5-Formyltetrahydrofolate Is an Inhibitory but Well Tolerated Metabolite in Arabidopsis Leaves*

, 5-Formyltetrahydrofolate (5-CHO-THF) is formed via a second catalytic activity of serine hydroxymethyltransferase (SHMT) and strongly inhibits SHMT and other folate-dependent enzymes in vitro . The only enzyme known to metabolize 5-CHO-THF is 5-CHO-THF cycloligase (5-FCL), which catalyzes its conversion to 5,10-methenyltetrahydrofolate. Because 5-FCL is mitochondrial in plants and mitochondrial SHMT is central to photorespiration, we examined the impact of an insertional mutation in the Arabidopsis 5-FCL gene (At5g13050) under photorespiratory (30 and 370 (cid:1) mol of CO 2 mol (cid:2) 1 ) and non-photorespiratory (3200 (cid:1) mol of CO 2 mol (cid:2) 1 ) conditions. The mutation had only mild visible effects at 370 (cid:1) mol of CO 2 mol (cid:2) 1 , reducing growth rate by (cid:1) 20% and delaying flowering by 1 week. However, the mutation doubled leaf 5-CHO-THF level under all TES, 2-{[2-hydroxy-1,1- bis(hydroxymethyl)ethyl]amino}ethanesulfonic acid; CHES, 2-(cyclo-hexylamino)ethanesulfonic acid; Tricine, N -[2-hydroxy-1,1-bis(hydroxy- methyl)ethyl]glycine. mutant plants (5 weeks old) were washed free of soil, and their root systems were severed under water, leaving (cid:1) 0.5 cm of the main root. The plants were placed in Petri dishes containing 0 or 10 m M 5-CHO-THF in 0.5 (cid:6) Hoagland’s nutrient solution and incubated in light (75 (cid:2) E m (cid:3) 2 s (cid:3) 1 ) at 22 °C for 24 h. The plants were then washed, frozen in liquid N 2 , and lyophilized. Samples (100–150 mg) were taken for GC-MS analysis as above.


5-Formyltetrahydrofolate (5-CHO-THF)
is formed from 5,10-methenyltetrahydrofolate (5,10-CHϭTHF) by a hydrolytic reaction catalyzed by serine hydroxymethyltransferase (SHMT) in the presence of glycine (1,2). Spontaneous chemical hydrolysis of 5,10-CHϭTHF may be a minor additional source (3). 5-CHO-THF is the most stable natural folate and the most enigmatic, for it is the only one that does not serve as a cofactor in one-carbon metabolism. Instead, 5-CHO-THF is a potent inhibitor of SHMT and most other folate-dependent enzymes in vitro (4,5). 5-CHO-THF probably acts as a stable storage form of folate in seeds and fungal spores (5-7), but it is not clear what role, if any, it plays in metabolically active tissues (8). This question is particularly pertinent for leaves. Leaf mitochondria have very high levels of SHMT and, during photorespiration, receive a massive influx of glycine (which leads to a matching SHMT-mediated glycine 3 serine flux) (9). Conditions in leaf mitochondria therefore favor 5-CHO-THF formation (Fig. 1). Indeed, 5-CHO-THF can comprise 50% of the folate pool in leaf mitochondria (10,11), which is far more than in mammalian mitochondria (12)(13)(14). Furthermore, 5-CHO-THF is reported to make up 14 -40% of the folate pool in leaves and other metabolically active plant organs (10, 15), a much higher proportion than the 3-10% typical of mammals and yeast (2,16). 5-Formyltetrahydrofolate cycloligase, EC 6.3.3.2 (5-FCL, also known as 5,10-methenyltetrahydrofolate synthetase), is the only enzyme known to recycle 5-CHO-THF to a metabolically active form, which it achieves by catalyzing irreversible, ATP-dependent conversion to 5,10-CHϭTHF (2,5). This enzyme is also something of an enigma. For one thing, despite the inhibitory effects of its substrate, 5-FCL is not essential in yeast: 5-FCL disruptants had 4-fold more 5-CHO-THF but no other new phenotype (2). For another, phylogenomic profiling (17,18) indicates that some bacteria lack 5-FCL even though they have SHMT. Lastly, 5-FCL overexpression in human cells lowered the folate level and raised folate turnover rate, suggesting that 5-FCL may have a second function as a folatedegrading enzyme (19).
