Analysis of the Antimalarial Drug Resistance Protein Pfcrt Expressed in Yeast*

Mutations in the novel membrane protein Pfcrt were recently found to be essential for chloroquine resistance (CQR) in Plasmodium falciparum, the parasite responsible for most lethal human malaria (Fidock, D. A., Nomura, T., Talley, A. K., Cooper, R. A., Dzekunov, S. M., Ferdig, M. T., Ursos, L. M., Sidhu, A. B., Naude, B., Deitsch, K. W., Su, X. Z., Wootton, J. C., Roepe, P. D., and Wellems, T. E. (2000) Mol. Cell6, 861–871). Pfcrt is localized to the digestive vacuolar membrane of the intraerythrocytic parasite and may function as a transporter. Study of this putative transport function would be greatly assisted by overexpression in yeast followed by characterization of membrane vesicles. Unfortunately, the very high AT content of malarial genes precludes efficient heterologous expression. Thus, we back-translated Pfcrt to design idealized genes with preferred yeast codons, no long poly(A) sequences, and minimal stem-loop structure. We synthesized a designed gene with a two-step PCR method, fused this to N- and C-terminal sequences to aid membrane insertion and purification, and now report efficient expression of wild type and mutant Pfcrt proteins in the plasma membrane of Saccharomyces cerevisiaeand Pichia pastoris yeast. To our knowledge, this is the first successful expression of a full-length malarial parasite integral membrane protein in yeast. Purified membranes and inside-out plasma membrane vesicle preparations were used to analyze wild typeversus CQR-conferring mutant Pfcrt function, which may include effects on H+ transport (Dzekunov, S., Ursos, L. M. B., and Roepe, P. D. (2000) Mol. Biochem. Parasitol. 110, 107–124), and to perfect a rapid purification of biotinylated Pfcrt. These data expand on the role of Pfcrt in conferring CQR and define a productive route for analysis of importantP. falciparum transport proteins and membrane associated vaccine candidates.

Malaria causes ϳ2.5 million deaths annually, mostly children. Four malarial species infect humans, the most deadly being Plasmodium falciparum. For decades, malaria has been treated effectively with the 4-aminoquinoline chloroquine (CQ) 1 ; however, CQ-resistant (CQR) strains of P. falciparum have spread with considerable success and continue to evolve. Mutations in two genes (pfmdr and cg2) were previously suggested to mediate CQR, but more recently CQR was found to be caused by mutations in the Pfcrt gene (1). Pfmdr1 protein likely plays a modulatory role that influences the drug resistance profile (2). Pfcrt protein is localized to the membrane of the parasite digestive vacuole (DV). The DV is the site of hemoglobin digestion, which is a principle source of food for the parasite during rapid intraerythrocytic development. Mutant Pfcrt may confer drug resistance by directly or indirectly lowering DV pH (3), which quite effectively titrates soluble drug target (heme released from hemoglobin) out of solution without compromising detoxification of heme to hemozoin (4,5). There are actually multiple mutant Pfcrt alleles that appear to cause drug resistance and that arose in a geographically distinct pattern (1).
Further elucidating the molecular mechanism of CQR is essential. Pfcrt plays a critical role in CQR, probably via some yet to be determined transport function. The protein resides in a subcellular membrane within an intracellular parasite, so to experimentally study Pfcrt transport function requires that transported substrates cross three membranes in a coordinated fashion. This is extremely difficult to manipulate experimentally. Since the technology required for fabricating membrane vesicles of various types (e.g. secretory, inside out plasma, right side out plasma, vacuolar) and for purifying and reconstituting polytopic integral membrane proteins is well developed for yeast, heterologous expression of Pfcrt in yeast would obviously greatly assist further analysis of Pfcrt transport function.
Unfortunately, a literature survey for successful heterologous expression of P. falciparum genes is not encouraging. There is but one successful report of low level heterologous expression of a native P. falciparum cDNA (dihydrofolate reductase-thymidylate synthase (DHFR-TS) protein (6)). A notorious feature of the P. falciparum genome is its very high A,T content (7,8), thus, P. falciparum genes reveal a markedly biased codon usage. Coding regions are ϳ70% AϩT, and are flanked by AϩT-rich regions as high as 86% (9 -12). The high AϩT content makes heterologous expression in both prokaryotic (bacterial) and eukaryotic (yeast, insect) systems extremely difficult if not impossible because: (a) these species have their own preferred codon usage that is not dominated by AϩT (13)(14)(15)(16), and (b) AϩT-rich sequences tend to form putative polyadenylation (poly(A)) sites as well as efficiency and posi-tioning elements in yeast that result in terminated, truncated, or poorly transcribed mRNAs (17)(18)(19)(20).
