C. The amplified products were electrophoresed in 1% agarose, excised, gel-purified, and sequenced directly with the 5Ј nested primer DV1585 (5Ј-GCCGAACCTCAATG-GAACAC). Reverse transcription PCR (RT-PCR) was done with 1 g of poly(A) RNA from human small intestine, placenta, or cerebellum (CLONTECH) with a antisense primer DV1537 (5Ј-TCCACAATGTCT-GTGATGACGTGC) and a cDNA cycle kit (Invitrogen) for the reverse transcription reaction followed by PCR with primers DV2375 and DV1537 with the same conditions as utilized for PCR of genomic DNA. We cloned the amplified DNA fragments in pCR II (Invitrogen) and transfected the recombinant molecules into DH5␣-competent cells. To quantify the ratio of the long and short forms of the inserts, we performed a nested PCR amplification with primers complementary to internal sequence DV1475 (5Ј-CCCAGCTGAGCCCAACAGTGACC) and DV2374 and bacterial transformants harboring recombinant plasmids as template DNA. The PCR conditions were as utilized for PCR of genomic DNA. The amplified DNA fragments were electrophoresed in 4% Nusieve agarose to resolve the long and short products differing in length by 6 bp.
DNA and Deduced Amino Acid Sequence Analysis-We sequenced all cDNA clones in both directions using the Sequenase version 2.0 DNA sequencing kit (U.S. Biochemical Corp.) according to the manufacturer's protocols. DNA sequences were aligned and compared using the MacVector (Eastman Kodak Co.) and the BLAST programs. We compared the deduced amino acid sequences of the putative full-length human P5CS cDNA clones to that of S. cerevisiae ␥-GP, ␥-GPR, and additional kinases with the MegAlign program (DNASTAR).
Ribonuclease Protection Assay on Human RNA-We cloned a PCR fragment of the HsP5CS cDNA corresponding to bp ϩ474 to ϩ852 of the long form into pCR II (Invitrogen) and sequenced the insert. To provide a template for synthesis of the riboprobe, we linearized the insert by digestion with EcoRI and utilized [␣-32 P]UTP, T7 RNA polymerase (Ambion) to make cRNA and the RPA II kit (Ambion) to conduct ribo-nuclease protection assay (RPA). The cRNA consists of a sequence 100% complementary to a 379-nucleotide fragment of the HsP5CS.long plus 12 nucleotides of the vector. We hybridized human poly(A) RNA overnight at 45°C with the cRNA (2 ϫ 10 5 cpm) and then digested with an RNase mixture (1 unit of RNase A and 200 units of RNase T1, RPA II kit, Ambion) at 37°C for 30 min. We determined the size of the protected fragments by electrophoresis on a 6% polyacrylamide sequencing gel in parallel with a sequencing ladder and determined the intensity with a PhosphorImager (Molecular Dynamics, Inc., Sunnyvale, CA).
Northern Blot Analysis-We hybridized two commercial human multiple tissue Northern blot filters (CLONTECH) under high stringency conditions (final wash in 0.1% SDS, 0.1ϫ SSC, 50°C for 40 min) as suggested by the manufacturer. Human ␤-actin was used as a probe to verify RNA quality and quantity. Hybridization and autoradiography were performed according to Hu et al. (30).
Construction of Human P5CS Yeast Expression Vectors-We used the 2-based p424GAL1 (32) E. coli/yeast shuttle vector plasmid containing TRP1-and Amp R -selectable markers and a multiple cloning site downstream of an S. cerevisiae galactokinase (GAL1) promoter. pTB26, a plasmid containing the S. cerevisiae PRO1 gene, was a gift from M. Brandriss. We subcloned the yeast PRO1 gene from pTB26 to p424GAL1 to form pScPRO1.
