Cloning and Molecular Characterization of Three Genes, Including Two Genes Encoding Serine Hydroxymethyltransferases, Whose Inactivation Is Required to Render Yeast Auxotrophic for Glycine*

The genes encoding both the cytosolic and mitochon- drial serine hydroxymethyltransferases (SHM2 and SHM1, respectively) and a third unidentified gene of the yeast Saccharomyces cerevisiae have been isolated and their nucleotide sequences determined. Analysis of the predicted amino acid sequence of the amino-terminal regions, sequence comparison with other genes encoding SHMT enzymes, and subcellular fractionation stud- ies all suggested that the SHMl gene encodes the mitochondrial SHMT, while the SHM2 gene encodes the cytosolic enzyme. The SHM2 gene but not the SHM1 gene has putative GCN4 sites upstream of the putative TATA box, suggesting regulation of its transcription by the general amino acid control system. Yeast mutants with disruptions at each SHM gene and in both genes were constructed and all mutants had the same growth requirements as the parental strains. Mutagenesis of the double-disrupted, shml shm2 yeast yielded strains of a single complementation group that are auxotrophic for glycine. Complementation of the glycine auxotrophy using a yeast genomic library retrieved the SHMl and SHM2 genes and a third gene designated GLY1. Gene disruption studies demonstrated that inactivation of SHM1,

* This work was supported by Grant MT-9822 from the Medical Research Council of Canada. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisernent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequencels) reported in this paper has been submitted and L22529.
Genes encoding SHMT (gZyA) have been isolated and sequenced from Escherichia coli (2,3) and several other bacteria (4-6). Eukaryotes have both mitochondrial and cytosolic SHMT isozymes, encoded by separate nuclear genes. The cytosolic gene from Neurospora crassa has been cloned and sequenced (7) and both isozymes from rabbit liver have been purified and their amino acid sequences determined (8,9). The rabbit cytosolic gene has been cloned and sequenced (10) and a plant mitochondrial gene has been cloned from pea (11). Recently, the cDNA sequences of the human cytosolic and mitochondrial genes have been published (12). There is a high degree of homology among all these proteins, particularly in the vicinity of residues known to be important for catalysis, such as the lysine residue which binds the pyridoxal phosphate cofactor.
Mutants of the E. coli glyA gene are auxotrophic for glycine (2). In Chinese hamster ovary cells, loss of SHMT activity in the mitochondria has been correlated with glycine auxotrophy (13,14). CHOAuxBl mutants lacking folate pools as a result of a mutation in folylpolyglutamate synthetase have a requirement for glycine, adenine, and thymidine (14-16). Complementation of the mutant with a bacterial folylpolyglutamate synthetase gene or a human cDNA lacking a mitochondrial leader sequence restores the adenine and thymidine deficiency but not the glycine deficiency (17), suggesting that mitochondrial folate metabolism is required for glycine synthesis in mammalian cells. In N. crassa, the for gene encoding the cytoplasmic SHMT complements a mutant which requires formate for growth (7). This finding suggests that in the absence of SHMT activity, C1 units enter the cytoplasmic folate pool as 10-CHO-H4PteGlu, in a reaction catalyzed by formyltetrahydrofolate synthetase, as had been proposed by Barlowe and Appling (18). The Neurospora for gene is under the control of the amino acid crosspathways control system, regulated by the transcription factor cpc-1. The analogous yeast regulatory system is the general amino acid control system regulated by GCN4.
In this report, we describe the cloning, DNA sequence determination, and chromosomal disruption of the genes encoding both the mitochondrial and cytoplasmic SHMT isozymes, SHMl and SHM2 and of a gene, whose disruption is required in a shml shm2 strain to yield a glycine auxotroph of the yeast Saccharomyces cereuisiae.
The yeast SHM and GLYl genes were isolated from genomic libraries constructed in the shuttle vector Yep13 (20) and Yep24 (21), respectively. The PCR-amplified fragments were cloned into pEMBL10' (22). The genes were subcloned into pUC18 and pUC19 (23) for DNA sequencing and bacterial expression studies.