Thus, expression of a synthetic DHFR-TS gene with optimized codon usage was found to increase the level of expressed protein (21). This result recently enticed others to construct genes with optimized yeast codon usage for pfsub1 (a subtilisinlike protease), pfmsp-1 (a merozoite stage-specific surface protein complex), and the antigen Pfs48/45 (22)(23)(24). These three studies have yielded some additional success; however, heterologous expression of P. falciparum genes remains extremely difficult, and no successful overexpression of a polytopic integral membrane protein has yet been reported. Polytopic integral membrane proteins are typically encoded by large genes that are more difficult to assemble synthetically. Also, aside from codon usage and poly(A) termination issues, the encoded proteins contain folding, membrane translocation, and membrane insertion sequences that can be species-and membranespecific. These are not well defined for malarial membrane proteins.
We have examined these issues and have constructed a synthetic Pfcrt gene that harbors appropriate base pair content and other necessary features. We have modified the N and C termini regions based on previous work wherein we were able to functionally express the human multidrug resistance protein (humdr1, P-glycoprotein) in S. cerevisiae (27), have subcloned the resulting constructs into appropriate yeast expression vectors, and have achieved high level inducible expression of Pfcrt and Pfcrt-biotin acceptor domain fusion (Pfcrt-bad) proteins in the plasma membrane of P. pastoris. More modest constitutive expression in S. cerevisiae was also achieved. Using the successfully expressed wild type gene as template, we have also created a CQR-associated mutant Pfcrt allele and have overexpressed the mutant Pfcrt-bad protein. Analysis of membrane vesicles from these yeast supports the earlier suggestion (1,3,5) that Pfcrt is involved in modulating H ϩ transport.

EXPERIMENTAL PROCEDURES
Materials-Cloned Pfu polymerase was from Stratagene (La Jolla, CA). Rabbit anti-Pfcrt IgG was a kind gift from Dr. T. Wellems (NIAID/ National Institutes of Health, Bethesda, MD). HRP-conjugated monkey anti-rabbit IgG, HRP-conjugated streptavidin and ECL detection reagents were from Amersham Biosciences. Prestained SDS-PAGE molecular markers were from Bio-Rad. Immobilized monomeric avidin resin was from Pierce. Yeast and bacteria growth media reagents were from Difco. Oligonucleotides were custom made by MWG-Biotech (High Point, NC). Dodecyl maltoside (DM) was from Calbiochem (San Diego, CA). Plasmids and sequencing primers relevant for subcloning and gene expression in P. pastoris yeast were purchased from Invitrogen (version L Picchia expression kit). All other reagents were reagent grade or better and were purchased from Sigma.
Designing the Synthetic (Yeast-optimized) Wild Type Pfcrt Gene-The Pfcrt-hb3 gene sequence (wild type Pfcrt) was obtained from Gen-Bank TM (www.ncbi.nlm.nih.gov) and we used the CODOP program (generously provided by Dr. Elisabeth P Carpenter, Division of Protein Structure, National Institute for Medical Research, London, UK (22)) to back-translate the encoded protein sequence. We allowed CODOP to back-translate Pfcrt into a theoretically optimized Pfcrt gene using a S. cerevisiae yeast codon usage table (www.kazusa.or.jp/codon). The process was repeated many times with different random seeding values. All the poly(A) and premature codon patterns in a selected sequence were then screened and destroyed by a second layer of codon engineering. We suspect disrupting poly(A) is more important than 100% optimization of codon usage (see "Results," Table I caption). A Kozak consensus (GC-CGCCACCAUGG) was included and adjusted at Ϫ3 (to A) and ϩ4 (to G). Finally, this theoretically optimized Pfcrt gene was divided up into a collection of 40 base fragments encoding both DNA strands and the melting temperature (T m ) for the 20-bp overlap regions in each primer set (see "Results") were calculated. All primer set T m were then adjusted to reside in the range 56 -64°C by yet a third round of codon adjustment. Vector NTI software was also used to check all primers for repeats, palindromes, hairpins, and dimers. In the final theoretical optimized Pfcrt gene AT% is reduced from 72 to 55% (see "Results"). The initial PCR program was one denaturation step at 94°C for 1 min, followed by 25 cycles of 94°C (30 s), 52°C (30 s), and 72°C (2 min) and then a final incubation at 72°C for 10 min. 10 l of this "assembly solution" was then diluted 10-fold in 100 l of similar PCR mixture, but with a 1 M concentration each of the 5Ј-and 3Ј-flanking primers (1 and 34 in the sequence of 66, respectively). The amplifying PCR program used was one denaturation step at 94°C for 1 min, followed by 25 cycles at 94°C (45 s), 68°C (45 s), and 72°C (5 min) and then a final incubation cycle at 72°C for 10 min.