To replace the S. cerevisiae PRO1 ORF with those of the long and short forms of human P5CSs, we used a homologous recombinationbased cloning strategy (33) to construct pHsP5CS.short.2. Briefly, we designed two oligonucleotide primers to amplify the entire human P5CS ORF flanking with S. cerevisiae PRO1 5Ј-and 3Ј-untranslated sequences. The 5Ј sense primer contained 67 nt of sequence identical to the sense strand of S. cerevisiae PRO1 beginning 67 nt 5Ј to the initiation ATG (Ϫ67 to Ϫ1) followed by 27 nt identical to the sense strand of the 5Ј-end of the human P5CS ORF (ϩ1 to ϩ27). The 3Ј antisense primer contained 60 nt of sequence corresponding to the antisense strand beginning 60 bp downstream of the stop codon of S. cerevisiae PRO1 ORF (ϩ1686 to ϩ1627) followed by 34 nt of sequence identical to the antisense strand of the 3Ј-end of the human P5CS.short ORF (ϩ2382 to ϩ2349). We used a combination of Pfu and Taq polymerases (Stratagene) to amplify HsP5CS.short. Homologous recombination in yeast was done by cotransfecting the BglII-and HpaI-linearized pScPRO1 and the yeast/human P5CS fragment into yeast strain DT1103 (see below). We synthesized pHsP5CS.long.2 by replacing the relevant region of P5CS.short with the counterpart of the long form containing the 6-bp insert. The sequence at the recombination junctions of these recombinant plasmids was vetted by sequencing.
Construction of Murine P5CS Expression Vectors-We used the pcDNA3 (Invitrogen) mammalian expression plasmid containing Neo Rand Amp R -selectable markers and a multiple cloning site downstream of human cytomegalovirus promoter. We digested the murine EST clone, designated as pMmP5CS.long.1 containing the full-length murine P5CS long form cDNA with EcoRI and ApaI and subcloned the insert into pcDNA3 to form pMmP5CS.long.2.
To introduce MmP5CS.short into pcDNA3, we first made an intermediate construct, designated as pMmP5CS.short.1, by replacing the region containing the 6-bp insert of the long form with the corresponding sequence of the short form. We took advantage of the unique restriction enzyme sites BsgI (ϩ482) and StuI (ϩ807) to engineer pMmP5CS.long.1. Initially, we amplified the short form cDNA by RT-PCR of murine small intestine poly(A) RNA (1 g) with a antisense primer DV2231 (5Ј-AGGAGGTGAAGAATGCGGTTGCTG) and a cDNA cycle kit followed by PCR with primers DV2967 (5Ј-CCTGCACTCAG-GACAGAACCATCTGAA) and DV2899 (5Ј-GGGCTGGCAATCTTC-CCTCTG) with the same conditions as utilized for PCR of genomic DNA (see above). We digested the amplified DNA fragments with BsgI and StuI and ligated them into BsgI/StuI-linearized pMmP5CS.long.1 to produce pMmP5CS.short.1. Subsequently, we subcloned the insert of pMmP5CS.short.1 into pcDNA3 to form pMmP5CS.short.2. We vetted the replaced segment of the recombinant plasmid pMmP5CS.short.2 by sequencing.
P5CS Enzyme Assay and End Product Regulation-We assayed P5CS activity radioisotopically as described by Wakabayashi et al. (8). In preliminary experiments (not shown) we verified that P5CS activity is dependent on the presence of ATP, MgCl 2 , and NADPH and increases linearly with added protein up to a concentration of 1 mg/ml and with time up to 60 min.

