The Role of the STAS Domain in the Function and Biogenesis of a Sulfate Transporter as Probed by Random Mutagenesis*

Sulfate transporters in plants represent a family of proteins containing transmembrane domains that constitute the catalytic part of the protein and a short linking region that joins this catalytic moiety with a C-terminal STAS domain. The STAS domain resembles an anti-sigma factor antagonist of Bacillus subtilis, which is one distinguishing feature of the SLC26 transporter family; this family includes transporters for sulfate and other anions such as iodide and carbonate. Recent work has demonstrated that this domain is critical for the activity of Arabidopsis thaliana sulfate transporters, and specific lesions in this domain, or the exchange of STAS domains between different sulfate transporters, can severely impair transport activity. In this work we generated a Saccharomyces cerevisiae expression library of the A. thaliana Sultr1;2 gene with random mutations in the linking region-STAS domain and identified STAS domain lesions that altered Sultr1;2 biogenesis and/or function. A number of mutations in the β-sheet that forms the core of the STAS domain prevented intracellular accumulation of Sultr1;2. In contrast, the linking region and one surface of the STAS domain containing N termini of the first and second α-helices have a number of amino acids critical for the function of the protein; mutations in these regions still allow protein accumulation in the plasmamembrane, but the protein is no longer capable of efficiently transporting sulfate into cells. These results suggest that the STAS domain is critical for both the activity and biosynthesis/stability of the transporter, and that STAS sub-domains correlate with these specific functions.

trudes into the cytosol of the cell (3,4). The STAS domain is joined to the catalytic domain of transporters by a linking (L) region, varies in length, and is not highly conserved relative to the catalytic domain. Some characteristics that make SLC26 transporters distinct from each other are likely to reside within unique features of L-STAS domains.
Interestingly, the STAS domain of SLC26 transporter proteins resembles the bacterial anti-sigma factor antagonist SpoIIAA of Bacillus (2). SpoIIAA is a transcription factor that associates with the anti-sigma factor SpoIIAB; this association causes release of the sigma factor from SpoIIAB, triggering sporulation-specific transcription (5). SpoIIAB is also a kinase that can phosphorylate and inactivate SpoIIAA. Recently, STAS domains have been shown to be part of polypeptides other than anion transporters (6), suggesting that STAS sequences may serve a general regulatory function.
STAS domains of SLC26 transporters appear to play a role in the function/regulation of transport activity, because mutations that alter this domain can cause loss of function resulting in the diseases diastrophic dysplasia, Pendred syndrome, and congenital chloride diarrhea (7)(8)(9)(10)(11)(12). Recent work, using a heterologous Saccharomyces cerevisiae system, has demonstrated that the STAS domain is critical for the activity of the Arabidopsis thaliana sulfate transporter Sultr1;2 (13,14). When the L-STAS region of the sulfate transporter was eliminated, most of the truncated protein either never reached the S. cerevisiae plasmamembrane and was rapidly degraded, either en route to or once associated with the plasmamembrane. In contrast, if the STAS domain was deleted (L region maintained), the transporter integrated into the plasmamembrane, but the integrated protein had no detectable uptake activity. Studies of the DRA (down-regulated in adenoma) Cl Ϫ -HCO 3 Ϫ antiporter (SLC26A3) also demonstrated a STAS domain requirement for transport function (15). Furthermore, altering A. thaliana Sultr1;2 for codons encoding two cysteines or a threonine of the STAS domain (the latter is analogous to the phosphorylated serine of SpoIIAA) resulted in a polypeptide no longer able to rescue the phenotype of a S. cerevisiae sulfate transporter mutant (14). These results suggest that the STAS domain contributes to both the catalytic activity and biogenesis of Sultr1;2 and probably has analogous functions in other SLC26 proteins.
The control of ion homeostasis in cells is likely to involve interactions among transporters, ion channels, regulatory proteins, proteins involved in ion assimilation and sequestration, and concentrations of ions and perhaps other small molecules in the cytoplasm and/or extracellular environment. For SLC26 family proteins, the STAS domain appears to participate in some aspects of homeostatic control. Recent studies with the DRA Cl Ϫ -HCO 3 Ϫ antiporter have demonstrated the importance of STAS domain interactions with the regulatory sequence of the cystic fibrosis transmembrane conductance regulator, a cAMP-regulated Cl Ϫ channel. This interaction modulates the activity of both the transporter and the channel (16).
To elucidate protein domain function, the generation and characterization of mutants can be invaluable. S. cerevisiae has served as a heterologous expression system for many eukaryotic genes, and mutant strains have been used to study the catalytic activity and kinetic features of several plant sulfate transporters (13,14,(17)(18)(19)(20). In this study we generated a yeast expression library of the A. thaliana Sultr1;2 gene that had random mutations in the L-STAS domain and selected individual, mutated Sultr1;2 genes by transforming them into CP154-7B, which is null for the two major S. cerevisiae sulfate transporters (21). Transformants were analyzed for their ability to grow on SO 4 2Ϫ as a sole sulfur (S) source, ability to transport SO 4 2Ϫ , and capacity to accumulate the altered Sultr1;2 polypeptide and properly target it to the plasmamembrane. Lesions in the L-STAS-encoding region of altered Sultr1;2 were characterized to determine which amino acids were critical to the functionality/biogenesis of the Sultr1;2 polypeptide. The consequences of mutations in the STAS domain are discussed and interpreted with the aid of a computer-derived structural model based on the crystal structure of SpoIIAA (14,22).