A sequence of PCR reactions was required to synthesize the fulllength optimized Dd2crt-bad gene using the optimized HB3crt-bad gene (wild type Pfcrt-tcbd fusion gene described above) as initial template. First, six PCR fragments (A-F) were amplified from HB3crt-bad using the six complimentary sets of 12 oligonucleotides that created the codon and restriction site mutations described above. These were purified, combined in pairs of two, and PCR-amplified to yield three larger fragments I, II, III (e.g. fragment I encoding codons 1-225 was created using fragments A and B as template), etc. By using this sequential strategy, the entire Dd2crt-bad gene was constructed via tiered PCR, subcloned into pYKM77 as described below, and colony-amplified and purified.
Construction of Recombinant Plasmids-Plasmid pYKM77 (kindly provided by Drs. K. Kuchler and J. Thorner) was isolated and restricted with PstI and HindIII. The synthetic HB3crt gene, the synthetic HB3crt-bad fusion gene, and the synthetic Dd2crt-bad gene were each trimmed with PstI and HindIII and ligated to the 5.4-kbp vector. Recombinant plasmids were isolated from bacterial transformants and analyzed by restriction. Candidate pHZHB3crtbad (encoding the wild type fusion protein under control of a modified Ste6 promoter), pHZDd2crtbad (encoding mutant crt fusion protein), and pHZHB3crt (encoding wild type Pfcrt) plasmids were sequenced in both directions using oligonucleotides from the gene assembly steps, the BigDye Terminator cycle sequencing program, and ABI prism 373 software. None of the dozen or so sequences that were fully sequenced in each case properly encoded the proteins of interest. Typically, two to three unwanted point mutations were found in the fully assembled recombinant clones. However, by combining restriction fragments from these constructs (using sites fortuitously placed during the initial gene design) our final synthetic genes encoding full-length proteins of the correct sequence were ligated together.
These genes were then also subcloned into the vector pPIC3.5 to create pPIC35hb3crt, pPIC35hb3crtbad, and pPIC35dd2crtbad for expression in P. pastoris. The pPIC vector harbors an inducible promoter activated by MeOH, as described under "Results." Yeast Transformation-Yeast were transformed by the lithium acetate method with 2 g of target plasmid and 10 g of carrier plasmid to enhance transformation efficiency. Transformants were plated on selective synthetic complete medium lacking uracil (S. cerevisiae) or MGM (P. pastoris) agar plates.
Isolation of Yeast Crude and Plasma Membranes-Yeast cells were grown to midlog phase, and crude cellular membranes were isolated via a glass bead lysis protocol (27) and stored at Ϫ80°C. Plasma membranes were purified via the acid precipitation method of Goffeau and Dufour (38), and clear plasma membrane pellets were resuspended in glycerol-containing solution and stored at Ϫ80 C.
Preparation of Inside-out Plasma Membrane Vesicles-Inside-out plasma membrane vesicles (ISOV) were isolated following the procedure described by Menendez et al. (39) with some modifications (27). In particular, ISOV were fabricated with 100 mM KCl inside the vesicles so that H ϩ pumping could be analyzed with symmetrical high Cl Ϫ on either side of the plasma membrane. When elevating [salt], [sucrose] was adjusted to conserve osmolality.
Yeast Vacuole Isolation-Our method is adapted from Ohsumi and Anraku (40) with some minor modifications.