Polymerase chain reaction amplification was performed using Taq polymerase (Boehringer Mannheim) as described (25) with the following modifications. To a microcentrifuge tube 0.1 pg of plasmid DNAfrom the pool of a yeast genomic library (20) was added. Samples were incubated in a DNA thermal cycler at 94 "C for 2 min before commencement of 30 cycles of amplification. Each cycle included 1 min denaturation (94 "C), 1.5 min annealing (55 "C), and 1.5 min of chain elongation (72 "C). After completion of the last cycle, the PCR products were purified by polyacrylamide gel electrophoresis, digested withEcoR1, and cloned into the phagemid vector pEMBL10'.
Plasmid Isolation, DNA Modification, Southern Analysis, a n d Bacterial Dansformation-Plasmid preparations were by the alkaline lysis procedure as described in Sambrook et al. (25). Restriction endonuclease digestions, nuclease digestions, and ligations were carried out according to the manufacturer's instructions. Colony blots were done as described (25). DNA was transferred to nitrocellulose or nylon membranes and hybridized by the Southern method as described in Ref. 25 with the appropriate probes at 65 "C. DNA restriction fragments used as probes were isolated from low melting temperature-agarose gels and labeled with dATP by the random primer procedure of Feinberg and Vogelstein (26) using Klenow fragment of DNA polymerase I. Transformation was by the method of Hanahan as described in Ref. 25. RNA Isolation a n d Northern Analysis-'Tbtal yeast RNA was isolated as described (27). Oligonucleotides SHMl (5'-GAAATTGGCGTTCCT-GACTT-3') and SHM2 (5-CTTCCTTACCGGTCTTAGGG-3') are specific for SHMl and SHM2 mRNAs, respectively, and were used as probes. Northern analysis was performed as described (28) using yeast cells synchronized by the a-factor arrest procedure (29).
Gene Disruption-The PCR-generated fragments were used to construct null mutants at the SHMl and SHMP loci using gene replacement (30). The 700-bp DNA fragments generated by PCR that were derived from the SHMl and SHMP genes were inserted into the EcoRI site of pEMBL10' to form pCRSHMl and pCRSHM2, respectively. The yeast HIS3 gene was inserted into pCRSHMl at the XhoI site within the SHMl gene fragment and the LEU2 gene was inserted into pCRSHM2 at the BglII site within the SHM2 gene fragment (see Fig.  3). The resulting plasmids, pCRSHMl::HIS3 and pCRSHM2::LEU2, were digested with EcoRI and HindIII, respectively, to release the HIS3 and LEU2 genes flanked by SHMl and SHMP, respectively. The linear DNA was used to transform recipient yeast W3031Ato histidine and W3031B to leucine prototrophy (31). Gene disruption at the GLYl locus was achieved through replacement of the 0.8-kb EcoRI fragment located within the cloned 3.0-kb SalI-XhoI yeast genomic fragment that includes GLYl (see Fig. 8). The resultant plasmid, pGLYl::URAB, was digested with Sal1 and BamHI and used to transform yeast to uracil prototrophy. Construction of the various null mutant strains was confirmed by Southern analysis. The probe used for Southern analysis of GLY1-disrupted strains was the 1.3-kb segment of DNA consisting of two HindIII fragments, as indicated in Fig. 8. DNA Sequencing-Plasmid DNA was isolated by the boiling method (25) and prepared for double-stranded sequencing by the method of Hsiao (32). The dideoxynucleotide termination method of Sanger et al. (33) was used with [ u -~~S I~A T P and T7 DNA polymerase. Chemical Mutagenesis-The SHM double-disrupted yeast strain YMO9 was treated with the mutagen ethyl methanesulfonate as described (34). Survivors were plated on SD medium supplemented with uracil, adenine, tryptophan, glycine (10 m), and formate (10) m.
Colonies were replica-plated onto SD medium supplemented with uracil, adenine, and tryptophan. Those yeasts which appeared to be a w otropic were replated to confirm their requirement for glycine and/or formate.
Cell Extraction4rude extracts of E. coli cells were prepared by sonication. For whole cell enzyme assays, yeast cells were either disrupted using glass beads or converted to spheroplasts using glusulase (Du Pont). The pelleted spheroplasts were resuspended in 25 m Tris-HCI, pH 7.6, 2 m EDTA, 10 m P-mercaptoethanol, 1 m phenylmethylsulfonyl fluoride, 10 m benzamidine followed by homogenization with a Teflon homogenizer. Extracts were centrifuged at 1,900 x g to pellet whole cells and debris. The supernatants were centrifuged at 13,000 x g to pellet mitochondria and this supernatant was taken as the soluble extract. Mitochondrial pellets were resuspended in extraction buffer and sonicated for use in enzyme assays. Enzyme Assay-SHMT activity was measured by a modification of the method of Geller and Kotb (35). [3-3HlSerine incorporation into tetrahydrofolate-bound products was measured. A typical assay contained 0.4 n m serine, 1 m H,PteGlu, and 2.5 m pyridoxal phosphate.
