thiBPQ Encodes an ABC Transporter Required for Transport of Thiamine and Thiamine Pyrophosphate inSalmonella typhimurium *

In Salmonella typhimurium, thiamine pyrophosphate (TPP) is a required cofactor for several enzymes in central metabolism. Herein we identify a new thi operon,thiBPQ (designated sfuABC in Escherichia coli), required for the transport of thiamine and TPP into the cell. Insertions in the operon result in strains that are phenotypically and biochemically defective in thiamine and TPP transport. Data presented herein show that this operon is transcriptionally repressed in the presence of exogenous thiamine, with TPP the likely regulatory molecule. This work represents the first identification of thiamine transport genes in bacteria and demonstrates the function of a proposed ABC transporter in E. coli.

Thiamine pyrophosphate (TPP) 1 is a required cofactor synthesized de novo in Salmonella typhimurium. The primary role for TPP is in central metabolism as an electron carrier and nucleophile for such enzymes as pyruvate dehydrogenase (EC 1.2.4.1), acetolactate synthase (EC 4.1.3.18), and ␣-ketoglutarate dehydrogenase (EC 1.2.4.2). Despite its importance in cellular physiology, neither the de novo biosynthetic pathway nor the salvage systems for thiamine are fully understood in any organism.
Thiamine monophosphate (TMP) is generated by the condensation of two independently synthesized moieties: 4-amino-5hydroxy-methyl pyrimidine pyrophosphate (HMP-PP) and 4-methyl-5-(␤-hydroxyethyl) thiazole phosphate (THZ-P). TMP is then phosphorylated by the action of thiamine monophosphate kinase, ThiL (1), to form the physiologically relevant form of the vitamin, TPP. Recently several studies in the enteric bacteria S. typhimurium and Escherichia coli have elucidated many steps in the formation of thiamine (2)(3)(4)(5), but much of the pathway remains unknown.
Mutants defective in various steps in de novo synthesis can be supplemented exogenously with THZ, HMP, thiamine, TMP, or TPP. These results suggested that S. typhimurium had the ability to take up and incorporate these compounds into the de novo thiamine biosynthetic pathway. It was demonstrated several years ago that thiamine was actively transported in E. coli, and this transport was shown to involve a thiamine-binding protein whose activity was repressed by excess thiamine (6 -9).
The transport of TPP was not addressed in these previous studies. The presence of the thiamine-binding protein led to the hypothesis that thiamine was transported via a periplasmic binding protein-dependent ABC-type transporter (10).
We report here the identification of an operon (thiBPQ) at centisome (Cs) 1.5 on the S. typhimurium and E. coli chromosomes involved in the specific translocation of thiamine and its phosphoesters across the inner membrane. Analysis of the E. coli sequence (designated sfuABC) in addition to phenotypic analysis in S. typhimurium suggested that thiBPQ encoded thiamine binding protein, inner membrane channel, and energytransducing ATPase, respectively. Transcriptional fusions in this operon were regulated in response to exogenous thiamine.

Bacterial Strains
All strains used in this study are derivatives of S. typhimurium LT2 and are listed in Table I. MudJ is used throughout the paper to refer to the MudI 1734 transposon, which has been described (11), and Tn10d(Tc) refers to the transposition defective mini-Tn10-(Tn10⌬16⌬17) (12).

Culture Media and Biochemicals
No-carbon source E media (NCE) supplemented with 1 mM MgSO 4 and 11 mM glucose was used as minimal media (13,14). Difco nutrient broth (NB, 8 g/liter) with NaCl (5 g/liter) added was used as rich medium. Difco BiTek agar (15 g/liter) was added for solid medium. Antibiotics were added as needed to the following concentrations in rich and minimal media respectively: kanamycin (50, 125 g/ml), tetracycline (20, 10 g/ml), and chloramphenicol (40, 4 g/ml). Radiolabeled thiamine (C2-14 C-THZ-thiamine) with a specific activity of 24 mCi/ mmol was purchased from Amersham Pharmacia Biotech (Arlington Heights, IL). All other chemicals were purchased from Sigma.

