Organization and Function of the YsiA Regulon of Bacillus subtilis Involved in Fatty Acid Degradation*

The organization and function of the Bacillus subtilis YsiA regulon involved in fatty acid degradation were investigated. Northern and primer extension analyses indicated that this regulon comprises five operons, i.e. lcfA-ysiA-B-etfB-A, ykuF-G, yhfL, yusM-L-K-J, and ywjF-acdA-rpoE. YusJ and AcdA, YsiB and YusL, and YusK presumably encode acyl-CoA dehydrogenases, 3-hydroxyl-CoA dehydrogenase/enoyl-CoA hydratase complexes, and acetyl-CoA C-acyltransferase, respectively, which are directly involved in the fatty acid β-oxidation cycle. In addition, LcfA and YhfL are likely to encode long chain acyl-CoA ligases. On gel retardation and footprinting analyses involving the purified YsiA protein, we identified cis-sequences for YsiA binding (YsiA boxes) in the promoter regions upstream of ysiA, ykuF, yusL, yhfL, and ywjF, the equilibrium dissociation constants (Kd) for YsiA binding being 20, 21, 37, 43, and 65 nm, respectively. YsiA binding was specifically inhibited by long chain acyl-CoAs with 14–20 carbon atoms, acyl-CoAs with 18 carbon atoms being more effective; out of long chain acyl-CoAs tested, monounsaturated oleoyl-CoA, and branched chain 12-metyltetradecanoyl-CoA were most effective. These in vitro findings were supported by the in vivo observation that the knock-out of acyl-CoA dehydrogenation through yusJ, etfA, or etfB disruption resulted in YsiA inactivation, probably because of the accumulation of long chain acyl-CoAs in the cells. Furthermore, the disruption of yusL, yusK, yusJ, etfA, etfB, or ykuG affected the utilization of palmitic acid, a representative long chain fatty acid. Based on this work, ysiA, ysiB, ykuF, ykuG, yhfL, yusM, yusL, yusK, yusJ, and ywjF can be renamed fadR, fadB, fadH, fadG, lcfB, fadM, fadN, fadA, fadE, and fadF.

In all organisms fatty acids are essential components of membranes and are important sources of metabolic energy. Thus, fatty acid degradation and biosynthesis pathways must be switched on and off based on the availability of fatty acids. The regulation of these pathways has been mainly studied in a model prokaryote, Escherichia coli (1). E. coli produces straight chain and unsaturated fatty acids, and E. coli FabH selectively uses acetyl-CoA to initiate the fatty acid biosynthesis pathway (2,3). In contrast, Bacillus subtilis mainly produces branched chain fatty acids and possesses two FabH isoenzymes (named FabHA and FabHB) that differ from the E. coli enzyme in that they are selective for branched chain acyl-CoAs (4). The second FabF-FabB class of condensing enzymes is responsible for the subsequent rounds of fatty acid elongation in the pathway (5). In B. subtilis, the FabF protein is the sole condensing enzyme able to carry out the subsequent elongation reactions in fatty acid synthesis (6). Fatty acids that are intracellularly formed or extracellularly supplied are degraded through ␤-oxidation when the cells are starved of carbon sources.
In E. coli, the transcription factor FadR functions as a coordinate switch that negatively regulates the machinery required for fatty acid ␤-oxidative degradation (fad) and positively regulates the key enzymes (3-ketoacyl-acyl carrier protein dehydratase (FabA) and 3-ketoacyl-acyl carrier protein synthase I (FabB)) in unsaturated fatty acid biosynthesis (5,(7)(8)(9). The fad genes are fadL encoding a fatty acid transporter, and fadD, fadE, fadB, fadA, and fadH involved in fatty acid ␤-oxidation encoding fatty acid-CoA ligase, acyl-CoA dehydrogenase, 3-hydroxyacyl-CoA dehydrogenase/enoyl-CoA hydratase complex, acetyl-CoA C-acyltransferase, and 2,4-dienoyl-CoA reductase, respectively (8,10). Besides, FadR activates expression of the iclR gene, which encodes the negative regulator of the glyoxylate shunt enzymes, and represses the universal stress protein gene (uspA). The expression of the FadR regulon members is regulated in a coordinated manner in the presence of long chain acyl-CoAs, which antagonize FadR as to its operator binding (11).
