Folate-dependent one-carbon metabolism is a vital function in all living cells. The one-carbon units generated in this pathway are essential for the synthesis of the purines and thymidylate and for the de novo synthesis of methyl groups. The interaction of the anticonvulsant phenytoin with folate has been a cause for concern for some time. Phenytoin administered at therapeutic doses has been shown to deplete plasma folate concentrations in humans (Pisciotta 1982) and liver folate concentrations in rats (Carl and Smith 1983). Chronic use of phenytoin can result in sufficient depletion of folate to cause megaloblastic anemia (Pisciotta 1982). Folate, on the other hand, administered at pharmacological doses, has been blamed for a decrease in the concentrations of phenytoin in the brains of treated rats (Smith and Carl 1981, Smith and Obbens 1979).

While the details of these interactions remain obscure, several potential mechanisms have been proposed as explanations. One of the first hypotheses suggested that the phenytoin was increasing the pH of the small intestine and inhibiting the intestinal conjugase activity (Hoffbrand and Necheles 1968). However, phenytoin causes folate depletion in rats on defined diets that do not contain folylpolyglutamates (Carl and Smith 1983) indicating that phenytoin-induced intestinal conjugase inhibition, if it occurs at all, is certainly not the only reason for folate depletion. The phenytoin-generated increase in gut pH may, however, decrease the driving force of the proton pump which supplies the energy for at least one folate transporter in the gut (Schron 1991). Other hypotheses include direct competition between folate and phenytoin for uptake sites (Rosenberg et al. 1979), inhibition of folate interconverting enzymes by phenytoin (Carl and Smith 1983), increased catabolism of folates by phenytoin induction of folate catabolic enzymes (Chanarin 1979), and inhibition of central appetite centers by phenytoin, decreasing food intake and thereby leading to decreased tissue folate concentrations (Hoppner and Lampi 1989). In addition, chronic treatment of rodents with phenytoin has been shown to affect the hepatic activity of methylenetetrahydrofolate reductase (Billings 1984, Carl and Smith 1983). Whether this effect is a direct effect of phenytoin on the enzyme or is an effect secondary to the phenytoin effect on folates is unknown.

To differentiate among these possibilities we treated rats with phenytoin for up to 8 wk and determined the composition of the folate pool in mucosa, liver, bile and brain at different times during the treatment. Using our adaptation (Carl and Smith 1995) of the method developed by Priest and co-workers (see Carl and Smith 1995 for review) we were able to examine the composition of the folate pools in these tissues with regard to both the pteridine form and the polyglutamate form.

METHODS

A portion of liver (~0.9 g) was homogenized in nine volumes of homogenizing buffer and subjected to subcellular fractionation. The homogenate was centrifuged at 1000 × g for 10 min at 2°C. The supernatant was layered onto a gradient (2.0 mL each of 5, 7.5 and 10 g Ficoll in 100 mL homogenizing buffer) in a SW41 centrifuge tube. The gradient was centrifuged at 100,000 × g for one hour in a SW41 rotor at 4°C. The top 2.0 mL of the supernatant from the gradient was transferred to a clean tube. This fraction was considered cytoplasm. The gradient layers were decanted and discarded and the tube was wiped out with a Kimwipe before the mitochondrial pellet was resuspended in 1.0 mL of cold homogenizing buffer. Half of the cytoplasm was diluted with an equal volume of heating buffer and heated at 90°C for 10 min. Half of the mitochondrial suspension was diluted with an equal volume of heating buffer and heated at 90°C for 10 min.

The section of small intestine was flushed with several volumes of cold saline solution. The mucosal cells were scraped from the intestinal muscle sheath using a microscope slide against a chilled glass plate. Fat and connective tissue were removed from the mucosa, and were then collected into a centrifuge tube and centrifuged at 400 × g for 5 min to collect the cells from the excess saline solution. The saline supernatant was discarded, and the mucosal cells were weighed and homogenized in four volumes of hot heating buffer (as described previously, Carl and Smith 1995).

Heated samples were chilled on ice and then centrifuged (1000 × g, 10 min, 4°C) to remove precipitated protein. The supernatants were transferred to plastic snap cap tubes, capped tightly and stored at 70°C until folates could be assayed.

The effect of treatment time on the concentration of each of the folates was evaluated by linear regression analysis using the individual folate concentrations and the days of treatment with phenytoin as the dependent and independent variables, respectively. The effect of time of phenytoin treatment on weight gain and food consumption was also evaluated by linear regression analysis using the treatment time as the independent variable.

