The Monocarboxylate Transporter Homolog Mch5p Catalyzes Riboflavin (Vitamin B2) Uptake in Saccharomyces cerevisiae*

Riboflavin is a water-soluble vitamin (vitamin B2) required for the production of the flavin cofactors FMN and FAD. Mammals are unable to synthesize riboflavin and need a dietary supply of the vitamin. Riboflavin transport proteins operating in the plasma membrane thus have an important role in the absorption of the vitamin. However, their sequences remained elusive, and not a single eukaryotic riboflavin transporter is known to date. Here we used a genetic approach to isolate MCH5, a Saccharomyces cerevisiae gene with homology to mammalian monocarboxylate transporters, and characterize the protein as a plasma membrane transporter for riboflavin. This conclusion is based on the suppression of riboflavin biosynthetic mutants (rib mutants) by overexpression of MCH5 and by synthetic growth defects caused by deletion of MCH5 in rib mutants. We also show that cellular processes in multiple compartments are affected by deletion of MCH5 and localize the protein to the plasma membrane. Transport experiments in S. cerevisiae and Schizosaccharomyces pombe cells demonstrate that Mch5p is a high affinity transporter (Km = 17 μm) with a pH optimum at pH 7.5. Riboflavin uptake is not inhibited by protonophores, does not require metabolic energy, and operates by a facilitated diffusion mechanism. The expression of MCH5 is regulated by the cellular riboflavin content. This indicates that S. cerevisiae has a mechanism to sense riboflavin and avert riboflavin deficiency by increasing the expression of the plasma membrane transporter MCH5. Moreover, the other members of the MCH gene family appear to have unrelated functions.

The proteins that require FMN or FAD as cofactors are termed flavoproteins. Mostly, they contain noncovalently bound flavin cofactors and are specific for either FAD or FMN. Some of them contain auxiliary groups such as pteridin, heme, iron-sulfur centers, molybdenum, or other metal ions or contain disulfides in their active site. Flavoproteins are involved in a wide range of biochemical reactions. They play a pivotal role in the dehydrogenation of metabolites in one-and two-electron transfer reactions from and to redox centers, in the activation of oxygen for oxidation, and in hydroxylation reactions (1).
The flavin cofactor FMN is produced from riboflavin by the action of riboflavin kinase. FAD derives from FMN and ATP, a reaction catalyzed by FAD synthetase. Riboflavin is a vitamin (vitamin B 2 ) for mammals and many other organisms. Thus, dietary riboflavin has to be taken up from the gut and then provided to every single cell in a multicellular organism. Plasma membrane riboflavin transporters are thought to play an important role in the distribution of riboflavin. However, their existence in many cell types up to now has only been demonstrated biochemically (reviewed in Ref. 2). Whereas passive uptake of riboflavin is commonly observed in riboflavin-sufficient conditions, riboflavin uptake at low concentrations follows saturation kinetics and displays high affinity for the substrate (K m ϭ 1 nM to 1 M (2)).
Riboflavin is not required by most fungal organisms (3), indicating that most of them produce riboflavin. Indeed, the yeast Saccharomyces cerevisiae is known to be an excellent dietary source of riboflavin (4) and to possess all enzymes required for riboflavin biosynthesis, which are encoded by the RIB genes (RIB1, RIB2, RIB3, RIB4, RIB5, and RIB7). Some fungal organisms are utilized in commercial riboflavin production processes. One of the best riboflavin producers is the filamentous hemiascomycete Ashbya gossypii that is capable of producing the astonishing amount of 15 g of riboflavin/liter of medium, a concentration at which riboflavin readily crystallizes (5). Other riboflavin overproducers include mutants of the yeasts Candida famata, Pichia guilliermondii, or engineered Bacillus subtilis strains (5,6).
Despite of being able to synthesize riboflavin, fungal organisms are also able to take up riboflavin from the culture medium. This activity was recently analyzed in an Ashbya gossypii rib5⌬ mutant defective in the last step of riboflavin biosynthesis (7). Riboflavin uptake was found to be a high affinity (K m ϭ 40 M) process with low overall activity. The activity was sensitive to competition by FAD and FMN and to inhibition by 2,4-dinitrophenol, an uncoupler of the transmembrane proton gradient (7). Riboflavin auxotrophic S. cerevisiae mutants were also found to take up extracellular riboflavin with high affinity (K m ϭ 15 M) and specificity, but no energy requirement was apparent (8). Riboflavin uptake has also been shown in riboflavin requiring mutants of the yeast Pichia guilliermondii (9). Despite this biochemical evidence, not a single eukaryotic riboflavin transporter is known to date. Here, we used a classical genetic approach and identified Mch5p, a yeast protein with homology to mammalian monocarboxylate transporters (MCTs), 3 as the plasma membrane transporter for riboflavin.