Construction and Screening of a Library of Sultr1;2 Containing Random Mutations in the STAS Domain
Intermediate Construct (pIMRNDM)-Plasmids used for expression of the C-terminal, hemagglutinin (HA)-tagged sulfate transporters were constructed in pYX222x (13) (Fig. 1A). This plasmid (a) was digested with EcoRI and BamHI and the resulting linearized vector (aЈ) in which the region encoding the catalytic domain plus part of the region encoding the L-STAS domain (to the BamHI site) was deleted. The deleted plasmid was ligated with a PCR-amplified EcoRI-BamHI fragment containing the sequence encoding Sultr1;2, extending from the initiator Met-1 to Phe-478 (the end of the catalytic region; this fragment is designated "b" in Fig. 1A). The primer set used to create b was forward (5Ј-ata gaa ttc atg tcg tca aga gct cac cct gtg gac-3Ј) and reverse (5Ј-ata gga tcc aaa gat gac gcc aaa gaa tgc tcc aat-3Ј). The resulting intermediate construct, pIMRNDM, was used to clone a randomly mutagenized L-STAS encoding region. For this purpose, we introduced a BamHI site at nucleotides 1495-1500 in the Sultr1;2 cDNA (GenBank TM accession no. AB042322), which resulted in a V479G substitution in TMD12 of the catalytic region of the transporter. This amino acid is not conserved among S. cerevisiae, plant, and human SLC26 family members (it could be Val, Leu, Gly, Ala, Ile, or Phe), and this region of the C-terminal TMD does not appear to directly influence substrate transport.
Construction of Sultr1;2 Library with Mutated STAS Domains-During the amplification of the Sultr1;2 cDNA, we identified an allele of the Sultr1;2 gene with the missense mutation I553V (a to g at position 1717) that arose as a consequence of the natural error rate of TaqDNA polymerase. This mutation eliminated the BamHI site in the Sultr1;2 cDNA. The CP154-7B strain transformed with the allele containing V479G/I553V (WTЈ) was rescued for the methionine auxotrophy phenotype and exhibited SO 4 2Ϫ transport activity similar to that of CP154-7B harboring the unaltered Sultr1;2 gene (WT, Fig. 1B). The Sultr1;2 cDNA encoding the I553V lesion was used as a template for PCR random mutagenesis; this enabled us to exploit the BamHI sites at the N-terminal end of the L-STAS domain for digesting (with BamHI/SalI) and cloning of PCR fragments into pIMRNDM deleted for the L-STAS region of Sultr1;2. Random mutagenesis of the L-STAS encoding region was performed by PCR amplification of the cDNA using the DiversityKit (Clontech, Palo Alto, CA) and the primer set RNDMFB (5Ј-ata gga tcc gtt gag atc gga ctt ctt att gcc gtc tcg-3Ј) and RNDMRS (5Ј-act aag tcg acc gac ctc gtt gga gag ttt tgg aca gca-3Ј). Two conditions with different MnSO 4 concentrations (320 M and 640 M, for "mild" and "severe" mutagenesis, respectively) were used to generate the mutagenized libraries. The mildly mutagenesis library had 2-3 mutation per 1000 bp, whereas the severely mutagenized library had 4 -6 mutations per 1000 bp. The PCR fragment of the mutated L-STAS region ("e" in Fig. 1A) was digested with BamHI and SalI and cloned into pIMRNDM digested with the same enzymes ("cЈ" in Fig. 1A).
Selection and Growth of Transformants-The cDNA L-STAS-mutagenized library was transformed into CP154-7B by the lithium acetate method (23), and transformants were selected for histidine auxotrophy. Each transformant was streaked onto low SO 4 2Ϫ SD medium (500 M SO 4 2Ϫ , 400 M methionine, and 96 M histidine) and high SO 4 2Ϫ SD medium (40 mM SO 4 2Ϫ , 143 M methionine, and 96 M histidine) to test for SO 4 2Ϫ -dependent growth. Growth of the transformants was monitored at 30°C for 2-3 days. DNA fragments coding for the L-STAS domains of the various Sultr1;2 alleles were amplified by colony PCR using a primer set that annealed to sequences bordering the L-STAS-encoding region (forward, 5Ј-c atc ctc gca gct atc atc at-3Ј and reverse, 5Ј-ctc agt tag cta gct gag-3Ј), and the amplified fragment was sequenced.