Another intriguing feature of 5-FCL is that its subcellular location differs among eukaryotes. Whereas the enzyme is largely if not solely cytosolic in yeast and mammals (2,16,20,21), it is exclusively mitochondrial in plants (4). Taken with the key role of mitochondrial SHMT in photorespiration, with the inhibition of SHMT by 5-CHO-THF, and with the presence of 5-CHO-THF in leaf mitochondria, the location of plant 5-FCL implies (i) that this enzyme governs mitochondrial 5-CHO-THF levels and (ii) that, via its effect on 5-CHO-THF levels, 5-FCL could regulate the in vivo activities of SHMT and other folatelinked mitochondrial enzymes (Fig. 1).
To investigate such regulatory functions for 5-FCL and 5-CHO-THF, we identified and characterized an Arabidopsis 5-FCL insertional mutant. This mutant had greatly elevated mitochondrial 5-CHO-THF levels and hyperaccumulated glycine under photorespiratory conditions. However, the mutant grew almost normally in ambient air, indicating that folate-dependent metabolic reactions in plants are substantially tolerant to 5-CHO-THF.
Plants and Growth Conditions-Arabidopsis thaliana plants were grown at 22-28°C in 12-h days (photosynthetic photon flux density 80 E m Ϫ2 s Ϫ1 ) in potting soil irrigated with water. Material was lyophilized to determine dry weight. In experiments at various CO 2 levels, plants were grown for 4 weeks in ambient air (ϳ370 mol of CO 2 mol Ϫ1 ), then either kept in ambient air or transferred to high CO 2 (3200 Ϯ 40 mol mol Ϫ1 ) or low CO 2 (30 Ϯ 20 mol mol Ϫ1 ) conditions for 5 days. CO 2 levels were monitored with a Vernier CO 2 sensor (Vernier Software and Technology, Beaverton, OR).
Arabidopsis Mutant-A 5-FCL mutant (28D07) was identified in the Syngenta T-DNA insertion collection (ecotype Columbia) (23). Segregants, wild type or homozygous for the mutation, were identified by PCR using gene-specific primers located 5Ј or 3Ј of the T-DNA insertion (5Ј-CTGAAGTGAGTGGCAACTACA-3Ј and 5Ј-GTCTCACTTCTTCTC-TTACCTT-3Ј, respectively) and the T-DNA-specific primer 5Ј-GCATCT-GAATTTCATAACCAATC-3Ј. DNA was extracted by the "Shorty" protocol available on the website of the University of Wisconsin Biotechnology Center. The insertion site was confirmed by sequencing.
Amino Acid Analysis-Leaf tissue (ϳ160 mg) was frozen in liquid N 2 , lyophilized, weighed, and pulverized. The resulting powder was extracted by shaking with 0.5 ml each of water and CHCl 3 and was then stored at Ϫ20°C for 24 h before centrifugation. Ribitol and ␥-aminobutyric acid were added as internal standards. For HPLC, 20 l of the aqueous phase was derivatized with AccQ⅐Fuor TM reagent (6-aminoquinolyl-N-hydroxysuccinimidylcarbamate; Waters, Milford, MA) in a final volume of 100 l, and a 20-l aliquot was analyzed by HPLCfluorescence according to Waters' recommendations. For GC-MS, aqueous phase aliquots equivalent to 0.625 mg dry weight were dried under N 2 and methoximated and trimethylsilylated in pyridine (final volume 50 l) as described (28). One l of the derivatized mixture was injected (pulsed splitless injection, Agilent 6890 series autoinjector, Agilent Technologies, Palo Alto, CA) onto a 60-m DB-5MS column (J&W Scientific, Palo Alto, CA). GC-MS analysis was performed with an Agilent 6890 gas chromatograph and a 5973 series Agilent quadrupole mass spectrometer as described (28).
Isolation of Mitochondria and SHMT Assays-Mitochondria were prepared from 15-30 g of leaves from 4-week-old Arabidopsis plants as described (29), with the following modifications. The pellet containing chloroplasts and mitochondria was suspended in 2 ml of a solution containing 20 mM TES-NaOH, pH 7.2, 0.25 M sucrose, 1 mM EDTA, 2 mM MgCl 2 , 0.1% bovine serum albumin, 14 mM ␤-mercaptoethanol, applied to a step gradient composed of 2.5 ml of 21%, 5.5 ml of 26%, and 3 ml of 47% (v/v) Percoll, and centrifuged at 65,000 ϫ g for 45 min in a swinging bucket rotor. Mitochondria were recovered from the 26 -47% Percoll interface, diluted 12-fold in 10 mM Tricine-NaOH, pH 8.0, 1 mM EDTA, 14 mM ␤-mercaptoethanol, and 0.25 M sucrose, and then centrifuged at 12,500 ϫ g for 20 min. This step was repeated twice. The final mitochondrial pellet was suspended in 300 l of 10 mM Tricine-NaOH, pH 8.0, 1 mM EDTA and stored under N 2 at Ϫ80°C until analysis. SHMT was assayed in mitochondrial extracts as described (4).