RESULTS

In contrast to the minimal effects of phenytoin on folates in the intestinal mucosa, the effect of phenytoin on liver folate was much greater, depleting total liver folate to 50% of control by 10 d after initiation of treatment (Fig. 1). After 10 d the liver folate concentrations seemed to rebound slightly and then level off at a concentration lower than control. All four of the different pteridine fractions measured U, D, F and M showed generally the same pattern (Figs. 2-5). However, the D fraction and the F fraction appeared to decline more quickly reaching their lowest concentrations within three days (Figs. 3 and 4) while it took the U fraction and the M fraction 10 d to reach their nadir (Figs. 2 and 5). The pentaglutamate of each of the different pteridine groups declined by a greater percentage than did other glutamate derivatives (Figs. 2-5). The di- and tri-glutamates in the liver were present in very small concentrations with the mono- and tetra-glutamates being present in somewhat higher concentrations. To make the graphs easier to read we combined the short chain (n = 1-4) folates as one point and we combined the hexa- and hepta-glutamates as one point. Heptaglutamates were a greater proportion of the M fraction in liver than they were of the other fractions, but that has been previously reported (Carl et al. 1995). It is interesting that the monoglutamate of the U fraction exhibited a significant positive correlation (r = 0.333, P = 0.02) with the time of phenytoin treatment (Fig. 6).

Surprisingly, when we fractionated the liver into subcellular fractions and measured the folate compositions of the mitochondrial and cytoplasmic fractions, we found that the U group folate was the only folate to exhibit a significant response to phenytoin treatment (data not shown). The cytoplasmic U folate had decreased significantly by day three of treatment but then gradually returned to normal. However, the mitochondrial U folate decreased precipitiously in the first 3 d of treatment and the concentrations remained low throughout the eight weeks of treatment. The cytoplasmic U folate ranged from 18 to 40 pmol folate/mg protein, and the mitochondrial U folate ranged from 20 to 50 pmol folate/mg protein. Again the pentaglutamate showed a greater proportionate decline than the other glutamate derivatives (data not shown). Neither the M group in the cytoplasm (total M_cyto = 95.9 ± 2.8 pmol folate/mg protein), where nearly all M group folate is found, nor the F group in the mitochondria (total F_mito = 29.3 ± 1.6 pmol folate/mg protein), where most of the F group folate is found (total F_cyto = 22.9 ± 1.5 pmol folate/mg protein), showed a significant response to the phenytoin treatment. The D group folate concentration was higher in cytoplasm (total D_cyto = 4.6 ± 1.0 pmol folate/mg protein) than in mitochondria (total D_mito = 2.4 ± 0.4 pmol folate/mg protein) but was relatively low in both subcellular fractions (less than 5% of the total folate) and showed no measurable response to phenytoin treatment (data not shown).

The purity of the cytoplasmic and mitochondrial fractions were monitored using the marker enzymes lactate dehydrogenase for cytoplasm and cytochrome oxidase for mitochondria. The ratio of the activity of lactate dehydrogenase in the cytoplasmic fraction of the liver of a rat to the activity of lactate dehydrogenase in the mitochondrial fraction of the liver of the same rat ranged from 600 to 2200 for the 48 rats used in the present study, indicating that there was little contamination of cytoplasm in the mitochondrial fraction. The ratio of the activity of cytochrome oxidase in the mitochondrial fraction of the liver of a rat to the activity of cytochrome oxidase in the cytoplasmic fraction of the liver of the same rat ranged from 14 to 4000. Since the ratio of the activities of the mitochondrial marker (cytochrome oxidase) in the mitochondrial fraction compared to the cytoplasmic fraction ranges as low as 14, then it would appear that mitochondrial contamination in cytoplasmic fractions was a greater potential source of error than cytoplasmic contamination in mitochondrial fractions. However, even at its worst the contamination would be less than 7%, and in most cases much less than 7%. Mitochondrial contamination in cytoplasmic fractions also was considerably more variable than cytoplasmic contamination in mitochondrial fractions. Moreover, because the cytoplasmic folate concentration was about twice the mitochondrial folate concentration, the worst contamination at a ratio of 1:14 (mitochondrial contamination in cytoplasm) would have to be halved so that the worst cross contamination of mitochondria in cytoplasm would have to be considered 1 part in 28 or about 3.5%.