Analysis of Plasmid-encoded Gene Products in Maxicells-E. coli strain DR1984 was transformed with various plasmids, UV irradiated, and plasmid-encoded proteins were labeled with [35Slmethionine and [35S]cysteine as described by Sancar et al. (37). SDS-polyacrylamide gel electrophoresis was performed using 13% polyacrylamide gels according to the method of Laemmli (38) and the protein bands were visualized by fluorography.

RESULTS
Isolation a n d Cloning of the S. cerevisiae SHM Genes-DNA fragments of 700 bp were amplified by PCR using primers GLYl and GLY3 (DNA sequences are described under "Experimental Procedures"). GLYl has a degenerate sequence coding for the amino acids NKYSEG, residues 63-68 in the Neurospora cytosolic SHMT (7). while GLY3 codes for PEFKEY, residues 315-320 of the Neurospora enzyme. The residues are shown in Fig. 6. The amplified fragments were digested with EcoRI and ligated into EcoRI-digested pEMBL10'. Plasmids containing inserts were identified and the insert DNA was sequenced from both ends to verify homology between the predicted amino acid sequence encoded by the insert DNA and SHMT enzymes characterized from other sources. Two distinct sequences were found, each containing a n open reading frame with a predicted amino acid sequence homologous to SHMT enzymes and the corresponding genes were designated SHMl and SHM2. Probes were prepared from these amplified fragments and used to screen a yeast genomic library in the plasmid Yepl3. One clone, pSH36 was found containing the intact S H M l gene on a 5.3-kbp insert (see Fig. 1). Two clones were found, pSH2 and pSH3, containing the SHM2 gene ( Fig.  1). Clone pSH2 contained a 6.3-kbp insert of yeast genomic DNA. Subcloning and DNA sequence analysis revealed that pSH2 contained a n incomplete SHM gene, lacking the initiator methionine. Rescreening of the yeast genomic library using a 0.5kbp DNA fragment from the 5' end of the pSH2 insert led to the isolation of pSH3, which contains the intact SHMZ gene within a 7.5-kbp insert.
Subcloning a n d Expression Studies-The pSH3 insert was subcloned into pUC18 as a 9-kbp PstI-BamHI fragment to form p3BP. Although the gene was in the correct orientation for transcription from the lac promoter, this plasmid failed to complement the GS245 glyA strain. A deletion of 3.2 kbp from the 5' end of the gene was done by digestion with SphI and religation. The resultant plasmid, termed p3BPS, complemented the glycine auxotrophy of the E. coli GS245 strain and extracts of the transformants had SHMT activity that was 90% that obtained with wild type E. coli, while GS245 alone had no detectable activity. This indicates a relatively low level of expression, considering the high copy number of the gene. A labeled plasmid-dependent protein product of M, 50,000 was observed in maxicells containing this plasmid (Fig. 2, lane 2 ) a s well a s some peptides of greater mobility, which may represent degradation products. The M, 50,000 protein band disappeared when the SHM2 gene present in the plasmid was disrupted by filling in a BglII site (see Fig. 3) within the coding sequence (not shown), verifying that the band corresponds to the product of the gene. A 5.6-kbp BamHI-Hind111 fragment from pSH36 was sub-
p3BPS, containing the S I I M 2 grne flnnr 2 I. p96RH flnnr .?I. or p36RHABg (/one 4 1. The latter two plasmids contaln thr SIIMI grne. Extracts were analyzed by rlectmphorrsis on SDS-137 polyacrylamide gels followed by autoradiography. The position of the SH.MT-sprclfic polypeptide is indicated by t h r arrow. The positions of cnrlcctrophorrsrd standards arr indicated hy their molecular wrights at thr / r f t .

FIG. 3. Sequencing strategy for the SHMl ( A ) and SHM2 (R)
genes. The open hoxrs represrnt thr oprn rrad~ng framrs for thv grnrs. The solid regions within the hoxrs rcpresrnt the SIf.M-sprcific fragments amplified by the polymerase chaln reaction and which were used as probes to isolate the intact genes. The s o l d arrows show strrtchrs of sequence obtained from a singlr template. Arrows heginning at wrttml bars a t restriction sites represent srqurnces ohtainrd from trmplates with deletions using the corresponding rrstriction rndonuclrasr.
