Expression of Escherichia coZi Folylpolyglutamate Synthetase in the Chinese Hamster Ovary Cell Mitochondrion*

Chinese hamster ovary (CHO) cell transfectants ex- pressing Escherichia coli folylpoly-y-glutamate synthetase (FPGS) activity solely in their cytosol lack mito- chondrial folylpolyglutamates and are auxotrophic for glycine. Addition of a mammalian mitochondrial leader sequence targeted E. coli FPGS to the mitochondria of these cells. Mitochondrial expression of FPGS restored mitochondrial folylpolyglutamate pools and overcame the glycine requirement. Pteroyltriglutamates func-tioned as effectively as the longer glutamate chain length folates found in wild type CHO cells in the metabolic cycle of glycine synthesis provided they were lo- cated in the mitochondria. Although folylpolygluta- mates cannot enter the mitochondria, mitochondrial folylpolyglutamates can be released without prior hydrolysis and CHO transfectants expressing E. coli FPGS activity solely in the mitochondria possessed normal cytosolic folylpolyglutamate pools. The proportion of cel- lular folate in the mitochondrion is governed by competition between mitochondrial and cytosolic FPGS activities.

I$ To whom correspondence and reprint requests should be addressed.
A U X B l transfectants ( A m -c o l i ) expressing the Escherichia coli FPGS gene (folC) metabolize folates primarily to the triglutamate derivative (2). These triglutamates were retained by the cell as effectively as the longer polyglutamate derivatives normally found in CHO cells (2) and also appeared to function as effectively in supporting the metabolic cycles of thymidine and purine synthesis (6). However, AUX-coli remained auxotrophic for glycine while glycine synthesis was restored in A U X B l transfectants expressing human FPGS (AUX-human) and containing folylpolyglutamates of chain length ranging from four to eight (6). Further studies showed that WTT2 cells a n d AUXh u m a n transfectants contained mitochondrial and cytosolic FPGS activity and that a significant proportion of cellular folate was located in the mitochondria of these cells, while AUXcoli, which only expresses a cytosolic FPGS activity, lacked mitochondrial folates despite possessing a normal cytosolic folate pool (7). The role of mitochondrial folate, metabolism is poorly understood. Mammalian cells possess cytosolic and mitochondrial isozymes of serine hydroxymethyltransferase, the enzyme responsible for glycine synthesis, and loss of the mitochondrial activity results in a glycine auxotrophy (8,9). The glycine-requiring phenotype ofAUX-coli may have been due to a lack of mitochondrial folate rather than an inability of pteroyltriglutamate to support glycine synthesis.
Wild type CHO cells (WTT2) were obtained from Dr. Sharon Krag, Johns Hopkins University. Mutant CHO cells (AUXBl), which lack FPGS activity and are auxotrophic for glycine, thymidine, and purines (3-61, were obtained from Dr. Victor Ling, Ontario Cancer Institute, Toronto. GlyB, a glycine-requiring CHO mutant (12), was obtained from Dr. L. Thompson, Lawrence Livermore National Laboratory. The generation of AUX-coli (D5-3A8) and Dl-1A10, AUXBl transfectants expressing high and low E. coli FPGS activity solely in the cytosol and auxotrophic for glycine, has been described previously (2). Cells were routinely cultured in DMEMiFBS + GHT under a water-saturated 5%) CO, atmosphere in a 37 "C incubator and were routinely passaged when they neared confluence as described previously (2,6). Cell number was determined using a Coulter counter.
Mammalian Expression Vectors pSV2 mammalian expression vectors, which were originally developed by Southern and Berg (13), were used in this study (Fig. 1). Their construction is briefly described below. Plasmids pSV2-p-globin (4.7 kb) (13) and pSV2(Bam)-p-globin, a derivative of pSV2-P-globin with a BamHI linker inserted at the PvuII site, were obtained from Dr. Re- Construction of folC Gene with Mitochondrial Leader Sequence imfolCi-A 1.4-kb HindIII-BamHI fragment from plasmid pAC5 (19), which contains the coding region of the E. coli folC gene (GenBank accession number J02808), was subcloned into HindIII-BamHI-treated pTZ19U and transformed into E. coli MV1190 (Bio-Rad) to generate pTZ19-foZC (4.3 kb, amp'). Single-stranded DNA was produced using helper phage M13K07, and an EcoRV cleavage site was generated between upstream nucleotides -3 and -4 (from start ATG, Fig. 2) by a C to T mutation at base -2 (20) to generate plasmid pTZlS-folC(RV). DNAwas sequenced by the method of Sanger et al. (21) using Sequenase 2 (U. S. Biochemical Corp.).