Parasite Digestive Vacuole Isolation-The P. falciparum DV was isolated following procedures described in Ref. 41, with some modifications. 5 ml of culture suspensions of twice synchronized midtrophozoites of the Sudan 106/1 strain (at ϳ12% parasitemia) were washed three times in PBS (pH 7.4). Each 5-ml sample was resuspended in PBS containing 0.15% saponin, incubated for 5 min, and centrifuged to collect trophozoites. The isolated trophozoites were washed repeatedly in ice-cold PBS until the supernatant was clear, resuspended in 10 volumes of ice-cold trituration buffer (0.25 M sucrose, 10 mM NaP i (pH 7.10), 0.5% streptomycin sulfate), and triturated three times on ice using a 27-gauge needle. The suspension was centrifuged (20,000 ϫ g/2 min/4°C), supernatant was discarded, and the pellet resuspended in 5 volumes of 2 mM MgSO 4 , 100 mM KCl, 10 mM NaCl, 25 mM Hepes, 25 mM NaHCO 3 , 5 mM NaP i (pH 7.10)). To 1 ml of the suspension 20 l of 5 mg/ml DNase I was added, and the suspension was incubated at 25°C for 5 min, followed by centrifugation as before. The supernatant was discarded, and the pellet was resuspended in 5 volumes of ice-cold trituration buffer and triturated once again. The suspension was layered on 7 ml of ice-cold 42% Percoll solution containing 0.25 M sucrose, 1.5 mM MgCl 2 , 10 mM NaP i (pH 7.10), and centrifuged (16,000 ϫ g for 30 min at 4°C). The bottom layer containing the purified food vacuoles was washed three times in the same ice-cold buffer, resuspended in 200 l of washing buffer, and stored at Ϫ80°C.
Western Blot and Biotin Detection-14% SDS-PAGE gels were run at 110 V for about 2 h, proteins transferred onto polyvinylidene difluoride membranes at 40 mAmp overnight, and Western or biotin detection blots were performed the next day. For Western blots, membranes were washed with PBS-Tween (PBS-T) solution for 30 min, blocked with 5% milk in PBS-T for 1 h, incubated with primary Ab for 1 h, washed three times with PBS-T (15 min each), incubated with secondary Ab (whole anti-rabbit IgG-HRP) for 1 h, and washed again. Biotin detection was essentially the same, with streptavidin-HRP in PBS-T (1:5000 dilution) as the primary (and only) Ab. Membranes were stripped with 62.5 mM Tris-HCl (pH 6.80), 2% SDS, 100 mM ␤-mercaptoethanol at 50°C for 30 min with gentle shaking.
ATPase Activity Assays-We followed the method of Chifflet and To fuse the tcbd biotin acceptor domain to the C terminus of Pfcrt, the tcbd fragment was amplified from YEp352-BIO6 using primers that created NotI and HindIII restriction sites and that removed a PstI site, purified, and then combined with purified Pfcrt in an additional PCR reaction (STEP THREE).
colleagues (42) with some modifications. Briefly, ISOV were mixed with ATP in an assay buffer containing 50 mM Tris-Mes (pH 7.5), 90 mM NH 4 Cl, 5 mM MgCl 2 , and 0.01% NaN 3 . The reaction was initiated at 37°C for 1 min and quenched by the addition of SDS and ammonium molybdate. Absorbance of the generated inorganic phosphate-molybdate complex was measured at 700 nm.
Proton Pumping Assay-Proton pumping assays were performed with ISOV essentially as described (27,43) with some modifications. Namely, 2 M acridine orange (AO) was used in the assay instead of 20 M, excitation was at 490 nm instead of 465 nm, and the transport buffer composition in most experiments was 330 mM sucrose, 100 mM KCl, 10 mM Mes-Tris, 4 mM MgCl 2 , 1 mM dithiothreitol. Some experi-

TABLE I
Comparison of malarial versus yeast codon preferences and codon usage for native versus yeast optimized Pfcrt Use per 1000 codons is listed for P. falciparum and S. cerevisiae (www.kazusa.or.jp/codon) (third and fourth columns). Codon usage in native pfcrt (fifth column) was compared with preferred usage for S. cerevisiae (fourth column). Subsequently, the optimized gene was analyzed for poly(A) and premature termination sequences, and the sequence was further adjusted. Thus, for example, even though the yeast preferred codon for lysine is AAA, all lysine residues were coded with AAG to avoid poly(A) sequences (compare last two columns) in the final optimized gene.  Fig. 1 is a summary of synthetic (yeast-optimized) wild type Pfcrt and wild type Pfcrt-tcbd gene assembly, and Table I shows the final codon composition of the constructed wild type Pfcrt gene. Once an optimized wild type Pfcrt gene was designed using CODOP and other computer-based algorithms (see "Experimental Procedures"), the two strands of the gene were divided into 66 40-mer oligonucleotides with 20-bp overlap regions. Sequences of the overlap regions were then adjusted to yield similar T m , which we find to be very important for efficient assembly of a gene this size in the first PCR step (cf. Fig. 1). After 25 cycles, trace levels of assembled full-length gene (not visible on a conventional agarose DNA gel) were then amplified in the second PCR step (Fig. 1). After DNA sequencing of recombinant clones (which confirms the error rate of the polymerase), a full-length optimized gene was assembled from several restriction fragments that contained no errors. As summarized in Table I, importantly, the overlap T m , stem loop structure, and poly(A) region modifications we made to the initially designed gene (see "Experimental Procedures") led to the final gene having a non-negligible number of yeast codons that are not "highly preferred." Therefore, since neither the native wild type or CQR-associated Pfcrt sequences nor a wild type sequence optimized at the N-terminal 50 codons is expressed in yeast, but the final synthetic genes are (see below), efficient expression of malarial genes in yeast requires conversion of most, but not necessarily all, "unpreferred" codons.