Media and Growth Conditions-Yeast media were YPD (2% peptone, 2% glucose, 1% yeast extract), SD-Glc (2% glucose, 0.67% yeast nitrogen base without amino acids (YNB) or SD-Gal (2% galactose, 0.67% YNB). SD media contain 200 g/liter riboflavin and were supplemented with additional riboflavin to give the desired final concentration. Riboflavinfree media were made from yeast nitrogen base without amino acids without vitamins (Bio 101) and contained all vitamins except riboflavin in standard concentrations. As a rule, only supplements that were required by the strains were added. YADE (2% yeast extract, 0.2% ammonium sulfate, 0.2% glucose, and 3% ethanol) was used for the preparation of mitochondria. S. pombe cells were grown in EMM, a standard synthetic medium for S. pombe, which is devoid of riboflavin (15). Suspensions of yeast cells for use in growth assays were prepared in 96-well plates. Serial 10-fold dilutions starting from A 600 ϭ 0.6 in adjoining wells were transferred to agar plates using a stainless steel replication device. Growth was recorded after incubation at 30°C.
Plasmids-An S. cerevisiae library in the multicopy vector YEp352 was used (16). As illustrated in Fig. 1A, YOR305W was isolated from a complementing library plasmid as BglII/SalI (polylinker site) fragment and ligated into the BamHI/SalI sites of the URA3 2 plasmid pRS426 (17). To isolate YOR306C (MCH5), we used a SalI/EcoRI restriction fragment, which was ligated into the same sites of pRS426.
For transport assays in S. cerevisiae, MCH5 was amplified from a complementing library plasmid and cloned downstream of the TDH3 promotor in plasmid p426-TDH (18). N-terminal tagging of MCH5 was performed in the low copy plasmid pRS316 by fusing a GAL1 promotor to a 9-Myc tag, followed by the MCH5 ORF lacking the ATG, followed by 224 bp of MCH5 downstream sequence. As assayed by growth of rib4⌬ and rib5⌬ mutants on low concentrations of riboflavin, both constructs encoded functional versions of Mch5p (data not shown). Reporter assays were performed with cells transformed with a construct containing 832 bp of MCH5 5Ј-sequences (obtained as a PstI/SphI cut PCR product), which were fused to the E. coli lacZ gene (a SphI/XbaI cut PCR product generated from genomic DNA) in the URA3/CEN plasmid YCplac33 (19). To express MCH1, MCH2, MCH3, MCH4, or MCH5 at similar levels in S. cerevisiae, we amplified each ORF from genomic DNA with primers that added restriction sites adjacent to the start and to the stop codon. The PCR products were ligated into pPCR-Skript Amp (Stratagene), sequence-verified, and finally cloned downstream of the galactose-inducible GAL1 promotor of the multicopy vector pYES2 (Invitrogen). S. pombe expression used the thiaminrepressible pREP3X vector (20), into which MCH5 was cloned as a PCR product.
Uptake Experiments-S. cerevisiae rib4⌬ for use in uptake experiments were grown overnight in SD media containing 20 mg/liter riboflavin. The cells were washed, suspended in SD lacking riboflavin for A 600 ϭ 0.2, and shaken for at least 8 h at 30°C. S. pombe cells were grown to midlogarithmic phase in EMM. All cells were washed with water, resuspended in 40 mM K 2 HPO 4 /KH 2 PO 4 , 10 mM KCl, pH 7.5, and stored on ice. Uptake experiments were performed at 30°C in a volume of 500 l containing five OD cells, 40 mM K 2 HPO 4 /KH 2 PO 4 , pH 7.5, 10 mM KCl, and 3.2 M [ 14 C]riboflavin (specific activity 5.54 MBq/mg; a gift of Prof. Dr. Reinhard Krämer, Universität zu Köln, Germany). Aliquots were removed at timed intervals, filtered on glass fiber filters, washed with cold water, and analyzed by scintillation counting. We observed a distinct day-to-day variability in overall riboflavin transport activity in both expression systems, which restricted us to only compare results obtained with the same batch of cells. Additional controls ensured that storage on ice preserved the transport activity of the cells. Glucose uptake was performed as in Ref. 21, biotin uptake as in Ref. 22.

MCH5 Shows Genetic Interactions with Riboflavin Biosynthetic
Genes-To isolate the S. cerevisiae plasma membrane riboflavin transporter, we first analyzed the growth phenotype of mutants with defects in riboflavin biosynthesis. The rib4⌬ and rib5⌬ mutants are defective in the penultimate and last step of riboflavin biosynthesis, respectively. When compared with wild-type cells, both mutants showed growth defects on plates containing low concentrations of riboflavin (Fig. 1B). Surprisingly, for two enzymes catalyzing consecutive steps in the same biochemical pathway, rib4⌬ and rib5⌬ cells behaved differently. On minimal plates, rib4⌬ cells displayed a milder phenotype and could grow with 2 mg/liter riboflavin. rib5⌬ cells, in contrast, required a 10-fold higher concentration of riboflavin for growth (Fig. 1B). Similarly, rib4⌬ could grow on complete media (YPD) without added riboflavin, whereas only a supplement of 20 mg/liter of riboflavin allowed growth of rib5⌬ (Fig. 1C). The milder phenotype of rib4⌬ is probably explained by the fact that the reaction catalyzed by Rib4p proceeds without enzymatic catalysis (27). Thus, although both mutants showed a different behavior, both had strong growth defects on low riboflavin plates. Importantly, both appeared to grow normally when their riboflavin requirement was met (Fig. 1, B and C). The strong growth defects of rib4⌬ and rib5⌬ mutants initiated a genetic screen in which the riboflavin transporter was searched as a multicopy suppressor of the rib mutants. To this end, both mutants were transformed with a multicopy genomic library, followed by plating on media containing 0.2 (rib4⌬) or 2 mg/liter (rib5⌬) riboflavin. Plasmid DNA was isolated from colonies that were able to grow on these media and retransformed into rib4⌬ or rib5⌬ mutants. Plasmids that allowed growth on plates containing no riboflavin were eliminated because they probably contained the riboflavin biosynthesis genes defective in the mutants. For random samples, this was confirmed by restriction analysis or PCR.
