The TRK1 Potassium Transporter Is the Critical Effector for Killing of Candida albicans by the Cationic Protein, Histatin 5*

The principal feature of killing of Candida albicans and other pathogenic fungi by the catonic protein Histatin 5 (Hst 5) is loss of cytoplasmic small molecules and ions, including ATP and K+, which can be blocked by the anion channel inhibitor 4,4′-diisothiocyanatostilbene-2,2′-disulfonic acid. We constructed C. albicans strains expressing one, two, or three copies of the TRK1 gene in order to investigate possible roles of Trk1p (the organism's principal K+ transporter) in the actions of Hst 5. All measured parameters (Hst 5 killing, Hst 5-stimulated ATP efflux, normal Trk1p-mediated K+ (86Rb+) influx, and Trk1p-mediated chloride conductance) were similarly reduced (5–7-fold) by removal of a single copy of the TRK1 gene from this diploid organism and were fully restored by complementation of the missing allele. A TRK1 overexpression strain of C. albicans, constructed by integrating an additional TRK1 gene into wild-type cells, demonstrated cytoplasmic sequestration of Trk1 protein, along with somewhat diminished toxicity of Hst 5. These results could be produced either by depletion of intracellular free Hst 5 due to sequestered binding, or to cooperativity in Hst 5-protein interactions at the plasma membrane. Furthermore, Trk1p-mediated chloride conductance was blocked by 4,4′-diisothiocyanatostilbene-2,2′-disulfonic acid in all of the tested strains, strongly suggesting that the TRK1 protein provides the essential pathway for ATP loss and is the critical effector for Hst 5 toxicity in C. albicans.

Salivary histatins (Hsts) 1 are structurally related histidinerich cationic proteins produced by acinar cells in human salivary glands and are key components of the nonimmune host defense system of the oral cavity (1,2). In vitro, Hst 5 (24 amino acids) kills pathogenic Candida albicans at physiological concentrations (15-30 M) (3)(4)(5) and possesses both fungistatic and fungicidal activities against a spectrum of other fungi including Candida glabrata, Candida krusei, Saccharomyces cerevisiae, Cryptococcus neoformans (6), and Neurospora crassa. 2 Hst 5 is also effective against azole-or amphotericinresistant strains of these fungi (7), which is of considerable practical importance and is of basic significance as well, suggesting a fundamental difference in their mechanisms of action.
Hst 5 has been shown not to act like conventional poreforming cationic peptides (8,9). Instead, it acts via a multistep process that includes initial binding of extracellular Hst 5 to a Candida surface protein Ssa1/2p (10), followed by subsequent entry of Hst 5 into the cytoplasm (11,12). That the primary target for Hst 5 is indeed intracellular has been verified by the observation that expression of a chromosomally encoded human-salivary histatin 5 gene, under the control of a regulated promoter, yields all of the customary toxic effects (13).
Whatever the primary target for Hst 5, its actions quickly become apparent at the plasma membrane and include noncytolytic efflux of cellular ATP and of potassium and magnesium ions (8,11) and elevated permeability to small cationic dyes such as propidium iodide (14) but not to larger anionic dyes such as calcein (15). These obvious membrane effects are accompanied by a rapid and irreversible loss of cell volume and ultimately by cell cycle arrest (16). However, all such toxic effects, including cell death, can be attenuated or blocked by a curious range of agents: by protonophores such as CCCP and DNP (8,17,18), by elevated extracellular calcium (19,20), by elevated external potassium (9,11), and (very surprisingly) by well known anion channel inhibitors such as niflumic acid, 5-nitro-2-(3-phenylpropylamino)benzoic acid, and DIDS (16).
Potassium ions appear to be especially important in these protective phenomena. In addition to their biochemical role as enzyme cofactors (especially in protein synthesis; see Ref. 21), they have several major physiological functions. Under normal circumstances, K ϩ accounts for ϳ40% of the cytosolic osmoticum and therefore establishes background conditions against which control of cell turgor must operate. Also, as the major cytoplasmic cation, K ϩ controls the ionic strength of cytoplasm, a property closely related to its function as a catalytic cofactor. Cytoplasmic K ϩ can also buffer the electric potential difference across the plasma membrane (V m , membrane voltage), which in turn supplies energy for the majority of "active" transport processes in fungal membranes. This role normally becomes important only when the membrane itself or its primary voltage generator, the H ϩ -ATPase (22,23), is damaged or down-regulated.
Tok1p, a cloned potassium channel which is the only certified K ϩ channel in Candida plasma membranes, may be the principal diffusion pathway for voltage stabilization (24 -26), but NSC1 (a protein whose encoding gene is not yet known) certainly contributes when extracellular divalent ions fall too low (especially Ca 2ϩ , ϳ10 M or below; see Refs. 27 and 28). Another protein involved in these potassium-related phenomena is the major K ϩ uptake system in Candida, Trk1p. This transporter, initially cloned in Saccharomyces by Gaber et al. (29), is believed to operate as a proton-coupled system, having similar properties to the high affinity K ϩ uptake system in Neurospora (30,31), and has also been reckoned as a voltage regulator on the basis of dye distribution measurements (32). Finally, it has recently been shown to mediate chloride currents through Saccharomyces plasma membrane, which are made easily visible by raising intracellular Cl Ϫ above ϳ10 mM (33).
The possible involvement of Tok1p in histatin toxicity was investigated by creating TOK1-disrupted mutants of C. albicans (34). Knockout of TOK1 (tok1⌬/tok1⌬) completely abolished the characteristic currents of Tok1p and also reduced (but did not abolish) Hst 5-induced efflux of ATP. However, these cells had only slightly increased residual viability after Hst 5 treatment. The fact that standard Hst 5 treatment still produced very substantial killing and ATP efflux, after knockout of both wild-type TOK1 alleles, clearly demonstrated that Tok1p channels could not be the primary site of Hst 5 action, although they do play a modulating role. NSC1, the putative nonselective cation channel that is responsible for very low affinity K ϩ uptake in Saccharomyces (28,35), is not yet accessible to mutational analysis. We therefore turned to the "active" K ϩ transporter, Trk1p, for further examination of potassium involvement in Hst 5-mediated effects on Candida.
The initial experimental objective was the same as that used with Tok1p (34), which was to assess the effects of Hst 5 on killing and ATP loss in Candida strains deleted of one and deleted of both TRK1 alleles. However, it proved impossible to delete both TRK1 alleles, and all such attempts resulted in retention of a wild-type allele, a phenomenon associated with essential candidal genes. We therefore decided to overexpress TRK protein by introducing an additional TRK1 gene, into the RP10 locus of wild-type (CaTK2(wt)) cells, for comparison of TRK1p-mediated Hst 5 effects. TRK1 transcription, assessed by quantitative RT-PCR, approximated a multiplicative dependence on the number of genes present, but functional assays (conventional transport and electrophysiological measurements) implied a saturable step in translation and/or protein function. The most important findings, however, were that (a) TRK1 is required for Hst 5-toxicity, which marks Trk1p as the principal effector for Hst 5 action in Candida, and (b) the well known anion channel inhibitor DIDS blocks native chloride conductance in Candida Trk1p, which suggests that Trk1p in fact provides the anion pathway for Hst 5-induced ATP loss.

EXPERIMENTAL PROCEDURES
Strains and Media-Escherichia coli strain DH5␣ (Invitrogen) was used as host for plasmids and was grown in LB medium (Difco). Ampicillin was added at a final concentration of 100 g ml Ϫ1 . Strain CAI4 of C. albicans (36) was used for transformation, and strain SC5314 (37) provided genomic DNA for the CaURA3 locus amplification. Cells for transformation (CAI4) were grown on a rich medium (YPD; Difco) but were transferred to a minimal medium (YNB; Qbiogene) plus 2% glucose and lacking uridine for selection of the URA ϩ transformants. Solid media were made with 1.5% Bacto agar (Difco). To select ura Ϫ auxotrophs from the transformants, cells were plated on solid YNB containing 50 g/ml uridine (200 M) and 1 mg/ml FOA (Sigma). The FOA-resistant colonies were picked after 3 days of incubation at room temperature.