Genetic Methods
Transduction Methods-All transductional crosses were performed by using the high frequency transducing bacteriophage P22 mutant HT 105/1 int-201 (15) as described (16). Transductants were purified and identified as phage-free by cross-streaking on green plates (17).
Mutant Isolation-Strains defective for TPP transport were isolated by insertional mutagenesis with one of two transposons, Tn10d(Tc) or MudJ. To facilitate mutant isolation, a pool of cells containing Ͼ80,000 independent insertions was generated as described elsewhere (18,19). A P22 lysate was grown on these cells to generate either a MudJ or Tn10d(Tc) phage pool.
To isolate Tn10d(Tc) insertion mutants, the Tn10d(Tc) phage pool described above was used to transduce a strain defective in de novo thiamine synthesis (either DM62 (thi-924::MudJ) or DM460 (thiH910::MudJ)) to tetracycline resistance (Tc r ) on NB-tetracycline plates. The Tc r transductants were screened for those that were able to grow with 1 M thiamine but not 1 M TPP. Putative insertion mutants defective in high affinity TPP transport (defining the thiP locus) were streaked for phage sensitivity and saved for further analysis.
Point mutations defective in high affinity TPP transport were isolated as described above with the following exception. The Tn10d(Tc) pool utilized had been mutagenized with hydroxylamine as described (13,20), resulting in the isolation of point mutations linked to a Tn10d(Tc) element.
After mapping, MudJ insertions in the thiP locus were identified * This work was supported by National Science Foundation Grant MCB9723830 (to D. M. D.) and by the Shaw Scientist Program of the Milwaukee Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Molecular Biology Techniques
DNA Sequencing-DNA was sequenced at the University of Wisconsin-Madison Biotechnology Center-Nucleic Acid and Protein Facility. DNA sequence analysis program BLAST (21) was used to compare this sequence with known sequences from the data base.
Chromosome Location of thiP Operon-The TPP transport-deficient mutations were mapped on the S. typhimurium chromosome via sequencing MudQ phage DNA from strain DM3468 (zac-8602::MudQ thi-995 ⌬) which had been generated from strain DM3403 (zac-8602::MudJ thi-995 ⌬) as described (22). The resulting locked-in P22 phage was induced, and DNA was isolated as described (23). The purified DNA was then used as a template for DNA cycle sequencing using a Sequitherm (Epicentre Madison, WI) kit. The primer used was MuR (5Ј-GAAACGCTTTCGCGTTTTTCGTGC-3Ј) which hybridizes to the right end of the MudQ insertion.
Mapping of Insertions by PCR-The location of four insertions in the thiP operon were determined via a PCR-based protocol (24). Amplification between the insertions was done using Vent (exo Ϫ ) polymerase (New England Biolabs, Inc., Beverly, MA) in a Thermolyne Temp-Tronic Thermocycler (Dubuque, IA). Reaction conditions were as follows: 95°C denaturation for 1 min, 55°C annealing for 1 min, and 72°C extension for 2 min. Primers used were: Tn10-I (5Ј-GACAAGATGTG-TATCCACCTTAAC-3Ј), which hybridizes to the 66-base pair inverted repeat Tn10 sequence; MuL (5Ј-ATCCCGAATAATCCAATGTCC-3Ј), which hybridizes to the left end of the MudJ insertion; and MuR (defined above). Additional MgSO 4 was added to all reaction mixes to a final concentration of 1 mM. Amplified products were visualized via agarose gel electrophoresis, purified using Qiaquick gel extraction kit (Qiagen, Chatworth, CA), and sequenced at the University of Wisconsin-Madison Biotechnology Center-Nucleic Acid and Protein Facility.

Growth Curves
Curves were done aerobically as described (16). Final concentrations of THZ, thiamine, TMP, and TPP were as indicated.