The remaining plasmids, 4 of 20,000 tranformants in rib4⌬ and 10 of 36,000 transformants in rib5⌬, did not allow growth on riboflavin-free plates and showed partial correction of the growth defects of the mutants. All complementing plasmids contained fragments from the same region of chromosome XV, and the smallest plasmid harbored only two ORFs, YOR305W and YOR306C (Fig. 1A). Only little information is available on YOR305W. The ORF appears to be present in other budding yeast species and cause slow growth when deleted, making it unlikely that it is a spurious ORF (28). YOR306C, in contrast, was previously characterized to be a yeast homolog of mammalian monocarboxylate transporters (MCTs; reviewed in Ref. 29) and hence named MCH5 (for monocarboxylate transporter homolog 5 (13)).
To analyze which of the two ORFs were responsible for the suppressor activity, we individually subcloned MCH5 and YOR305W from the library plasmid into the multicopy plasmid pRS426 (Fig. 1A). We transformed these plasmids into rib4⌬ and rib5⌬ mutants and analyzed the riboflavin requirement of the transformed cells (Fig. 1B). Of the two genes present on the library plasmid, only MCH5 was able to suppress the growth phenotype resulting from loss of RIB4 or RIB5 (Fig. 1B). Expression of MCH5, however, did not fully restore growth of rib4⌬ and rib5⌬ back to wild-type levels. Whereas wild-type growth rates were typical for strains that contained the authentic RIB genes on library plasmids, overexpression of MCH5 did not allow growth in the absence of riboflavin and allowed somewhat slower growth on plates containing 0.2 mg/liter riboflavin ( Fig. 1B and data not shown).
We also transformed the MCH5 containing plasmid into rib2⌬, rib3⌬, and rib7⌬ deletion mutants. Before transformation, these strains behaved like rib5⌬ and required 20 mg/liter riboflavin for growth. After transformation with pRS426-MCH5, all of them were able to grow on plates containing 2 mg/liter riboflavin (data not shown). In conclusion, our experiments establish that the suppressor activity identified in the screening is carried by MCH5 and not by YOR305W. Moreover, MCH5 was able to suppress all rib mutants when overexpressed.
To analyze the interaction of MCH5 with riboflavin biosynthesis genes in greater detail, we generated mch5⌬ rib4⌬ and mch5⌬ rib5⌬ double mutants and analyzed their riboflavin requirements (Fig. 1C). Deletion of MCH5 in a wild-type strain did not lead to a phenotype. In contrast, deletion of MCH5 increased the riboflavin requirements of rib4⌬ and rib5⌬ (Fig. 1C). Surprisingly, mch5⌬ rib4⌬ and mch5⌬ rib5⌬ strains showed identical growth on YPD plates containing 20 mg/liter riboflavin (Fig. 1C), whereas of the corresponding single rib mutants, rib4⌬ showed much better growth. Although the reason for this is not known, this may indicate that the mch5⌬ rib4⌬ double mutant can perform the chemical reaction that bypasses Rib4p only with reduced efficiency. In conclusion, overexpression of MCH5 rescues the growth defects of rib mutants, whereas their phenotype is exacerbated by deletion of MCH5.
The Function of MCH5 Is Unique within the MCH Gene Family-Mch5p is a 521-amino acid protein with 12 hydrophobic regions probably corresponding to 12 transmembrane domains (30). It is a member of a yeast protein family consisting of five members (Mch1p-Mch5p (13)) and is most similar to Mch4p with which it shares 45% identical amino acids (31). Of the MCH genes, MCH5 appears to have the highest expression level, followed by MCH4 (13). To analyze if the function of MCH5 was shared by any of the other MCH genes, we individually expressed the MCH1, MCH2, MCH3, and MCH4 ORFs in rib5⌬. To achieve similar levels of expression, all ORFs were amplified from genomic DNA and cloned downstream of the GAL1 promotor in the multicopy plasmid pYES2. Analysis of the transformants revealed that only MCH5 allowed growth of rib5⌬ cells on low riboflavin concentrations (Fig. 1D). Moreover, suppression by MCH5 was not seen on glucose-containing plates, indicating that expression of MCH5 was necessary for its activity as a multicopy suppressor (Fig. 1D).