In order to simplify referencing of experimental results to the underlying TRK1 gene configuration, we have adopted a transparent nomenclature for our principal isogenic C. albicans strains as follows.
CaTK2(wt) is wild-type strain CAI4, possessing two chromosomal alleles of TRK1; CaTK1 is the single-allele deletion strain derived from CaTK2(wt) that carries only one allele of TRK1; CaTK2 is a complemented strain constructed from CaTK1, in which one copy of TRK1 has been placed at the RP10 locus; and CaTK3, constructed from CaTK2(wt), has a third copy of TRK1, also inserted at the RP10 locus.
DNA Manipulations-All synthetic oligonucleotides used in strain constructions are listed in Table I and were synthesized by Integrated  DNA Technologies, Inc. (Coralville, IA). Molecular cloning, DNA hybridizations, and electrophoresis were carried out as previously described (13). CAI4 cells were transformed with linearized DNA using the LiAc procedure as we have previously described (13). Plasmid DNA was isolated from E. coli by the alkaline lysis method, employing the QIAprep Spin Miniprep Kit (Qiagen). Purification of restriction digest mixes and PCR mixes and isolation of DNA fragments from agarose gels were done with QIAquick PCR purification and QIAquick gel extraction kits, respectively (both from Qiagen) using the manufacturer's recommended conditions. Restriction endonucleases were purchased from Fermentas, and the Fast-Link DNA ligation kit came from Epicenter Technologies. C. albicans genomic DNA was isolated as previously described (13).
Amplification of the Candida URA3 Locus-The CaURA3 gene to be used as a selectable marker was first isolated by complementation of ura3 mutants in S. cerevisiae and of pyrF mutants in E. coli (37,38), but our experience using the 1.37-kb DNA fragment yielded very low rates of transformation in C. albicans, irrespective of the targeted locus. It seemed likely that the low transformation frequencies might result from too short flanking untranslated regions in the commonly used transforming fragment (424 bp upstream, 129 bp downstream), so we tested transformation with a much longer fragment, 2.37 kb, in which the URA3 gene is flanked by 1.04 kb upstream and 0.52 kb downstream. This length was selected to assure inclusion of the complete URA3 promoter, along with additional potential regulatory sequences. Experiments quickly found transformation frequencies with this larger fragment to range from 30-to 600-fold higher than with the standard 1.37-kb fragment 3 ; therefore, we have used this long fragment of CaURA3 DNA for plasmid constructs in the CaTRK1 disruption cassette.
Design of the TRK1 Disruption Cassette-The design principle for the CaTRK1 disruption cassette is shown in Fig. 1. Fig. 1A depicts the TRK1 locus, including the ORF (shaded bar, arrowhead), three target regions (5ЈTR, 3ЈTR, and 3ЈR-2), of which the two 3Ј regions are within the ORF; and two short flanking sequences (• and f) are used for control PCR. Fig. 1B depicts the plasmid bearing the disruption cassette (pDBT1U3), which consists, in order, of 5Ј-TR from the TRK1 locus, cloned as a SalI-BamHI fragment; 3ЈR-2, cloned as a BamHI-KpnI fragment; the wild-type URA3 gene; and 3ЈTR, cloned as a SacI-ApaI fragment. Since 3ЈR-2 naturally occupies a region downstream of 3ЈTR in the TRK1 locus (these fragments are described below), integration of the disruption cassette into the genome via homologous recombination ( Fig. 1C) results in a pair of directed repeats, upstream and downstream of the selectable marker. This construct affords a simple way, via intrachromosomal recombination and upon FOA screening (39,40), to delete the URA3 gene and 3ЈTR (Fig. 1D), leaving only the remnants 5ЈTR, 3ЈR-2, plus (in this case) 221 terminal bp of the TRK1 ORF. The principal advantages of this design are (a) that the resulting disrupted allele is no longer a target for the same disruption cassette, leaving only the surviving intact allele as target; (b) that no foreign sequences are left in the recipient locus or genome; and (c) that the selectable marker cannot be recycled if, by chance, the cassette integrates ectopically.
The 5ЈTR is the 247-bp fragment spanning positions Ϫ287 to Ϫ41 upstream of the translational start codon of the CaTRK1; 3ЈTR is the 297-bp fragment spanning positions 2291-2587 within the CaTRK1 ORF; and 3ЈR-2 is the 304-bp fragment spanning positions 2659 -2962 a Lowercase letters f and r indicate that the respective primer is forward or reverse. b All sequences are given in the conventional 5Ј-to-3Ј direction. c Restriction sites introduced into primer sequences for cloning purposes are in lowercase letters.
FIG. 1. Construction of C. albicans strain CaTR1 (trk1/TRK1). A, diagram of the CaTRK1 locus indicating three segments selected as components of the disruption cassette: 5ЈTR, the 5Ј-target region; 3ЈTR, the primary 3Ј-target region; and 3ЈR-2, a second 3Ј-target for intrachromosomal recombination. The shaded arrow (between the filled circle and filled square) represents the entire TRK1 open reading frame. B, the integrating fragment, harbored by plasmid pDBT1U3 (see "Experimental Methods"), that incorporates the TRK1 disruption cassette. Restriction sites introduced for cloning purposes are indicated. The open arrow represents the selectable marker URA3. C, diagram of allelic exchange via a double crossover event, that replaces one allele with the disruption cassette. Direct repeats are formed by the two 3ЈR-2 segments. D, diagram of FOA-induced intrachromosomal recombination to produce both excision of the selectable marker (and its recycling) and elimination of 3ЈTR (making this allele immune to reinsertion of the same cassette). E, agarose display of the PCR-amplified constructs, with the lower band containing the disrupted region. Wild-type and disrupted alleles were separately PCR-amplified, and the resulting samples were combined into a single gel lane. Filled circles throughout designate the position of the TRK1ctrl-f control primer, and filled squares locate the 2TRK1ctrl-r control primer (Table I).
of the ORF and ending 218 bp above the stop codon. Thus, the deleted region is delimited between the 3Ј-end of 5ЈTR and the 5Ј-end 3ЈR-2 so that 221 bp of the ORF, plus the 3Ј-untranslated region remain. This region was retained, because sequence analysis in the Candida genome data base (available on the World Wide Web at www-sequence.stanford.edu/group/candida/) identified a potential IMP4 gene close to the 3Ј-region of the TRK1 locus, with the 3Ј-untranslated regions overlapping between the two genes. In S. cerevisiae, IMP4 is an essential gene, since Imp4p is a part of the U3 small nucleolar ribonucleoprotein that is required for pre-18 S rRNA processing (41). The putative C. albicans Imp4 protein shows 66% identity with the corresponding S. cerevisiae Imp4p, so it very likely is essential in C. albicans as well.
Construction of Plasmids pUC18A and pDBU3-Because it was not usable for our purposes, the EcoRI site present at the 3Ј end of the MCS in pUC18 was replaced by a unique ApaI site, and the plasmid was renamed pUC18A. For this exchange, pUC18 was linearized by EcoRI digestion and used as template for divergent PCR, via primers (A)PUCf/(A)PUC-r, which carried the ApaI recognition sequence at their 5Јends (Table I). PCR cycling conditions were as follows: initial denaturation for 3 min at 94°C, followed by 30 three-step cycles consisting of 15-s denaturation at 94°C, 15-s annealing at 55°C, and 3-min elongation at 72°C. This was followed by a 5-min final elongation at 72°C. After purification, the resulting PCR fragment was ApaI-digested and self-ligated to yield pUC18A. The 2.37-kb DNA fragment bearing the CaURA3 gene was amplified from Candida genomic DNA (strain SC5314), via the primer pair (K)URA3-f/(Sc)URA3-r (Table I). Sequences for these primers were derived from the sequence of the URA3 locus in the C. albicans database (available on the World Wide Web at www.ncbi.nlm.nih.gov/BLAST). PCR cycling conditions were as follows: initial denaturation for 3 min at 94°C, followed by 30 three-step cycles consisting of 15-s denaturation at 94°C, 15-s annealing at 55°C, and 2-min elongation at 72°C. This was followed by a 5-min final elongation at 72°C. The nucleotide sequence of the fragment was confirmed by DNA sequencing. After purification and digestion with KpnI/SacI, this fragment was ligated to KpnI/SacI-digested pUC18A, generating pDBU3, the basic vector for construction of C. albicans gene-disruption cassettes. As a part of the pUC18A sequence, two convenient restriction sites, Eco31I and Eam1105I, are available for linearizing constructs containing the C. albicans gene-disruption cassettes.