The optimized wild type Pfcrt gene was subcloned into a yeast expression vector based on pYKM77 (26), which was previously modified for high level expression of the human mdr1 protein in yeast (27,43), to create pHZHB3crt. Also, the optimized gene was fused in-frame to a 270-nucleotide region of the P. shermanii transcarboxylase gene (Fig. 1, STEP 3) that encodes the minimal biotin acceptor domain (37) and similarly subcloned to create pHZHB3crtbad. The wild type Pfcrt-bad gene was also used as template to construct pHZDd2crtbad encoding CQR-associated mutant crt (see "Experimental Procedures"). S. cerevisiae were transfected with each construct, as well as similar constructs harboring the native P. falciparum HB3 (CQS) and Dd2 (CQR) crt alleles. As shown in Fig. 2, the optimized HB3 and Dd2 gene sequences were constitutively and efficiently expressed, whereas the native gene sequences were not.
In addition, Pfcrt, HB3crt-bad and Dd2crt-bad genes were appropriately restricted and subcloned into the vector pPIC3.5 (Invitrogen). This nonfusion vector integrates at the AOX1 (alcohol oxidase) locus when linearized prior to transformation, followed by histidine selection. Thus, expression of the cloned gene is induced by a convenient metabolic switch (glycerol to MeOH as the sole carbon source). As shown in Fig. 3, optimized HB3crt (A and B, lane 3) and optimized HB3crt-bad (A and B, lane 4) were even more efficiently expressed in P. pastoris, and inducible expression plateaued in ϳ12 h (Fig. 3C). Dd2crt-bad is expressed to similar levels in P. pastoris, but the time course for efficient plasma membrane insertion is slower relative to HB3crt-bad (Fig. 4). Whether this result implies a physiologically significant alteration in processing of the CQR-associated mutant within the native P. falciparum environment requires additional detailed study.
Since the levels of expression were higher in P. pastoris and at plateau were equal for HB3crt-bad versus Dd2crt-bad, subsequent functional analysis was performed with P. pastoris membranes harvested 16 h postinduction. In our analysis pre-sented below, our negative controls were isolated after identical MeOH induction treatments performed side-by-side.
Previously (27) we developed a convenient assay for monitoring pH gradient formation in yeast ISOV preparations. Since mutant Pfcrt protein appears to be involved in modulating the pH gradient across the DV membrane (3), we analyzed pH gradient formation in ISOV from control KM71/ pPIC3.5 versus KM71/pPIC35HB3crtbad versus KM71/ pPIC35Dd2crtbad strains (Fig. 5). Importantly, the experiments shown in Fig. 5 were with ISOV fabricated with high (100 mM) [Cl Ϫ ] inside so that we better approximated the conditions of the native P. falciparum DV (likely high internal [Cl Ϫ ]). Interestingly, Fig. 5 shows that in the presence of symmetrical 100 mM [Cl Ϫ ] the magnitude of ⌬pH formed in ISOV harboring Dd2crt-bad (light gray lines) is conspicuously higher than in control (black line), but ⌬pH for ISOV harboring HB3crt-bad (dark gray line) is similar. In Fig. 5, A and C, we show two comparisons between six different independent ISOV preparations to demonstrate reproducibility (see also Fig. 6, left-hand side).