A HindIII-PuuII fragment (0.26 kb) of pSPOTC2, containing the cDNA for the leader sequence region of ornithine transcarbamoylase, was ligated to HindIII-EcoRV-treated pTZlS-folC!RV) and pTZ19-mfolC (4.4 kb) was isolated. This plasmid contains the folC gene with the ornithine transcarbamoylase leader peptide coding region ligated to its amino-terminal sequence (mfolC, Fig. 2). AHindIII-EcoRI fragment of pTZ-mfolC containing the mfolC gene was isolated and cloned into similarly treated pSP72 to generate pSP72-mfolC (3.99 kb1, which contains the mfolC gene followed by a downstream BglII site.

Cloning of Hygromycin-resistance Gene in Mammalian Expression
Vectors-A 0.95-kb fragment containing the E. coli hyg gene was obtained by BamHI digestion of pLG89. The p-globin gene was removed from pSV2(Bam)-p-gZobin by HindIII-BglII digestion, and the remaining fragment (4.2 kb) and the hyg fragment were treated with phosphatase and blunt-ended with DNA polymerase I (Klenow fragment). Ligation of the two blunt-ended fragments resulted in plasmid pSV2-hyg (5.13 kb, Fig. 1) with elimination of the HindIII, BglII, and BamHI cloning sites. The orientation of the inserted hyg gene was checked by fragment sizes after digestion with EcoRI and BamHI.
Cloning of mfolC Gene into Mammalian Expression Vectors-pSV2-72 was constructed by ligation of HindIII-BglII-digested pSP72 (2.4 kb) to similarly treated pSV2-p-globin (4.2 kbJ. This removed the p-globin gene with its internal BamHI site from pSV2-p-globin and resulted in a plasmid with a single BamHI site which simplified further subcloning. ABamHI fragment (2.15 kb) of pSVa-hyg, containing all the elements required for expression of the hyg gene in mammalian cells (SV40 promoter, polyadenylation signal, and t-splice site) was ligated to BamHI and calf intestinal alkaline phosphatase-treated pSV2-72 to generate pSV2-72-hyg (8.75 kbJ. The pSP72 stuffer fragment (2.4 kb) of pSV2-72-hyg was removed by HindIII-BglII digestion, and the remaining hyg-containing fragment (6.36 kb) was ligated to the mfolC gene fragment from a HindIII-BglII digest of pSP72-mfolC to give pSV2-hyg-mfolC (7.92 kb, Fig. l J , which contains both the mfolC and hyg genes. A plasmid containing only the mfolC gene in a mammalian expression vector (pSV2-mfolC, 5.8 kb, Fig.  1) was made by ligation ofHindII1-BglII-digested pSP72-mfolC to similarly treated pSV2(Bam)-p-globin.
Transfection of CHO Cells AUX-coli and AUXBl cells (2 x lo5) were plated on 60-mm dishes and cultured for 24 h in DMEWdFBS + GHT. Duplicate plates were then transfected with pSV2-hyg, pSV2-mfolC, or pSV2-hyg-mfolC ( 5 pg of DNA) or mock-transfected, by a modification (22) of the calcium phos-phate precipitation method of Graham and van der Eb (23) as described previously (2). Following transfection and glycerol shock, the cells were rinsed once with 5 ml DMEWdFBS + GHT + pedstrep and then incubated in the same medium (5 ml).
At 48 h after transfection, the remaining cells were transferred to duplicate 100-mm plates containing selective media (10 ml) consisting of DMEWdFBS supplemented with GHT + hygromycin (0.4 mg/mlj, HT + PteGlu (1 PM), or HT + PteGlu + hygromycin. After 8 days (with media changes at 4 days), one plate from each duplicate was fixed with methanol and stained with methylene blue, and the number of colonies were counted. Six distinct colonies from each remaining plate were transferred, using a cloning cylinder, to separate 35-mm dishes containing the same selective media, and the selection was continued for a further 8 days. Cells were then resuspended and diluted in selective media, and aliquots, equivalent to less than one cell per well, were transferred to 96-well microtiter plates. Colonies resulting from single cells were transferred after 7 days to selective media in 35-mm dishes. The transfectants were selectively cultured to near confluence in transition from 35-mm to 60-mm to 100-mm plates (approximately 1 month). Individual clones were then routinely passaged in DMEWdFBS + GHT or stored at -80 "C.