In B. subtilis, a central mediator such as FadR that acts to balance the anabolic and catabolic fatty acid pathways has not been reported so far. Instead, a transcription regulatory protein for FapR was recently described (12) that controls the expression of many fap regulon members involved in fatty acid and phospholipid metabolism, including fapHA and fapHB, and fabF encoding the enzymes condensing branched chain acyl-CoAs with malonyl-acyl carrier protein and subsequent chain elongation, respectively. The FapR protein is controlled by the cellular pool of malonyl-CoA, which senses the status of fatty acid biosynthesis and adjusts the expression of the fap regulon members accordingly. B. subtilis possesses several paralogous genes that are most likely involved in the ␤-oxidation of fatty acids, such as lcfA, yhfL, yhfT, ysiB, yusLKJ, acdA, yhfS, and mmgABC (13-16) (see Fig. 1), yet mmgABC is a E -dependent * This work was supported by grants-in-aid for Scientific Research on Priority Areas, Scientific Research (B), and the High Tech Research Center Project for Private Universities from the Ministry of Education, Culture, Sports, Science, and Technology of Japan. 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. 1  operon (13). Nevertheless, a transcription factor such as FadR that functions as a coordinate switch that regulates fatty acid degradation negatively and fatty acid biosynthesis positively has not been described so far.
In the framework of the Japan Functional Analysis Network for B. subtilis (bacillus.genome.jp/), we have performed comprehensive DNA microarray analysis of hundreds of DNAbinding transcription regulatory proteins. DNA microarray analysis involving the wild type strain and a disruptant as to ysiA, encoding one of the helix-turn-helix transcription regulatory proteins, indicated that the YsiA protein negatively regulates more than 10 genes, the majority of which most likely participate in fatty acid ␤-oxidation, which presumably plays a central regulatory role in fatty acid degradation. In this communication, we describe the organization and function of the YsiA regulon comprising the lcfA-ysiA-B-etfB-A, ykuF-G, yhfL, yusM-L-K-J, and ywjF-acdA-rpoE operons. YsiA represses the transcription of the genes, except lcfA and yusM, belonging to these operons through its binding to their YsiA boxes. The disruption of etfA, eftB, ykuG, yusL, yusK, and yusJ affected cell growth on fatty acid as the sole carbon source. The in vivo as well as in vitro experiments suggested that long chain acyl-CoAs are most likely inducers (or ligands) that antagonize YsiA as to its binding to YsiA boxes.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Their Construction-The B. subtilis strains used in this work are listed in Table 1. Strain FU788, provided by K. Kobayashi (Nara Institute of Science and Technology, Japan), was constructed in the framework of the Japan Functional Analysis Network for B. subtilis (JAFAN; bacillus-.genome.jp) through the transformation of strain 168 with a PCR product prepared by means of long flanking homology PCR (19) (the first run of PCR with primer pairs of ysiA-F1/ ysiA-R1, ysiA-F2/ysiA-R2, and pUC-F(cat)/pUC-R(cat) ( Table  2), using strain 168 DNA and the cat gene (20) as templates, and the second PCR with ysiA-F1/ysiA-R2 (Table 2) and the above three PCR products as templates, respectively). Strains YKUFd and YHFLd were constructed as described previously (21), using primer pairs YKUFd-H/YKUFd-B, and YHFLd-H/YH-FLd-B (Table 2), respectively. DNA Microarray and Northern Analyses-DNA microarray analysis was performed similarly to as described previously (22)(23)(24). RNA samples were prepared from the cells of strains 168 (wild type) and FU788 (ysiA::cat) that had been grown in LB medium (25) and harvested in the middle of the logarithmic growth at A 600 ϭ 0.5.
Primer Extension Analysis-Primer extension analysis was performed as described previously (26). Total RNA was extracted and purified from cell pellets as described above. Reverse transcription was initiated from the EysiA-R, EykuF-R, EyhfL-R, EyusL-R, and EywjF-R primers, corresponding to nucleotides ϩ122 to ϩ141, ϩ146 to ϩ165, ϩ102 to ϩ121, ϩ141 to ϩ160, and ϩ111 to ϩ131 (ϩ1 is the transcription initiation base for each of the YsiA-regulated promoters, which was identified in this work), respectively (Table 2); they had been labeled at their 5Ј end by use of a Megalabel kit (Takara Bio) and [␥-32 P]ATP (GE Healthcare). Templates for the dideoxy sequencing reactions for ladder preparation starting from the same end-labeled primer were prepared by PCR using the respective primer pairs (Table 2; EysiA-F/EysiA-R, EykuF-F/EykuF-R, EyhfL-F/EyhfL-R, EyusL-F/EyusL-R, and EywjF-F/ EywjF-R) and DNA from strain 168 as a template.