Based on these results, we next tested another prediction of one hypothesis that has been offered for Pfcrt function, namely, that it might perform some type of anion transport (perhaps Cl Ϫ ) to assist maintenance of the high chemical gradient in H ϩ that likely exists across the DV membrane (44). Presumably, similar to other endo membranes that maintain high ⌬pH, dissipation of the electrical potential difference caused by H ϩ influx, via counter balanced diffusion of Cl Ϫ , is required to maintain high DV membrane ⌬pH. Fig. 5, B and D, show that upon substituting Cl Ϫ in the pH gradient assay buffer with equimolar glutamate there is nearly complete normalization of ⌬pH formation for the three ISOV preparations. Again, two comparisons using six independent preparations are shown to demonstrate reproducibility. Therefore, increased ⌬pH formation provided by mutant Pfcrt-bad is Cl Ϫ -dependent. As shown in Fig. 6, the increased ⌬pH for Dd2crt-bad ISOV is also more sensitive to verapamil (VPL) relative to control or HB3crt-bad ISOV. This is consistent with partial reversal of resistance due Since the amplitude of ATP-dependent ⌬pH formation was increased via expression of Dd2crt-bad, we next analyzed liberation of inorganic phosphate from these ISOV preparations (Fig. 7). ISOV from KM71/Dd2crt-bad yeast fabricated with high [Cl Ϫ ] inside (dotted bars) showed elevated release of inorganic phosphate relative to control (closed bars) or KM71/ HB3crt-bad (hatched bars) ISOV. This effect was again reversed by VPL (Fig. 7, center) and partially reversed by FCCP (Fig. 7, right-hand side). Release of P i was inhibited by vanadate with similar K i for the two preparations (not shown), consistent with either a stimulation or up-regulation of the endogenous (presumably PMA1-like (45)) H ϩ -ATPase found in the P. pastoris plasma membrane. To test the latter, we analyzed H ϩ -ATPase levels in the plasma membranes of control versus Dd2crt-bad versus HB3crt-bad yeast via Western blot. As shown in Fig. 8 (for two independent sets of preparations, the same preparations as those used for experiments in Fig. 5), to the best of our ability to ascertain, the plasma membrane ATPase levels are nearly identical for the three strains even though the apparent H ϩ translocation ability and ATPase activity are higher for Dd2crt-bad.
Finally, obviously, additional detailed studies that expand upon these initial observations would be greatly assisted by purification of the biotinylated enzyme. Fig. 9 shows that a combination of membrane purification, membrane wash, solubilization with DM, and monomeric avidin chromatography results in a rapid (at least 2000-fold, see caption) purification of HB3crt-bad protein.

DISCUSSION
Our results may be summarized as follows. 1) The P. falciparum gene critical for evolution of CQR, Pfcrt, was optimized for expression in the yeast S. cerevisiae. Expression was detected, and a higher expression of the same constructs was found in the yeast P. pastoris. To our knowledge, this is the first successful demonstration of overexpression of a polytopic integral membrane protein from Plasmodia in yeast. 2) Wild type (HB3) and CQR-associated mutant (Dd2)crt-bad are found in the plasma membrane of both strains of yeast. Levels are not affected by adding the bad domain, but the rate of folding or of membrane incorporation in P. pastoris appears slower for Dd2crt-bad versus HB3crt-bad. 3) Initial studies indicate that when ISOV are created with high internal [Cl Ϫ ] and diluted into equimolar Cl Ϫ , Dd2crt-bad stimulates release of inorganic phosphate and H ϩ -pumping ability via the PMA1 H ϩ -ATPase, in a Cl Ϫ -dependent fashion. This effect is also inhibited by VPL. 4) Rapid purification of Pfcrt-bad is accomplished via DM solubilization and avidin-biotin affinity chromatography.
These results have a variety of important implications. There is tremendous interest in the transport biology of P.  Fig. 2, KM71 Pfcrt-bad membranes show an immunoreactive and biotinylated band that runs at a higher molecular weight because of the added biotin acceptor domain. Importantly, fusion of the tcbd-encoded domain to Pfcrt (Fig. 1, STEP 3) does not affect levels of Pfcrt expressed (A, compare lanes 3 and  4). Comparison of lanes 1-4 in B reveals expected endogenously biotinylated yeast proteins that are useful for verifying equivalent protein loading per lane and the consistency of membrane (ISOV) preparations. C shows the time course of HB3crt-bad MeOH induction in P. pastoris. Each lane contains 50 g of membrane protein, and Pfcrt-bad is detected with avidin-HRP. Odd numbered lanes are KM71/pPIC3.5 (control), even numbered lanes are KM71/HB3crt-bad, and the time points examined for each (proceeding left to right) are 0 (no induction) 1, 3, 6, 9, 12, 24, and 48 h. HB3crt-bad is fully induced in P. pastoris within 12-24 h. Cell fractionation (see "Experimental Procedures") followed by similar gel analysis reveals that most expressed HB3crt and HB3crt-bad is found in the crude membrane fraction, with ϳ50% of this localized to P. pastoris plasma membrane (not shown). More detailed subcellular fractionation work will be reported elsewhere.