Cell Growth i n Media Lacking Products of One-Carbon Metabolism
As a n initial screen of the transfectants, near-confluent folate-de- The folate requirement of transfectants cultured in media lacking products of one-carbon metabolism was assessed a s described previously (6). Folate-depleted cells cultured in DMEWdFBS + GHT medium were resuspended in various test media and replated a t 1 x lo5 cells per 60-mm dish. The test media were DMEWdFBS + GHT in which either G, H, or T was omitted and the media were supplemented with various concentrations of PteGlu (0-100 pMj. Duplicate sets of plates were removed for cell counting a t 24-h intervals, and media were replaced in the remaining plates a t 48-h intervals. Cell doubling times were calculated by best fitting exponential growth curves to the data.

Assay of FPGS
FPGS activity in cell and subcellular extracts was measured in duplicate as described previously (2)

Cellular Uptake and Metabolism of Folate
Folate-depleted cells, cultured for several weeks in DMEWdFBS + GHT medium, were replated in the same medium a t 5 x lo5 cells per 100-mm dish or 5 x lo6 cells per 150-cm' flask (for subcellular fractionation). After culturing for 24 h, the medium was replaced with the same medium containing various concentrations of ["HJPteGlu, 5-formyl-H,I:'HIPteGlu, or 5-methyl-H,[3H1PteG1u (2 n"20 p~) with duplicate samples for each concentration. After the labeling period (usually 24 h), the cells were washed with ice-cold PBS (4 x 5 mlj, and 0.1% trypsin in PBS (0.5 ml) was added (room temperature, 22 "C). When the cells had detached from the plate (1-2 min), they were resuspended in cold PBS containing 0.01%. soybean trypsin inhibitor (10 ml). Aliquots of the suspension were removed for counting cell number and total radioactivity. The remaining suspension was centrifuged at 200 x g for 5 min, and labeled folate in the cells extracted or subcellular fractions were prepared as described below. Polyglutamate chain length distributions were determined a s described previously (2).

Subcellular FPGS and Folate Distributions
CHO cells were cultured as described previously in 150-cm2 flasks (2, 6, 7). The monolayers were washed twice with ice-cold PBS (2 x 10 ml) and dislodged with 0.1% trypsin-EDTA. Cells were transferred with cold PBS containing O.Ol%> trypsin inhibitor (2 x 5 ml) to 50-ml plastic conical tubes. Aliquots were removed for cell counting, and the remaining cells were pelleted by centrifugation (500 x g, 10 min). Cell pellets were washed once with cold PBS (10 mlj and then resuspended in 10 ml of homogenization buffer (0.25 M sucrose, 1 m M EDTA, pH 6.9), and aliquots were removed for assessing labeled folate accumulation. The remaining cells were homogenized, and crude subcellular fractions were prepared by differential centrifugation as described previously (71. In some cases, postnuclear supernatants were fractionated on a self-forming continuous Percoll density gradient as described previously (7). Organelle marker enzymes were assayed as described previously (7).

Identification of the mfolC Gene Product by Western Blotting
Mitochondrial and cytosolic fractions of AUXBl transfectants were prepared by differential centrifugation. The cytosolic fraction was concentrated using a microconcentrator (Centricon 10, Amicon). Polyacrylamide gel electrophoresis in sodium dodecyl sulfate was carried out using a 5% stacking gel and a 10-15% gradient separating gel in a slab gel apparatus using the discontinuous buffer system of Laemmli (24). Proteins were transferred to nitrocellulose in a TransBlot Semi-Dry apparatus (Bio-Rad) at 4 "C for 1 h at 15 V, according to the manufacturer's instructions.
Further manipulations were carried out a t room temperature in a plastic dish on an oscillating rocker. The nitrocellulose was blocked for 1 h using PBS containing 3% non-fat milk. Anti-FPGS serum (16 pgirnl) was added to the blocking solution and allowed to adsorb for 12 h at 4 "C. The blot was then washed four times with 25 ml of PBS + Tween 20 (0.05%) for 15 min each. Peroxidase-conjugated antibody was added at a 1:2000 dilution in blocking solution and allowed to adsorb for 2 h, and the blot was washed as before. The blot was then rinsed four times with 25 ml of PBS (no Tween), and 4-chloro-1-naphthol (0.6 mg/ml) in PBS containing 20% methanol and 0.02% H,O, was added. Color development was stopped after 20 min by rinsing the blot with deionized water.