Arrows originating a t solid squaws rrprc'sent sequrncrs ohtainrd using specific primers. The remaining sequrncrs were ohtalned from the ends cloned into pUC18 and 19. The Hind111 site was derived from the pBR322 portion of the Yep13 vector, while the RamHI site appears to be a t t h e insert-vector boundary. The pSHB6 subclones in pUC (p36HB and p36BH) did not complement the GS245 strain ofE. coli in either orientation nor did a derivative of the pUCl9 subclone (p36BHABg) with a RamHI-RglII deletion removing 600 bp of upstream noncoding region and bring-

Y L M S D M A H I S G L V A A N V V P S P F E H S D I V T T T T H K S L R G P R
ATATTTGATGAGTUlTATCACACATATCCGOTTTGOT~CAGCCMTG~TCCCATCTCCA~GMCArTCC~TATAGTTACCACMCCACTCACAAGTCC~GAGAGGCCCMG * * * * * * * * * * * *

N G Y K L V S G G T D N W L J V I D L S G T Q V D G A R V E T I L S A L N I A A N K N T I P G D K S A L F P S G L R I G T P A R T T R G F G R E E F S Q V A K Y
* * * * * * * * * * * *

T M C A A G M C A C C A T C C C A G O T t A T M~T G C T C T~T C C C C T C~G O T C T A A~T C G O T A C T C C A G C M T G A C C A C G A~G A~~C C G T G A~A G T~C T C M G T C G~G T A
* * * * * * * * * * * * ing the open reading frame to within 300 bp of the lac promoter. There was no detectable SHMT activity in the transformants of these plasmids. No plasmid-dependent protein band of a size c o~s p o n d i n g to SHMT was observed in maxicells transformed with either p36BH or p36BHA~g (Fig. 2, lanes 3 and 4 1. A small amount of intact SHM1 product may be present if it migrates anomalous~y at M, 55,000 but if so, it is inactive.

DNA Sequence ~~e r~~n~~i o n
of the SHM Genes-The sequencing strategies for the SHM1 and SHM2 genes are shown in Fig. 3. Subclones were obtained for sequencing by restriction enzyme digestion and, in some cases exonuclease IIVS1 nuclease digestions, as indicated in the figure legend. The nucleotide sequence of the SHMI gene as well as 868 bp of 5'-noncoding sequence and 76 bp of 3"noncodin~ sequence is shown in Fig. 4. All of the reported sequence was determined on both strands, except the first 250 bp. The SHMl gene includes an open reading frame of 1470 bp coding for a predicted protein of 490 amino acids with a M , of 53,000. The ATG sequence corresponding to the predicted initiation codon is preceded by an A at the -3 position and followed by a dT at the +6 position, in agreement with the consensus sequences compiled fmm expressed genes in yeast (39). The ATG codon at +34 does not have these features, suggesting that it is not the in viuo start codon. There is a sequence of 5'-TATAAA-3' iocated at position -154 to -149 relative to the initiation codon, which conforms to a consensus TATA box sequence.

Q G G P H N H T I A A L A T A L K Q A A T P E F K E Y Q T Q V L K N A K A L E S
* * * * * * * * * * * *

N I A L N K N S I P G D K S A L V P G G V R I G A P A H T T R G M G E E D F H R
1234 ATTGTTCMTACATTMCMGGCTGTAGMTTCGCTCMCAAGTTCMC~GCTTCCAAAGGATGCTTGTAGATTAMGGACTTCMAGCCMGGTCGACGAAGGCTCTGATGTTTTG * * * * * * * * * * * * two differences from the consensus sequence have been shown to be functional in general amino acid control of other yeast genes (40). An amino acid sequence alignment that includes the yeast SHMT enzymes and other SHMTs whose sequences have previously been reported is shown in Fig. 6. The S H M l and SHM2 products have 59.5% amino acid sequence identity with each other. The SHM2 product is even more similar to the N . crassa cytosolic SHMT (7) with 70.6% identical amino acids than to the S H M l product, while 55% of the residues of the S H M l product are identical to the Neurospora cytosolic enzyme. This suggests that the SHM2 product is the yeast cytosolic SHMT and the S H M l product is the mitochondrial enzyme. The similarity of the yeast enzymes compared to the rabbit enzymes is about the same as their similarity compared to the pea mitochondrial enzyme ( 5 2 4 7 % ) . Both the yeast cytoplasmic and mitochondrial enzymes are slightly more similar to the rabbit cytoplasmic than to the rabbit mitochondrial enzyme, although the cytoplasmic enzymes have the highest degree of similarity (55%). The pea mitochondrial SHMT is also equally similar to the rabbit cytosolic and mitochondrial enzymes (11). These results agree well with the phylogenetic scheme shown by Garrow et al. (121, which shows that the mammalian SHMT isozymes are more closely related to each other than to plant or fungal enzymes.