Production and Purification of YsiA-To produce the YsiA protein in E. coli, the ysiA region was amplified by PCR using a primer pair (YsiA-N/YsiA-H, Table 2) and DNA of strain 168 as a template. After digestion with NdeI and HindIII of the PCR product and vector pET-22b(ϩ) (Novagen), the resulting fragments were ligated and then used for the transformation of E. coli BL21 (DE3). Correct cloning of ysiA in plasmid pET-22b(ϩ)-YsiA was confirmed by nucleotide sequencing.
The YsiA protein was overexpressed in E. coli BL21 (DE3) bearing pET-22b(ϩ)-YsiA by the addition of isopropyl-␤-Dthiogalactopyranoside (IPTG) 2 to the medium at 1 mM. The cells were harvested, washed with 50 mM Tris-Cl buffer (pH 8.0) two times, and then suspended in 50 mM Tris-Cl buffer (pH 8.0) containing 1% (w/v) glycerol. The cells were disrupted by sonication, and a supernatant was obtained by centrifugation (10,000 ϫ g, 10 min). The YsiA protein was purified with an anion exchange column chromatography (TOYO PEARL (DEAE-650M) (TOSOH Corp., Japan)) to near homogeneity. The purified YsiA protein was subjected to gel filtration column chromatography on Sephacryl S-200 HR (GE Healthcare) using 50 mM Tris-Cl buffer (pH 8.0) containing 1% (w/v) glycerol and 100 mM NaCl to determine its molecular mass. Gel Retardation and DNase I Footprinting Experiments-Gel retardation and DNase I footprinting experiments were performed as described previously (27). For gel retardation analysis, the labeled probe DNAs carrying cis-sequences for YsiA binding (YsiA boxes) of the ysiA, yhfL, ykuF, yusL, and ywjF promoters were PCR products amplified from DNA of strain 168 using primer pairs (EysiA-F/EysiA-R, EyhfL-F/EyhfL-R, PykuF-F/PykuF-R, PyusL-F/PyusL-R, and EywjF-F/PywjF-R; Table 2) in the presence of [␣-32 P]dC⌻P (GE Healthcare), respectively. For DNase I footprinting, the probe DNAs were prepared by PCR amplification using the same respective primer pairs, either of the primers having been labeled at the 5Ј terminus with a Megalabel kit and [␥-32 P]ATP (GE Healthcare).
Chemical Synthesis of Branched Long Chain Acyl-CoAs-12-Metyltetradecanoyl-CoA and 13-metyltetradecanoyl-CoA were chemically synthesized from 12-metyltetradecanoic acid and 13-metyltetradecanoic acid (Larodan Fine Chemicals AB) by the method of Seubert (28) with minor modifications. The branched long chain fatty acids were reacted with oxalyl chloride to form acyl chlorides. The acyl chlorides were converted with CoA to acyl-CoAs. The amounts of the formed acyl-CoAs were determined by using gas chromatography.
CGATCTCAAGATCATCGTCGT Primer extension, gel retardation, and footpriting analyses Underlined sequences are the sites of restriction enzymes used for strain construction.

Fatty Acid Degradation Genes in B. subtilis
Cell Growth and ␤-Galactosidase (␤-Gal) Assay-Cells of pMUTIN (29) disruptants of the YsiA regulon members (Table  1), were grown at 30°C on tryptose blood agar base (Difco) plus 10 mM glucose containing erythromycin (0.3 g/ml) overnight. The cells were inoculated into 50 ml of LB medium with and without 1 mM IPTG and then incubated at 37°C. During growth, 1-ml aliquots of the culture were withdrawn, and the ␤-Gal activity in the cells was spectrophotometrically assayed, as described previously (21).
To examine the utilization of fatty acid by the pMUTIN disruptant as the sole carbon source, the cells pregrown on tryptose blood agar base plus glucose plates as described above were suspended in S6 medium (30) containing tryptophan, and then the cells (A 600 ϫ ml ϭ 0.5) were spread on S6 solid medium containing 50 g/ml tryptophan and 4.35 mM sodium palmitate. The cells grown on each plate at 37°C were suspended in 5 ml of S6 medium containing tryptophan to measure A 600 . The ␤-Gal activity in the cells was assayed as described above.