Growth, Subcellular Location, and Sulfate Uptake
CP154-7B transformants were grown at 30°C in SD-His-Met liquid medium, containing 500 M MgSO 4 , or SD-His liquid medium, containing ϳ40 mM SO 4 2Ϫ . Transformants were assayed for rates of growth, SO 4 2Ϫ transport activity, and subcellular location of Sultr1;2. Cell growth was evaluated by measuring the optical density of cultures at 600 nm (A 600 ) in a DU640 spectrophotometer (Beckman Coulter, Palo Alto, CA). Cells in mid-logarithmic phase of growth (A 600 ϭ 0.2-0.3) were used for fractionation of cell membranes and polypeptide analysis (13). SO 4 2Ϫ uptake was assayed in exponentially growing cells (A 600 Ͻ 0.4) (13).

RESULTS
Generating Mutagenized L-STAS of Sultr1;2-Random mutations were introduced into the Sultr1;2 sequence encoding the L-STAS domain (from the end of TMD12 to the end of the polypeptide; described under "Experimental Procedures"). The mutant Sultr1;2 cDNA library was transformed into S. cerevisiae CP154-7B (21), and transformants were tested for methionine-independent, SO 4 2Ϫdependent growth on solid medium ( Fig. 2A). This assay provided an evaluation of the effectiveness of the introduced Sultr1;2 gene in rescuing the methionine auxotrophy phenotype of CP154-7B. Sultr1;2 alleles carried by transformants unable to grow in the absence of methionine were mostly from the "mild condition" mutagenized library (except for alleles "Y517CϩV616E," "N545I," and "S588R"), because the "severe condition"-mutagenized library often gave multiple mutations, making it difficult to interpret the phenotypes; these alleles were categorized as "nonfunctional." We also identified "functional" alleles in transformants for which the mutated Sultr1;2 was able to rescue the methionine auxotrophy phenotype of CP154-7B ( Fig. 2A). These functional alleles were isolated using the severe condition mutagenized library, because the majority of changes in functional alleles identified from the mild mutagenesis library were silent (the amino acid sequence was not altered). Colony PCR was performed to amplify DNA fragments encoding the L-STAS domain of Sultr1;2 in transformants, and amplified fragments were directly sequenced to identify mutated bases. For nonfunctional mutants, only alleles containing a single missense mutation were chosen for further analyses, except for the allele with both Y517C and V616E substitutions. Table 1 shows specific changes in STAS domain 2Ϫ uptake rate in CP154-7B cells expressing wild-type Sultr1;2 (WT) and Sultr1;2 with V479G and I553V mutations (WTЈ). Assays were performed as described under "Experimental Procedures," and uptake rates were normalized to dry weight (DW).
sequences from transformants expressing functional or nonfunctional Sultr1;2, provides a qualitative view of accumulation of Sultr1;2 in cells, and indicates whether or not the transporter localizes to the plasmamembrane (see below). A DNA fragment encoding the L-STAS peptide from each mutant allele was recloned into pIMRNDM to generate full-length Sultr1;2, the ligated insert was sequenced to ensure that no modification occurred during the cloning procedure, and the plasmid with the recombinant gene was transformed into CP154-7B; re-transformation was performed to confirm that the phenotype was solely a consequence of a lesion in the L-STAS region of the polypeptide.
The previously characterized A. thaliana sel1-3, sel1-7, and sel1-8 mutants have lesions in the L-encoding region of Sultr1;2 (G509E, P503L, and I511T, respectively); these lesions make the plant resistant to selenate (27). SO 4 2Ϫ uptake rates in the sel1-8 mutant of A. thaliana were reported to be lower than in the wild-type plants (27). None of these altered Sultr1;2 alle-les could rescue the methionine auxotrophy phenotype of CP154-7B, as shown in Fig. 2A (compare sulfate, 40 mM, with sulfate, 500 M, in the top row of each grid). These mutant alleles do not have the lesions incorporated into the constructs for cloning purposes (V479G and I553V).
Phenotypic Analysis of S. cerevisiae Cells Expressing Sultr1;2-SO 4 2Ϫ -dependent growth characteristics of strains newly transformed with plasmids containing WT, WTЈ, and the randomly mutagenized Sultr1;2 alleles were monitored for growth in high and low SO 4 2Ϫ medium. CP154-7B transformed with unaltered Sultr1;2 or with the Sultr1;2 allele with the V479G and I553V changes doubled in ϳ2. 8  Accumulation of Nonfunctional Sultr1;2-Sultr1;2 protein accumulation in nonfunctional transformants was evaluated in whole cell extracts by Western blot analyses using antibodies specific for the HA tag (positioned at the C terminus of the Sultr1;2 protein). The fusion of the HA epitope to the C terminus of the introduced protein does not affect the ability of Sultr1;2 to rescue the CP154-7B mutant phenotype (13). An antibody against Pma1 was used as a control to identify the plasmamembrane fraction and to normalize the HA signal. The level of the HA signal relative to that of Pma1 is shown in Fig. 3A and quantified in Fig. 3B. Some transformants with nonfunctional Sultr1;2 had significantly less Sultr1;2 protein than cells transformed with WTЈ Sultr1;2. Transformants for which the ratio of Sultr1;2 to Pma1 signal was 20% or less relative to that measured in transformants harboring the WTЈ Sultr1;2 construct were S546P, Q573P, I577N, E578K, and L607W. Low level protein accumulation in these strains is likely to explain their inability to grow on low SO 4 2Ϫ medium and to efficiently transport SO 4 2Ϫ ; these mutant alleles are classified as Class I in Table 1.