I V Q Y I N K A V E F A Q Q V Q Q K L P K D A C R L K D F K A K V D E G S D V L N T U K K E I Y D U A G E Y P L
The predicted S H M l product has 16 additional amino acids at its amino-terminal than does the predicted SHM2 product. This amino-terminal sequence is also more extended than the corresponding regions in the rabbit SHMT enzymes. The sequence of the rabbit mitochondrial enzyme was determined from amino acid sequencing of the mature protein and does not include the mitochondrial leader sequence. The first 20 amino acids of the S H M l gene product have features typical of mitochondrial leader sequences: the presence of neither negatively charged amino acids nor extensive hydrophobic stretches of amino acids. Amino acids 15-20 of the yeast SHM1, VHRRGL, 2 .

MFPIWSAWIKCHATVHRRGLLTSGAQSLVSi(P'ISEGDPEMFDILQQERHRQKHSITL1PSEN MPYTLSDAHHKLITSHLV3TDPEVDSIIK3EIERQKHSIDLIASEN
3 . conform to the consensus sequence of hydrophobic-polar-lys-Arg-small-bulky hydrophobic, proposed by Schmidt et al. (41) to be present at the cleavage sites of some proteins having mitochondrial leader sequences. Since

A
Leu" aligns with leucine residues conserved in 4 out of 5 eukaryotic SHMT enzymes, we would predict that the cleavage site of the mitochondrial SHMT is immediately following Leu".
Characterization of SHM Null Mutants-The chromosomal SHMl and SHM2 genes were inactivated by gene disruption. The 700-bp PCR-generated DNA fragments derived from the SHMl and SHM2 genes were cloned into pEMBL10' and designated pCRSHMl and pCRSHM2, respectively. A DNA fragment containing the S. cereuisiae HIS3 gene was inserted into a XhoI site within the SHMl gene-specific DNA in pCRSHMl (see Fig. 3A ). A linearized DNA fragment containing the HIS3 gene flanked by S H M l DNA was used to transform yeast strain W3031A to histidine prototrophy. Similarly, a LEU2 gene-containing DNAfragment was inserted into pCRSHM2 at a unique BglII site located in the SHM2 gene-specific DNA fragment (Fig. 3B). The LEU2 gene flanked by SHMB DNA, a s a linearized DNA fragment, was used to transform strain W3031R to leucine prototrophy. Southern analysis confirmed the insertion of the HIS3 and LEU2 sequences into the coding rekons of chromosomal S H M l a n d SHMB (Fig. 7 ) . In p a n d A, l a n e s 2-5 all show disruption mutations at SHMl. Inpanel B, lanes 2 and 4 show disruptions of SHM2, while lanes 3 and 5 have wild type SHM2, presumably due to integration of the selectable marker at the LEU site. There are faint bands in both panels A and 3, indicating cross-hybridization of the probe with the heterologous SHM gene and showing that those genes are wild type.
Both YM06 (MATa shml:BZS3) and YM07 (MATa shm2:: LEU21 possessed phenotypes indistinquishable from the parental strains, indicating that disruption of a gene that codes for one of the yeast SHMT isozymes does not yield a glycine or formate auxotroph. Northern analysis was performed to confirm that insertion of the HIS3 and LEU2 genes obviated the expression of wild type SHMl and SHM2 mRNA (not shown).
A yeast strain containing a double disruption at the SHMl and SHM2 loci was constructed through mating of YM06 and YM07 and sporulation of the resultant diploid (YM08). YMO9 is a haploid product of YM08 sporulation, which is prototrophic for histidine and leucine, indicating disruption at the SHMl and SHM2 genes. This was confirmed by Southern (Fig. 7,   panel C, lanes 5-7) and Northern analysis (not shown). Inactivation of both SHMl and SHMZ did not result in any additional growth requirements for YMO9.