5-CHO-THF Feeding-Three wild type and three mutant plants (5 weeks old) were washed free of soil, and their root systems were severed under water, leaving ϳ0.5 cm of the main root. The plants were placed in Petri dishes containing 0 or 10 mM 5-CHO-THF in 0.5ϫ Hoagland's nutrient solution and incubated in light (75 E m Ϫ2 s Ϫ1 ) at 22°C for 24 h. The plants were then washed, frozen in liquid N 2 , and lyophilized. Samples (100 -150 mg) were taken for GC-MS analysis as above. In photorespiration, glycine coming from peroxisomes is converted to serine in the mitochondria by the concerted action of the glycine decarboxylase complex (GDC) and SHMT. SHMT also mediates formation of 5-CHO-THF from 5,10-CHϭTHF; the reverse reaction is catalyzed by 5-FCL, which is solely mitochondrial in plants (4). Possible SHMT inhibition by 5-CHO-THF is shown by a dashed line. Plant mitochondria also contain an isoform of the bifunctional 5,10-methylene-THF dehydrogenase/5,10-CHϭTHF cyclohydrolase that interconverts 5,10-CH 2 -THF, 5,10-CHϭTHF, and 10-CHO-THF and an isoform of 10-formyl-THF synthetase; there are also extramitochondrial isoforms of both enzymes (15).

5-Formyltetrahydrofolate in Arabidopsis
confirmed the presence of an insert close to the 3Ј end of the sixth intron, which is located within the protein-coding part of the gene (Fig. 2A). Plants homozygous for the mutation and their wild type siblings were identified by PCR and subjected to Southern analysis using a T-DNA sequence (a fragment of the bar gene) as probe (Fig. 2B). Only the mutant plants gave hybridizing bands, establishing that the T-DNA is inserted only at the 5-FCL locus. The multiple banding pattern in Fig.  2B indicates that several concatenated T-DNA copies are present at this locus. Northern analysis of leaf RNA showed no detectable 5-FCL transcript in the mutant (Fig. 2C), indicating a knock-out mutation. The homozygous mutants and their wild type siblings were therefore propagated for further work. When grown in soil at ambient levels of CO 2 (ϳ370 mol of CO 2 mol Ϫ1 ), visible differences between mutant plants and their wild type siblings were modest. The growth rate of mutant plants was ϳ20% lower (Fig. 2D), and they showed a flowering delay of about 1 week (Fig. 2E). There was no difference in leaf color or form (Fig. 2E).
Folates in Leaves Exposed to Various CO 2 Concentrations-Plants grown for 4 weeks in ambient air were transferred for 5 days to air containing 30 or 3200 mol of CO 2 mol Ϫ1 or kept in ambient air. The lower CO 2 concentration, which is beneath the CO 2 compensation point, stimulates photorespiration (and hence the glycine 3 serine flux rate in mitochondria) whereas the higher one suppresses it (32). CO 2 concentration had rather little effect on folate profiles so that the divergences between wild type and mutant leaves were generally similar in all three atmospheres (Fig. 3). As might be expected, 5-CHO-THF levels were higher in the mutant (2.1-to 2.6-fold, significant at p Ͻ 0.05). Less expectedly, 10-CHO-/5,10-CHϭTHF levels were also much higher in mutant leaves exposed to 370 and 30 mol of CO 2 mol Ϫ1 . Added together, these differences made the total folate content of the mutant significantly higher at the two lower CO 2 levels.
Mitochondrial Folates-Because mitochondria in photorespiring leaves are expected to be the main site of 5-CHO-THF formation and the only site of its removal by 5-FCL (Fig. 1), we investigated the impact of the 5-FCL mutation on mitochondrial folate levels of plants grown in ambient air. Four separate mitochondrial preparations were made from wild type or mutant plants (Fig. 4). Despite some variability among the preparations, mitochondria from the mutant clearly showed massive 5-CHO-THF accumulation relative to wild type (on average 8-fold, significant at p Ͻ 0.01) and a 120% increase in total folate. The mitochondrial 5-CHO-THF content rose from a mean value of 15% of total mitochondrial folate in the wild type to a mean of 72% in the mutant. The mitochondrial contents of 5-CH 3 -THF and 10-CHO-/5,10-CHϭTHF did not change significantly in the mutant, nor did that of 10-formyldihydrofolate (which forms readily from 10-CHO-THF in isolated mitochon-
SHMT Activity in Mitochondria-The intramitochondrial 5-CHO-THF buildup in mutant plants, coupled with their fairly normal growth, led us to measure mitochondrial SHMT activities to check for a possible compensatory increase in the mutant. No such increase was found; SHMT activities in extracts of wild type and mutant mitochondria were 340 Ϯ 32 and 336 Ϯ 41 nmol min Ϫ1 mg Ϫ1 of protein, respectively (means Ϯ S. E., n ϭ 3).