Most transformants with nonfunctional Sultr1;2 had polypeptide levels similar to that of the WTЈ transformant (Table 1; Class II). Some transformants for which the CP154-7B mutant phenotype was not rescued by the introduced Sultr1;2 gene, even though the Sultr1;2 protein in these strains appeared to accumulate, may experience difficulties in routing the expressed polypeptide to its site of function on the plasmamembrane. In a previous study (28), a G678V mutation in the STAS domain of DTDST (SLC26A2) abolished plasmamembrane targeting of the protein. Therefore, we examined all transformants determined to have nonfunctional Sultr1;2, but that accumulated a range of Sultr1;2 levels, for distribution of the transporter protein among the different cellular membrane fractions (Fig. 4). Transformants with low overall levels of Sultr1;2 polypeptide showed either no (S546P and L607W) or very low (V537D, V537A, Q573P, I577N, and E578K) accumulation in the plasmamembrane (Fig. 4, fractions 10 -13 from the gradient). In some cases (V537D, I577N, and L607W) a significant level of Sultr1;2 was observed in membrane fractions lighter than the plasmamembrane. These results demonstrate that some of the aberrant Sultr1;2 proteins cannot reach their site of function in the plasmamembrane (or reach it less efficiently, or the polypeptide is not stable once in the plasmamembrane), explaining their inability to rescue CP154-7B for methionine auxotrophy. However, the majority of transformants that were unable to grow with SO 4 2Ϫ as the sole S source (e.g. I496N, P503L, G509E, I511T, I511N, Y517C/V616E, R518S, Q522K, Y523H, V537A, Y542C, F543Y, N545I, V549F, S588R, N611K, N611S, and G630V) showed a relatively normal distribution of Sultr1;2 among the membrane fractions, with much of it being localized to the plasmamembrane. The Sultr1;2 proteins synthesized in these latter transformants (Class II) are likely to be defective for transport function.
Transformants Synthesizing Functional Sultr1;2-Transformants for which the mutated Sultr1;2 was able to rescue the methionine auxotrophy phenotype of the CP154-7B are listed in Table 1 as Class III. Some of these mutants had more than one lesion within the L-STAS region. The doubling time of these strains in both low and high SO 4 2Ϫ medium is presented in Fig. 2B. Relative accumulation of Sultr1;2 in these strains and their ability to take up SO 4 2Ϫ are shown in Fig. 5, A and B, respectively. We also calculated a K m and V max for the uptake of SO 4 2Ϫ by each strain, as shown in Table 2. Growth rates of individual strains varied and generally showed a positive correlation with the V max for SO 4 2Ϫ uptake ( Table 2; growth rate data are from Fig. 2B); the growth of E560G, which has a large standard deviation, is the only exception. The decreased rate of SO 4 2Ϫ uptake in many of the strains harboring functional Sultr1;2 mutant alleles was mostly explained by decreased protein accumulation, with a few exceptions. T527A/ L615M/T637A and A540S/V549I/I608N show decreased SO 4 2Ϫ uptake relative to WTЈ (Fig. 5B), whereas protein accumulation (Fig. 5A) and distribution in plasmamembrane are comparable to that of WTЈ (data not shown). V549F also shows very limited uptake; its V max was 10% of WTЈ, whereas protein accumulation was approximately half of WTЈ.

TABLE 1 Summary of mutations analyzed in this work
Class I includes alleles of the mutated Sultr1;2 gene for which the protein accumulates to significantly lower levels (Ͻ20%) than that of the unaltered Sultr1;2 polypeptide but does not rescue the methionine auxotroph phenotype of CP154-7B. Class II includes alleles of mutated Sultr1;2 gene for which the protein is expressed and accumulates in the plasmamembrane, but it does not rescue the methionine auxotroph phenotype of CP154-7B. Class III includes alleles of mutated Sultr1;2 gene for which the protein confers varying SO 4 2Ϫ uptake activity to CP154-7B, but the transformants may not grow as rapidly in low SO 4 2Ϫ medium as CP154-7B transformed with the unaltered Sultr1;2 gene. Complementation of the growth phenotype of CP154-7B on SO 4 2Ϫ is based on data shown in Fig. 2: Ϫ, no growth on SO 4 2Ϫ ; ϩ and ϩϩ, slow growth on SO 4 2Ϫ (ϩ, doubling time Ͼ10 h; ϩϩ, 6 -10 h in low SO 4 2Ϫ medium); ϩϩϩ, growth on SO 4 2Ϫ is comparable to that of CP154-7B transformed with the unaltered Sultr1;2 gene. Protein accumulation measurements are based on immunological signals from the fused C-terminal HA epitope relative to the Pma1 signal, as shown in Fig. 3: ϩ, Ͻ20% of wild type; ϩϩ, between 20% and 70% of wild type; ϩϩϩ, Ͼ70% of wild type. Accumulation in the plasmamembrane is based on data presented in Fig. 4.