Phenytoin treatment had no significant effect in bile on the U fraction folates (Total U folate = 0.830 ± 0.041 µmol folate/L), the D fraction folates (Total D folate = 0.131 ± 0.020 µmol folate/L) or the F fraction folates (Total F folate = 1.122 ± 0.080 µmol folate/L) individually (data not shown). However, there was a significant but transient increase in the mono- and di-glutamate derivatives of the M fraction folates (Fig. 7), with these derivatives returning to normal after 10 d of treatment. In addition, the distribution of glutamate derivatives in bile was different for the M fraction than for the other pteridine derivatives (U, D and F fractions) of the folate pool (Fig. 8). The concentrations of the diglutamate and triglutamate in the M fraction were in the same range as the concentration of the monoglutamate, while the concentrations of the monoglutamate of the other pteridine derivatives (fractions U, D and F) were much higher than the concentrations of the di- and tri-glutamates of the same pteridine fraction. In fact the change in the mono- and di-glutamates of the M fraction with phenytoin treatment caused the M fraction to temporarily look more like the other fractions. However, with continued treatment the M fraction returned to its normal distribution showing a distribution after 8 wk almost identical to the control (Fig. 7). The distribution of the different pteridine derivatives in bile of the untreated rat was 30% U, 8% D, 36% F and 26% M. 

Analysis of changes in brain folate pools in response to phenytoin treatment showed that the glutamate chain lengths of the U forms of folate increased with time of treatment (Fig. 9). Specifically, the concentrations of the short chain (1-4 glutamates) U folates showed a significant negative correlation with time of treatment (r = 0.605, P = 0.00001), while the concentrations of the long chain (4-7 glutamates) U folates showed a significant positive correlation with time of treatment (r = 0.479, P = 0.00086). The correlations were significant whether combined into the groups as shown (Fig. 9) or calculated for the individual chain lengths. The D, F and M folate fractions did not appear to be affected by phenytoin treatment either in the distribution of the glutamate chain length or in the concentration of the pteridine moiety (data not shown). However, the variability in the measurement for D, F and M fractions is greater because of the method of estimation. Therefore greater differences would have to be present in the D, F and M fractions than in the U fraction in order to be detected statistically by the method used.

The phenytoin containing diet used here was identical to the control diet used in an earlier study (Critchfield et al. 1993). The mean plasma concentration of phenytoin in that study was 11.3 mg/L (41.2 µmol/L) more than double the plasma concentration of 4.4 mg/L (16 µmol/L) of phenytoin in the rat model for continuously protective, non-toxic treatment of rats with phenytoin by gavage (Carl and Smith 1983). In the study by Critchfield et al. (1993) no toxicity in terms of weight gain or food consumption was observed and, since the same procedure was used here, no attempt was made to monitor the plasma phenytoin.

For the present experiment, however, linear regression analysis of weight gain of the rats versus time of treatment yielded a correlation coefficient of 0.6782 which was nearly significant (P = 0.0645). And, there was a nearly significant negative correlation between food consumption and time of treatment with phenytoin (correlation coefficient = 0.6767; P = 0.0653) with the group treated for 8 wk with phenytoin consuming an average of 89% of the food consumed by the untreated group. While the treatment appears to cause a minor decrease in food consumption, it is unlikely that the differences seen in hepatic folate concentrations are due simply to the apparently decreased food consumption.

DISCUSSION

In the liver, folate was depleted from each of the four pteridine fractions, U, D, F and M, in a similar pattern indicating that the mechanism leading to phenytoin-induced folate depletion was nonspecific, i.e. all folates were affected rather than just specific folates. However, there appeared to be minor differences between the fractions, and the pentaglutamates of all fractions were affected to a greater extent than folates of other chain lengths, confirming an earlier observation using a different methodology (Carl et al. 1991). The F and D fractions declined faster than the U and M fractions, but the F fraction did not decline as much as the other fractions. The observation that the F fraction was somewhat more resistant to folate depletion than the other fractions may be related to the previous observation that the primary localization of F folates is in mitochondria (Cook and Blair 1979, Carl et al. 1995) which may be more resistant to folate depletion than the cytoplasm. The more rapid depletion of the F and D folate forms from hepatic cells suggests that these forms are more expendable to the cell than are the U and M forms, a conclusion supported by the fact that F and D folates are involved in the synthesis of nucleic acids which are not as essential in nondividing tissues such as liver. But the differences among the folate groups in the rates of depletion and the preference for pentaglutamates in the depletion process may be the reaction of the cell to decreased levels of folate. A clue to the cause may lie in the increase in the concentration of U folylmonoglutamates. Such an increase is consistent with a decrease in the synthesis of polyglutamate derivatives of the folates.