Zsolation of Glycine Requiring Yeast-The glycine prototrophy of the SHM double-disrupted YMO9 indicated that inactivation of a third gene may be required to render yeast a glycine auxotroph. Treatment of YMO9 with ethyl methanesulfonate allowed the isolation of a collection of yeast Gly-mutants. Two classes of Gly-yeast were observed. The two classes are distinguishable by the ability or inability to grow when exogenous glycine is replaced with 10 m M formate. Those yeast which remain Gly-in the presence of 10 m~ formate were studied further. Mating and complementation studies indicated that among eight independent Gly-yeasts examined a single complementation group exists (data not shown). The mutant allele for this complementation group has been designated d Y 1.
Zsolation of the S. cereuisiae GLYl Gene-% isolate yeast sequences that complement the glyl allele, Gly-yeast (YM09loa, see Table I) were transformed with the Yep24 (URA3) yeast genomic library (21). Plasmid DNA isolated from Ura+ Gly' transformants was used to retransform YM09-10a to glycine prototrophy. Based upon restriction enzyme and Southern analysis three types of yeast sequences were isolated (data not shown). ' b o of the sequences retrieved were the yeast SHMl and SHM2 genes. This observation infers that the isolation of a Gly-yeast requires inactivation of both yeast genes that encode SHMT. The third yeast sequence isolated is not homologous to either SHM gene, and a partial restriction map of the yeast fragment in pGLY1-loa3 is shown in Fig. 8. Various plasmids were constructed by cloning portions of the yeast genomic fragment retrieved from the complementing plasmid pGLY1-loa3 into Yep24. Each derivative was used to attempt complemen~tion of YM09-10a to glycine prototrophy (Fig. 8). A 1.5-kb SphI-Sal1 fragment was shown to be sufficient to complement the Gly-strain. The DNA sequence of this portion of the yeast genome is shown in Fig. 9. This sequence is identical to a sequence present in the EMBL Library data bank, which represents a portion of yeast chromosome 5 (accession no. L10380). Although portions of the sequence were determined on only one strand, all initial discrepancies with the EMBL sequence were resolved by sequencing both strands and running the labeling reactions with dITP to remove compressions. The fragment contains one open reading frame (designated ORF 35 in the EMBL citation) beginning at nucleotide 1437, which encodes a protein of 387 amino acids and a predicted molecular mass of 42,797 daltons, extending beyond the Sal1 site for an additional 41 base pairs. Our complementation studies suggest that the product is functional with the COOHterminal 14 codons deleted. An inspection of GenBank did not identify any known gene products with significant homology to the GLYl gene product.
Characterization of Yeasts with a Chromosomal Deletion at GLYl-Yeast with disruptions at GLYl were constructed to confirm that inactivation of the chromosomal GLY1, SHM1, and SHM2 genes, is required to render yeast auxotrophic for glycine. An 0.8-kb EcoRI fragment that includes the GLYl promoter and the sequence encoding the translation initiation codon, located within the 3.0-kb XhoI-Sal1 yeast genomic fragment which complements glyl, was replaced with the URA3 gene (Fig. 8). Linearized URA3 DNA and flanking yeast sequences were used to transform W3031B (SHM1 SHMZ), YM06 (shrnl SHM2), YM07 (SHM1 shm21, andYM09 (shml shm2f to Ura'. Southern analysis identified those Ura' yeast that have a disruption at the GAY1 locus (Fig. 7 D ) . The phenotype of each recipient and GLY1-disrupted yeast was examined through observation of growth rates in medium with and without glycine (Fig. 10). Strain YM13 (shml shm2 glyl::URA3) was completely auxotrophic for glycine and did not grow on medium without this supplement (panel A ) but grew normally in the presence of glycine {panel B). Strains YM11 (shml SHM2 glyl::URA3) and YM12 (SHM1 shm2 glyl::URA3) are partial glycine auxotrophs and grow at similar reduced rates, indicating that both SHM genes are functional and can synthesize glycine. Strain YMlO (SHM1 SHM2 gly1::URAS) is prototrophic for glycine but with a reduced doubling time compared to the parental yeast strain, W3031B. Interestingly, all strains disrupted at the SHM2 gene saturated at a higher cell concentration than those containing the intact gene, however, those shm2 strains also disrupted in GLYl required glycine supplementation to show this increased growth (Fig. 10).  Cell Cycle Analysis-Because of its role in the generation of 1-carbon units for DNAprecursor biosynthesis, SHMT plays a n important role in DNA replication. In S. cerevisiae, most DNA replication genes are subject to transcriptional regulation during the GI to S phase transition in the cell cycle (reviewed in Ref. 42). This particular pattern of transcription is conveyed by MluI cell cycle box elements, which have a consensus sequence ofACGCGTNA (29). These MluI cell cycle box elements are not present in the 5'-flanking sequences of SHMl or SHM2, suggesting that neither of these genes is regulated in the cell cycle. This was confirmed by Northern blot analysis of mRNA specific for SHMl and SHM2 in synchronously dividing cultures (Fig.  11). The levels of both mRNAs were found to remain constant throughout all phases of the cell in contrast to the TMPl transcripts, encoding thymidylate synthase, which are cell cycle regulated (43).