Construction of pMP2-U3 and pMP2-5Јtrk1-U3-The rationale for construction of pMP2-U3 was to generate a vector specifically designed for displacement of the promoter of a second gene copy under study. We have used this strategy to inactivate a second allele that could not be deleted with the disruption cassette used for the first allele (34). A KpnI-SacI URA3 fragment was cloned via KpnI-and SacI-digested pMalPr (13) to yield pMP2-U3. Then the fragment bearing the first 1120 bp of TRK1, including the putative translational start codon, was PCR-amplified using primers (H)TRK1-f/(B)TRK1-r (Table I). After digestion with HindIII-BamHI, this fragment was ligated into HindIII/ BamHI-digested pMP2-U3 to generate pMP2-5Јtrk1-U3.
Construction of C. albicans Strain CaTK1-Strain CaTK2(wt) was first transformed with Eco31I-linearized pDBT1U3 in order to inactivate the first TRK1 allele (Fig. 1C). Transformation frequency was in the range of 30 -50 URA3 ϩ colonies/g of transforming DNA. Correct allelic replacement was confirmed by PCR of four randomly selected URA ϩ transformants using the primer pairs TRK1ctrl-f/(Sc)URA3-r and (K)URA3-f/2TRK1ctrl-r and by Southern analysis (data not shown). One of them was grown overnight (ϳ14 h) in 5 ml of liquid YPD medium, supplemented with 50 g/ml uridine. Then serial dilutions were made, and the suspensions were plated on YNB containing 50 g/ml uridine and 1 mg/ml FOA. Four FOA-resistant colonies were picked and PCR-analyzed for loss of the marker and the major portion of the first TRK1 allele (Fig. 1, D and E). Thus, the trk1/TRK1 heterozygous strain CaTK1 was obtained.
Attempts to Delete the Second TRK1 Allele-The CaTK1 trk1/TRK1 strain was subjected to a second round of transformation with the TRK1 disruption cassette. Transformation frequency in this second round was lower compared with the first allele integration events and was in the range of 5-10 URA3 ϩ colonies/g of transforming DNA. Twenty-four URA ϩ transformants from three independent experiments were analyzed by PCR and by Southern analysis, and all displayed the transforming cassette in the remaining allele but also retained a wild type third allele (data not shown), a phenomenon previously observed by us and other laboratories (e.g. see Refs. 34 and 42). The molecular basis of this third allele maintenance is currently unknown.
Therefore, a second strategy was attempted, by means of a conditional mutation of the second allele via a regulatable promoter. This approach succeeded previously for inactivating the second TOK1 allele and obtaining a functional tok1-null strain (34). Spheroplasts of CaTK1 were transformed with the intact pMP2-5Јtrk1-U3 plasmid, and 24 URA ϩ transformants from three independent experiments were analyzed. All transformants displayed the new construct in the remaining allele but again also retained a wild type allele (data not shown). The failure of both approaches to produce a trk1-null strain indicates that disruption of the second TRK allele is a lethal event and suggests that the TRK1 gene in C. albicans is essential for viability.
Construction of Strains CaTK2 and CaTK3-Complementation of the TRK1 heterozygous strain was accomplished by introducing the wild type TRK1 gene into the RP10 locus (13) of CaTK1. This strain, possessing two wild-type alleles and the disrupted trk1⌬ site, was named CaTK2. Another strain (CaTK3) expressing three wild-type TRK1 alleles, was made by introducing the extra copy of TRK1 into the wild-type strain CaTK2(wt), again at the RP10 locus. To construct both of these strains, a 4,636-bp fragment bearing TRK1 was PCR-amplified via primers (H)TRK1L-f and (B)TRK1L-r ( Table I). The resulting DNA was digested with HindIII and BamHI and ligated into the large Hin-dIII-BamHI fragment of pDBT (34) to generate plasmid pDBT1L, which bears both the URA3 selectable marker and the target RP10 locus sequence. Following linearization with NcoI (which cuts within the plasmid-borne RP10 target sequence), pDBT1L was transformed into CaTK1 and CaTK2(wt), making strains CaTK2 (trk1/TRK1/rp10::TRK1/ URA3) and CaTK3 (TRK1/TRK1/rp10::TRK1/URA3), respectively. The correct integration of the constructs into the RP10 locus was verified by PCR analysis and Southern blotting of the respective genomic DNA preparations (data not shown).
RNA Isolation and cDNA Synthesis for TaqMan RT-PCR-Total RNA from each strain was isolated as previously described (13) with the following modifications. C. albicans strains were grown overnight under the same conditions as for functional assays (34) in 7 ml of YNB medium, 2% glucose with the addition of uridine when required. Total RNA was isolated using RNeasy Mini Kit from Qiagen. The samples were both in-column-treated with DNase I using the RNase-Free-DNase set from Qiagen, and off-column-treated using the TURBO-DNA-Free set from Ambion. The absence of genomic DNA contamination was confirmed by PCR using the following primers: EFB1-5Ј/ EFB1-3Ј for elongation factor gene EF-1␤ (CaEFB1) and TRK1-5Ј/ TRK1-3Ј for CaTRK1. Following DNase I treatment, 2 g of total RNA was used per reaction for the first strand synthesis (cDNA) via the RETROscript kit from Ambion as previously described (13).
TaqMan RT-PCR Assay-Quantitative real time RT-PCR experiments were carried out as described previously for yeast (43). Primers and probes (forward primer EFB1, 5Ј-AAG ACT TGC CAG CCG GTA AAG-3Ј; reverse primer EFB1, 5Ј-TCA GCT TCT TCA TCA ACT TCA TCA-3Ј; forward primer TRK1, 5Ј-CGT CTT CAT CTC CTC AGT CAT CTT-3Ј; reverse primer TRK1, 5Ј-GGC TCA CTA TGG TGC TCT ATA TCA-3Ј; probe EFB1, 5Ј-CCG CTT CTG GTT CTG CTG CTG CCG-3Ј; probe TRK1, 5Ј-AAC AAG CCA GCC GGT TCT GAC AGC-3Ј) were designed via the Primer 3 Input program (available on the World Wide Web at frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi) and confirmed via Amplify 1.2 Software for PCR. Primers were custom synthesized by Integrated DNA Technologies, Inc. Reporter dyes and quencher dyes for the probes were 6-carboxyfluorescein and Black Hole Quencher, respectively. The T m of primers was 54.4 -57.7°C, and the T m for probes was 66.5°C for EFB1 and 64.0°C for TRK1, respectively. TaqMan PCR conditions were as follows: 2 min at 50°C plus 2 min at 95°C and then 40 cycles of 30 s at 95°C, 30 s at 55°C, and 30 s at 72°C. Amplification and detection were carried out in 96-well plates on an iCycler iQ real time detection system (Bio-Rad). Each reaction mix contained duplicates of the test cDNA templates, negative RNA controls, no-template controls, and positive genomic DNA controls. All samples contained 1ϫ IQ Supermix (Bio-Rad), 50 nM TaqMan probe, and 150 nM concentration of both the forward and the reverse primer. Fluorescent data were collected and analyzed with iCycler iQ software. The threshold cycle (Ct) value for each sample was calculated by determining the point at base line. Each sample was tested in duplicate, and the results were averaged to obtain the final Ct (i.e. the point at which sample fluorescence rises above the background level). Sensitivity was verified by analysis of ⌬Ct values with various template dilutions (slope of Ͻ0.10). The ⌬Ct value was calculated by obtaining the difference (⌬Ct) between Ct values of the target (TRK1) and the normalizer (EFB1). ⌬Ct for CaTK2(wt) was used as a reference (base line) for comparison of each strain. ⌬⌬Ct values were then calculated as the difference between each sample's ⌬Ct and the base line's ⌬Ct and were transformed to absolute values (2 Ϫ⌬⌬Ct ) to calculate comparative expression levels.