We also deleted RIB5 in individual strains carrying null alleles of MCH1, MCH2, MCH3, MCH4, or MCH5 and found synthetic defects only for the combined deletion of MCH5 and RIB5 (data not shown). These findings were corroborated by growth assays performed with a strain deleted for all five MCH genes (13). Because this strain was from the CEN.PK background, we generated a series of isogenic strains deleted for RIB5, RIB5 and MCH5, or RIB5 and all MCH genes and used them in growth assays (Fig. 1E). We found that mch1-5⌬ showed wildtype growth, but growth was affected by deletion of RIB5. As expected from our findings in the BY strain background (Fig. 1C), deletion of MCH5 further reduced the growth of rib5⌬. Importantly, a strain deleted for all MCH genes and for RIB5 showed no growth defects beyond the defects seen in mch5⌬ rib5⌬ (Fig. 1E). This demonstrates that MCH5 performs a unique function within the yeast MCH gene family. This function, however, appears to be different from the human monocarboxylate transporters. Whereas the human proteins act as transporters for pyruvate, lactate, ketone bodies, and other monocarboxylates (29), these substances were ruled out as possible substrates of the yeast Mch proteins (13). A, graphic representation and partial restriction map of the insert of a library plasmid that acted as a multicopy suppressor. The insert included two complete ORFs, MCH5 (YOR306C) and YOR305W, that were individually subcloned for further testing into pRS426 as indicated. B, S. cerevisiae wild-type cells (BY strain background) or isogenic strains with deletions in RIB4 or RIB5 were transformed with an empty vector (Ϫ), pRS426-MCH5, or pRS426-YOR305W. The cells were grown on SD plates containing the indicated concentrations of riboflavin and photographed after 3 days at 30°C. C, yeast cells (BY strain background) with the indicated deletions were grown on YPD plates supplemented with the indicated concentration of riboflavin. Growth was recorded after 2 days at 30°C. D, S. cerevisiae rib5⌬::kanMX mutants (BY strain background) were transformed with pYES2 based plasmids containing no insert (Ϫ) or the indicated MCH gene. Transformants were plated on SD plates containing glucose or galactose and riboflavin as indicated. Growth was recorded after 3 days at 30°C. E, yeast strains from the CEN.PK strain background carrying the indicated deletions were grown on YPD plates supplemented with the indicated concentrations of riboflavin. Growth was recorded after 3 days at 30°C.
Riboflavin Uptake in S. cerevisiae DECEMBER 2, 2005 • VOLUME 280 • NUMBER 48 rib⌬ Strains Develop Extragenic Suppressor Mutations-In the course of our experiments, we noted that strains with rib⌬ mutations frequently segregated colonies that display improved growth (see Figs. 1, C and D). This behavior has been observed by others for rib4⌬ and was interpreted to be due to the appearance of extragenic supressors (32). Initially, we identified suppressor mutants only in rib4⌬, but this finding was likely influenced by the higher riboflavin content of rib4⌬, relative to the other rib⌬ mutants. Closer investigation of this phenomenon provided evidence that colonies with improved growth characteristics arise at similar frequencies in all riboflavin biosynthetic mutants tested. 4 Whereas this high frequency of spontaneous mutations in rib⌬ mutants is probably explained by the involvement of the riboflavin biosynthetic pathway in the detoxification of 8-oxo-GTP, a spontaneously produced mutagenic substrate for DNA synthesis (33,34), the precise mechanism by which the suppressor mutations improve growth is currently unknown. Suppressor mutations also arise in rib⌬ mch5⌬ strains, indicating that suppression is not caused by a mutation in MCH5 (Fig.  1C). Due to this phenomenon, the strains used for the analyses described below were taken from plates that were free of suppressors and their growth in liquid media was limited to the minimal time required to perform the experiment. Moreover, we carefully monitored the growth rates to exclude cultures that grew faster then expected from further analysis. A detailed analysis of the suppressor mutants will be presented elsewhere.
Deletion of MCH5 Causes Inactivation of FAD-dependent Processes-To assess the consequences caused by inactivation of RIB5 and MCH5, we determined the activity of FAD dependent processes located in different cellular compartments. It has been demonstrated before that the formation of disulfide bonds in proteins depends on riboflavin biosynthesis (23) and is catalyzed by the FAD-dependent oxidase Ero1p localized in the ER lumen (35,36). A reduced capability to oxidize protein thiols results in the ER retention of disulfide-containing proteins after treatment of cells with reducing agents such as dithiothreitol (23). CPY, a vacuolar protease containing five disulfides, is a convenient marker to monitor protein oxidation, because disulfide formation is required for the folding and ER export and because its electrophoretic mobility reflects its cellular localization (23,35,37).