RESULTS
Candidate YsiA-regulated Genes Detected on DNA Microarray Analysis-The YsiA protein is a member of the TetR family of bacterial helix-turn-helix transcriptional regulatory proteins according to Pfam (31). Within the framework of the Japan Functional Analysis Network for B. subtilis, we performed DNA microarray analysis to find candidate YsiA-regulated genes through comparison of the transcriptomes from cells of the ysiA disruptant (strain FU788) and parental wild type strain 168 grown to the logarithmic growth phase in LB medium. The analysis revealed candidate YsiA-repressed genes of ykuF, G, yusM, L, K, J, ywjF, acdA, rpoE, yhfL, lcfA, ysiB, etfA, etfB, and ysiA itself. All of these genes except yhfL are clustered in the respective orders of ykuF and G; yusM, L, K, and J; ywjF, acdA, and rpoE; and lcfA, ysiA, B, etfB, and A genes. A BLASTP sequence similarity search (17) for these gene products revealed their probable functions to be as follows (Fig. 1). The YusL and YsiB, YusK, and YusJ and AcdA proteins exhibited high sequence similarities to the 3-hydroxyacyl-CoA dehydrogenase/enoyl-CoA hydratase complexes, acetyl-CoA C-acetyltransferases, and acyl-CoA dehydrogenases of many Grampositive low GC bacteria, respectively. The LcfA and YhfL proteins exhibited high similarity to long chain fatty acid-CoA ligases. Moreover, the EtfA and EtfB proteins exhibited high sequence similarities to the ␣and ␤-subunits of electron transfer flavoproteins, respectively. These enzymes are known to be involved in fatty acid ␤-oxidation. The products of the genes (LcfA, YhfL, AcdA, YusJ, YusL, YsiB, YusK, EtfA, and EtfB) could be well assigned in each of the enzymatic steps in the ␤-oxidation of fatty acids according to their probable functions ( Fig. 1). In addition, YkuF showed significant sequence similarity to the 2,4-dienoyl-CoA reductases of Gram-positive bacteria belonging to Bacillus and related genera. The YwjF protein, a 4Fe-4S ferredoxin protein, might possibly be involved in electron transfer associated with acyl-CoA dehydrogenation, and the YkuG protein possessing a peptidoglycan-binding domain might possibly participate in fatty acid transport as described below. However, we currently cannot explain how RpoE (RNA polymerase ␦-subunit) and YusM similar to proline dehydrogenases participate in fatty acid degradation. Operon Organization of YsiA-regulated Genes-The YsiA regulon appears to be involved in fatty acid degradation. To reveal the operon organization of YsiA-repressed genes, their transcripts were analyzed by Northern blotting using the respective DNA probes for these genes and total RNA from cells of strains FU788 (ysiA::cat) and 168 grown to the logarithmic growth phase in LB medium (Figs. 2 and 3). A 3.5-kb transcript was detected with either the ykuF or ykuG probe only in the lanes of RNA derived from the ysiA disruptant, indicating that the two genes constitute an  (17). The mmgABC genes in parentheses are supposed to encode acetyl-CoA C-acyltransferase, 3-hydroxyacyl-CoA dehydrogenase/enoyl-CoA hydratase complex, and acyl-CoA dehydrogenase, respectively (13). But these genes are transcribed by E -RNA polymerase (13). FEBRUARY 23, 2007 • VOLUME 282 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 5183
operon, whereas a 1.7-kb one was observed with the yhfL probe, indicating that it is monocistronically transcribed (Fig. 2). As expected, we found palindromic structures, which might act as -independent terminators, just downstream of each of ykuG and yhfL. 5.5-and 7.0-kb transcripts were commonly detected only in the lanes of RNA derived from the ysiA disruptant with the yusL, yusK, or yusJ probe, indicating that the yusL, yusK, and yusJ genes might constitute an operon. Another palindomic sequence was found just downstream of yusJ, so the 5.5-kb transcript covered the yusL, yusK and yusJ genes. The 7.0-kb transcript, detected with the yusL, K, and J probes, appeared to carry the yusM gene besides the yusL, yusK, and yusJ ones (Fig. 2). A 4.2-kb transcript was detected with the ywjF, acdA, and rpoE probes only in lanes of RNA derived from the ysiA disruptant, and a short 0.7-kb transcript was observed in the lanes of RNA derived from both strains FU788 (ysiA::cat) and 168 only with the rpoE probe (Fig. 2). Because an -independent terminatorlike sequence was found just downstream of rpoE, the 4.2-kb transcript likely covered the ywjF, acdA, and rpoE genes, whereas the 0.7-kb transcript covered only rpoE.