Western Blotting-Anti-Trk1p antibody was prepared to the synthetic oligopeptide (c)SIRRTNVYEEQS found in Trk1p and Trk2p from S. cerevisiae. This corresponds to the sequence 842 SVRRTNVYEEQS 853 in C. albicans, predicted to span half of the intracellular loop M2c-M1d in the Durell and Guy (44) structural model of the fungal TRK proteins. Trk1p peptide was synthesized in the Keck Biotechnology Resource Center at Yale Medical School, verified by mass spectrometry, and conjugated to mcKLH via its N-terminal cysteine, using the maleimideactivated conjugation kit (catalog no. 77607) from Pierce. Rabbit anti-Trk1p serum was produced by the animal immunization service at Yale Medical School, after six triweekly injections with 100 g of the mcKLH-conjugated peptide. Trk1p antibody was purified on a peptide affinity column by means of the SulfoLink Kit (catalog no. 44895 from Pierce) and subsequently used at a 100-fold dilution for Western blots. Membrane and cytosolic fractions from each C. albicans strain were prepared as we have previously described (10,15). Briefly, cells were glass bead-disrupted in the presence of protease inhibitors and centrifuged at 3,200 ϫ g for 5 min to remove unbroken cells and organelles. The supernatant was then centrifuged at 100,000 ϫ g for 1 h at 4°C to separate cytosolic and membrane fractions. The cytosolic fraction was collected in the supernatant, and the total crude membrane fraction was recovered from the pellet. Soluble membrane proteins were extracted from the membrane fraction with 10 mM sodium phosphate buffer containing 2.5% Triton X (with protease inhibitors at 4°C) and were recovered in the supernatant following centrifugation at 100,000 ϫ g for 1 h at 4°C. Equal amounts of protein (100 g) from each fraction were separated using 7.5% SDS-PAGE under reducing conditions, transferred to polyvinylidene difluoride membranes, and probed with anti-Trk1 antibody, followed by goat anti-rabbit peroxidase-conjugated secondary antibody (diluted 1:2,500). Fractions were also probed with yeast anti-Pma1 (kindly provided by Dr. Amy Chang, Albert Einstein College of Medicine) and yeast anti-Gas1 (kindly provided by Dr. Randy Schekman, University of California at Berkeley) to verify identity and purity of membrane fractions. Reactive proteins were visualized with the 1-Step TMB detection system (Pierce) according to the manufacturer's protocol.
Candidacidal Assay-Killing of Candida TRK mutant strains by Hst 5 was measured by the microdilution plate assay, as previously described (34). Assays were performed in triplicate for each strain and each test condition, after cells had been preincubated, at a density of 2.5 ϫ 10 5 cells/ml in 10 mM sodium phosphate buffer (pH 7.2), with the test concentration of Hst 5 for 60 min at 37°C. For DIDS experiments, cells were preincubated with 1 mM DIDS for 1 h before use in the candidacidal assay. Cell death was calculated as (1 Ϫ (number of colonies recovered from Hst 5-treated cells/number of colonies from control cells)) ϫ 100.
ATP Bioluminescence Assay-Histatin 5-induced release of ATP from C. albicans cells was measured as described (34) but with the following slight modifications. For each assay, cells were mixed with 31 M Hst 5 at a density of 10 6 cells/ml and incubated at 37°C, with shaking. At intervals of 5, 10, 20, 30, 45, and 60 min after exposure to Hst 5, the suspension was centrifuged (6,000 ϫ g, 3 min), and 25 l of supernatant was assayed for ATP by luminometry, using the FL-AA ATP assay kit (Sigma). Results are expressed as nmol of ATP released/10 6 cells.
Rubidium-86 Influx Measurements-The general design for assessing K ϩ transporter function described for Saccharomyces (32) was used with the following modifications for Candida. Cells were grown at 37°C to A 600 ϭ 0.8 -1.0, in 15 ml of YNB medium (Qbiogene) supplemented with 150 mM KCl and with 200 M uridine (when required). The resulting suspensions were spun down, and the cells were washed twice with glass-distilled water and then resuspended at the same density (2-5 ϫ 10 7 cells/ml) in 1 M sorbitol containing 2.5% glucose. Cells were potas-sium-starved in this solution for 5 h on a rotary shaker (250 rpm) at room temperature. For uptake measurements, the starved cells were centrifuged, washed, and resuspended at 5 ϫ 10 8 cells/ml in transport buffer (50 mM Tris-succinate, 2.5% glucose, pH 5.9). Uptake of 86  Electrophysiological Methods-Measurements of electric currents through the Candida plasma membrane were made on spheroplasts, using the whole-cell recording configuration of the patch clamp technique, as described previously for Saccharomyces (45), with slight modifications for Candida (34). All C. albicans strains were grown in shaking (250 rpm) liquid YPD medium, to A 600 ϭ 0.8 -1.2, at 30°C. All media used for preparing and recording from Candida spheroplasts are listed in Table II. To make spheroplasts, ϳ4 OD units of cells were harvested by centrifugation (500 ϫ g for 5 min), washed twice with 3 ml of Buffer A, and incubated in the same buffer plus 0.2% ␤-mercaptoethanol, at 30°C on a slow orbital shaker (64 rpm), for 30 min. The suspension was then recentrifuged, and the pellet was resuspended in 6 ml of Buffer B (spheroplasting buffer) plus 0.6 units/ml zymolyase 20T (catalog no. 320921; ICN Biomedicals Inc., Irvine, CA) and incubated for 45 min at 30°C. The resulting spheroplasts were spun down (500 ϫ g for 5 min), gently resuspended in Buffer C (osmotic stabilizing buffer), and incubated stationary, at room temperature (ϳ23°C), until use. Spheroplasts were prepared daily and used for patch recording over a 6 -8-h period. Occasionally, in order to maintain cleaner spheroplasts over longer times (thus facilitating more stable gigaohm seals), zymolyase (0.3 units/ml) was added to the spheroplast suspension in Buffer C ϳ15 min before use.
Patch pipettes with tip inner diameters of ϳ1 m were drawn as described (45) and filled with Buffer G (intracellular solution). When dipped into Buffer D (sealing buffer, extracellular recording solution), these electrodes had tip resistances of 3-5 megaohms. In experiments with low intracellular chloride (0.1-1.0 mM), pipette chloride salts were replaced with gluconate Ϫ salts, resulting in slightly higher tip resistances (7-10 megaohms). The reference electrode, consisting of a chlorided silver wire (Ag-AgCl) immersed in 1 M KCl, was connected to the efflux end of the recording chamber via a 1 M-KCl-agar bridge.
Depending on the strain and cell density, 1-10 l of spheroplast suspension was injected into the recording chamber containing ϳ700 l of Buffer D, gently mixed, and then allowed to settle for 10 min, to permit a small number of spheroplasts to adhere lightly to the chamber bottom. Light suction on the patch pipette, positioned near a clean spheroplast, drew the cell onto the pipette tip and yielded a seal of 10 -35 gigaohms in ϳ5 min, in ϳ60% of attempts. Whole-cell recording was subsequently achieved after breaking the membrane patch in the pipette tip, with combined light suction and a brief high voltage pulse (ϳ750 mV for 100 s). A ϳ10-min period was allowed for equilibration between pipette contents and spheroplast cytoplasm before recording was begun.