DISCUSSION
This study focused on determining residues of the L-STAS domain critical for biogenesis and functionality of Sultr1;2. We performed random mutagenesis of the Sultr1;2 sequence encoding the L-STAS domain, replaced the unaltered L-STAS encoding region with a population of mutagenized L-STAS encoding regions, and identified lesions that resulted in both functional and nonfunctional Sultr1;2 polypeptides (Table 1 and Fig. 6). Among the alleles encoding nonfunctional Sultr1;2, some accumulated little protein (Class I), most of which was unable to properly integrate into the plasmamembrane, and others accumulated normal protein levels with seemingly proper plasmamembrane localization (Class II). We also identified alleles with lesions in the L-STAS domain that could complement the CP154-7B phenotype (Class III), although usually not fully restore the growth rate in low SO 4 2Ϫ medium. To help interpret the effect of amino acid substitutions in mutated Sultr1;2, we used a STAS domain model, based on the SpoIIAA crystal structure (14). The STAS domain is thought to consist of four ␤-strands, forming a ␤-sheet, surrounded by five ␣-helices, as shown in Fig. 7. The ␤-sheet, in association with hydrophobic surfaces of the ␣-helices, forms a hydrophobic core that is not readily accessible to the external medium. In contrast, the externally exposed surfaces of the ␣-helices, and the loops between ␣-helices and ␤-strands, are predicted to be exposed and available for interactions with molecules in the solution environment. Although the L region was shown to be of importance for transporter function, it could not be modeled because it is not very similar to other known structures in the Protein Data Bank (PDB).
Mutations That Affect Biogenesis of Sultr1;2-␤-Sheet-Class I mutations diminished Sultr1;2 protein accumulation in the plasmamembrane (surrounded by blue boxes in Fig. 6), and some Class III mutations decreased protein accumulation, with some activity still retained (surrounded by black boxes in Fig. 6). Interestingly, most of the lesions for which the protein shows little accumulation are in or immediately contiguous to ␤-strands 3 and 4 that are part of the ␤-sheet structure (Figs. 6 and 7); the mutations in or immediately adjacent to ␤1 and ␤2 were nonfunctional, but the protein accumulated in the plasmamembrane.
The Q573P, I577N, and E578K substitutions in ␤3 all alter the charge/hydrophobic character of the domain (Pro poten- A, lysates were prepared from CP154-7B expressing various mutated forms of Sultr1;2. The polypeptides of the lysates were resolved by SDS-PAGE, transferred to polyvinylidene difluoride membranes, and the Sultr1;2 and Pma1 proteins were detected using antibodies, as described under "Experimental Procedures." B, signal strengths of antibody reactions were quantified using Image J (National Institutes of Health, rsb.info.nih.gov/ij/), and those for Sultr1;2 (anti-HA antibodies) were normalized with respect to those for Pma1 (anti-Pma1 antibodies). *, allele "Y517C" also has a V616E substitution. FIGURE 4. Subcellular localization of mutant Sultr1;2 polypeptides. Cell lysates (see "Experimental Procedures") were fractionated on 30 -60% sucrose gradients, fractions were separated by SDS-PAGE, and Sultr1;2 immunodetected by anti-HA antibodies (HA). Distributions of the plasmamembrane marker Pma1 (PM) and the ER marker Dpm1 (ER) were visualized by Western blot analyses using monoclonal antibodies. The plasmamembrane was mainly in fractions 10 -13. *, Tyr-517 allele also has a V616E mutation. The "L" indicates low accumulation of Sultr1;2 in whole cell extracts; **, no Sultr1;2 was detected in the plasmamembrane.
tially causes a kink in the polypeptide); polypeptides with these amino acid substitutions do not accumulate in CP154-7B (Fig.  3). As shown in the logo diagram (Fig. 6), Ile-577 and Glu-578 are highly conserved, whereas Gln-573 is not. Substitution of the conserved, hydrophobic Leu-607 to Trp in ␤4 also caused marked reduction in polypeptide accumulation (Fig. 3).