T Q I Y R S E S T E V D V D G N A I R E I K T Y K Y
Assays were done to determine SHMT activity in whole cell crude extracts of the null mutants and their parental strains. The shm2 extracts had 2-15% SHMT activity, compared to the parental strain, while the shml extracts had SHMT activity comparable to that of the parental strain, indicating that the cytoplasmic enzyme is the predominant SHMT in the yeast cell. Specific activities obtained were 2 0 4 0 nmol/h/mg for wild type and shml strains and 1-5 nmol/h/mg for shm2 strains. Spheroplasts were made and homogenates were subjected to differential centrifugation to separate the mitochondrial and cytoplasmic fractions. The shml extracts lacked SHMT activity in the mitochondrial fraction, while the shm2 extracts lacked SHMT activity in the cytoplasmic fraction and no SHMT activity was detected in extracts of the double mutants (Table 11). DISCUSSION We have assigned the SHMl protein as the mitochondrial SHMT and the SHM2 protein as the cytoplasmic enzyme. These assignments were based on: 1) the distribution of activities in the cytoplasmic and mitochondrial fractions after differential centrifugation of extracts of mutants disrupted at the SHMl or the SHM2 genes. 2) The greater similarity in predicted primary amino acid sequence between the SHM2 protein and known cytoplasmic SHMT proteins, particularly the Neurospora enzyme. 3) The extended amino-terminal sequence found in the SHMl protein, which is consistent in amino acid  (25). and samples were taken a t IO-min intervals after u-factor release as indicated. RNA was prepared from each sample and equal amounts were electrophoresed through a native agarose gel. The gels were dried and hybridized with probes specific forSIIM1. SHM2, TMP1, and LEU2 transcripts as described under "Experimental Procedures." The TMP 1 and LE112 determinations serve as controls for cell synchronization and equal loading of RNA. respectively. composition with a mitochondrial leader sequence. 4 ) Our results suggest that the mitochondrial SHMT may be expressed in E. coli but is inactive. The inactivity may be due to interference with the enzyme activity by a n uncleaved leader sequence. There has been a recent report of the sequencing of the gene corresponding to S H M l and its localization to yeast chromosome 2 (44). These authors identify the gene as encoding the cytosolic SHMT because the leader sequence does not correspond to the mitochondrial consensus. Their sequence differs from ours by one fewer A residue 81 bp upstream of our predicted initiation codon. We suggest that this difference results in a much longer predicted protein product due to 5' extension of the open reading frame into nonsense sequences. The additional residues are not homologous to known SHMT proteins and do not resemble a mitochondrial leader sequence. Expression would then require the use of i n Llitro mutagenesis to delete the leader sequence codons and insert a Shine-Dalgarno sequence followed by a n ATG codon a t t h e predicted site of cleavage. The cytosolic enz-yme is also relatively poorly r xpressed in E. coli. The level of expression we obtainrd is very similar to that found for the human SHMT enzymes using plasmids of similar copy number 12). We are conducting mutagenesis experiments to place the SHMB gene under the con-~l o~i~ of Genes ~n a c t i u a t e~ in a Yeast Glycine Auxotroph trol of a more efficient bacterial promoter and include a bacterial ribosome binding site to optimize expression of this gene product in E. coli.