Pulse-chase experiments revealed that wild-type cells matured CPY very quickly ( Fig. 2A). Although the CPY precursors (p1, corresponding to the ER form, and p2, corresponding to the Golgi form) were more abundant, the vacuolar (m) form of CPY could already be detected 10 min after dithiothreitol removal. Precursors were only present in trace amounts at later time points, where the mature form accumulated as expected ( Fig. 2A). rib5⌬ mutants showed a slower maturation of CPY. After a 10-min chase, no mature CPY was present and only p1 and p2 forms were observed. In rib5⌬, precursor and mature forms showed roughly equal abundance after a 30-min chase, whereas at 60 min and later, mCPY was the dominant form, and precursors had a lower abundance. Maturation of CPY was further slowed in rib5⌬ mch5⌬ double mutants where mature forms appeared only after a 60-min chase. It was also evident from Fig. 2A that rib5⌬ mch5⌬ do not produce the p2 form of CPY. This can be explained by assuming that only CPY molecules that are properly folded are allowed to exit the ER. Because protein oxidation and folding is slowed by deletion of RIB5 and MCH5, this also reduces the number of CPY molecules exiting per time. If the next step, processing in the Golgi, proceeds with normal speed, the number of CPY molecules in the Golgi (p2 form) might be below the limit of detection. As expected, CPY finally appears to accumulate in the vacuole of rib5⌬ mch5⌬, which is its final destination. Another striking observation was that rib5⌬ and rib5⌬ mch5⌬ strains did not accumulate mCPY to the same extent as wild-type strains. This possibly indicates degradation of the misfolded protein by ER quality control pathways. In summary, maturation of CPY, a process requiring FAD-dependent enzymes in the ER, is slower in rib5⌬ mutants and is even more delayed in rib5⌬ mch5⌬ double mutants. These findings are similar to the drastic reduction of apolipoprotein B-100 secretion observed in human cell lines after growth in riboflavin-deficient media (38).
To analyze if other cellular processes are affected by deletion of MCH5, we performed activity assays for succinate dehydrogenase, a FAD-dependent mitochondrial enzyme of the citric acid cycle (Fig. 2B).  We found that the activity of succinate dehydrogenase was reduced to 60% of wild-type levels upon deletion of RIB5, but due to the large variation in the assay, this difference was not statistically significant. Succinate dehydrogenase activity was reduced even further in rib5⌬ mch5⌬ double mutants, where it amounted to only 20% of wild-type levels (Fig. 2B). In summary, combined deletion of MCH5 and RIB5 affects FAD-dependent processes in multiple cellular compartments, making it unlikely that Mch5p is an organelle-specific transport protein.
Mch5p Localizes to the Yeast Plasma Membrane-To be able to directly analyze the subcellular localization of Mch5p, we generated different tagged versions of the protein. We noticed that fusion of green fluorescent protein to the C terminus destroyed its activity as a multicopy suppressor, and fusion of green fluorescent protein to the N terminus created unstable proteins (data not shown). However, a version of MCH5 containing an N-terminally added 9-Myc epitope encoded a stable protein. When expressed from a galactose-inducible promotor in a centromeric plasmid, the fused gene suppressed rib5⌬ mutants on galactose-but not on glucose-containing plates, demonstrating that the N-terminal 9-Myc epitope did not interfere with the function of Mch5p (data not shown). When we analyzed the distribution of 9-Myc-Mch5p by sucrose density gradient centrifugation, we found the protein to be predominantly present in the top and bottom fractions of the gradient. This distribution pattern was identical with the distribution of the plasma membrane ATPase Pma1p. On the other hand, 9-Myc-Mch5p did not co-fractionate with the ER protein Wbp1p or the Golgi protein Mnn9p (Fig. 2C). Previously, the Mch3 protein was localized to mitochondria (13), but Mch5p showed a different pattern of distribution when compared with the mitochondrial inner membrane protein Tim23p or the outer membrane porin Por1p (Fig. 2C). The same study localized the paralogous Mch4 protein to the vacuolar membrane (13), but again we found no evidence for a cofractionation of Mch5p with the vacuolar marker Vam3p. We conclude that Mch5p localizes to the yeast plasma membrane and affects processes in multiple compartments, consistent with a function in riboflavin transport across the plasma membrane.
Mch5p Is a Transporter for Riboflavin-Although all data presented above are consistent with Mch5p being a plasma membrane riboflavin transporter, this was difficult to demonstrate directly by uptake experiments. In this respect, Mch5p is another example that demonstrates that growth assays have a much higher sensitivity than uptake experiments, an observation we have made before in our studies of pyridoxine and biotin uptake (39,40). The approach that demonstrated the activity of Mch5p employed strong overexpression of MCH5 in rib4⌬ mutants and depletion of intracellular riboflavin by incubation in riboflavin-free media. Once these conditions had been established, cells transformed with the MCH5 overexpression vector showed activity in the uptake of [ 14 C]riboflavin, whereas control cells harboring an empty vector showed no uptake (Fig. 3A). Although this experiment provided the first evidence that MCH5 encoded a functional plasma membrane riboflavin transporter, another possible explanation was that a spontaneous suppressor mutation allowed the cells with the MCH5 plasmid to take up riboflavin more rapidly. To address this issue, we isolated a rib4⌬ strain that carried a suppressor mutation, transformed it with the MCH5 overexpression plasmid or with an empty control plasmid, and repeated the uptake experiments. Similar to the measurements presented in Fig. 3A, riboflavin uptake in suppressor strains required overexpression of MCH5 and was absent from the vector control cells (data not shown). This demonstrates that the unknown suppressor mutation that led to improved growth of rib4⌬ did not cause a measurable increase in ribo-flavin uptake and substantiates our interpretation that MCH5 encodes the plasma membrane riboflavin transporter.