Isolation and Characterization of Human and Murine P5CS
cDNAs-We utilized the C-terminal 200-amino acid sequence of Vigna (mothbean) P5CS to perform a BLAST searches of the GenBank TM EST data base and identified a single human candidate clone (GenBank TM accession no. R13516). The deduced amino acid sequence of this clone had 59% identity to the corresponding region of mothbean P5CS (data not shown). A second BLASTP search utilizing the EST as probe retrieved several cDNAs including Arabidopsis P5CS, mothbean P5CS, a Caenorhabditis elegans EST sequence (GenBank TM accession no. D35184) and ␥-GPR genes for S. cerevisiae and several prokaryotes (data not shown). We obtained this human putative P5CS EST clone, sequenced the entire insert for confirmation, and utilized it to probe human kidney and small intestine cDNA libraries. One cDNA clone (P5CS.K5) had an insert size of 2726 bp extending from ϩ472 (where ϩ1 is the A of the initiation methionine codon) to a poly(A) tail. The second clone (P5CS.G1) lacked the 3Ј-most 503 bp of P5CS.K5, had a 2225-bp overlap with the 5Ј-end of P5CS.K5, and extended an additional 529 bp 5Ј to Ϫ57. Using a similar strategy, we also obtained a murine putative P5CS EST clone (GenBank TM accession no. AA1042889) isolated from 13-day-old mouse heart cDNA library had an insert size of 3343 bp extending from Ϫ53 (where ϩ1 is the A of the initiation methionine codon) to a poly(A) tail. It has a 53-bp 5Ј-UTR, 2385-bp ORF, and 805-bp 3Ј-UTR. The mouse ORF encodes a polypeptide with 95% sequence identity to its human counterpart.
The overall amino acid sequence of the human putative P5CS has 43 and 40% identity to those of Arabidopsis thaliana and V. aconitifolia P5CSs, respectively. Examination of the human P5CS sequence indicates that the enzyme is bifunctional. The N-terminal half has 36% identity to S. cerevisiae ␥-GK, while the C-terminal half has 48% identity to S. cerevisiae ␥-GPR (Fig. 1). The N-terminal half of human P5CS also has strong similarity to lower eukaryotic and prokaryotic ␥-GKs, and the C-terminal half is similar to lower eukaryotic and prokaryotic ␥-GPRs (data not shown). As expected for a mitochondrial inner membrane protein, the N terminus of both the human and the murine P5CS sequences has features characteristic of a mitochondrial targeting sequence with frequent basic and no acidic residues (36). Based on the consensus for cleavage of mitochondrial targeting sequences (36 -38), there are potential cleavage sites following amino acids 56 and 64 (Fig. 1). Using BLASTP and Blocks data base (39) searches, we identified a 32-amino acid motif (between Asp 245 and Pro 276 ) with similarity to a sequence thought to be the active site motif in several other kinases in the aspartokinase family (see Figs.  1 and 2). In this motif, two aspartate residues (Asp 245 and Asp 258 ) are conserved in a comparison of eight aspartokinases including one from Vigna, two from yeast, and five from prokaryotes. Five other residues are identical in at least seven of the examples. Additionally, the human and murine P5CS sequences have three conserved residues (Asn 191 , Asp 193 , and Phe 194 ) predicted by homology to Vigna P5CS and E. coli ␥-GK to be involved in proline feedback regulation of P5CS activity (7,38,40) (Fig. 1). Also, there are four putative N-glycosylation sites and a putative NADP(H) binding domain (41) (Fig. 1).
Comparing the sequences of the various human putative P5CS cDNAs, we noted that some had a 6-bp insert following bp ϩ711 while others did not. To determine if this insert was the result of some form of alternative splicing, we amplified genomic DNA with primers complementary to cDNA sequences flanking the region of the insert (5Ј DV1475; 3Ј DV2374). Amplification of cDNA with these primers produces a 74-bp fragment. However, with genomic DNA as template, the product was ϳ500 bp. Sequence analysis showed that this fragment contains a portion of a 5Ј exon followed by a 400-bp intron and a portion of a 3Ј exon. We sequenced the amplified P5CS structural gene fragment and found that the 6-bp insert derives from the 5Ј-end of an intron with two potential donor splice sites in tandem (Fig. 3). Alternative utilization of these two donor splice sites generates two different mature transcripts: a short form, utilizing the most 5Ј donor site and lacking the insert (Fig. 3), and a long form, utilizing the most 3Ј donor site and containing the 6-bp insert following position ϩ711.