Current-voltage (I-V) data were generated via a staircase voltage clamp protocol, consisting of 15 sequential square pulses, each 2.5 s long and delivered from a holding voltage of Ϫ40 mV, to ϩ100 and ϩ80 mV and down to Ϫ180 mV, in 20-mV decrements, with 0.5 s at the holding voltage after each pulse (see voltage protocol in Fig. 5). Pulses were delivered, and the resultant membrane currents were recorded via an EPC9 patch clamp amplifier (Heka Elektronik, Lambrecht, Germany)  (33). Preliminary data analysis was performed via the Heka Pulse software, but more detailed analysis was carried out with Microsoft Excel and/or Igor Pro (WaveMetrics, Inc.).

RESULTS
The C. albicans TRK1 Gene-A 4,783-bp nucleotide sequence from C. albicans (strain ATCC 10261) was cloned by phenotypic complementation of a trk1⌬ trk2⌬ strain of S. cerevisiae. The enclosed ORF would translate into a 1,059-amino acid protein with ϳ36% sequence identity to each of the Saccharomyces proteins, Trk1p and Trk2p, and ϳ50% similarity allowing conservative substitutions. The corresponding comparisons between the two Saccharomyces proteins are ϳ53% sequence identity, and ϳ65% similarity with conservative substitutions (46). Accordingly, this ORF was assigned the gene name Ca-TRK1 and registered with NCBI under accession number AF267125. 4 Details of cloning this gene and characterization of the expressed protein will be published elsewhere. Sequencing of the Candida genome (strain SC5314) has now been completed, and contig 19 (available on the World Wide Web at www-sequence.stanford.edu:8080/btcontigs19super.html) lists two alleles, 19-10057 and 19-20057, which differ slightly from each other and somewhat more from the sequence at AF267125 but clearly represent small variations on the same gene. All of our BLAST searches from AF267125 or its translation identified sequences in contig 19-10057 or  or in the translations thereof but failed to identify any other related Candida sequences. Therefore, C. albicans appears to harbor only a single protein with significant identity to the Saccharomyces TRK1 and TRK2 proteins. However, our searches did reveal an open reading frame, in overlapping contigs 6-2069 and 6-1944, that potentially encodes an 808-amino acid potassium transporter in the HAK/KUP family.
In order to examine the possible role of the TRK1 protein (Trk1p) in killing of C. albicans by the cationic salivary protein Hst 5, we constructed a series of isogenic C. albicans TRK1 mutants from wild-type strain CAI4, here designated CaTK2(wt), which is itself a ura Ϫ derivative of SC5314. Singleallele disruption of the TRK1 locus produced the heterozygous trk1/TRK1 strain CaTK1; however, two independent strategies failed to produce a trk1-null strain, suggesting that disruption of both TRK alleles is lethal in C. albicans. Gene-dose modulation for Hst 5 effects was therefore devised by transforming an additional copy of the wild-type gene into the Candida RP10 locus to produce strain CaTK3 (TRK1/TRK1/rp10::TRK1/ URA3) that has three functional copies of TRK1. Complementation of the hemideleted strain CaTK1 yielded strain CaTK2 (trk1/TRK1/rp10::TRK1/URA3) that carries two functional copies of TRK1. All four strains (CaTK2(wt), CaTK1, CaTK2, and CaTK3) had similar growth characteristics in both YPD and YNB medium, and none required supplementation with extra K ϩ for normal growth. Also, all grew robustly on glycerol as sole carbon source, demonstrating the presence of normal oxidative metabolism.
Message and Protein Expression Levels of C. albicans TRK1 Mutants-Quantitative real time RT-PCR (TaqMan TM ) was performed to measure TRK1 transcript levels in all four Candida strains. For relative quantification of gene expression, transcript levels were normalized to those of the EF␤1 gene by comparative Ct (⌬⌬Ct). TRK1 transcripts in CaTK1 (TRK1/ trk1) were found to be ϳ15% of those in the wild-type strain CaTK2(wt) (Fig. 2A, gray bar; ⌬⌬Ct ϭ Ϫ7.0 Ϯ 1.6, n ϭ 3). Such lack of dominance of the wild-type allele is a common phenomenon found with many Candida genes (47,48). Complementation of the TRK1 gene, at the RP10 locus (trk1/TRK1/rp10:: TRK1), restored TRK1 transcripts in CaTK2 to the same level as in wild-type cells ( Fig. 2A, white bar; ⌬⌬Ct ϭ 1.0 Ϯ 0.5, n ϭ 6); but three TRK1 genes, in strain CaTK3 (TRK1/TRK1/ rp10::TRK1), yielded 10-fold the wild-type level of TRK1 transcripts ( Fig. 2A, black bar; ⌬⌬Ct ϭ 10.0 Ϯ 4.5, n ϭ 4). These results indicate a high order of cooperativity in transcription of TRK1 genes in Candida, a phenomenon that will be explored separately. Fig. 2B shows Western blots used to estimate the levels of TRK1 protein in membrane and cytosolic fractions from the same four strains of Candida, for direct comparison with the message/transcript levels. Trk1p was found in both cytosolic and membrane fractions of CaTK2(wt) (Fig. 2B, left two lanes), at an apparent molecular mass of ϳ120 kDa (cf. CaTrk1p, 119.7 kDa) but was barely detectable either in the cytosol or in the membrane fraction from CaTK1. This is in agreement with the 7-fold reduction of TRK1 expression as assessed by quantitative RT-PCR. Complementation of the TRK1 gene, in strain CaTK2, produced levels of Trk1 protein that were nearly equivalent to the wild-type levels in both the cytosolic and the membrane fractions. However, some experiments also revealed the CaTK2-membrane fraction to contain a faint additional band, at ϳ210 kDa, as shown in Fig. 2B (CaTK2 M, upper  band). This band, as well as the cytoplasmic band at ϳ120 kDa, was substantially enhanced in the overexpressing strain CaTK3, but the 120-kDa band in the membrane fraction of CaTK3 was only slightly enhanced with respect to that in wild-type Candida (Fig. 2B, band CaTK2(wt) M).
These results suggest that the membrane density of Trk1p is optimal, or near maximal, with two functional alleles in wildtype C. albicans and that overexpressed protein is shunted into cytoplasmic aggregates, perhaps the vesiculotubular clusters described upon overexpression of Trk2p in Saccharomyces (49), and into a membrane complex with other proteins (the 210-kDa immunopositive band), which could still be functional (see "Discussion"). To investigate whether the membrane-bound Trk1p detected by immunoblotting represents functional protein, we examined the abilities of all four strains to carry out the normal transport processes of TRK1 protein.
Trk1p Function Measured by Rubidium Uptake-The best known biological function of TRK in fungi is uptake and accumulation of potassium. Unidirectional influx of potassium is cumbersome to measure, because of the short half-life (12.4 h) of its principle radioisotope, 42 K ϩ . Rubidium, the closest chemical homologue of potassium, is far more convenient to measure (half-life of 86 Rb ϭ 18.7 days) and is handled by many membrane transporters almost as if it were potassium. Since the TRK proteins in Saccharomyces (Trk1p and Trk2p) have been extensively characterized by isotopic as well as chemical measurements of Rb ϩ flux (32, 50, 51), we made the simpler measurements, of 86 Rb ϩ influx, in Candida.