To gain independent proof for the function of Mch5p and to eliminate the complication caused by the suppressor mutations arising in S. cerevisiae rib⌬ strains, we heterologously expressed MCH5 in wild-type cells of the fission yeast S. pombe. The genome of S. pombe does not contain genes with high similarity to MCH5, and we found no evidence that wild-type S. pombe cells are capable of riboflavin uptake. S. pombe wild-type cells grow well in riboflavin-free media and contain high scoring orthologs of all S. cerevisiae RIB genes, indicating their proficiency to synthesize riboflavin (3,41). To express MCH5, we used a thiamineregulated promotor and grew the cells in standard EMM media. Uptake Riboflavin Uptake in S. cerevisiae DECEMBER 2, 2005 • VOLUME 280 • NUMBER 48 assays demonstrated that expression of MCH5 in thiamine-free media caused S. pombe cells to take up riboflavin, whereas growth in repressing conditions did not allow riboflavin uptake above background levels seen with control cells (Fig. 3B). Thus, overexpression in S. cerevisiae rib4⌬ mutants as well as expression in S. pombe wild-type cells demonstrates that MCH5 encodes a riboflavin transporter.
Characterization of the Activity of Mch5p-Expression of MCH5 from a plasmid in S. pombe produced higher riboflavin uptake rates then overexpression in S. cerevisiae (Fig. 3). In addition, S. pombe cells proved genetically more stable than S. cerevisiae rib4⌬ mutants, leading us to prefer the heterologous expression system for a detailed analysis of the activity of Mch5p. Riboflavin uptake via Mch5p was saturable and displayed an apparent K m of 17 M (Fig. 3C). The pH optimum was at pH 7.5, and about 70% of the activity was detectable at pH 7.0 and 8.0 (Fig.  3D). These values are in excellent agreement with the kinetic parameters determined earlier for S. cerevisiae riboflavin auxotrophic mutants that displayed an identical pH optimum and a K m of 15 M (8). We also investigated several substrate analogs and found lumichrome and acriflavin, two sugarless riboflavin degradation products (42), to potently reduce riboflavin uptake (TABLE ONE). We also observed that the addition of a 2-fold excess of unlabeled riboflavin reduced the uptake of labeled riboflavin. The remaining activity (52%), however, was somewhat lower than the expected activity calculated from the Michaelis-Menten equation, which is 76% (at K m ϭ 17 M, the velocity increases by a factor of 2.28 when the substrate concentration is raised from 3.2 to 9.6 M; because only one-third of all molecules are labeled at 9.6 M riboflavin, the expected overall increase is a velocity of 76%). Although the reason for this discrepancy is not known, this shows that it is, in principle, possible to compete the uptake of labeled riboflavin by adding authentic substrates.
Uptake assays in the presence of protonophores were used to analyze whether riboflavin is transported by proton symport. We demonstrated before that protonophores cause a severe reduction in the activity of known proton-vitamin symporters (22,40,43). We repeated some of these experiments and found that biotin uptake, for example, is reduced to 25% of controls by the addition of 0.1 mM carbonyl cyanide p-(trifluoromethoxy) phenylhydrazone (data not shown). The addition of protonophores, however, had a much weaker effect on the riboflavin uptake activity of Mch5p (TABLE TWO), where the activities amounted to 65-85% of uninhibited controls. To check if this reduction caused by protonophores is significant, we analyzed glucose uptake in S. cerevisiae.
Glucose transport in S. cerevisiae occurs by facilitated diffusion (44) and was reduced to 69% when 0.1 mM carbonyl cyanide p-(trifluoromethoxy) phenylhydrazone was present (data not shown). The 15-35% reduction of riboflavin transport caused by protonophores thus is comparable with the inhibition of glucose uptake, a facilitated diffusion process, whereas typical proton symport processes such as biotin uptake show a much stronger reduction in activity.
To gain additional information about the energy requirements of Mch5p for riboflavin transport, we analyzed the effect of glucose or ethanol on whole cell riboflavin uptake. Glucose was earlier shown to directly activate the plasma membrane proton ATPase, leading to an increased proton motif force across the plasma membrane (45). The addition of ethanol also activates the proton-ATPase and leads to a concomitant increase in the activity of proton-dependent transport processes (46). Importantly, we found that neither glucose nor ethanol significantly stimulated riboflavin uptake by Mch5p (Fig. 3E).
Together, our findings show that Mch5p operates as a facilitator of riboflavin uptake and not as an active carrier. This also explains why it was necessary to deplete the cellular riboflavin pools to detect the activity of Mch5p. Although, as a facilitator, Mch5p would allow substrate fluxes in both directions, intracellular compartmentalization of riboflavin or conversion to FMN and FAD, two substances not transported by Mch5p (TABLE ONE), could provide directionality.