To examine the tissue expression of the two P5SC isoforms, we conducted RT-PCR and RPAs with poly(A) RNA isolated from human small intestine, placenta, and cerebellum. The former has the advantage that individual clones can be scored to give a quantitative result; the latter has the advantage that it avoids PCR. In the RT-PCR assay, we cloned the amplified fragments and performed a second PCR amplification with a pair of nested primers surrounding the region containing the insert. In clones containing the insert, the amplified product is 80 bp; in clones without, the amplified product is 74 bp. We found that the ratio of the short to long forms in small intestinal RNA was 0.3 (58/19), whereas the ratio in placental (15/62) and cerebellum (15/59) RNAs was ϳ0.25 (Fig. 4). The RPA data are generally consistent with this result. The antisense riboprobe (cRNA) used for RPA is 391 bp and has 6 bp of vector sequence at each end flanking 379-bp HsP5CS.long sequence complementary to bp ϩ474 to ϩ852. When the probe hybridizes to the mature HsP5CS.long transcript, a 379-bp fragment is protected. Alternatively, when the probe hybridizes to a P5CS.short transcript, a single-stranded loop structure will form at the site of the insert yielding two products of 238 and 135 bp following RNase digestion. As seen in Fig. 4, this assay confirms the short form is predominant in the gut, whereas the converse is true in placenta and cerebellum. PhosphorImager analysis of the intensity of these bands gave short to long ratios of 3.75 in gut, 0.83 in placenta, and 0.55 in cerebellum.
Tissue Distribution and Size of Human P5CS Transcripts-Northern blots of mRNA from multiple human tissues showed a single predominant P5CS transcript of about 3.6 kb (Fig. 5). Pancreas, ovary, testis, and kidney had the highest expression, followed by colon, small intestine, placenta, heart, and skeletal muscle.
Functional Complementation of ␥-GKand ␥-GPR-deficient Yeasts-To confirm the identity and function of the two P5CS cDNAs, we constructed composite human P5CS cDNA clones, designated HsP5CS.short.1 and HsP5CS.long.1, respectively. The HsP5CS.short.1 cDNA is 3256 bp in length including 57 bp of 5Ј-untranslated sequence, a 2379-bp ORF, and 820 bp of 3Ј-UTR extending to the poly(A) tail. The HsP5CS.short.1 ORF encodes a 793-amino acid protein with a predicted M r of 90. HsP5CS.long.1 is identical to HsP5CS.short.1 except that it includes the 6-bp insert following position ϩ711, resulting in a 2385-bp ORF encoding a 795-amino acid protein. To express these in yeast, we utilized PCR and homologous recombination to construct yeast/human P5CS chimeric minigenes (pHsP5CS.short.2 and pHsP5CS.long.2), comprising the S. cerevisiae PRO1 promoter and 5Ј-and 3Ј-UTR sequences surrounding the intact human P5CS.short and P5CS.long ORFs. Previous work by ourselves and others has shown that substitution of yeast 5Ј-and 3Ј-UTR sequences may improve the expression of mammalian genes in yeast (30,42). DT1103, a proline auxotroph with a partially deleted ␥-GK gene (pro1), is unable to grow on minimal medium without proline. We transfected this strain with either pHsP5CS.short.2 or pHsP5CS.long.2 and selected transformants on medium lacking both proline and tryptophan (MGAtrp Ϫ ). Both pHsP5CS.short.2 and pHsP5CS.long.2 restored the PROϩ growth phenotype in DT1103 (see strains HV13 and HV14; Fig. 6) although not as efficiently as a construct containing the yeast ␥-GK gene (pScPRO1; see HV10, Fig. 6). To confirm that both P5CS cDNAs encode bifunctional enzymes, we performed a second complementation experiment, transfecting pHsP5CS.short.2 and pHsP5CS.long.2 into DT1102, a mutant yeast strain auxotrophic for proline because it has a partial deletion of the ␥-GPR gene (pro2). As shown by the DT1102 transformants, HV15 and HV17, pHsP5CS.short.2 and pHsP5CS.long.2 also complement deficiency of ␥-GPR (Fig. 6). As predicted by the S. cerevisiae requirement for the combined activities of ␥-GK and ␥-GPR for proline prototrophy, pScPRO1 does not complement the proline auxotrophy of DT1102 (see strain HV16, Fig. 6). We also measured the growth rates of these strains in proline-deficient liquid medium. The doubling time was 2.7 h for strain HV10 and about 6 h for strains HV13, HV14, HV15, and HV17. DT1103 and DT1102 transfected with vector alone do not grow in this medium (data not shown). Thus, as predicted by the sequence, these results confirm that both human P5CS cDNAs encode bifunctional enzymes with ␥-GK and ␥-GPR activities.