Free Amino Acids in Leaves Exposed to Various CO 2 Concentrations-To establish whether the mitochondrial accumula-tion of 5-CHO-THF inhibits SHMT in vivo, we compared the glycine and serine contents of the leaves of wild type and mutant plants exposed for 5 days to 30, 370, or 3200 mol of CO 2 mol Ϫ1 (Fig. 5). As CO 2 concentration declined, wild type leaves showed small increases in glycine whereas mutant leaves showed much larger ones, so that the glycine content of mutant leaves was 19-fold higher than the wild type at 370 mol of CO 2 mol Ϫ1 and 46-fold higher at 30 mol mol Ϫ1 . There was also a small accumulation of serine in the mutant relative to the wild type (1.4-fold at 370 mol of CO 2 mol Ϫ1 , 1.8-fold at 30 mol mol Ϫ1 , both significant at p Ͻ 0.01).
Besides serine, amino acids such as glutamate and alanine can act as amino donors in the formation of glycine from glyoxylate in the photorespiratory pathway (34). The levels of glutamate, alanine, and also aspartate were substantially reduced in the mutant compared with wild type at 30 mol of CO 2 mol Ϫ1 (Table I) but not at the other CO 2 concentrations (not shown).
Effect of Supplied 5-CHO-THF on Glycine and Serine Content-Because exogenous 5-CHO-THF is taken up by Arabidopsis and enters mitochondria (35), we examined the effect of feeding 5-CHO-THF via the transpiration stream to illuminated wild type and mutant plants. Leaf glycine and serine levels were measured after 24 h of continuous light (Fig. 6). Consistent with in vivo inhibition of SHMT, 5-CHO-THF feeding raised glycine levels in wild type and mutant plants significantly (p Ͻ 0.05). The increase in glycine was larger in the mutant (9.4 versus 2.5 mol g Ϫ1 dry weight). DISCUSSION Our data demonstrate that 5-CHO-THF can inhibit the activity of mitochondrial SHMT in vivo and thus provide support for the view that 5-CHO-THF regulates one-carbon metabolism (5). Despite its important implications, there is rather little evidence for or against this view, and it remains controversial (8). However the most striking aspect of our findings is that 5-FCL ablation and the ensuing 5-CHO-THF buildup in mitochondria had so little impact on plant performance. A priori, this impact seemed likely to be devastating, above all to photorespiring leaves (4). We therefore conclude that plants are surprisingly tolerant of 5-CHO-THF, as presaged by earlier reports of remarkably high levels of this folate in leaf mitochondria (10,11). Several mechanisms could contribute to this tolerance.
The first may be the ability of an expanded glycine pool to offset inhibition of SHMT by 5-CHO-THF. Assuming a matrix volume/protein ratio of 1-2 l mg Ϫ1 (36), the concentration of 5-CHO-THF in leaf mitochondria can be estimated from the data in Fig. 4 to be ϳ0.25-0.5 mM in the wild type and ϳ2-4 FIG. 3. Folate profiles of wild type and mutant leaves at three CO 2 levels. Arabidopsis plants were grown in ambient air and transferred for 5 days to 30 or 3200 mol of CO 2 mol Ϫ1 or kept in ambient air (370 mol of CO 2 mol Ϫ1 ) before analysis. Data are means Ϯ S.E. for three independent leaf samples. Folic acid was detected in some samples but never exceeded 1% of total folates; these data have been omitted for clarity. FIG. 5. Glycine and serine levels of wild type and mutant leaves at three CO 2 levels. Arabidopsis plants were grown in ambient air and transferred for 5 days to 30 or 3200 mol of CO 2 mol Ϫ1 or kept in ambient air (370 mol of CO 2 mol Ϫ1 ) before extraction and AccQ⅐Fluor TM HPLC amino acid analysis. Data are means Ϯ S.E. for three independent leaf samples, which were from the same type of experiment as Fig. 3.

