(Received for publication, March 13, 1997, and in revised form, April 28, 1997)
From the Laboratory of Biophysics, School of Medicine, Teikyo University, Hachioji, Tokyo 192-03, Japan
A novel, systematic approach was used to identify amino acid residues responsible for substrate recognition in the transmembrane 10 region of the Gal2 galactose transporter of Saccharomyces cerevisiae. A mixture of approximately 25,000 distinct plasmids that encode all the combinations of 12 amino acids in transmembrane 10 that are different in Gal2 and the homologous glucose transporter Hxt2 was synthesized. Selection of galactose transport-positive clones on galactose limited agar plates yielded 19 clones, all of which contained the Tyr446 residue found in Gal2. 14 of the 19 clones contained Trp455 found in Gal2, whereas the other 5 contained Cys455, a residue not found in either Gal2 or Hxt2. When Tyr446 of Gal2 was replaced with any of the other 19 amino acids, no galactose transport activity was observed in the resulting transporters, indicating that Tyr446 plays an essential role in the transport of this sugar. Replacement of 2 amino acids of Hxt2 with the corresponding Tyr446 and Trp455 of Gal2 allowed the modified Hxt2 to transport galactose. The Km of galactose transport for the modified transporter was 8-fold higher than that of Gal2. These results and other evidence unequivocally show that Tyr446 is essential and Trp455 is important for the discrimination of galactose versus glucose.
Site-directed mutagenesis has been extensively used in attempts to determine functional sites in transporters (1, 2). This approach is limited, however, by the fact that it is usually not possible to mutate every amino acid and replacements that are made often yield results that are negative in nature (3, 4). As an alternative method, the use of chimeras to identify functional domains of transporters has proved highly fruitful (5-9). We have used chimeras to analyze two homologous sugar transporters in the yeast Saccharomyces cerevisiae (3, 4): Gal2, a high affinity galactose transporter (10) that was unexpectedly found to transport glucose with nearly the same affinity (3), and Hxt2, a major glucose transporter that does not transport galactose (3, 10). These two transporters belong to the Glut transporter family, the largest known organic solute transporter family comprising more than 80 transporters found in prokaryotes through mammals (11, 12). Creating chimeras between the Gal2 and Hxt2 transporters gave us an opportunity to study the galactose recognition site in Gal2 and to gain insights into the substrate recognition sites in Glut family transporters in general. To unequivocally determine the substrate recognition site, we have taken two steps. In the first step (3), three types of systematic chimeras were made using the Escherichia coli homologous recombination system. The site responsible for differentially recognizing galactose and glucose was localized to a 101-amino acid region that includes the transmembrane 10 (TM10),1 TM11, and TM12 segments and the proximal half of the C-terminal hydrophilic tail. In the second step (4), the 101-amino acid region was subdivided into the above four regions by introducing five restriction enzyme sites into the corresponding segments of each gene without changing the amino acids encoded. By analyzing plasmids containing all the possible combinations of these segments inserted into the corresponding parts of Hxt2, we identified TM10 as the domain where galactose and glucose are differentially recognized. TM10 contains 35 amino acid residues, of which only 12 are different between Gal2 and Hxt2. Thus, it is reasonable to assume that the amino acid residue(s) essential for the substrate recognition can be found among these 12 residues. We employed a new comprehensive approach and found that 2 amino acid residues in TM10 are important for substrate recognition.
A DNA fragment containing GAL2 was cut out by PmaCI and EcoRI and ligated to SmaI and EcoRI sites in a multicloning site of pTV3, a YEp vector (3). The nucleotide sequence immediately following the initiation codon was modified from ATGGCAGTTGAG to ATGGCAGAATTC to create an EcoRI site, which changed the deduced amino acid sequence from Met-Ala-Val-Glu to Met-Ala-Glu-Phe. The nucleotide sequence immediately following the termination codon, TAATGCGTT, was modified to TAATCGATT to create a ClaI site. These two restriction enzyme sites were used to replace the open reading frame of GAL2 with HXT2. To do this, the nucleotide sequence immediately following the termination codon of HXT2 was modified from TAAGAGATT to TAATCGATT to create a ClaI site. Because HXT2 has an EcoRI site extending from the 7th to the 12th nucleotides of the coding sequence, the EcoRI site and the aforementioned ClaI site were used to replace GAL2 with HXT2. Plasmids were introduced into LBY416 (MATa hxt2::LEU2 snf3::HIS3 gal2 lys2 ade2 trp1 his3 leu2 ura3) (3).