Protein Assay
Protein concentration was measured by a deoxycholate-trichloroacetic acid precipitation modification of the Lowry procedure (25) using bovine serum albumin as the standard.

Thnsfection of AUX-coli and AUXBl Cells with the mfolC
Gene-To investigate the role of mitochondrial FPGS activity on folate metabolism, AUX-coli, which express E. coli FPGS activity solely in the cytosol, and A m 1 cells, which lack FPGS activity, were transfected with the E. coli folC gene containing a leader peptide ligated to its amino-terminal end (rnfolC). If the mfolC product is transported into the mitochondria and the ornithine transcarbamoylase leader sequence is processed normally, the mitochondria of the transformants should contain a modified E. coli FPGS protein with an additional five amino acids added at the amino terminus (Fig. 2). A s it was not known whether the mfolC construct would be expressed in CHO cells or if the protein would be transported into mitochondria and expressed in an active form, nor whether the expected triglutamate products would support mitochondrial glycine synthesis, various plasmids ( Fig. 1 ) were used to allow for selection by phenotypic complementation (-G + HT + Pte-Glu ( 1 VM) medium) and/or by antibiotic resistance (+hygromycin). G418 resistance could not be used as AUX-coli cells express the neo gene (2). Transfection efficiencies were assessed after an 8-day culture under selective conditions. In AUX-coli and A m 1 cells transfected with plasmids containing the hyg gene and selected in +GHT + hygromycin medium, colonies appeared at a rate of The numerous colonies obtained in -glycine medium after transfection with plasmids carrying the mfolC gene suggested that the gene was expressed and active product might be present in the mitochondria. No colonies were observed with mocktransfected cells in -glycine (reversion frequency of less than lO"9 or plus hygromycin medium. Growth of Dansfectants-Growth rates and nutritional requirements were initially assessed for 32 of the cloned transfectants. These transfectants were selected under -G or -G + hyg conditions and then cloned and cultured for several months under nonselective conditions. Ten out of fourteen AUXcolimcoli transfectants, AUX-coli cells transfected with pSV2-hyg-mfolC or pSV2-mfolC, retained the ability to grow in medium supplemented with PteGlu (1 PM) and lacking glycine with growth rates ranging from 14 to 125% that of WTT2. AUX-coli did not grow in this medium. As expected, all AUXcoli-mcoli transfectants grew in medium supplemented with glycine and folate but lacking purines and thymidine as did the parent AUX-coli line. All 18 AUX-mcoli transfectants, A m 1 cells transfected with pSV2-hyg-mfolC or pSV2-mfolC, retained the ability to grow in medium lacking glycine with growth rates ranging from 13 to 84% of that of WTT2. Surprisingly, these transformants also grew in medium lacking purines and thymidine with growth rates ranging from 17 to 77% of that of WTTZ while AUXB1 was unable to grow under these conditions. All cells transfected with pSV2-hyg-mfolC were able to grow in the presence of hygromycin regardless of whether they were originally selected in -G or -G + hyg medium although the fastest growth rates were observed in transfectants initially selected in -G + hyg medium.

Subcellular Distribution of FPGS Activity in Dansfectants-
With the exception of one cell line, all transfectants expressed enzyme activity with characteristics ofE. coli FPGS. H,Pte was a substrate (dihydrofolate synthetase activity), and enzyme activity was increased at 200 mM versus 20 mM KC1 (2,26). One cell line expressed FPGS activity with characteristics of mammalian enzyme and lacked dihydrofolate synthetase activity and was probably a revertant. Two transfectants of each class were chosen for more detailed study. The morphology of AUXcoli-mcoli-1 and -2 and AUX-mcoli-1 appeared identical with AUX-coli, and these cells appeared slightly larger than WTT2 cells. AUX-mcoli-2 was smaller than the other transfectants and grew at a faster rate.