Regulation of MCH5-All earlier studies on riboflavin uptake in yeast and other fungi made use of riboflavin auxotrophic mutants (7-9), and it was speculated that intracellular flavins repress the biosynthesis of the carrier (8). To investigate whether riboflavin has an effect on the expression of MCH5, we fused MCH5 upstream sequences to the lacZ reporter gene and transformed the construct into S. cerevisiae wildtype, rib5⌬, and rib5⌬ mch5⌬ cells. Reporter gene assays were performed after growth of the transformants in varying concentrations of riboflavin. These assays demonstrated that the activity of MCH5 in wild-type cells was not under control of riboflavin (Fig. 4). In contrast, rib5⌬ or rib5⌬ mch5⌬ cells showed increased reporter gene expression with decreasing riboflavin concentrations (Fig. 4). Although the data for rib5⌬ mch5⌬ were more scattered, they seemed to indicate that MCH5 is more strongly expressed in rib5⌬ mch5⌬ than in rib5⌬. Together, these data show that yeast cells have the ability to sense the cellular flavin content and increase the expression of MCH5 when riboflavin is scarce. These analyses complement our biochemical analysis of the Mch5 protein and support our conclusion that MCH5 is the first known gene of a riboflavin transporter.

DISCUSSION
FAD and FMN, the active forms of riboflavin, are indispensable as redox catalysts for many metabolic reactions. Although almost all fungi, unlike mammalian cells, are prototrophic for riboflavin, both fungal and animal cells possess riboflavin transporters in their plasma membranes. In order to find a protein involved in riboflavin uptake, we performed a genetic screen in the yeast S. cerevisiae and identified MCH5, a gene with homology to mammalian monocarboxylate transporters.
This function of Mch5p in riboflavin uptake is supported by growth assays (Fig. 1), protein localization data (Fig. 2), and uptake assays (Fig.  3). Most importantly, expression of MCH5 in a heterologous system, S. pombe, induced riboflavin uptake across the plasma membrane. Although this does not formally exclude the possibility that Mch5p is a positive regulator of riboflavin transport, the simplest explanation of our data is that Mch5p itself is the transporter for riboflavin. If Mch5p were a regulator, it would be capable of positively regulating an elusive plasma membrane riboflavin transporter in S. pombe that has almost identical kinetic characteristics as the S. cerevisiae riboflavin transporter. Moreover, the evolutionary distance between both yeast species and the lack of MCH genes in S. pombe would make it highly surprising if Mch5p would act as a regulator in S. pombe. Direct proof of a transport function of Mch5p will afford purification of the protein and reconstitution into lipid vesicles and probably requires the establishment of a more potent expression system for MCH5. The generation of tagged versions of Mch5p represents a first step toward this goal.
Riboflavin transport in S. cerevisiae has been investigated before (8). Uptake activity required the use of riboflavin auxotrophic mutants and specific conditions such as harvest in the midlog phase of anaerobic growth (8). Even when we precisely repeated these experiments, we failed to detect riboflavin uptake in our rib4⌬ and rib5⌬ mutants. The strains used by Perl et al. (8) were not available to us, preventing a direct comparison. Upon overexpression of MCH5, however, we were able to measure riboflavin transport activity that precisely matched the kinetic properties described before. We characterize Mch5p as a high affinity (K m ϭ 17 M) carrier with a narrow range of substrates that has a pH optimum at pH 7.5. Riboflavin uptake by Mch5p was not severely affected by protonophores (TABLE TWO), and the addition of a protonophore inhibited glucose uptake, a known facilitated diffusion proc-ess, and riboflavin uptake to a similar extent (ϳ35%). This low level of inhibition was in stark contrast to the 75% reduction observed for biotin uptake by yeast Vht1p, a known proton symporter (22). Moreover, glucose and ethanol did not increase the rate of riboflavin uptake (Fig. 3E). Both findings support a facilitated diffusion mechanism for riboflavin uptake and are consistent with earlier reports that showed that riboflavin uptake does not require metabolic energy (8). As a facilitator, Mch5p enables riboflavin uptake only if the cytoplasmic concentrations are below the outside concentrations. Since incoming riboflavin can be further metabolized to FMN and FAD or become compartmentalized in the vacuole or mitochondria (7,47), even an equilibrating system such as Mch5p would allow a net uptake of riboflavin. In situations where more riboflavin is produced than can be converted to FMN, expression of a facilitator like Mch5p would lead to a leakage of riboflavin into the surrounding medium. Indeed, it is known that many cells have the capability to overproduce riboflavin and export it into the medium (7,8,48). The routes for riboflavin export are not well characterized, but in the cases investigated, import and export catalysis were attributed to different proteins (7,8). Nevertheless it is possible that expression of MCH5 in riboflavin production strains provides a means to increase riboflavin export into the medium. S. cerevisiae wild-type cells do not show high expression of MCH5 (Fig. 4), indicating that Mch5p will not lead to huge losses of riboflavin in cells that synthesize the vitamin. In contrast, MCH5 is strongly expressed when the cytoplasmic concentrations of riboflavin are low, such as when rib5⌬ or rib5⌬ mch5⌬ mutants are grown in low riboflavin media (Fig. 4). These mutants presumably have only very little cytoplasmic riboflavin, and expression of MCH5 will result in riboflavin uptake rather than loss. Thus, loss of riboflavin through the facilitator Mch5p is prevented by low expression of the gene when riboflavin uptake is not required. Our results corroborate earlier findings on the regulation of riboflavin transport by mutations that cause riboflavin dependence and by the riboflavin supply (8). They also demonstrate that yeast cells possess mechanisms to detect riboflavin deficiency and respond by increasing the import of riboflavin across the plasma membrane.