Both Murine P5CS cDNA Isoforms Complement the CHO-K1 Cell Growth Phenotype-Although the human P5CS cDNAs complement the growth phenotype of the ␥-GKand ␥-GPR-deficient yeast strains, we were not able to measure P5CS activity in extracts of these yeast transformants. We reasoned that this might reflect the heterologous environment. To do biochemical studies, therefore, we transfected murine P5CS cDNAs into a subclone (NC5) of CHO-K1 (CHO-K1-NC5) cells previously shown to be auxotrophic for proline (43) because of deficiency of both P5CS and OAT activities (22,23). We selected stable transformants in medium lacking proline and assayed P5CS activity on individual clones. To rule out reversion at the hamster P5CS locus, we confirmed that the P5CS expressed by the transformants was murine rather than hamster by RT-PCR and sequence analysis (not shown). We found that both murine P5CS cDNAs expressed RNA transcripts and complemented the growth phenotype of CHO-K1-NC5 cells. Assay of P5CS activity in extracts of stable transformants expressing either MmP5CS.short or MmP5CS.long showed that both constructs conferred measurable P5CS activity. Consistent with the role of P5CS in ornithine and citrulline synthesis, previous studies using extracts of rat small intestine showed that P5CS activity is noncompetitively inhibited by ornithine (44,45). To determine if both isoforms of murine P5CS had similar sensitivity to ornithine inhibition, we assayed their activity in the presence of 5 mM ornithine. We found that P5CS.short, the predominant isoform in gut, is inhibited Ͼ75% by 5 mM ornithine with a K i in this assay of ϳ0.25 mM (Fig. 7). In agreement with the earlier work, we obtained similar results using mouse small intestine as the source of P5CS activity (Fig. 7). By contrast, P5CS.long, the predominant isoform in peripheral tissues, is completely insensitive to 5 mM ornithine (Fig. 7). DISCUSSION Utilizing homology to Vigna P5CS, we isolated two splice forms of cDNAs encoding human P5CS: a short form (P5CS.short) encoding a 793-amino acid protein and a long form (P5CS.long), with an additional 6-bp insert following bp ϩ711, encoding a 795-amino acid protein. The human P5CS polypeptides have high sequence identity to P5CS of plants and to ␥-GK and ␥-GPR enzymes of lower eukaryotes and prokaryotes. Expression of human P5CSs restores proline prototrophy in two mutant yeast strains, one deficient in ␥-GK (pro1), the other in ␥-GPR (pro2), confirming the bifunctional nature of the mammalian enzyme. To study the biochemistry and regulation of mammalian P5CS isoforms, we expressed murine P5CS cDNAs in CHO-K1 cells that lack endogenous P5CS activity. Both murine P5CS cDNAs restored proline prototrophy, and stable transformants expressed measurable P5CS activity. Using extracts of these stable transformants, we showed that the two isoforms differ dramatically in their sensitivity to inhibition by ornithine.
During the course of our studies, Aral et al. (27) also reported the sequence for a human P5CS cDNA. As we did, they identified a partial length EST clone containing the 3Ј 438 bp of the ORF. They used the sequence of this EST to initiate a series of 5Ј-rapid amplification of cDNA end reactions followed by direct sequencing of the products to extend the cDNA sequence 138 bp into the 5Ј-UTR. The nucleotide sequence of our P5CS cDNAs has 14 single bp differences with that of Aral et al. (27), all in the ORF and resulting in 13 amino acid changes. Some of these differences may represent polymorphisms; others may reflect sequence errors. Aral et al. (27) utilized direct sequencing of amplified material. Our composite sequence derives from sequencing multiple cDNA clones in both directions. The over- lapping sequences of our various cDNAs are 100% identical (for example, clones HsP5CS.G1 and HsP5CS.K5 have a 2225-bp overlap in perfect agreement).