Generation of [␤-32 P]TPP
[ 32 P]TPP was generated using cell-free extracts of a strain overproducing ThiL as described (1) with the following exceptions. The ThiL reaction was initiated with the addition of 15 l of ATP (10 l of 100 mM ATP ϩ MgCl 2 and 5 l of [␥-32 P]ATP (specific activity 6000 Ci/mmol)).
Radiolabeled [␤-32 P]TPP was purified from the ThiL reaction mix via column chromatography as described by Matsuda and Cooper (27) with the following exceptions. Twenty fractions (3 ml each) from the 60 ml of 0.1 M citrate buffer (pH 3.5) were collected. The TPP elution profile was tested by bioautography with strain DM1683 (thiL933::Tn10d(Tc)), which qualitatively determined the fractions containing significant TPP. To determine the radiochemical purity and concentration of the TPP, high pressure liquid chromatography analysis was performed on an aliquot of this fraction, as described previously (1,28). The high pressure liquid chromatography fraction containing the TPP peak was collected and scintillation counted for 1 min in a Packard Instruments Model 4530 Scintillation Counter (Downers Grove, IL), demonstrating that the TPP accounted for ϳ80% of the label. The specific activity of the TPP was calculated to be 9400 Ci/mmol.

Uptake of Radiolabeled TPP and Thiamine
The protocol for the uptake assay used for thiamine and TPP was a combination of previously described methods (7,29) and is summarized below. Overnight cultures grown in NB were pelleted and resuspended in an equal volume of 0.85 M NaCl. 0.5 ml of resuspended cells were inoculated into 10 ml of minimal medium and incubated with shaking at 37°C until the optical density at 560 nm was ϳ0.4. Cultures were then pelleted, resuspended in 2 ml of minimal medium, separated into 1-ml aliquots, and stored on ice until needed. The cultures were equilibrated at 37°C for 10 min, and assays were initiated by the addition of radioactive substrate (final concentration of 230 nM for [ 32

Isolation of TPP Transport Defective Mutants-Forty-seven
independently isolated mutations causing a similar thi phenotype were identified, including 2 point mutations, 5 Tn10(d), and 40 MudJ insertions. Phage P22 co-transduction analysis genetically mapped all of the insertions to the same locus, designated thiP (Ͼ90% linked). Growth curve analysis, represented in Fig. 1, revealed two significant points. Double mutants defective in both de novo synthesis (thiH) and the thiP locus required 1000-fold more TPP (100 nM versus 100 M) for maximal growth than the single thiH mutant (Fig. 1, B and D), and yet these strains could reach optimal growth rates when supplied with Ͼ1 M exogenous thiamine (Fig. 1, A and C). Additionally, thiP mutations in a wild-type background had no observable growth defects in minimal medium.
Physical Mapping of Insertions-The location of the thiP locus on the S. typhimurium chromosome was determined by sequencing the flanking DNA of a MudQ insertion (zac-8602::MudQ) known to be linked to thiP. BLAST (21) computer data base analysis determined that the insertion was 3Ј to the end of araC at Cs 1.5 based on sequence similarity to the E. coli genome sequence. Co-transductional analysis with P22 determined that thiP was Ͼ95% linked to the leucine biosynthetic operon, confirming that thiP resided at Cs 1.5 in S. typhimurium and that the gene order was consistent with the predicted order from the E. coli chromosome.
Four insertions in the thiP locus were physically mapped with a PCR-based protocol. Primers designed to the ends of the insertions were used to amplify DNA in strains containing both a MudJ and Tn10(d) insertion in the thiP locus. Amplified products (ϳ400 base pairs) from strains DM3925 and DM3931 were purified and sequenced. Comparison with the E. coli data base using BLASTN determined that the two insertions in strains DM3925 were in the sfuA homologue (Tn10(d) at predicted amino acid residue 171 and MudJ at residue 262), while the insertions in strain DM3931 were in the sfuC homologue (MudJ at predicted amino acid residue 36 and Tn10(d) at residue 192).
The sfuABC genes had been identified by the E. coli genome sequencing project between the leucine biosynthetic genes and arabinose utilization genes. This cluster was designated sfu after Serratia ferric uptake, to reflect the significant sequence similarity to the Serratia marcescens ferric uptake operon (an ABC transporter) and other ABC transporters. Here we designate the S. typhimurium genes thiBPQ to more clearly reflect the involvement of this gene cluster in thiamine transport (Fig.  2). Comparison to known ABC-transporters suggested that thiBPQ encoded a thiamine binding protein, an inner membrane channel, and an energy-transducing ATPase, respectively.
Data base searches were performed to determine how widespread the ThiBPQ transport complex was in biology. Since ThiP and ThiQ, the channel and ATPase, respectively, shared