Multiple Mixed MutagenesisTwo set of degenerate PCR
primers (see Fig. 1), a forward primer containing 84 nucleotides, and a
reverse primer containing 56 nucleotides were synthesized to create a
mixture of 24,576 (29·3·42) clones that
encode all the possible combinations of amino acid residues in the TM10
region that are different in Gal2 and Hxt2. When designing PCR primers,
the codons were changed to reduce the number of distinct primers: 1) at
locations where the codons for the same amino acid in Gal2 and Hxt2
were different, only one of the codons was adopted, and 2) at locations
where corresponding amino acids were different, the codons that
required the synthesis of the minimal number of degenerate primers were
employed. This is a modification of the method of random mutagenesis
using degenerate oligonucleotides (13, 14). PCR reactions using
Taq polymerase were performed (2400, Applied Biosystems)
without adding a DNA template because the two degenerate PCR primers
possessed a 10-nucleotide-long complementary overlap. The reactions
were started with a preincubation for 4 min at 94 °C, and this was
followed by 25 cycles of denaturation at 94 °C for 1 min, annealing
at 55 °C for 5 min, and elongation at 72 °C for 2 min and with
incubation at 72 °C for 4 min. After cutting both ends with
SacI and MluI, the resulting PCR products were
used to replace the corresponding TM10 region in HXT2 in the
plasmid Hxt2-0 (4). After amplification in E. coli, the plasmid mixture was introduced into LBY416 and the galactose
transport-positive clones were selected after incubation for 4-5 days
at 30 °C on galactose limited agar plates containing a synthetic
medium (15) supplemented with 200 µg/ml galactose. When random
mutagenesis of 1 or 2 amino acids in Gal2 or Hxt2 was necessary, a
degenerate PCR primer was prepared in which nucleotides were mixed to
produce codons for 20 amino acids at the appropriate position(s) in the sequence.
Other Assays
The nucleotide sequence of the replaced fragment and the surrounding regions in each of the clones used in this study was verified by sequencing both strands using a DNA sequencer (373A, Applied Biosystems). Immunoblotting of yeast homogenates (4) and transport of galactose and glucose in yeast cells was performed as described previously (3, 4), except that the transport assay medium used was 50 mM MES·NaOH (pH 6.0) containing 2 mM MgSO4, and the transport reaction was stopped using assay medium containing 0.5 mM HgCl2. No appreciable differences in transport characteristics were noted between the two assay media.
We have devised a novel method to systematically identify amino acid residue(s) in TM10 that are critical for the substrate recognition. We have tentatively named this method "multiple mixed mutagenesis." We first used PCR to prepare a mixture of approximately 25,000 distinct sequences that encode all the possible combinations of different amino acid residues between Gal2 and Hxt2. The PCR products were substituted for the corresponding TM10 region of HXT2, and these constructs were introduced into the galactose transport-negative strain LBY416. Comparison of the deduced amino acid sequences of 19 clones selected as galactose transport-positive clones revealed that Tyr446 was the only location where an amino acid residue from Gal2 was always found (Fig. 1). In addition, the Trp455 of Gal2 was present in 14 of 19 clones, with Cys present at this location in the remaining 5 clones. LBY416 cells harboring each of the 19 clones showed galactose transport (Fig. 1) at levels 8-36% of the original Gal2 transporter, a level of transport comparable with the 17% observed for the chimeric transporter Hxt2-8 (4) containing TM10 derived from Gal2 and the remaining parts from Hxt2.
To confirm that Tyr446 and Trp455 are important
for galactose recognition, several experiments were performed. 1) A
random mixture of Gal2 clones encoding 20 different amino acid
substitutions at Tyr446 and Trp455 was
subjected to galactose transport-positive selection. Each of the 15 clones picked up possessed Tyr446 and Trp455.
2) Similarly, when Phe431 and Tyr440 of Hxt2
(corresponding to Tyr446 and Trp455 of Gal2,
respectively) were randomly substituted with 20 amino acids, 13 galactose transport-positive clones were found to contain Tyr431. By contrast, 5 Trp440, 3 Cys440, 4 Thr440, and 1 Leu440
clones were observed among the galactose transport-positive clones. 3)
In addition, random mutagenesis of Tyr440 in a modified
Hxt2 containing Tyr431 yielded 13 Trp440, 3 His440, and 1 Met440 galactose
transport-positive clones. 4) Tyr446 of Gal2 was replaced
with the other 19 amino acids, and galactose and glucose transport
activities were measured (Fig. 2). Only the
Tyr446 clone was active in galactose transport, whereas
several clones including Gal2 containing Phe446,
Trp446, and Tyr446 were active in glucose
transport. Expression of all the 20 clones in this series was confirmed
by immunoblotting of cell homogenates (Fig. 3). Thus,
with the change of Tyr446 to Phe, the Gal2 transporter is
changed to a glucose-specific transporter, the activity of which in the
presence of 0.1 mM glucose is about half that of the Hxt2
transporter. 5) Hxt2 containing Tyr431 and
Trp440, designated Hxt2 (Tyr-Trp), showed almost the same
characteristics of galactose transport as Hxt2-8: the
Km and Vmax of galactose
transport in Hxt2 (Tyr-Trp) cells were 39 ± 6 mM
(mean ± S.E., n = 3) and 1.0 ± 0.7 nmol/107 cells/5 s, and those of Hxt2-8 cells were 41 ± 2 mM and 0.69 ± 0.03 nmol/107 cells/5
s. By contrast, galactose transport in Gal2 cells has a
Km of 5.3 ± 0.3 mM and a
Vmax of 0.97 ± 0.1 nmol/107
cells/5 s (4). The fact that Hxt2 (Tyr-Trp) showed almost the same
Km as that of Hxt2-8, which was 8-fold higher than
that of Gal2, suggests that there is some other region(s) contributing
to galactose recognition in Gal2 but that there are no other critical
amino acid residues within TM10. No significant differences in
substrate specificity were found in these cells (Fig.