In addition to the overall similarity of the N-terminal half of HsP5CS to ␥-GKs from a variety of organisms (e.g. 36% identity with S. cerevisiae ␥-GK), BLASTP and Blocks data base (39) searches utilizing the ␥-GK domain of the human P5CS as a probe revealed a 32-amino acid motif (from Asp 245 to Pro 276 ) in the N-terminal half of human P5CS with similarity to a conserved sequence characteristic for members of the aspartokinase family including aspartokinases, uridylate kinases, and carbamate kinases from a variety of species (46 -49). Aspartokinases catalyze phosphorylation of substrates with a carboxyl group as the nucleophilic acceptor (50). The conserved sequence overlaps with and extends more 5Ј of a region of sequence similarity between E. coli and Serratia marcescens aspartokinases and ␥-glutamyl kinases identified by Omori et al. (46), who suggested it is the active site of these related enzymes. Three residues in this motif corresponding to Asp 245 , Asp 260 , and Pro 276 in HsP5CS.long are conserved in all ␥-GKs and nearly all members of the aspartokinase family (Fig. 2). Another region likely to be important for HsP5CS function is a cluster of three amino acids (Asn 191 , Asp 193 , and Phe 194 ) about 50 residues N-terminal of the putative aspartokinase motif. These are conserved with Vigna P5CS, where mutagenesis experiments have shown that they are essential for feedback regulation of P5CS activity by proline (40) (Fig. 1).
We identified two forms of P5CS cDNAs differing only by a 6-bp insertion at position ϩ711. Two lines of evidence indicate that this insert is generated by a variant of alternative splicing called "exon sliding" (51). First, the insert sequence is colinear with the genomic sequence of the human P5CS structural gene and has the consensus sequence of a 5Ј splice site. The same colinearility is also present in murine P5CS (data not shown). Thus, the choice between the two tandem splice donor sequences determines whether the mature transcript will contain the 6-bp insert. The fact that we see markedly different ratios of short/long transcripts in various tissues suggests either that there is tissue-specific stability of the two products or that the choice of splice donor is somehow regulated. How this choice might be regulated is unclear. Based on the calculation of the consensus value of the 5Ј splice donor site (28,(52)(53)(54), both of these tandem donor sites are relatively "weak" with the consensus values of 77.7 and 73.2, respectively, where the value for a typical donor site ranges from 0.7 to 1.0 with the mean of 0.85. This suggests that there might be tissue-or developmentspecific factor(s) influencing the relative use of the two competing 5Ј splice sites. Multiple mechanisms including cis elements and trans-acting factors influence 5Ј splice site selection (53)(54)(55)(56)(57)(58)(59)(60)(61). Exonic splicing enhancers are positive acting, cis RNA sequences capable of inducing the assembly of the splicesome at weak 5Ј splice sites and are generally purine-rich. Interestingly, we note several purine-rich sequences in both HsP5CS and MmP5CS cDNAs 20 -500 bp 3Ј of the 6-bp insert, for example, a GGAAATGAAAA 47 bp 3Ј (bp ϩ765 to ϩ775), a GGAGGGGAAGAAG 314 bp 3Ј (bp ϩ1032 to ϩ1045), and an AGCAGGGAGAAA 376 bp 3Ј (bp ϩ1094 to ϩ1105) of the 6  short (Ͻ20 codons) inserts in mature transcripts by "exon sliding" have been described in the Drosophila Ultrabithorax gene (62), porcine urokinase-like plasminogen activator (63), human ␤-adducin gene (64), chicken growth hormone-releasing hormone gene (65), and human presenilin-1 gene (66).