FPGS activities are shown in Table  I. FPGS activity was higher in transfectants with H,Pte compared to H,PteGlu as a substrate, when assayed under standard conditions, and enzyme activity was significantly increased by increased KC1 (200 mM) and glutamate (2 mM), which are characteristic properties of the E. coli protein (2,26). WTT2 cells lacked activity with H,Pte as a substrate. As shown previously (7), enzyme activity was approximately equally distributed between the mitochondrial and cytosolic fractions in WTT2 cells, and the mitochondrial enzyme was of higher specific activity, while AUX-coli only contained cytosolic enzyme. All AUXcoli-mcoli transfectants expressed E. coli FPGS of very high specific activity in their mitochondria, and the mitochondrial fraction accounted for 50 to 70% of total cell enzyme activity. AUX-mcoli transfectants also expressed E. coli FPGS in their mitochondria but of lower specific activity than in the AUXcoli-mcoli transfectants. Some FPGS activity was found in the cytosolic fraction from AUX-mcoli transfectants, but the specific activity was very low and the distribution of enzyme activity was similar to that of the mitochondrial marker enzyme glutamate dehydrogenase. Similarly, the specific activity of FPGS was increased in the cytosolic fraction from AUXcoli-mcoli transfectants compared to AUX-coli, but the increase was consistent with mitochondrial enzyme contamination of this fraction. Although these data cannot eliminate the possibility of trace levels of expression of cytosolic FPGS from the mfolC construct, practically all, if not all, of the active FPGS resulting from expression of this gene in the transfectants was located in the mitochondria.
Western analyses of AUXcoli-mcoli transfectants using antibody to E. coli FPGS indicated the presence of two mitochondrial FPGS protein bands of similar intensity and differing in size by about 1 kDa (data not shown). The smaller size band appeared identical in size with the single bands observed with cytosolic extracts from AUXcoli-mcoli, whole cell extracts from AUX-coli, and pure E. coli protein. The larger size band may have been the unprocessed precursor protein but its size suggested that it represented a processing intermediate (27). The two mitochondrial protein bands were also detected in AUXmcoli transfectants, but no cytosolic protein was observed with these cells. These data also suggest little, if any, cytosolic FPGS in AUX-nzcoli transfectants and eliminate the possibility that accumulation of a partially active unprocessed precursor protein in the cytosol might result in some cytosolic FPGS.
Uptake and Subcellular Distribution of Folate by Dansfectant Cells-Cells were cultured in medium containing a physiological level of L3H]folinate (7 nM) for 24 h, and postnuclear supernatants from the CHO cells were fractionated on a selfforming continuous Percoll density gradient (Fig.  3). Labeled folate from WTT2 cells (not shown), AUXcoli-mcoli, and AUXmcoli cells was associated with heavy and light mitochondrial fractions and with the cytosolic fraction, while no organelle peaks were observed with AUX-coli extracts (7). All transfectants metabolized folinate primarily to the triglutamate, the characteristic major end product of E. coli FPGS, and expression of E. coli FPGS in the mitochondria of AUXcoli-mcoli and AUX-mcoli restored the ability of the cells to accumulate mito-  chondrial folates (Table 11). Mitochondrial folylpolyglutamate chain length distributions were longer than cytosolic distributions in AUXcoli-mcoli transfectants with a higher proportion of tetraglutamate (Table  111, consistent with the high specific activity of mitochondrial FPGS in these cells. However, only 6 to 14% of total cell folate was associated with the mitochondria in these transfectants. A much higher proportion of cell folate was associated with mitochondria of AUX-mcoli transfectants (30 to 48%). However, despite the apparent lack of cytosolic FPGS activity, 50 to 70% of the cellular folate in AUX-mcoli was located in the cytosol, suggesting the possibility of release of folylpolyglutamates from the mitochondria. The presence of normal cytosolic folate pools in these transformants explains the earlier observation that AUX-mcoli were capable of growth in folate-supplemented medium lacking thymidine and purines.
Effect of Medium Folate Concentration on Folate Accumulation a n d Metabolism-The ability of WTT2 cells and transfectants to accumulate labeled medium folinate acid is shown in Fig. 4. The concentrating ability of the cells is expressed as the ratio of intracellular to extracellular folate concentrations and assumes 1 pmoV10' cells is equivalent to an intracellular concentration of 1 V M (2). Intracellular folate concentrations for the transfectants may be slightly overestimated due to the slightly larger size of these cells. At physiological levels of medium folinate (2 to 20 nM), WTT2 cells and the transfectants accumulated similar levels of labeled vitamin. As the medium folinate increased to pharmacological levels (up to 20 PM), the ability of the various transfectants to accumulate folate was similar and all accumulated very high levels of folate (about 600 VM) while the ability of WTT2 to accumulate folate was more limited. Under physiological conditions, folinate accumulation by WTTZ and AUX-coli is limited by transport (6). At pharmacological levels (20 PM), accumulation by WTT2 is limited by FPGS activity while accumulation by AUX-coli cells, which express high levels of FPGS, remains limited by influx rates (6). The ability of AUX-mcoli to accumulate high levels of folinate as effectively as AUX-coli and AUXcoli-mcoli demonstrates that transport of folate into the mitochondria in AUXmcoli is not limiting under these conditions. Similar studies with labeled folic acid, which is transported, and consequently concentrated, less effectively than folinate by CHO cells, showed qualitatively similar effects (data not shown).