5-Formyltetrahydrofolate in Arabidopsis
mM in the mutant. Because 5-CHO-THF levels as low as 0.05-0.1 mM significantly reduce mitochondrial SHMT activity in vitro (4), the levels reached in the wild type would be expected to be inhibitory and those in the mutant even more so. In line with this prediction, the 5-FCL mutant accumulated more glycine than the wild type. The higher glycine level presumably compensated for the lower SHMT activity and drove a near normal flux through the vital glycine 3 serine reaction of photorespiration because, had it not done so, the 5-FCL mutation would have been lethal in ambient air, as are glycine decarboxylase or mitochondrial SHMT mutations (37,38). We were able to rule out the possibility that SHMT activity increased in the 5-FCL mutant.
Another facet of tolerance to 5-CHO-THF may be its sequestration in mitochondria. Assuming that the soluble protein content of leaves is ϳ10 mg g Ϫ1 fresh weight and that mitochondrial protein is ϳ4% of the total (39), it can be estimated from the data of Figs. 3 and 4 that the mitochondria contain ϳ50% of the total 5-CHO-THF in wild type leaves and Ն80% in mutant leaves. The basis for intramitochondrial retention of 5-CHO-THF may be that it is polyglutamylated, as are the bulk of mitochondrial folates (40), and that mitochondrial folate transporters prefer monoglutamates, as do most other folate carriers (41). Keeping much of the 5-CHO-THF inside mitochondria clearly reduces its potential to interfere with cytosolic or plastidial isoforms of folate enzymes. However, this is only a partial solution because plant mitochondria themselves contain various folate-dependent enzymes (42,43) that, in other organisms, are sensitive to 5-CHO-THF (44). These include dihydrofolate reductase and thymidylate synthase and isoforms of methionyl-tRNA transformylase, 5,10-CHϭTHF cyclohydrolase, 5,10-methylene-THF dehydrogenase, and the purine synthesis enzyme aminoimidazole-4-carboxamide-1-␤-D-riboside transformylase.
It is therefore hard to escape the conclusion that another mechanism of plant tolerance to 5-CHO-THF is either relative insensitivity of mitochondrial enzymes to this compound or metabolic plasticity in the form of flux redistribution from mitochondrial to extramitochondrial isoforms, where the latter exist. Some support for flux redistribution comes from the accumulation of 10-CHO-/5,10-CHϭTHF in photorespiring mutant plants. This must have been outside mitochondria as the mitochondrial 10-CHO-/5,10-CHϭTHF pool did not change. Perhaps the extramitochondrial pool expands because of interference with mitochondrial one-carbon metabolism and then drives compensating fluxes through non-mitochondrial reactions. Although 10-CHO-THF and 5,10-CHϭTHF could not be distinguished analytically, it seems probable that 10-CHO-THF predominates in vivo because the chemical equilibrium between the two forms lies strongly (ϳ95%) toward 10-CHO-THF at physiological pH (30).
Apparent insensitivity could be due, in part at least, to the sheer abundance of mitochondrial folate-dependent enzymes: SHMT and glycine decarboxylase alone constitute up to 40% of soluble mitochondrial protein (45). At such high abundances, it can be calculated from the data of Fig. 4 that even in mutant mitochondria, containing ϳ4 nmol of 5-CHO-THF mg Ϫ1 of protein, folate binding sites probably still outnumber 5-CHO-THF molecules. Furthermore, binding of 5-CHO-THF to SHMT and glycine decarboxylase may decrease the inhibition of other folate-dependent enzymes in mitochondria.
Because 5-CHO-THF is quite chemically stable and 5-FCL is the only enzyme known to metabolize it, the action of SHMT in a 5-FCL mutant would in the long run be expected to convert much of the cellular folate pool to 5-CHO-THF, especially under photorespiratory conditions. That the total leaf 5-CHO-THF pool increased no more than 2.6-fold (to a maximum of 31% of total folate) suggests that there may be another, unknown, way to metabolize 5-CHO-THF, or in effect to detoxify it, and that this contributes to tolerance. The same may hold true of yeast, where the 5-CHO-THF level increased only 4-fold in a 5-FCL knock-out (2), and of those bacteria that apparently lack 5-FCL genes (17,18). Unlike the 5-FCL reaction, which salvages 5-CHO-THF as an intact folate, any alternative detoxification route seems likely to entail cleavage of the p-aminobenzoate-glutamate bond or the pteridine-p-aminobenzoate bond (2,46). In support of the latter possibility, exploratory work in our laboratory has shown a large (ϳ15-fold) accumulation of a novel pteridine in the leaves of 5-FCL mutant plants.
Finally, we saw no evidence that 5-FCL moonlights as a folatedegrading enzyme in plants, as has been suggested for the mammalian enzyme (19). Although total folate content indeed increased in the 5-FCL mutant, it did so only under photorespiratory conditions, suggesting that the increase was not due to the mutation per se but rather to its metabolic sequelae.