Some mutations predicted to be in the ␤-sheet vicinity, as shown in Fig. 7, also caused the polypeptide to accumulate to a lesser extent than WTЈ Sultr1;2. Sultr1;2 with N611K or N611S (end of ␤4) substitution cannot rescue the CP154-7B phenotype. These strains exhibit a significant amount of protein accumulation (considered to be Class II), but lower than in WT, suggesting that the N611K/S lesions negatively affect protein biogenesis as well as transport function. Asn-611 is analogous to Asn-81 in SpoIIAA of B. sphaericus. The side chain of Asn-81 forms a hydrogen bond with Ala-102 in ␤5 in SpoIIAA, extending the ␤-sheet structure (22). Thus, Asn-611 of Sultr1;2 STAS may form an analogous hydrogen bond with Leu-636 (corresponds to Ala-102 in SpoIIAA) and thereby help stabilize the ␤-sheet structure of the polypeptide. CP154-7B transformed with Sultr1;2 with the G630V substitution (between ␣3 and ␣4, Class III), predicted to be in the region corresponding to ␤5 in SpoIIAA, exhibited reduced SO 4 2Ϫ uptake activity (Fig. 5B), which also correlated with reduced protein accumulation (Fig. 5A and Table 2), suggesting that the G630V substitution (shaded gray in Fig. 7) also affects protein biogenesis/turnover rather than function.
In addition, the V537A and V537D mutations, considered Class II, are in the C-terminal region of ␤2 (Fig. 6, surrounded by purple boxes); the former lesion did cause significant reduction in polypeptide accumulation, whereas the polypeptide level in the latter remained high (Fig. 3). The V537D substitution appears to result in a defect in plasmamembrane localization of Sultr1;2 (Fig. 4). These results suggest that Val-537 may be involved in the biogenesis of Sultr1;2 as well as transport function. Next to Val-537, a D538G substitution generated a Class III mutant; the Sultr1;2 protein accumulates to low levels in CP154-7B with low SO 4 2Ϫ transport activity. However, this allele has two other lesions, R550G (␣1) and I617M (␣3). The Sultr1;2 protein with a single I617M substitution accumulated in CP154-7B transformants to a level comparable to that of the unaltered protein and exhibited near normal levels of SO 4 2Ϫ transport activity. Thus D538G and/or R550G are likely to be responsible for the defect in protein biogenesis.
Together, these findings demonstrate that lesions within or close to the STAS domain ␤-sheet of Sultr1;2 can significantly change protein stability or biogenesis. This appears to be especially true when the lesions are in ␤3 and ␤4. The ␤-sheet seems to serve as a core structure of the STAS domain and lesions within this structure may disrupt proper STAS packing, which could destabilize the entire Sultr1;2 polypeptide.
Other Destabilizing Lesions-Although most modifications of the L region do not significantly alter protein accumulation, FIGURE 5. Analysis of strains expressing functional Sultr1;2 in CP154-7B. A, Sultr1;2 accumulation was determined in the same manner as described for Fig. 3. B, SO 4 2Ϫ uptake rates (see "Experimental Procedures") of CP154-7B expressing mutant and wild type (WTЈ) Sultr1;2 were determined. The uptake assay was performed with cells grown in SD-His-Met medium containing 500 M SO 4 2Ϫ at an A 600 ϭ 0.1ϳ0.3. The standard deviation for uptake values for all mutants was small, except for the Q521L strain, for which standard deviation values are shown. Mutant alleles marked with asterisks contain additional lesions. T527A* represents Sultr1;2 with the substitutions T527A, L615M, and T637A. Other alleles of Sultr1;2 with multiple mutations are A540S* (A540S, V549I, and I608V); E560K* (E560K, I608V, and P614S); and D538G* (D538G, R550G, and I617M).

TABLE 2
Phenotypes associated with Class III Sultr1;2 alleles V max and K m are based on data shown in Fig. 5B, relative protein accumulation is from Fig. 5A, and doubling time in low SO 4 2Ϫ liquid medium is from Fig. 2B. Standard deviations are in parentheses.

Function of STAS Domain in Sulfate Transporters
the I496N lesion is a notable exception; this mutant polypeptide accumulates to ϳ25% of the level of unaltered Sultr1;2 (the lowest level among Class II mutant alleles), with some being localized to the plasmamembrane (Figs. 3 and 4). Hence, the L region may play some role in the biogenesis/stabilization of Sultr1;2. Recently, the motor protein prestin (SLC26A5), altered at a position corresponding to Ile-496 in Sultr1;2 (V499G in prestin), with an additional Y501H change, exhibited normal protein accumulation with diminished functionality (29). Position Ile-496 is represented by hydrophobic amino acids in most sulfate transporters (Fig. 6, I496N mutations are  surrounded by a purple box).