4). Thus, changing two critical amino acids
(Phe431 to Tyr and Tyr440 to Trp) causes the
Hxt2 transporter to recognize galactose in addition to glucose.
Previous studies on the Glut family transporters have overlooked the
importance of the amino acid situated at Tyr446 in Gal2 and
the equivalent residues in other transporters. When TM10 is depicted as
forming an -helix (Fig. 5), Tyr446 is
situated in the middle of the amino acids that are different in Gal2
and Hxt2, whereas the other side of the
-helix consists of amino
acid residues that are common to both Gal2 and Hxt2. Tyr446
is not conserved in Glut family transporters that also transport galactose in other organisms including GalP in E. coli (16), STP1 in Arabidopsis thaliana (17), HUP2 in Chlorella
kessleri (18), and SGTP1 in Schistosoma mansoni (19).
This suggests that there are subtle changes in substrate recognition
sites among these homologous transporters. The importance of
Trp388 of the mammalian Glut1 transporter (corresponding to
Trp455 of Gal2) has been pointed out previously (20-23).
Replacement of Trp with other amino acids changed several aspects of
transporter physiology including expression levels (20), targeting
(20), reduction in intrinsic activity under certain conditions (21), and forskolin binding (22, 23). It is of particular interest to
determine whether the replacement of Trp with other amino acids brings
any change in various functional aspects of Gal2 in addition to the
substrate recognition. The role of Tyr446 and
Trp455 of Gal2 in substrate recognition may be interpreted
in many ways. These amino acid residues may relieve steric hindrance,
so that galactose that is excluded from Hxt2 may be accepted by Gal2. However, the fact that Gal2 transporters possessing different amino
acids at residue 446 show more strict selectivity for galactose transport compared with glucose transport (Fig. 2) may not be consistent with this idea. It is possible that the two aromatic side
chains of Tyr and Trp form hydrogen bonds with galactose. With this
interpretation, the structure around C4 of galactose may form a bond
primarily with Tyr. Another possibility is that these aromatic residues
lie close to galactose ("stacking effect") as is the case in
galactose binding lectins (24). The possibility that these residues
play an indirect but crucial role in determining the conformation of
amino acid residue(s) directly interacting with galactose is not
excluded by this study. It is also possible that two amino acids may
not form a single recognition site but instead function independently.
Considerable evidence supports the notion of two substrate binding
sites in Glut family transporters (2).
Recently Arbuckle et al. (9) constructed eight chimeras of human Glut2 and Glut3 and used them to show that TM7 is important for substrate recognition. The reason why the region identified for substrate recognition is different from ours is not known. The difference in the results of their study and ours may reflect difference in homologous transporters in human and yeast or differences in the transport substrates used; Arbuckle et al. used fructose as a discriminating sugar, whereas we used galactose. If the latter possibility proved to be the case, it would suggest that both regions are necessary for the recognition of sugar molecules. It is also possible that TM7 and TM10 contribute to two different binding sites that have different substrate specificities, TM7 for exofacial binding site and TM10 for endofacial binding site (2).
Using a three-step chimera approach, we were able to identify the amino acid residues responsible for the differential recognition of galactose and glucose. At each step we have tried to avoid making assumptions about particular locations. It seems reasonable to expect that this multiple mixed mutagenesis method will be generally applicable to other classes of proteins that have homologous counterparts. This method is an alternative to site-directed mutagenesis and is an appropriate approach for determining not only the substrate recognition site but also other functionally important sites.
We thank H. Ronne (Ludwig Institute for Cancer Research, Uppsala, Sweden), L. Bisson (University of California, Davis), and J. Nikawa (Kyushu Institute of Technology, Iizuka, Japan) for providing us plasmids and yeast cells. Thanks are also due to D. W. Saffen (University of Tokyo) for critical reading of the manuscript.