Extracellular chemical Rb ϩ was set at 1 mM, which is appropriate for the moderate to high affinity K ϩ uptake system in wild-type Candida cells, and the measuring interval was kept short, 0 -3 min after the addition of 86 Rb ϩ to the cell suspension, in order to maintain linearity. The measured isotope content of Candida began to "bend" after 5-6 min (data not shown), reflecting the rise of significant isotope efflux. The resulting Rb ϩ influxes are shown in Fig. 3 for all four strains. In all cases, there appeared to be an abrupt uptake of rubidium at the start of measurement, which averaged 3.5 nmol/10 8 cells. This was almost certainly an artifact, probably representing residual Rb ϩ left in the cell wall, despite the rinsing of harvested pellets with MgCl 2 . The jump was ignored in calculating the TRK-dependent fluxes.
Rubidium influx in CaTK1 cells occurred at only 2.7 nmol/ 10 8 cells/min, compared with 13.3 nmol/10 8 cells/min in wildtype CaTK2(wt), a functional ratio that is not far from the corresponding transcript expression ratio (1:5; cf. 1:7). No significant difference was found on a point-to-point basis between 86 Rb ϩ content of the complemented strain CaTK2 and that of wild-type cells, and the computed uptake rate (11.9 nmol/10 8 cells/min) was compatible with the Western blots, confirming that placement of the TRK1 gene at the RP10 locus, together with the single remaining normal allele, can restore wild-type protein expression levels (Fig. 2B). Furthermore, the addition of a third TRK1 gene at the RP10 locus to wild-type Candida, in strain CaTK3, augmented rubidium influx ϳ1.5-fold to 19.5 nmol/10 8 cells/min, as if in direct proportion to gene dosage, compared with the wild-type. Although this coincidence of gene dosage and function may be post-translationally regulated, it is probably best regarded as fortuitous, in view of the much elevated (10-fold) transcript levels in CaTK3. Evidently, the majority of Trk1p resulting from hyperproduction of TRK1 message is nonfunctional and sequestered into a cytosolic compartment. Nevertheless, overexpression of TRK1 did apprecia-bly enhance potassium (rubidium) transport in CaTK3 cells, in agreement with elevation of Trk1p detected in the membrane fraction. These results were supported further by assessment of the novel chloride transport properties of Trk1 proteins in these cells.
A Second Functional Assay: The Chloride Channel through Trk1p-A completely unexpected property of TRK proteins was recently identified by patch clamp experiments on whole protoplast membranes of S. cerevisiae; they conduct large inward currents when the cytoplasm is loaded with chloride ions (33,52). The molecular mechanism underlying such implied chloride effluxes is unknown, but it apparently serves as a voltagedriven escape for excess cytoplasmic chloride ions. Although the fluxes themselves can be very large, an order of magnitude larger than maximal potassium fluxes, the TRK-mediated Cl Ϫ permeability is fixed and independent of chloride concentration. Similar TRK-related chloride currents have recently been observed in Candida, and typical current records from spheroplasts of all four Candida strains are compared in Fig. 4. These sets of superimposed traces display the amplitude and time courses of membrane current required to step the membrane voltage from Ϫ40 mV (holding voltage) to ϩ100 mV, ϩ80 mV, etc., to Ϫ180 mV and hold it there for 2.5 s. Positive currents (so-called outward currents; upward traces) reflect the noisy opening and closing of Tok1p channels, which let K ϩ out of the cells and which were previously shown (34) not to be critical to . Ct values were defined as the absolute value of the difference between the Ct of the target RNA (TRK1) and the elongation factor EF-1␤ RNA (EFB1). The copy numbers for the TRK1 genes were normalized to the copy numbers obtained for EFB1 by dividing the mean of the duplicates for the test genes by the mean of the duplicates of EFB1 for each individual sample. Here, the CaTK2(wt) ⌬Ct value is used as a base line (x axis). To obtain the comparison expression levels (2 Ϫ⌬⌬Ct ) for each strain, the ⌬Ct value for CaTK2(wt) strain (used as a standard) was subtracted from the ⌬Ct values for CaTK1 (filled bar, light), CaTK2 (clear bar), or CaTK3 (filled bar, dark) in each experiment. All reactions were performed in triplicate. B, levels of TRK1 protein in membrane and cytosolic fractions from all four strains were determined by Western blotting, probed with Trk1p antibody. Trk1p (ϳ120 kDa) was detected in both cytosolic and membrane fractions of CaTK2(wt) but was barely detectable in the cytosol of CaTK1 and was nearly absent from that of the membrane fraction. The complemented strain, CaTK2, and the wild type, CaTK2(wt), had equivalent levels of Trk1p, and the protein was also substantially elevated in CaTK3 in the cytosolic fraction and slightly elevated in the membrane fraction. Additionally, a high molecular mass band (ϳ210 kDa) was detected in the CaTK3 membrane fraction.
Hst 5 effects. The positive currents varied randomly in amplitude from strain to strain, unrelated to the presence or absence of TRK genes.
The inward, negative currents (downward in the tracing sets of Fig. 4) were dependent upon both TRK1 and high intracellular chloride, as is made clear by comparison of the top left panel (CaTK2(wt)) with the panels directly to the right (CaTK1) and directly below (0.1 mM Cl Ϫ ), showing that replacement of intracellular (pipette) chloride with gluconate (Buffer GL), leaving both intracellular and extracellular K ϩ unchanged, nearly abolished the negative currents. In Candida, these Cl Ϫ currents increased with a half-time of about 20 s, stabilizing after ϳ2 min. Comparison of the records for the four different Candida strains (Fig. 4, top row) makes clear that the relative magnitudes of chloride currents, under standard conditions in the different strains, had roughly the same proportion as the rubidium fluxes shown in Fig. 3. This was true when either the time courses of current or the corresponding steadystate currents were compared.
The record sets in Fig. 4 were selected as representative (viz. near average for the data collected from each strain). Security in the above statements about relative amplitudes is reinforced by Fig. 5, plotting the averaged negative currents (last half of each tracing), for all measurements on each strain, against the clamped membrane voltage. Because data at positive membrane voltages are irrelevant (Tok1p), the plots were truncated at ϩ60 mV. These so-called current-voltage plots (I-V plots) display a characteristic and important parameter for each I-V plot: the limiting slope conductance at large negative voltages, indicated by the dashed line (Fig. 5, top left panel). Comparing the four strains, these limiting conductances defined the same relationship to the number of expressed TRK1 genes as did the rubidium influxes in Fig. 3. This fact is demonstrated by the phase plot in Fig. 5, inset, showing that Cl Ϫ conductance is linearly proportional to Rb ϩ influx.
Susceptibility of C. albicans TRK1 Mutants to Killing by Histatin 5-Previous experiments (34) had demonstrated that the inwardly rectifying potassium channel, Tok1p, in Candida only slightly modulates the killing of C. albicans by Hst 5. Very different results emerged with Trk1p, the primary "active" K ϩ transporter. Averaged results from a homogeneous set of experiments are shown in Fig. 6. The wild-type strain CaTK2(wt) showed normal sensitivity to Hst 5, with ϳ95% of the cells killed at 125 M Hst 5, ϳ81% at 62 M, and an estimated half-killing concentration of ϳ20 M. Significantly, strain CaTK1, having a 7-fold reduction of TRK1 transcripts and a 5-fold reduction of Trk1p function, was quite insensitive to Hst 5, which produced only ϳ11% killing at 125 M (data not shown), with an apparent half-killing concentration of Ͼ600 M (n ϭ 6). This result was not due to reduced cellular uptake of Hst 5, since we found that CaTK2(wt) and CaTK1 had equivalent total cellular levels of Hst 5 following 30-and 60-min incubation with fluorescein isothiocyanate-Hst 5, as assessed by quantitative FACScan analysis (data not shown). Thus, the CaTK1 strain proved profoundly resistant to histatin 5 killing despite normal uptake of peptide.