The effect of media folate on intracellular pteroylpolyglutamate distributions is shown in Table 111. There was little difference in the type of folylpolyglutamate that accumulated in the various transfectants over a wide range of medium folate or folinate concentrations. Extending the culture time from 24 to 72 h increased the amount of tetraglutamate that accumulated, but triglutamate was still the major intracellular derivative. Increases in medium folate cause a shortening of intracellular folate glutamate chain lengths in WTT2 cells and complete loss of very long chain length derivatives (6).
Effect of Medium Folate Concentration on Subcellular Folate Distribution-The effects of medium folinate concentration and culture time on subcellular folate accumulation are shown in Table IV. Total cell folate accumulation was fairly linear with time at both low (7 n M ) and high (2 VM) medium folinate, increasing about 4-fold in WTT2 between 5 and 24 h and about 4 to 6-fold in AUXcoli-mcoli-l and AUX-mcoli-1. At low folinate, the proportion of cellular folate in mitochondria of WTTZ cells decreased slightly between 5 and 24 h. At high folinate, the proportion of folate in the mitochondria was much lower but did not change between 5 and 24 h. Mitochondrial folate accumulation by AUXcoli-mcoli-1 cells was not affected by culture time, but the relatively low proportion of cell folate in the mitochondria of these cells was decreased further at high medium folinate. As mitochondria of AUXcoli-mcoli-1 cells accumulate more folate than WTT2 mitochondria under high medium folinate conditions, the low proportion of mitochondrial folate cannot be due to a limitation in mitochondrial transport, Folylpolyglutamate Synthetase  but appears to be due to the very high capacity of cytosolic FPGS in these cells to polyglutamylate folate and thus prevent its transport into mitochondria. Mitochondrial folate accumulation by AUX-mcoli-1 cells remained high at both low and high medium folinate, but the proportion did decrease between 5 and 24 h. Although total cellular folate accumulation by AUXmcoli-1 cells under low and high medium folinate conditions increased about 5-fold between 5 and 24 h, cytosolic folate accumulation increased 10-and %fold, respectively. These data again suggest that the cytosolic folate pool in these cells is derived from the efflux of mitochondrial folate.
Folate accumulation by AUX-coli is limited by transport rather than by FPGS activity because these cells express high levels of FPGS (61, and folate accumulation only becomes proportional to FPGS levels in AUX-coli transfectants that express less than 30% of the FPGS activity of the AUX-coli transfectant used in this study (2). Folate accumulation was reduced in Dl-IAIO cells (Table IV), an AUX-coli transfectant expressing about 10-fold less FPGS activity than the AUX-coli transfec-tant used in this study (2). If the AUX-mcoli transformants express any cytosolic FPGS activity, the maximum level, after allowing for mitochondrial contamination of the cytosolic extract (Table I), can be calculated to be less than 1% of that of AUX-coli and 10% of that of Dl-IAIO. However, under low and high medium folinate conditions, AUX-mcoli transfectants accumulated 70 to 80% as much cytosolic folate as AUX-coli and higher levels of cytosolic folate than were found in Dl-IAlO (Tables I1 and IV), again suggesting that the cytosolic folate pool in these cells was derived from mitochondrial folylpolyglutamates. GlyB, a glycine-requiring CHO mutant (12) that is defective in mitochondrial folate transport,' accumulated pharmacological levels of folate as effectively as WTTZ, despite the lack of mitochondrial folate metabolism (Table IV). However, FPGS levels in the mitochondria and cytosol of GlyB are about 2-fold elevated compared to WTT2 cells.'