A decreased hepatic synthesis of polyglutamate derivatives of the folates might also explain the transient appearance of increased concentrations of M folylmono- and di-glutamates in the bile. A decreased synthesis of polyglutamate derivatives may be caused either by direct inhibition of folylpolyglutamate synthetase (FPGS), the enzyme responsible for adding glutamates to the folates, or by the inhibition of the demethylation of the 5CH₃H₄PteGlu₁. 5CH₃H₄PteGlu₁ is the primary folate source to cells, and it is a poor substrate for the FPGS (Cichowicz and Shane 1987). The product of demethylation of 5CH₃H₄PteGlu₁ is H₄PteGlu₁, which is a much better substrate for the FPGS. If inhibition of FPGS were responsible, then we might expect an accumulation of the U folylmonoglutamate as observed (Fig. 6) with the possible build up of M folylmonoglutamate due to product (H₄PteGlu₁) inhibition. On the other hand, it is possible that M folylmonoglutamate build up might be caused by inhibition of the demethylation of the 5CH₃H₄PteGlu₁ by phenytoin. Indeed, such an inhibition would slow the conversion of 5CH₃H₄PteGlu₁, the folate imported into the cell, to H₄PteGlu₁, the folate substrate for FPGS. The enzyme responsible for this conversion is methionine synthetase. Inhibition of methionine synthetase would cause 5CH₃H₄PteGlu₁ to accumulate, but, unless the FPGS were also inhibited, we would not expect the H₄PteGlu₁ to also accumulate over time. Therefore, it appears most likely from the present data that phenytoin causes a decrease in the activity of the hepatic FPGS resulting in a decreased capacity of the hepatic cell to retain folate.

The data presented here do not exclude the possibility that treatment with phenytoin may induce increased catabolism of the folates. However, the data are suggestive that nonspecific catabolic mechanisms do not seem to be involved in the phenytoin-induced depletion of folate concentrations. If we assume that the folylmonoglutamates are the primary targets of hypothetical, nonspecific catabolic enzymes such as cytochrome P-450, then it is difficult to explain why the pentaglutamate derivatives decrease most rapidly and why, in fact, in liver the monoglutamate derivatives actually increase, at least the U forms and probably the M form. It is safe to assume that the monoglutamates are the primary targets of any nonspecific catabolic enzymes because the polyglutamates are, very likely, entirely bound to enzyme complexes which should protect the folylpolyglutamates from nonspecific degradation. Schirch and Strong (1989) calculated that the total hepatic folate concentration was approximately equal to the total concentration of folate binding proteins, and since the folylpolyglutamates bind more strongly to most of these proteins than the folylmonoglutamates, we may assume that most of the folylpolyglutamates will be bound and will therefore be protected from catabolism. Only the folylmonoglutamates entering the cell and the short chain folylglutamates formed by the hydrolysis of the folylpolyglutamates would be susceptible to nonspecific catabolism. If the shorter chain folates are more susceptible to catabolism, we might expect these folates to be depleted first. The data presented here indicate that the pentaglutamate concentrations are most affected by phenytoin. These data suggest that a phenytoin-induced increase in folate catabolism is probably not responsible for the depletion of folates observed.

No effect of phenytoin on brain folates has been previously reported. The increased concentrations of long chain folates (particularly in the U and F fractions) in brain with phenytoin treatment may simply have been a response by the brain to decreased concentrations of available folate. It has been known for some time that the average glutamate chain length increases in tissues as folate supplies are depleted from the body (Richardson et al. 1979, Cassady et al. 1980, Ward and Nixon 1990, Varela-Moreiras and Selhub 1992). The long chain increases were quantitatively small but consistent. Short chain concentrations declined consistently and significantly only in the U fraction, but the variability in the U fraction was much lower than the variabilities in the D, F and M fractions.