FIG. 2. Relative location of four insertions in the thiP locus.
Four insertions in the thiP locus were physically mapped via a PCR-based protocol. Sequence comparison to E. coli determined that the two insertions in strain DM3925 were in sfuA, whereas the insertions in strain DM3931 were in sfuC. S. typhimurium amino acid sequences are depicted, residues conserved with E. coli are in bold. We designate these genes thiBPQ to more clearly define involvement in thiamine salvage. As shown, thiBPQ maps to Cs 1.5 on the S.typhimurium and E. coli chromosomes in between the leucine biosynthetic genes and arabinose utilization. significant homology with all ABC transporters, ThiB was used as a marker. BLASTP (21) analysis with the E. coli proteins determined that all three components of the transport complex were only present in Hemophilus influenza. No other significant homologs for ThiB were found in any other organism listed in GenBank TM .
Motif searches were also performed on the complex components using the PROSITE data base (30). These analyses determined that ThiB and ThiQ contained a bacterial binding protein motif and one nucleotide binding site, respectively.
Mutations in thiBPQ Are Defective in TPP Transport-To test whether the thiBPQ operon encoded the predicted transporter, mutants defective in the operon were tested for their ability to transport radiolabeled thiamine and TPP (Fig. 3). These data clearly showed that the insertions in thiBPQ caused a defect in the transport of both thiamine and TPP. LT2 yielded rates of uptake of 2.9 Ϯ 0.07 pmol of TPP/min/A 560 nm and 6.3 Ϯ 0.34 pmol of thiamine/min/A 560 nm , whereas insertions in thiB or thiQ had rates that were Յ 0.
As stated earlier, mutants blocked in both de novo synthesis and thiBPQ were able to grow in the presence of 1 M thiamine but not TPP. This result suggested there was an additional mechanism for transport of thiamine that was independent of ThiBPQ. To confirm that thiamine accumulated in a thiB mutant under these conditions, uptake assays were performed with 460 nM, 4.6 M, and 23 M thiamine (data not shown). Slight rate increases in accumulation of thiamine could be seen (no transport at 460 nM to 2.03 Ϯ 0.32 pmol of thiamine/min/ A 560 nm at 23 M).
thiBPQ Comprise a TPP-regulated Operon-Although an operon structure was expected based on the E. coli sequence, a strain containing a MudJ transcriptional fusion in thiQ was used to confirm this assumption. Strains DM3926 (thiQ1054::MudJ) and DM3931 (thiQ1054::MudJ thiB1012::Tn10d(Tc)) produced ϳ30 and ϳ4 units of ␤-galactosidase activity, respectively, when grown in minimal medium, consistent with an operon structure for the thiBPQ gene cluster.
Data presented in Table II demonstrate that thiBPQ belonged to the growing number of TPP transcriptionally regulated genes in S. typhimurium (4,31). This experiment was complicated by the fact that the transcriptional fusions in the genes of interest were defective for the transport of thiamine and TPP. We were able to circumvent this problem since we had determined that thiamine could accumulate in a thiB mutant when provided at concentrations Ͼ1 M. As shown in Table II, ␤-galactosidase activities varying from 70 to 20 units were obtained in minimal medium using different insertions in the operon; however, in the presence of exogenous thiamine (1 mM), expression was reduced 3-4-fold (Strains DM3781 and DM3782). To address whether this response was to thiamine or TPP, the previously characterized thiL927 point mutation was utilized (1). Strains containing this point mutation and a thi reporter are completely repressed by TPP, but not thiamine due to altered thiamine monophosphate kinase, ThiL (EC 2.7.4.16), activity. Two insertions thiB1062::MudJ and thiQ1054::MudJ were transduced into a thiL927 point mutation background, generating strains DM3671 and 3670, respectively. As shown in Table II, strains containing the thiL927 point mutation were not repressed