falciparum, because alterations in that transport are linked to drug resistance which causes millions of deaths annually. Recently, mutations in the Pfcrt gene were linked to CQR in P. falciparum (1), but the function of the polytopic integral membrane protein encoded by this gene is unknown. Since 1) the protein product is localized to the DV membrane, 2) a high ⌬pH is physiologically required across this membrane, 3) lowered accumulation of weak base antimalarials is found in resistant parasite DV, and 4) mutations in Pfcrt have been linked to possible changes in DV pH (3)(4)(5), it is likely that Pfcrt performs some type of molecular membrane transport function. Elucidation of this function within the native DV membrane (a subcellular membrane within a cell within another cell) would be exceedingly difficult, if not impossible. Thus, heterologous expression in yeast followed by biochemical characterization of well defined yeast membrane vesicle preparations offers one conspicuously attractive route for further analysis. Unfortunately, as described earlier, the base composition of malarial genes makes this impossible unless the gene is optimized to reflect preferred yeast codon bias and to eliminate deleterious mRNA structures. Only a few examples of gene optimization have been reported, with varying levels of success, and yeast optimization of no genes the size of Pfcrt (to our knowledge) has yet been attempted. Moreover, since polytopic integral membrane proteins are notorious for exhibiting processing and membrane insertion difficulties when expressed heterologously, it is justifiably thought by most investigators that attempting heterologous expression of malarial genes encoding polytopic membrane proteins is a very risky proposition. Our results show that, assuming the gene is designed as described, it is less risky than initially anticipated.
There are two reasons why this simple result is very important. The first is, again, tremendous interest is focused on membrane transport pathways of P. falciparum. For example, nucleotide transport (46), sugar transport (47), and H ϩ transport by unusually intriguing pyrophosphate-hydrolyzing pumps (48) are all catalyzed by specific transporters, and all offer unique insight into Plasmodia biology and (hopefully) therapeutic intervention. With the continued growth of malaria as a shockingly severe global health threat, developing meth-  (27) was used to assess formation of a ⌬pH in ISOV via the yeast plasma membrane H ϩ -ATPase for KM71/pPIC3.5 (black line each panel), KM71/HB3crt-bad (dark gray line), and KM71/Dd2crt-bad (light gray line). As the inside of the vesicle acidifies upon addition of 2 mM ATP (first arrow; a slight blip upon addition of ATP has been removed for clarity), AO is trapped via the weak base effect and forms aggregates, thus strongly quenching fluorescence (27,40). After plateau, high [NH 4 Cl], 5 M nigericin, or 100 M vanadate similarly dissipate the formed ⌬pH (second arrow (27,40)). 25 g of ISOV (quantified by Amido Black assay (27)) are used in each experiment. A and C, as well as B and D, compare two independent control versus two independent HB3crt-bad versus two independent Dd2crt-bad ISOV preparations to demonstrate reproducibility. A and C are in experiments performed in the presence of high symmetrical KCl (100 mM each side of the membrane) whereas B and D are the same ISOV but diluted into equimolar potassium glutamate (no symmetrical Cl Ϫ , but a high Cl Ϫ gradient oriented outward). ods for rapidly analyzing these transporters is vital.
Second, a number of integral membrane proteins expressed by Plasmodia at various stages of development could (in theory) serve as antigens in vaccine development. They are externally disposed and hence accessible to immune surveillance. Better methods for efficiently expressing and purifying these proteins would presumably be welcome. We believe this detailed, methodical, yeast-based approach is one such method, and our initial data with Pfcrt support this notion.
Beyond these technical issues, our results are also interesting because they suggest ways in which mutant Pfcrt proteins (e.g. Dd2crt) cause drug resistance in P. falciparum. Clearly we are just beginning a molecular level analysis of wild type versus various mutant Pfcrt function, and there is much yet to learn, but under conditions of high symmetrical [Cl Ϫ ] across the membrane, Dd2crt: 1) stimulates release of P i from yeast plasma membranes, and 2) appears to increase the ⌬pH that is formed in yeast ISOV. Moreover, the ⌬pH effects are inhibited by either addition of VPL or the withdrawal of Cl Ϫ , and the ATPase enhancement in ISOV is similarly inhibited by VPL and also by FCCP. Since CQR has previously been associated with alterations in DV pH (3,5), it is not difficult to envision how these effects could contribute to CQR (5). One reason the ISOV system is of interest in this context is because in certain ways it mimics the endogenous physiology of the DV (e.g. bioenergetics dominated by a H ϩ -ATPase with an extrafacially disposed ATP hydrolysis site).