It is interesting to note that the pattern of change of the folate pools in the intestinal mucosa, an initial slight decrease in the concentrations of the short chain folates followed after 7-10 d by a gradual recovery of short chain folate concentrations, was very similar to the pattern of folate depletion and recovery seen in liver but on a much smaller scale, smaller both in terms of concentrations and percentages. It would appear that the changes in the short chain folate pool observed in the intestinal mucosa were a result of changes in the liver folate pool rather than direct effects of phenytoin on the mucosal cells. The long chain folates, represented by the F folyltri-, tetra-, penta- and hexa-glutamates which constituted anywhere from 15 to 40% of the total mucosal F folates, were not affected by phenytoin treatment. These long chain F folates may constitute the active pool of mitochondrial folates in the intestinal mucosa while the small concentrations of long chain M folates detected may constitute the active pool of cytoplasmic folates. The short chain folates, on the other hand, may be folates in the process of being absorbed and may reflect more closely the folate intake of the rat than would the more permanent long chain pools. Or expressed another way, the long chain folates may be the functional folates of the intestinal mucosa while the shorter chain folates would be the dietary or bile components passing through the mucosa on their way to the portal system.

The overall composition of the bile folate pool is somewhat puzzling in that the M fraction composition with respect to the glutamate derivatization is quite different from the U, D and F fractions of folate (Fig. 8). As expected the bile folates were short chain folates with detectable levels of tetraglutamates but no measurable concentrations of any longer chain glutamates. However, the M folates had a much greater representation of di-, tri- and tetra-glutamates than did the U, D and F folates. This may be a function of access indicating that the cytoplasmic folates, primarily the M group, have greater access to the bile than do the other groups allowing higher concentrations of longer chain folates to pass into the bile than for other groups which may have to pass more than one membrane to gain access to the bile. On the other hand, the M group distribution in bile may be a function of the binding properties of the folylpolyglutamyl synthetase (FPGS). 5CH₃H₄PteGlu₁ is 20% as effective a substrate for FPGS as is H₄PteGlu₁. However, once the glutamate is added, the 5CH₃H₄PteGlu₂ is less than 2% as effective a substrate (Cichowicz and Shane 1987) while the H₄PteGlu₂, the 5,10CH₂H₄PteGlu₂, and 10CHOH₄PteGlu₂ are 63%, 18% and 13% as effective, respectively. Therefore, especially the diglutamate folates from the U and the F forms would be more likely to form polyglutamates and be retained by the cells than would the M folates. This hypothesis is consistent with the data showing an increase in the mono- and di-glutamates of the M form for several days after the initiation of phenytoin treatment. If the phenytoin is inhibiting the FPGS, then those folates that are the poorest substrates for FPGS would be the most likely to be excreted into the bile.

The effect of adding phenytoin to the diet on the food consumption of the rats was rather surprising. In a previous study (Critchfield et al. 1993) we pair-fed one group of untreated rats using a phenytoin-treated group as the control to determine the quantity of food for presentation while feeding another group of rats ad libitum. We found no significant differences among the three groups of rats in food intake or in weight gain. In that study, in which we were looking at the effects of anticonvulsants on trace element levels, we used Wistar rats instead of Sprague-Dawley rats. It is possible that the Sprague Dawley rats are much more sensitive to the effects of phenytoin than are the Wistar rats. But neither the decreased food consumption nor the decreased weight gain (both about 11%) could account for the precipitous loss of folate (about 50%) in the livers of the treated animals. Indeed, with a decreased intake of only 11%, we would not expect to see a significant effect on folate absorption from the diet, and the data from the intestinal mucosa indicate that there was very little effect of phenytoin on the uptake of folate by the mucosal cells, whether it was by direct inhibition or by inhibition of food consumption. These data indicate that the hypothesis of Hoppner and Lampi (1989) does not explain the phenytoin effect on folate in rats. The hypothesis proposed by Hoppner and Lampi states that a phenytoin-induced decrease in food intake is responsible for the observed decrease in liver folate concentrations.

The data presented here are more consistent with a mechanism of phenytoin-induced folate depletion involving generalized folate processing. Phenytoin inhibition of intestinal conjugase activity could be considered a generalized effect on folate processing, but conjugase inhibition does not explain the folate depletion observed here or in earlier studies (Carl and Smith 1983) in which defined diets containing only PteGlu₁ were given to the rats. A phenytoin inhibition of the general processing enzyme, folylpolyglutamate synthetase (FPGS), is consistent with the data presented here. However, these data do not eliminate the possibility that there may be a phenytoin-induced increase in folate catabolism. The data presented here, combined with other evidence, suggests that catabolism does not explain the phenytoin effect on folate, but this specific hypothesis was not directly examined in the present study.

ACKNOWLEDGMENTS