Complementation of CaTK1 by insertion of a second wildtype TRK1 gene at the RP10 locus fully restored sensitivity to Hst 5 in CaTK2 with no statistically significant difference in Hst 5 killing between CaTK2(wt) and CaTK2 at any dose. Thus, deletion of the TRK1 gene itself, not some unrelated or incidental change in the CaTK1 strain, was responsible for the diminished histatin sensitivity. However, a second and very surprising result emerged from killing tests on the TRK1 overexpression strain, CaTK3. This strain proved significantly less susceptible to Hst 5 killing than the wild type, as judged by a fitted half-killing concentration of 31 M Hst 5, again computed on the basis of multiple (from three to seven) independent determinations. The extrapolated maximal killing at high Hst 5 doses could be taken as 100% for both CaTK2(wt) and CaTK3. Although the Western blots (Fig. 2B) showed total 120-kDa membrane-bound Trk1p in CaTK3 to be only slightly elevated from the wild-type level, cytosolic Trk1p was greatly elevated. This suggests that a large cytosolic reservoir of Trk1p could sequester Hst 5 and diminish its free concentration, that is accessible to plasma membrane sites. Experiments are in progress to investigate this possibility.
Correlated Loss of ATP by Hst 5-treated Candida-The fact that Hst 5 kills Candida without overt cell lysis or without itself forming pores in the plasma membrane has been well understood for more than 5 years (8,9). However, a cardinal feature of the process was shown to be the net loss of cytoplasmic ions and small molecules, especially ATP (8,11,18), at a rate (for ATP) that correlated well with the killing effectiveness of Hst 5 (53). It was therefore important to determine whether the same relationships between killing effectiveness and TRK1 transcript and protein levels also held for ATP loss in the four C. albicans strains.
The answer to this question was unequivocally positive, as is shown by the data in Fig. 7. All data plots for ATP release during the first hour could be reasonably fit by straight lines, revealing the following quantitative relationships. Standard Hst 5 treatment (31 M) released ATP from wild-type CaTK2(wt) cells at 0.20 nmol/10 6 cells/min and from CaTK1 cells at 0.045 nmol/10 6 cells/min (ϳ22%). Although this residual flux was larger than the corresponding residual killing (ϳ10%, Fig. 6), many processes could dispel an exact correspondence, including the metabolically cumulative effects of sustained ATP loss. Complementation of CaTK1 with the TRK1 gene at the RP10 locus (strain CaTK2) reestablished Hst 5-induced ATP loss at 0.18 nmol/10 6 cells/min, again essentially identical with the wild-type value. Overexpression of TRK1, in strain CaTK3, had qualitatively the same effect on ATP efflux as upon Hst 5-induced killing, in that it actually FIG. 3. Trk1p-mediated rubidium influx rises with the TRK1 copy number. •, C. albicans strain CaTK2(wt); f, CaTK1; E, CaTRK2; OE, CaTK3. For uptake assays, K ϩ -starved cells (10 8 /ml; see "Experimental Procedures") were suspended in transport buffer (50 mM Tris-succinate at pH 5.9 plus 2.5% glucose) and equilibrated for 10 min, after which 1.07 mM Rb ϩ labeled with 0.1 Ci of 86 Rb was added. Cells were harvested and washed by vacuum filtration at intervals of 20 s and 1, 2, and 3 min, and 86 Rb was assayed by liquid scintillation spectrometry. Each data point represents the mean Ϯ S.D. of at least three independent experiments. The four sets of data were fitted by linear regression, forced to a common ordinate intercept. Calculated slopes (influxes, in units of nmol/(10 8 cells ϫ min)) were 19.5, 13.3, 11.9, and 2.7, read from top to bottom.
reduced the rate of ATP loss, compared with that in wild-type cells, to 0.11 nmol/10 6 cells/min, or ϳ55% of the control value. Thus, here again, ATP release seemed to be the hallmark of the degree of lethality of Hst 5. These data imply a complex kinetic relationship between expressed functional TRK1 protein and the process of ATP efflux. Patch clamp records from spheroplasts of C. albicans. Top row, whole-cell current traces, recorded with chloride-filled pipettes, from spheroplasts expressing 1-3 copies of the wild-type TRK1 gene, as indicated. Bottom row, left, whole-cell currents from two strains recorded with low chloride pipettes. Bottom row, right, voltage protocol imposed to produce the current traces in all six panels; a staircase of 2.5-s pulses was applied to the membrane, clamping to ϩ100 mV, ϩ80 mV . . . down to Ϫ180 mV, each from a holding voltage of Ϫ40 mV. Current traces for each test, as well as the voltage protocol, are shown superimposed. The upward currents (all six panels) represent K ϩ efflux through Tok1p, the K ϩ channel (34). Downward currents (top row) represent (Ͼ95%) chloride efflux through Trk1p, the K ϩ transporter (33,52). Vertical comparison (two left columns) demonstrates that reduction of intracellular chloride to submillimolar range nearly abolished the downward currents. All record sets are representative (near average) for several replicates obtained from independent experiments. . I-V plots were corrected for small leakage currents, as already described (33). Inset, phase plot of Rb ϩ fluxes (from Fig. 3) against maximal slope conductances to demonstrate equivalence of the two methods for measuring functional protein in the four strains. The regression line is forced through the origin.

Simultaneous Blockade of Hst 5-induced Killing and Trk1p
mediated Chloride Conductance-Discovery of the fact that Hst 5-induced killing of C. albicans was accompanied by major leakage of ATP (53), suggested that histatin must produce either greatly enhanced anion permeability or generalized breakdown of the plasma membrane. Attention was further focused on anion permeability by the fact that anion-channel blockers, such as DIDS, protected C. albicans cells against Hst 5-induced killing as well as against ATP loss. Bringing those observations together with the Trk1p-mediated chloride con-ductance prompted us to examine the effects of DIDS on the chloride currents through Trk1p. As is shown in Fig. 8, both the large inward (downward) currents in CaTK2(wt) and the small inward currents in CaTK1 were essentially completely blocked by 0.1 mM intracellular DIDS. Quantitatively, the currents at Ϫ180 mV were reduced from (Ϫ)175 pA/cell (CaTK2(wt)) or (Ϫ)25 pA (CaTK1) to the range of (Ϫ)3-5 pA, as can be seen by comparing the summary I-V plots in the two left panels of In an attempt to identify other anion channels that might also be affected by DIDS and related inhibitors, we searched the C. albicans genome data base (available on the World Wide Web at www.ncbi.nlm.nih. gov/sutils/genom_tree.cgi) and identified open reading frames for two potential C. albicans chloride channel proteins: one residing on contig 6-2409 and encoding a protein of 983 amino acids and the second residing on contig 6-2190 and encoding a protein of 879 amino acids. Both show significant sequence identity with mammalian CLC3, CLC4, and CLC5 chloride channels, but the actual expression and subcellular locations of these proteins are unknown. Both sequences differ greatly from ScGef1p, the chloride channel protein in S. cerevisiae, which is localized in endomembranes (the C. albicans GEF1 locus resides on contig 6-2519). 5 Testing of C. albicans single allele deletion strains for these two loci showed them to be at most 15-20% more resistant to Hst 5 than the wild-type, 6 observations which argue that, like the potassium channel Tok1p (34), the two identified chloride channels could play a modulating role, but not a primary role, in Hst 5 killing of C. albicans.
Trk1p-mediated chloride conductance modulated by DIDS is a thoroughly remarkable result, which not only confirms that Trk1 protein is the critical effector of Hst 5 toxicity in Candida but also proffers a mechanism: that Hst 5 binding to Trk1p produces a leakage pathway through either that protein itself or a larger complex involving Trk1p.

DISCUSSION
Mediation of Hst 5 Toxicity by Trk1p-We have constructed an isogenic set of C. albicans strains expressing one, two, or three copies of the TRK1 gene, in order to investigate possible functional roles of TRK1 protein in toxic actions of the cationic antimicrobial peptide, Hst 5. Although the quantitative relationships among TRK message, expressed protein, and functional protein (Rb ϩ influx, chloride conductance) proved complex, stark evidence for a critical role of Trk1p in Hst 5 toxicity is provided by the finding that the single allele deletion strain (CaTK1) displayed greatly reduced levels (less than 0.25ϫ normal) of TRK transcript and functional Trk1p, paralleling similarly reduced toxic effects of Hst 5 (killing, ATP efflux). That conclusion is further strengthened by restoration of essentially wild-type values for all parameters upon complementation of CaTK1.