Folylpolyglutamate Chain Length Specificity of One-carbon
Metabolism-The ability of pteroyltriglutamates to support metabolic cycles of one-carbon metabolism was assessed by culturing cells in medium lacking various products of one-carbon metabolism and determining the intracellular folate concentration that supported half-maximal growth rates. For most of the transfectants, the maximal growth rates obtained in the different media were slower than for WTT2 but similar to the parent A m 1 mutant (Table V). Maximal growth rates of AUX-mcoli-2, which was smaller than the other transfectants, were similar to WTT2. Table V shows medium PteGlu concentrations that supported half-maximal growth rates for the different cell lines under the different nutritional conditions. The data are typical of that found in two or three separate trials. The high folate requirement of AUX-coli in medium lacking glycine was reduced for the AUXcoli-mcoli transfectants and was reduced further in AUX-coli transfectants. All transfectants grew well in medium lacking thymidine or purines. A s shown previously, A U X B l cells did not grow in medium lacking glycine or purines, and growth in the absence of thymidine required very high medium PteGlu (Table V).
Subcellular folate concentrations that supported half-maximal growth rates for the different cell lines under the different nutritional conditions are shown in Table VI. The data were derived from growth requirement ( Table V) and cellular labeling experiments with PteGlu (similar to data shown in Fig. 4), and from subcellular distribution studies (Tables I1 and IV      support growth in medium lacking glycine was reduced in the in all the transfectants and in WTT2 cells, indicating that AUXcoli-mcoli transfectants and was reduced further in AUXpteroyltriglutamates function as effectively as the longer chain coli transfectants to levels that were similar to WTTB cells. length folylpolyglutamates in WTTX cells in the metabolic However, mitochondrial folate concentrations that supported cycles of glycine synthesis, provided the folate is in the mitohalf-maximal growth rates under these conditions were similar chondria. Similarly, no significant differences were observed in the cytosolic folate levels that supported thymidylate and purine synthesis between the various transfectants and WTT2 cells. AUX-mcoli-2 consistently required lower intracellular folate levels than AUX-mcoli-1 to support these metabolic cycles. However, the folate concentration in the cytosol and mitochondria ofAUX-mcoli-2 would be very similar to AUX-mcoli-1 due to the smaller size of AUX-mcoli-2. DISCUSSION Although it has been known for some time that mammalian cells and tissues contain mitochondrial and cytosolic folylpolyglutamate pools (7,12,(28)(29)(30), the role of mitochondrial folate metabolism in cellular one-carbon metabolism is poorly understood. In the current study, we describe model CHO cells which were developed for assessing the metabolic role of mitochondrial folate metabolism. The presence of mitochondrial FPGS activity in the transfectants demonstrated that the E. coli FPGS protein, targeted by the human ornithine transcarbamoylase leader sequence, could be successfully transported into the CHO cell mitochondria and be expressed as a n active enzyme. In the construct used, there are 5 additional amino acid residues between the normal cleavage site of the ornithine transcarbamoylase leader sequence and the start methionine of the mature FPGS protein. Although we did not test whether proteolytic processing occurred at the normal proOTC cleavage site, Western analysis demonstrated a mitochondrial band indistinguishable in size from mature FPGS as well as what appeared to be a processing intermediate. proOTC is normally processed via a two-step proteolytic cleavage (27). The active FPGS in mitochondria implies that the extra amino-terminal residues on the FPGS protein did not significantly affect enzyme activity, and that this prokaryotic protein seems to refold correctly after unfolding and translocation across the mitochondrial membrane. These transfectants, expressing E. coli FPGS activity with different ratios of mitochondrial to cytosolic activity, provide a potentially useful model for studying the role of mitochondrial folate metabolism in one-carbon metabolism.
It has been suggested, based on in vitro studies, that folatemediated reactions in the mitochondria may generate one-carbon units for utilization in cytosolic processes, and a novel pathway has been proposed for a one-carbon shuttle between the two compartments using serine and formate (31). GlyA, a CHO mutant which lacks mitochondrial serine hydroxymethyltransferase activity, is auxotrophic for glycine (8) although it is not clear why the cytosolic hydroxymethyltransferase, which normally accounts for about 25% of the total cell enzyme in CHO cells, cannot meet its glycine requirement. Our results also directly demonstrate this requirement for mitochondrial glycine synthesis in the CHO cell. AUX-coli cells, which lack mitochondrial folates despite possessing normal cytosolic folate pools, were auxotrophic for glycine when cultured in medium containing physiological levels of folate, while transfectants expressing mitochondrial FPGS activity and containing mitochondrial folate pools were prototrophic for glycine.