The S546P substitution is the only Class I modification characterized that is in an ␣ helix ( Table 1). Introduction of a Pro residue in this ␣ helix may cause extreme conformation perturbation by introducing steric constraints into the protein and disrupting the ␣1 structure, which in turn could lead to a marked decrease in protein accumulation. Furthermore, CP154-7B expressing Sultr1;2 with a E560G lesion exhibits slow growth and low rates of SO 4 2Ϫ uptake, which correlates with reduced protein accumulation ( Table 2). These results suggest a potential role of Glu-560 in protein biogenesis/stability rather than in function (it is shaded gray in Fig. 7). Glu-560 is highly conserved among A. thaliana Sultr polypeptides and FIGURE 6. Sequence logos representing residue frequencies. Frequencies were determined using twelve A. thaliana sulfate transporters, five human SLC26 family members, and PDB code 1h4z (SpoIIAA). Features of the presentation were previously described (Ref. 24; available at weblogo.berkeley.edu/). The y axis indicates how often a particular amino acid is present at the location. Basic amino acids (K, R, and H) are blue, acidic amino acids (D and E) are red, polar amino acids (G, S, T, Y, and C) are green, N and Q are purple, and hydrophobic amino acids (A, V, L, I, P, W, F, and M) are black. Only amino acids with a certain degree of conservation are included in the output of the logo diagram. Unconserved amino acids do not appear in the output. The amino acid sequence of Sultr1;2 is below the sequence logos, and specific mutations analyzed are indicated below the Sultr1;2 sequence. Blue boxes mark nonfunctional Class I substitutions that cause a decrease in protein accumulation. Red boxes mark nonfunctional Class II substitutions that allow for high level protein accumulation, and proper localization of the protein to the plasmamembrane; exceptions are I496N and V537A/D (surrounded by purple boxes), which affect protein accumulation more severely than the other Class II lesions (see text). Black boxes mark Class III mutations, which often result in reduced protein accumulation, but the Sultr1;2 that remains confers some uptake activity to transformants. Asterisks indicate alleles with multiple mutations. a, allele V616E* has Y517C; b, allele T527A* has L615M and T637A; c, allele D538G* has R550G and I617M; d, allele P614S* has E560K and I608V; and e, allele I608V* has A540S and V549I. Two mutations introduced for cloning purposes during the construction of mutagenized library are also indicated (V479G and I553V).
part of an acidic cluster located at the C-terminal end of ␣1, or possibly at the beginning of the variable loop (2).
Mutations That Affect Activity of Sultr1;2 but Not Biogenesis-Class II mutations are defined as lesions that eliminate the functionality of Sultr1;2, although the aberrant polypeptide accumulates to relatively normal levels in the plasmamembrane (surrounded by red boxes in Fig. 6). These lesions are mostly in the L region and the protein surface formed by the N termini of ␣-helices and some contiguous sequences. These subdomains appear to be crucial for maintaining transporter activity.
L Region-Most of the lesions in the L region of Sultr1;2, including P503L, G509E, I511T, I511N, Y517C/V616E, and R518S, caused little change in either protein abundance or plasmamembrane localization, suggesting an involvement of the L region in sulfate transporter function and not biogenesis. All of these positions are highly conserved among SLC26 family members. Pro-503 is in a predicted loop that may separate two ␤-strands or a ␤-strand and ␣-helix, based on various algorithms used for protein structure predictions (PELE Protein Structure Prediction at workbench.sdsc.edu) and is fully conserved in SLC26 family members (Fig. 6). This Pro may undergo cis-trans isomerization, which may specify the physical distance between the catalytic moiety of the transporter and the STAS domain to, in some unknown way, regulate transport activity. Substitution by a Leu would markedly change this distance.
Specific STAS Surface Critical for Activity-Sultr1;2 with individual Y542C, F543Y, or N545I substitutions appeared completely nonfunctional, even though there was normal accumulation of the polypeptide in the plasmamembrane (Figs. 2-4). These residues are located in the N terminus of ␣1 and the loop adjacent to ␣1 (Figs. 6 and 7), regions that are well conserved in transporters of the SLC26 family (Fig. 6). It is noteworthy that a Cys is observed at the position analogous to Tyr-542 of Sultr1;2 only in low affinity type sulfate transporters (Sultr2;1 and Sultr2;2), and a Tyr at the position analogous to Phe-543 of Sultr1;2 in human SLC26A proteins, but not in those of plants (Fig. 7). This suggests strict Tyr and Phe requirements at these positions in some SLC26 transporter sub-types. In addition, Sultr1;2 with three amino acid substitutions, A540S, V549I, and I608V, accumulates to normal levels in transformed CP154-7B but has reduced SO 4 2Ϫ uptake activity (Fig. 5  and Table 2, Class III). Although it is not known which among the three substitutions leads to reduced transport activity, A540S is most likely to have the greatest effect on the structure/function of the polypeptide, because the V549I and I608V substitutions are likely to have little effect on protein structure/activity: they have highly similar character to Val and Ile (therefore, a red box was placed around A540S in Fig. 6). Position 540 is occupied by a nonpolar amino acid (Pro, Ala, or Leu) in most SLC26 family members (Fig. 6). Substitution by Ser, which is polar, may disturb molecular interactions required for function. The results presented above suggest that the N terminus of ␣1 plus the adjacent loop are critical for maintaining Sultr1;2 function (not accumulation or localization). This conclusion is also supported by the finding that the I544N allele of human DRA (HCO 3 Ϫ -Cl Ϫ exchanger) is functionally inactive and causes congenital chloride diarrhea (15,30); Ile-544 in human DRA corresponds to Ile-541 in Sultr1;2.