completely in response to thiamine, while the isogenic strain exhibited normal repression. These data are consistent with repression of the thiBPQ operon occurring in response to TPP, as are other previously described thiamine biosynthetic operons. DISCUSSION Work presented here identifies a new ABC transporter (thiBPQ) in S. typhimurium. We show here that the TPPregulated thiBPQ operon at Cs 1.5 is responsible not only for the transport of thiamine but also TPP in S. typhimurium. This agrees with previous work in E. coli which demonstrated that the thiamine transport was energy-dependent, independent of de novo biosynthesis, and repressed by thiamine (possibly TPP) (8,32). More recently, it was shown in E. coli that the ThiB protein has high affinity for the binding of thiamine, TMP, and FIG. 3. ThiBPQ mediates thiamine and TPP uptake. Uptake assays were performed using both radiolabeled 14 C-thiamine (A) and [ 32 P]TPP (B) as described under "Experimental Procedures." In panel A, thiamine uptake was tested in an LT2 (f) and DM3922 (‚) (thiB1062::MudJ) background. Data, including standard deviations for three independently performed thiamine uptake assays are shown. In panel B, TPP transport was tested in LT2 (f), DM3922 (‚), and DM3926 (q) (thiQ1054::MudJ). Representative data from more than three independently performed TPP uptake assays are shown. TPP, 2 suggesting that thiBPQ is also responsible for TMP transport. This three-gene operon is present in E. coli at the same chromosome location; therefore, we propose that the gene designations be changed from sfuABC to thiBPQ to more clearly define the role of this operon in thiamine metabolism. A recent publication (33) has identified the thiamine transport gene in Saccharomyces cerevisiae. Unlike the transport system for thiamine and TPP in E. coli and S. typhimurium (ABC tranporter), this gene is a member of the major facilitator superfamily (MFS) of transporters (34). The MFS class of transporters differs from ABC-type transporters in two significant ways: the translocation complex is encoded by a single gene, and the energy for translocation does not come from ATP hydrolysis but is carrier mediated. S. cerevisiae is unable to transport TPP (35). It is interesting that a different mechanism for thiamine transport has evolved and that these two systems differ not only in structure but in substrate recognition.
The fact that strains containing mutations in thiBPQ and a de novo block in thiamine biosynthesis were corrected with 100-fold less thiamine (1 M) than TPP (100 M) suggested that there was another mechanism for thiamine uptake in S. typhimurium. In fact we could show that, at concentrations Ͼ4.6 M, thiamine did accumulate in a thiB mutant. We propose that this accumulation was due to a low affinity thiamine transport system or nonspecific transport by another translocation complex.
The transcriptional regulation of this operon was addressed using MudJ insertions in thiB and thiQ. These analyses determined that this operon was repressed in response to excess, exogenously supplied thiamine (Ͼ1 mM). Additional experiments showed that TPP was the likely effector, as has been shown for other transcriptionally regulated thi genes in S. typhimurium. thiBPQ represents the third operon shown to be regulated in response to TPP in S. typhimurium. These genes map throughout the S. typhimurium chromosome: thiBPQ, thiMD, and thiCEFSGH at Cs 1.5, 46, and 90, respectively, and represent genes involved in salvage, biosynthesis, and transport.
Recently it has been shown that many genes involved in thiamine metabolism in bacteria have a highly conserved 39 base pair region called the thi-box 5Ј to the start of translation (36). Analysis of the thiBPQ sequence from E. coli determined that this operon also contained this element. The presence of the thi box in 5Ј to translation in all genes found to be regulated by TPP predicts the existence of a TPP-responsive regulatory protein.