Further insights into the molecular mechanisms underlying Hst 5 toxicity are provided by two surprising findings: (a) that the 3-TRK1 strain (CaTK3) displays a 50% increase of Trk1p function in the plasma membrane (Figs. 4 and 6), compared with the wild type, but a 40 -50% decrease of Hst 5 sensitivity, and (b) that the recently discovered anion permeability (Cl Ϫ ) (33) of Trk1p and the Hst 5-induced efflux of cytoplasmic ATP are equally inhibited by disulfonic stilbenes.
Anion Channeling through Cation Transporters-Hst 5-induced potassium loss, along with the reduction of Hst 5 sensitivity observed in elevated extracellular K ϩ , initially brought our attention to the K ϩ channel, Tok1p (34) and then to this work on Trk1p. However, reduction of Hst 5 toxicity by anion channel inhibitors such as DIDS seemed unrelated until the 5 Table II); intracellular (pipette) solutions were Buffer G Ϯ 0.1 mM DIDS in 1% Me 2 SO. Data are representative of four different protoplasts for each strain and condition tested. I-V plots were corrected for small leakage currents. ability of TRK proteins to conduct anions was recognized (33). In retrospect, chloride conductance through TRK-like proteins almost certainly underlay earlier observations in fungi, especially the chloride-mediated depolarization of Neurospora plasma membranes (54), but the general phenomenon of anion channeling through cation transporters only began to emerge a decade ago, in animal systems such as the EAAT proteins, which are responsible for sodium-coupled uptake of excitatory amino acids (55)(56)(57).
Peculiarly in Candida, but not in Saccharomyces, the TRKdependent chloride conductance is blocked by classic anion channel inhibitors, especially DIDS. However, most importantly, in protecting the organism from Hst 5 toxicity, DIDS reacts with the same protein identified by the expression level experiments as the likely target for primary action of Hst 5. Our current working hypothesis to account for these facts is that the structural path for Cl Ϫ through Trk1p is distorted by Hst 5 binding so that it permits the efflux of larger anions, such as ATP, but DIDS "seals" the channel against Hst 5 action. Trk1p-dependent loss of ATP and other large anions initiates cellular shrinkage and eventual cell cycle arrest that are thus blocked by DIDS (16). Moreover, as evidenced in Fig. 8A, the end result produced by DIDS is identical in the two strains (CaTK2(wt) and CaTK1), suggesting that elimination of the small residual Hst 5 sensitivity of CaTK1 by DIDS mimics the likely behavior of true trk1-null strains and re-emphasizes that the TRK1 transporter system qualifies as the critical target/ effector for Hst 5-mediated killing of C. albicans.
Overexpression of TRK1-The unexpected conflict between elevated Trk1p function and diminished Hst 5 toxicity observed in the overexpression strain, CaTK3, has at least two possible explanations. Most simply, cytoplasmically localized Trk1p is likely to be sequestered within endomembranes (karmellae or vesicular-tubular clusters) (49), as a device to counter the threat of excessive membrane permeability. Then, if direct interaction between Hst 5 and Trk1p occurs as postulated above, elevated endomembranal Trk1p could also bind significant Hst 5, decreasing the free concentration available to the effector site, which is plasma membrane-localized Trk1p.
An alternative possibility is that plasma membrane-bound Trk1p may have different oligomeric states, only some of which are susceptible to Hst 5. This hypothesis can be formalized in a simple kinetic scheme, for example with three distinct oligomers of the transporter: (a) naked Trk1p (T), which does not transport and does not bind Hst 5; (b) Trk1p bound to an activator (A), to make AT, which does transport and does bind Hst 5; and (c) Trk1p doubly bound to A, to make TAT, which transports (through both Ts) but does not bind Hst 5.
As the membrane concentration of T rises, the concentration of A is fixed and becomes limiting. Formally, this scheme appears as follows, with the equilibrium constants shown below. The total concentration of A in the membrane is fixed, independently of Trk1p, so that the following is true.
Then the concentration of AT in the membrane is given by the following expression, where the concentration of T can vary among the different strains. The 210-kDa protein seen in Western blots (Fig. 2B) is a candidate for at least the AT component of this set. That Trk1p indeed normally functions in oligomeric associations is suggested by a variety of additional information. Purely on the basis of structural modeling, Durell and Guy (44) postulated that yeast TRK proteins (in S. cerevisiae and Schizosaccharomyces pombe) should form symmetric dimers or tetramers. Without such clustering, the monomers, which fold like potassium channels (49), would have unusually well conserved surface residues, exposed primarily to membrane lipids. Hetero-oligomerization of TRK proteins in Saccharomyces is also a strong possibility because of the presence of three genes/proteins of unknown function but having strong sequence homology with plant potassium-channel ␤-subunits. Furthermore, bacterial TRK proteins, such TrkH in E. coli, are well known to be assisted by accessory subunits: TrkA, a NAD ϩ -dependent dehydrogenase (58), and SapD (TrkE), an ATP-binding protein (59). Thus, it seems likely that metabolic regulation of Candida Trk1p will involve accessory binding proteins, which should act by modulating the kinetic properties of expressed protein, poised at "normal" with the wild-type ratio of functional Trk1p to accessory protein. Further direct studies on Candida TRK protein will be required to test both of these models.
Energetic Factors-The well established ability of protonophores, such as CCCP and DNP, to reduce Hst 5 killing of C. albicans cells (8) can also be mediated in part via Trk1p. Although it is generally assumed that these classical uncoupling agents permeabilize plasma membranes as well as mitochondrial membranes, the hard evidence for such effects on plant and fungal cells is sparse, and where it does exist, the window between pure uncoupling effects and gross depletion of ATP is very narrow (60). In other words "uncoupler" effects on intact fungal cells are as likely to result from ATP depletion as from plasma membrane permeabilization, unless ATP levels and membrane conductance, explicitly, are measured. Also, sustained ATP depletion usually triggers down-regulation of membrane transport, by decreased membrane conductance (or leakiness), for cells that are not killed (60). Therefore, CCCP and DNP could depolarize Candida plasma membranes, with or without "tightening" them.
Since we have already noted that different doses of wild-type TRK1 have no significant effect on intracellular accumulation of Hst 5, uptake of this polycation, presumably voltage-driven, is certainly mediated by other proteins, such as NSC1. Thus, two distinctly different routes exist for uncoupler-mediated protection against Hst 5 toxicity: decreased uptake of the peptide because of membrane depolarization and down-regulation of Trk1p function, following from ATP depletion, which would be expected to choke its DIDS-sensitive anionic pathway. As suggested by Gyurko et al. (17), mitochondrial de-energization could play a role in both of these phenomena, via diminished ATP synthesis in the presence of uncouplers. However, differences in energy metabolism cannot account for the reduced susceptibility of CaTK1 to Hst 5, because all four strains were screened to verify the presence of normal oxidative metabolism. The fact that CaTK1 cells have typical respiration and oxidative phosphorylation but are almost completely resistant to Hst 5 killing demonstrates that the mitochondria themselves cannot be the critical locus for Hst 5 cytotoxicity, contrary to previous suggestions (14).
Our findings show that decreased expression of TRK1 and its product results in profound resistance to actions of the antifungal cationic protein Hst 5 and identifies the TRK1 potassium uptake system as critical for Hst 5 candidacidal activity. Whether other antimicrobial peptides/proteins may also exert their activities through similar effectors remains to be discovered. Trk1p and/or other potential cognate activator proteins represent a specific target site by which Hst 5 exerts cytotoxic effects against C. albicans, to provide nonimmune protection in the human-host defense system.