The polypeptide surface defined by the N terminus of ␣1 can be extended toward the top end of the ␤-sheet structure and the N terminus of ␣2. These regions also contain Class II lesions, and the side chains of all of these residues are in close proximity in the three-dimensional structural model (delimited by dotted lines in Fig. 7). The Q522K and Y523H substitutions, at the N terminus of ␤1 (Fig. 6), result in accumulation of nonfunctional Sultr1;2 in the plasmamembrane (Figs. 3 and 4). Sultr1;2 with V537A/D (C-terminal end of ␤2) or S588R mutations (N-terminal end of ␣2) are also nonfunctional; these mutant proteins accumulate in cells to a somewhat lower level than the WTЈ protein (Figs. 3 and 4). A T587A substitution (potentially a phosphorylation site) was previously shown to eliminate Sultr1;2 activity and is a potential site of phosphorylation (14). The Ser-588, Thr-587, and Asp-586 side chains have the potential to form charge-dipoles, similar to the interactions demonstrated for SpoIIAA (22).
Overall consideration of the polypeptide structure predicts that residues Gln-522, Tyr-523, Val-537, Ala-540, Tyr-542,  Fig. 6. Note that E560G and G630V (substitutions in gray) cause decreased accumulation of the Sultr1;2 polypeptide, but the remaining protein exhibits some activity (suggesting that Glu-560 and Gly-630 are important in Sultr1;2 biogenesis). The mutations that allow Sultr1;2 protein accumulation but diminish its function are mostly clustered on the STAS surface, which is delimited in both STAS domain orientations shown by dotted lines. This surface includes Ala-540, Tyr-542, Phe-543, and Asn-545, which are in or contiguous to the N terminus of ␣1, Thr-587 and Ser-588 in the N-terminal end of ␣2, Gln-522 and Tyr-523 in the N-terminal end of ␤1, and Val-537 in the C-terminal end of ␤2.
Phe-543, Asn-545, Thr-587, and Ser-588 are located on the same STAS surface (see Fig. 7). Hence, lesions on this surface appear to generate a protein with severely compromised function. This STAS surface is analogous to a region of SpoIIAA that allows SpoIIAB binding and subsequent SpoIIAA phosphorylation (31). Although there is no evidence of phosphorylation of STAS domains of SLC26A family proteins, accumulation of critical amino acid residues along the STAS surface associated with protein interactions in SpoIIAA raises the possibility of molecular interactions that affect sulfate transporter activity in this analogous STAS sub-domain. Recently, preliminary results have demonstrated an interaction between L-STAS regions (these regions appear to form a homomeric oligomer), and that both the L and the STAS domains are involved in this interaction; more work is required to determine the precise requirements for this interaction.
Furthermore, there seems to be a requirement for a Val at position 549 of Sultr1;2, because the V549F substitution (toward the middle of ␣1) shows only residual transporter activity, although the protein accumulates to nearly the same extent as the WTЈ protein (Fig. 5) and is localized in the plasmamembrane (data not shown). A substitution by Phe, another hydrophobic amino acid, but with a bulky side chain, causes a loss of activity. Interestingly, Phe is present at this position in some human but not in plant SLC26A proteins (Fig.  6). Examination of the predicted three-dimensional structure of the Sultr1;2 STAS domain suggests that Val-549 is either part of or contiguous to the putative interacting surface discussed above (see Fig. 7) and that a V549F substitution may distort the surface, compromising specific interactions.
The results for some mutated proteins are more difficult to interpret because of the presence of multiple lesions. The T527A(␤1) plus L615M(␣3) plus T637A (between ␣3 and ␣4) strain accumulates Sultr1;2 to approximately the same level as that of WTЈ but has ϳ30% lower SO 4 2Ϫ uptake activity (Fig. 5), suggesting that none of these changes significantly alters Sultr1;2 biogenesis but that they have some effect on transport function.
Lesions That Do Not Affect Biogenesis or Function-Although Class III mutations were identified to find residues of the STAS domain that were not essential residues for Sultr1;2 activity, most mutations in this class decreased protein accumulation and/or function to some extent, except for Q521L and I617M. Sultr1;2 with a Q521L or I617M substitution rescued the CP154-7B growth (Fig. 2) and SO 4 2Ϫ transport ( Fig. 5B and Table 2) phenotypes to the same extent as WTЈ. Gln-521 is at the most C-terminal residue of the L region, Ile-617 is in ␣3, and neither of them is highly conserved among family members (Fig. 6, mutations surrounded by black boxes). These results suggest that there is no absolute requirement for Gln and Ile at these positions. Other mutations identified within ␣3 include P614S, L615M, and V616E (Fig. 6). Because the alleles with these mutations all contain other lesions, definitive conclusions on the importance of ␣3 are difficult to draw at this time.