However, first, it is important to point out that multiple interpretations of these data are formally possible. The first, which we do not favor, is that Dd2crt does not affect ⌬pH under symmetrical [Cl Ϫ ], but promotes active translocation of the probe we use to measure ⌬pH (acridine orange) at a rate faster than passive diffusion. If this is the case, the transport is directly or indirectly Cl Ϫ -dependent. Active translocation of AO at a rate faster than the (already extremely rapid) passive translocation under initial rate conditions would be quite surprising (5). Also, this model does not easily explain the elevated ATPase activity found in Dd2crt-bad yeast ISOV under these conditions (the assay is performed in the absence of AO, antimalarials, or any other small molecule putative substrate that might be envisioned via this model). On the other hand, AO does share some structural homology with the antimalarial quinicrine. The AO transport model requires energy, and since  Fig. 6, we present the average (ϮS.E.) from at least two trials with at least two independent ISOV preparations of each type. The reactions were all initiated by addition of 1 mM ATP and allowed to proceed for either 30 or 0 min in the presence and absence of 0.5 mM vanadate. In each case, the results are presented as percent of the vanadate-inhibitable activity (27) exhibited by the control ISOV at pH 7.5. At pH 7.5, in our hands, the vanadate-inhibitable specific activity (mol of P i /mg of protein/min) as determined by P i standard curves is 16.1 for control ISOV. This agrees reasonably well with activity reported for S. cerevisiae ISOV (39). To our knowledge, no ATPase activity has previously been quantified for P. pastoris ISOV.

FIG. 8. Quantification of PMA1 expressed in ISOV.
Western blot using anti-PMA1 (anti-S. cerevisiae plasma membrane H ϩ -ATPase) antibody generously provided by Dr. Carolyn Slayman (Yale) was performed as described under "Experimental Procedures." We show results from two independent sets of ISOV preparations; lanes 1-3 contain 30 g of highly purified plasma membrane protein (ISOV protein) for one set, and lanes 5-7 contain 20 g of ISOV protein for another set. These two independent sets of preparations are the same as those used for the experiments shown in Fig. 5. Lanes 1 and 5 are KM71/pPIC3.5, lanes 2 and 6 are KM71/HB3crt-bad, and lanes 3 and 7 KM71/Dd2crt-bad. Elevated PMA1 expression is not the explanation for the increased H ϩ pumping found for KM71/Dd2crt-bad ISOV (Fig. 5). Comparing densitometry of these bands versus protein assay quantification, we calculate ϳ2000-fold purification of Pfcrt-bad protein via a combination of membrane isolation, membrane wash, DM solubilization, and monomeric streptavidin chromatography.
Pfcrt does not harbor recognizable ATP hydrolysis domains, the only energy available is the ⌬H ϩ generated by the P. pastoris plasma membrane ATPase. Thus, if this is the case, Dd2crt might catalyze H ϩ /AO antiport. Clearly, testing this model requires an extensive and detailed array of additional kinetic and thermodynamic solute transport assays.
The second possibility is that Dd2crt protein directly interacts with H ϩ -ATPase, stimulating both hydrolysis of ATP and H ϩ pumping in the presence of high symmetrical [Cl Ϫ ]. This would be very interesting, but somewhat curious, since the H ϩ -ATPase present in these membranes is presumably a P2-type H ϩ -ATPase, whereas the H ϩ -ATPase that is expected to be the partner of Pfcrt in DV membrane bioenergetics is presumably a V-type H ϩ -ATPase (49). V-and P2-type Hϩ-ATPases do not share sequence homology other than in the ATP-hydrolyzing domains. Clearly, co-expression of relevant ATPases and Pfcrt would be an avenue for future testing of this model, as would purification and co-reconstitution of Pfcrt and PMA1 versus Pfcrt and vacuolar ATPase.
The third explanation, which we currently favor, is that Pfcrt protein directly or indirectly mediates passive movement of Cl Ϫ that is perhaps gated by some antimalarial drugs. We suggest that upon mutation of Pfcrt to confer CQR (e.g. Dd2crt) and under appropriate ionic conditions, the mutant Pfcrt catalyzes more efficient passive Cl Ϫ transport under certain conditions such that even higher ⌬pH values are possible under these conditions. It is common (if not essential) for proteins that mediate the passive movement of Cl Ϫ to exist in endomembranes across which a high ⌬pH is maintained, and binding/ inhibition of various ion transporters by amphipathic hydrophobic drugs is actually quite common. Such a function would assist formation of a higher ⌬pH as found in this study, stimulate H ϩ -ATPase activity of PMA1 in ISOV as found, be reversed by agents that also reverse resistance (e.g. VPL) as found, and obviously exhibit a simple dependence on Cl Ϫ as found. Future studies with these preparations and others will help distinguish between the possibilities.