Surprisingly, inactivation of either one, or both yeast SHM genes did not result in any additional auxotrophic requirement. This is in contrast with results from other systems. Bacterial strains deficient in SHMT, and also CHO cells defective in mitochondrial SHMT exhibit an auxotrophy for glycine (2,13,14), while Neurospora strains lacking cytoplasmic SHMT require formate for growth (7). The s. cerevisiae gene disruption mutants are not leaky, since no enzyme activity can be detected in the double mutants, and both single mutants, particularly shm2, show decreased enzyme activity. Northern blots showed that the mRNA specific for the disrupted SHM gene was not detected in the shml or shm2 mutant strains. Mutagenesis of the double disruption mutants led to the isolation of glycinerequiring mutants, some of which could grow with formate in place of glycine. The GLYl gene which complements the glycine auxotrophs is at a separate locus from the SHM genes. These studies indicate that yeast has an additional pathway, besides the SHMT reactions for glycine synthesis. One possible pathway for glycine synthesis is from glyoxalate using an aminotransferase. Our GenBank search for proteins homologous to GLYl did not detect homology to tran~minases. This does not exclude the possibility that GLYl encodes an amino transaminase since few residues are conserved in all known transaminases (48, 49), or the reaction in yeast may be catalyzed by a different family of aminotransferase. The glycine produced by this pathway could meet cellular requirements for protein synthesis and supply C1 units for the mitochondrial folate pool via the glycine cleavage system. The cytoplasmic folate pool would be provided with C1 units from formate using the 10-CHO-H,folate synthetase activity of the cytoplasmic C,-H,folate synthase. Unlike Neurospora, yeast must be able to generate sufficient formate from sugars to supply the Cl pool, since shml shm2 double mutants do not require exogenous formate. The production of this formate may be mediated by the product of the gene inactivated in our formate or glycine-requiring mutants. In this mutant, the glycine produced by the GLYl pathway is insufficient to supply the C1 pools, since the supply of formate is now disrupted. Exogenous glycine can be cleaved to form 5,10-CHz-H,folate in the motochondria and the C1 units converted to formate by the mitochondrial Cl-H4folate synthase, through reversal of its 10-CHO-H~folate synthetase activity. The formate can pass into the cytoplasm and be incorporated into the cytoplasmic C1 pool as proposed by Barlowe and Appling (18). We are presently characterizing the formate or glycine-requiring mutants and will study the effect of all our mutations in strains that also lack functional ADE3 or the MIS1 products to verify their roles in supplying C1 units in the absence of SHMT activity.
The growth studies suggest that the GLYl pathway is the major source of glycine in yeast, since disruption of GLYl alone affects the growth rate, whereas disruption of both SHM genes does not. The equal growth rates observed with the YMll (shml SHMZ glyl::URA3) and YM12 (SHM1 shm2 g2yl::URAS) (Fig. 10) strains show that in yeast glycine synthesis i s not confined to the mitochondria but the cytoplasmic SHMT makes a significant contribution to glycine synthesis.
SHMl and SHM2 belong to a small class of S. cerevisiae genes that are involved in DNA precursor biosynthesis but are not subject to cell cycle regulation at the transcriptional level. Included in this class are DFRl, encGding dihydrofolate reductase, DCD1, encoding dCMP deaminase, and DUT1, encoding dUTP p~p h o s p~a~s e (45). Our results, however, do not preclude cell cycle regulation of SHM gene expression by posttranslational processes.
Analysis of the 5"noncoding regions of the SHM genes suggested that expression of SHMZ but not SHMl may be regulated by the GCN4 transcription factor. This result was anticipated, since the Neurospora cytosolic SHMT is under the control of the amino acid cross-pathways control system (71, which is homologous to the yeast general amino acid control. The GCN4 factor-mediated general amino acid control pathway regulates amino acid biosynthesis. SHMT may be regulated by this pathway since the enzyme is involved in glycine biosynthesis and can be used to generate serine. GCN4 control is also likely involved in regulating the expression of two other folatedependent yeast enzymes, NAD-dependent 5,lO-methylenetetrahydrofolate dehydrogenase (46) and dihydrofolate reductase (47). It is somewhat surprising that the mitochondrial enzyme is not under general amino acid control, since the major site of glycine biosynthesis is mitochondrial in other systems and our studies with the glycine auxotroph triple mutants discussed above suggest that deficiency of the mitochondrial SHMT, combined with the second site mutation results in growth stimulation by glycine. We are presently performing deletion and mutagenesis studies to identify the sequences required for expression of both genes.