Interestingly, the 2-amino acid insert (Val 238 and Asn 239 in P5CS.long) is immediately N-terminal of the putative aspartokinase motif (amino acids 245-276) and 43 amino acids Cterminal of the cluster of residues (amino acids 191-194) predicted to be involved in the feedback regulation of plant P5CS by proline (Fig. 1). This location in or near the active site and the apparent variation in the ratio of short versus long isoforms in different tissues suggests that the presence or absence of the insert may play a role in the regulation of P5CS activity. Previous work utilizing extracts of rat small intestinal mucosa showed that mammalian P5CS is inhibited by ornithine with a K i of about 0.4 mM (44,45). We found similar results for the isoform encoded by the predominant gut transcript (P5CS.short). In mice and humans, plasma ornithine is about 75 M with a distribution ratio of intracellular/extracellular ornithine of about 5, predicting an intracellular ornithine concentration of ϳ0.37 mM (15). This suggests that physiologic concentrations of ornithine are in the range predicted to regulate P5CS.short activity. By contrast, we found the long P5CS isoform is insensitive to ornithine inhibition. These results correlate functional differences in the protein with the choice of splice sites. As far as we know, this is the first demonstration of functional differences in protein isoforms generated by exon sliding.
The P5CS reaction serves two physiologic functions in mammals: proline biosynthesis and arginine biosynthesis. In tissues with a high protein synthetic requirement for proline (e.g. cartilage and possibly central nervous system), P5CS catalyzes the first committed step in the two-step pathway synthesizing proline from glutamate (5). In prokaryotes and plants, this pathway is regulated by allosteric inhibition of P5CS by the end product, proline (7,40,67,68). In contrast to prokaryotes, yeast, and plants, the mammalian P5CS reaction is also the first committed step in the pathway of de novo arginine biosynthesis. In this pathway, the product of the P5CS reaction is converted to ornithine by OAT and then to citrulline and arginine by the enzymes of the urea cycle (15,69,70). Arginine synthesis is active in the epithelium of small intestine, particularly in early postnatal life (69 -75). De Jonge et al. (71) showed that genes encoding arginine synthetic enzymes including OAT, carbamyl phosphate synthetase, ornithine transcarbamylase, argininosuccinate synthetase, and argininosuccinate lyase are highly expressed in rat small intestine during perinatal development, and the pattern of expression of these genes is spatiotemporally and zonationally regulated as follows. mRNAs for OAT, carbamyl phosphate synthetase, and ornithine transcarbamylase are concentrated at the base of the villi and developing crypts, while mRNAs for argininosuccinate synthetase and argininosuccinate lyase are detected only in the upper part of the villi, suggesting that the basal enterocytes synthesize citrulline and that the enterocytes in the upper half of the villus synthesize arginine. The physiologic significance of this arginine synthetic pathway is highlighted by the observations that mice with targeted disruption of OAT die as a result of arginine deficiency in the neonatal period and that human infants with inherited deficiency of OAT develop severe hypoargininemia (76). The variable sensitivity to end product inhibition of the two isoforms of mammalian P5CS (P5CS.short and P5CS.long) provides a mechanism for this differential regulation. Additionally, mammalian P5C reductase, which catalyzes the first committed step in proline biosynthesis, is sensi-tive to inhibition by proline (77). Combining our results with this information suggests a tissue-specific model for regulation of the pathway converting glutamate to either arginine or proline. In gut, where the emphasis is on arginine biosynthesis, P5CS.short is the predominant isoform, and ornithine regulates the pathway by its inhibitory effect on P5CS. In peripheral tissues where proline biosynthesis is important, P5CS.long predominates, and the pathway is insensitive to ornithine but is regulated at the level of mammalian P5C reductase by proline (Fig. 8).
Aral et al. (27) and Kamoun et al. (78) describe two siblings with joint hyperlaxity, skin hyperelasticity, cataracts, and mental retardation who are the product of a consanguineous union. Their biochemical phenotype (low plasma proline, citrulline, and ornithine) suggests P5CS deficiency, and both were shown to be homozygous for the P5CS missense mutation, L396S. The functional consequence of this mutation on P5CS activity was not tested. Availability of the human P5CS cDNA sequences and the expression systems provided by P5CS-deficient CHO-K1 cells and yeast strains provides the molecular resources to study the P5CS genes of these and similar patients.