A surprising consequence of our analysis is that the yeast monocarboxylate transporter homologs do not transport the same substrates as their mammalian counterparts from the MCT family. The MCT proteins transport negatively charged molecules such as pyruvate, lactate, and acetate, and these substrates are transported in yeast by the unrelated lactate-proton symporter JEN1 (49) and have been excluded as substrates for the Mch proteins (13). The closest human homolog of Mch5p, MCT8 (also called XPCT (50)), was recently found to catalyze the uptake of the thyroid hormones thyroxine, 3,3Ј,5-triiodothyronine, 3,3Ј,5Ј-triiodothyronine, and 3,3Ј-diiodothyronine, all of which are negatively charged monocarboxylates (1,51,52). Riboflavin, in contrast, is a neutral compound. MCTs and the Mch proteins also differ in their mode of action, with the mammalian proteins acting as proton symporters (29) and Mch5p acting as facilitator. Although having a common origin, this suggests that the mammalian MCT and yeast Mch proteins have different substrate specificities and use different catalytic mechanisms.
Whereas our studies identify Mch5p as a plasma membrane protein (Figs. 2C and 3), other yeast Mch proteins were earlier reported to localize to intracellular membranes. In Makuc et al. (13), MCH3 and MCH4 were C-terminally tagged with a hemagglutinin epitope or with green fluorescent protein, and protein localization was studied by fractionation and microscopy. Since the function of these proteins is not known, the fusion constructs could not be tested for functionality. Our experiments with MCH5 indicate that only short tags in N-terminal FIGURE 4. MCH5 promotor activity depends on the availability of riboflavin. The promotor of MCH5 was fused to the lacZ reporter gene and transformed into S. cerevisiae wild-type, rib5⌬, or rib5⌬ mch5⌬ cells. The cells were grown in media containing, from left to right, 20, 10, or 0.2 mg/liter riboflavin, permeabilized, and assayed. ␤-Galactosidase activity was measured in triplicates with ortho-nitrophenyl-␤-D-galactoside as a substrate and is reported in Miller units (58). Columns represent means, and error bars indicate SD values of three independent assays. positions are compatible with the function of the protein. Specifically, no functional green fluorescent protein-tagged versions could be generated. In the light of these findings, it is unclear if the reported localization of Mch4p to the vacuole and Mch3p to mitochondria (13) will provide useful leads to assign a function to these proteins.
In addition to identifying the first known riboflavin transporter, our study also provides new insights into the riboflavin biosynthetic pathway, which starts from GTP and ribulose-5 phosphate (see Ref. 53 for a recent review). The first unusual finding is that rib5⌬ cells display a higher requirement for riboflavin than rib4⌬ cells (Fig. 1, C and D). The reaction catalyzed by Rib4p (lumazine synthase) is the condensation of 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione and 3,4-dihydroxy-2-butanone 4-phosphate to form 6,7-dimethyl-8-ribityllumazine (lumazine), the immediate biosynthetic precursor of riboflavin. It is known from experiments in vitro that the formation of lumazine occurs spontaneously in aqueous solution at room temperature and at neutral pH (i.e. close to physiological conditions) (27). Although the in vitro reaction is rather slow compared with the enzyme-catalyzed reaction and proceeds with low regiospecificity (27), our findings demonstrate that spontaneous formation of 6,7-dimethyl-8-ribityllumazine indeed occurs in vivo and makes a significant contribution to the intracellular riboflavin available to rib4⌬ mutants.
The second surprising finding is that rib mutants frequently segregate colonies with improved growth (Fig. 1). These suppressor mutations appear at high frequency in all rib strains investigated 4 and have occasionally been observed before (32,34). This phenomenon is probably explained by the fact that the riboflavin biosynthetic pathway contributes to the detoxification of 8-oxo-GTP, a mutagenic substrate for DNA synthesis arising from spontaneous oxidation of GTP (33). Suppressor strains also appear in rib⌬ mch5⌬ double mutants, indicating that suppression is not caused by a mutation in MCH5 that increases gene expression or changes the properties of the protein. We do not know the nature of the suppressor mutations, but they may lead to a bypass of the riboflavin biosynthetic pathway, allow more efficient utilization of riboflavin for FMN and FAD generation, or allow uptake across the plasma membrane via transporters other than Mch5p. Since all rib mutants segregated extragenic suppressor mutants at similar frequency, we took several precautions to prevent the use of suppressed strains in our experiments. Moreover, we demonstrated that a culture consisting of only suppressor mutants did not show riboflavin uptake unless transformed with a MCH5 multicopy plasmid. Spontaneous suppressor mutants arising in rib4⌬ and rib5⌬ also did not preclude the identification of MCH5 as a gene that causes multicopy suppression of the riboflavin deficiency. Taken together, it is highly unlikely that the small number of suppressor mutants that may have been present in our experiments has significantly influenced their results.
In summary, this is the first report that identifies a plasma membrane riboflavin transporter. Although it appears that mammalian and yeast riboflavin transporters do not possess related structures, the yeast MCH5 mutants generated here may provide a route to characterize mammalian proteins with related function.