Although different levels of medium folate supported halfmaximal growth rates of the AUXcoli-mcoli and AUX-mcoli transfectants and WTT2 cells, and intracellular folate levels also varied under these conditions, mitochondrial folate concentrations in all these cells were essentially the same reflecting that this pool supplied folate cofactor for glycine synthesis and that pteroyltriglutamates were as effective a s longer polyglutamate derivatives in glycine synthesis. The glycine auxotrophy ofAUX-coli cells, but not AUXBl cells, could be overcome by very high medium folate concentrations. AUX-coli cells accumulate very high cytosolic pteroyltriglutamate levels under these conditions, and it is possible that this allows sufficient glycine synthesis to occur in the cytosol. However, although pteroyltriglutamates do not enter the mitochondria, pteroylmono-and possibly diglutamates would be expected to be transported but would not be concentrated by AUX-coli mitochondria due to the lack of FPGS activity. We have previously shown that AUX-coli mitochondria contain trace amounts of pteroylmonoand diglutamates, reflecting that a small proportion of the cytosolic folates in this cell line are also mono-and diglutamate derivatives (7). Under conditions of very high cytosolic folate levels, sufficient pteroyldiglutamate may be present in the mitochondria of AUX-coli cells to allow glycine synthesis. Halfmaximal growth rates of AUX-coli cells in the absence of glycine required about 100 pmol of folate/106 cells, and pteroyldiglutamate accounted for about 4 pmoV1O' cells. If 25% of this was in the mitochondria, this would be similar to the mitochondrial folate concentrations that supported half-maximal growth rates in the other transfectants. Very high levels of medium folate do not support growth of glyA cells under these conditions, which would also suggest that elevated cytosolic folate is not sufficient to allow glycine synthesis.
The ability of AUX-mcoli transfectants to support the cytosolic cycles of thymidylate and purine synthesis was unexpected as it was anticipated that these cells would lack cytosolic folates. However, these transfectants contained cytosolic folylpolyglutamates. Cytosolic folate concentrations that supported similar rates of thymidylate or purine synthesis were similar for all the transfectants and for WTT2 cells and, as has been previously reported (6), these metabolic cycles also appear to function effectively with pteroyltriglutamates.
We have previously shown that folate accumulation by AUXcoli, which expresses high levels of E. coli FPGS, is limited primarily by folate influx rather than by FPGS activity (6). AUXcoli-mcoli transfectants were selected by their ability to grow in the absence of glycine, and all of these transfectants, including the two which were characterized in detail in this report, expressed extremely high levels of mitochondrial FPGS activity although the proportion of cellular folate in the mitochondria of these cells was relatively low. This could not be due to a limitation in mitochondrial folate transport as, under similar culture conditions, AUX-mcoli transfectants accumulated higher absolute levels of mitochondrial folate despite possessing lower mitochondrial FPGS activities. Subcellular folate accumulation in AUXcoli-mcoli transfectants was governed by competition between cytosolic and mitochondrial FPGS, and very high levels of mitochondrial FPGS were required to enable mitochondrial folate accumulation and to overcome the ability of the cytosolic enzyme in these cells to trap all entering folate. Although FPGS is a very low abundance protein in cultured mammalian cells, these levels are normally sufficient to enable retention of all transported folate under physiological conditions (6). One metabolic reason for FPGS to be present at such low levels would be to allow mitochondrial accumulation of folate. Overexpression of human FPGS in the cytosol of mammalian cells results in an inability to accumulate mitochondrial f01ate.~ AUX-mcoli transfectants appeared to lack cytosolic FPGS activity although the possibility of trace levels of cytosolic enzyme activity could not be excluded. No accumulation of a precursor protein was detected in the cytosol, and the half-life of most mitochondrial precursor proteins is very short (32).
Although a large proportion, in some cases the majority, of cellular folate was located in the mitochondria of these cells, they also contained a normal cytosolic folylpolyglutamate pool. Various lines of evidence indicated that the cytosolic folate pool was AUXcoli-mcoli cells accumulated pharmacological levels of folinate as effectively as AUX-coli cells, which is consistent with cellular folate influx being rate-limiting for folate accumulation by these cells. AUX-mcoli cells were at least as effective at accumulating pharmacological levels of folinate, indicating that mitochondrial transport of folate was not limiting folate accumulation under these conditions. All transfectants accumulated much higher levels of folinate than did WTT2 cells, reflecting that FPGS activity is rate-limiting for accumulation of high levels of folinate in cells expressing mammalian FPGS activity (6). Pharmacological levels of folinate also reduced the proportion of cellular folate in the mitochondria of all transfectants although the absolute level of mitochondrial folate was high. This reduced proportion may reflect increased efflux under these conditions. ing the EcoRV site in the folC gene and Drs. Arthur Horwich (Yale) and