The lcfA, ysiA, B, etfB, and A genes are clustered in this order. The plasmid pMUTIN integrant in the ysiA gene (strain YSIAd (ysiA::pMUTIN)) ( Fig. 4) was constructed to examine ysiA expression. As seen in gene organization of the vicinity of the pMUTIN-disrupted ysiA (Fig. 4), plasmid pMUTIN carrying the 5Ј-side inside portion of ysiA was integrated into it through single cross-over event; the integration not only disrupts the ysiA gene but also replaces it with lacZ as reporter of ysiA expression (29). Strain YSIAd constitutively synthesized ␤-Gal (ϳ100 nmol/min/A 600 ) during cell growth in LB medium, indicating that the ysiA gene is autorepressed. Hence, we performed Northern blotting using 2-fold more RNA of strains FU788 (ysiA::cat) and 168 with prolonged 32 P exposure of the imaging plates (Fig. 3). So 5.1-, 3.2-, and 0.7-kb transcripts were only detected in the lane of the wild type RNA with the ysiA probe, whereas the 5.1-and 3.2-kb ones were only observed in the lane of the wild type RNA with the ysiB probe, the 0.7-kb   Fig. 2 were subjected to Northern analysis using the lcfA, ysiA, and ysiB probes (indicated beneath the panels). The analysis was performed as described in the text. Each lane contained 20 g of total RNA, two times more than that used in Northern blottings (Fig. 2). The imaging plates were exposed for five times longer than those in the analysis (Fig. 2) transcript likely coding only ysiA. The 5.1-and 1.8-kb transcripts were only detected in the lane of the wild type RNA with the lcfA probe, the 1.8-kb one appearing to carry only lcfA (Fig.  3). A probable -independent terminator was found immediately downstream of etfA, so the 3.2-kb transcript probably covers the ysiA, ysiB, etfB, and etfA genes, and the 5.1-kb transcript likely carries the lcfA gene immediate upstream of ysiA besides these four genes. The overall results of Northern blotting indicated that the YsiA regulon comprises five operons (lcfA-ysiA-ysiB-etfB-etfA, ykuF-ykuG, yhfL, yusM-yusL-yusK-yusJ, and ywjF-acdA-rpoE).
To locate the transcription initiation sites for the 3.2-kb ysiA-B-etfB-A, 3.5-kb ykuF-G, 1.7-kb yhfL, 5.5-kb yusL-K-J, and 4.2-kb ywjF-acdA-rpoE transcripts, the synthesis of which was highly repressed by YsiA, we performed primer extension analysis using the same RNA samples as those used for Northern analysis (Figs. 2 and 3). As shown in Fig. 5, a specific band of run-off cDNA from each of the primers for the transcripts was strongly repressed by YsiA, so it was only detected with the total RNA from the ysiA disruptant; the RNA fragment corresponding to the 5Ј-portion of the 3.2-kb ysiAB-etfBA and 0.7-kb ysiA transcripts was supposed to be abundant in this ysiA disruptant (strain YSIAd). This identification of the transcription initiation bases upstream of the ysiA, ykuF, yhfL, yusL, and ywjF genes allowed us to locate the YsiA-repressive promoters presumably recognized by A (Fig. 5). (This identification of the initiation base of the yusL-K-J transcription shifted the initiation codon of YusL 78 bp downstream compared with the previous assignment (14).) Then we electronically searched for candidate palindromic sequences in the regions upstream of the ysiA, yhfL, ykuF, yusL, and ywjF genes, which might be YsiA boxes. This search revealed palindromic sequences upstream of each gene, the sequences of which are very similar to each other (Fig. 6). It was notable that the putative YsiA boxes for repres-  (29) carrying the 5Ј-side inside portion of each target gene into them through single cross-over event to result in strains LCFAd, BFS2426, BFS2427, ETFAd, ETFBd, BFS1341, BFS1347, BFS1346, BFS1345, BFS1246, ACDAd, YKUFd, BFS1835, and YHFLd, respectively. Gene organization of the vicinity of the disrupted genes (ysiA::pMUTIN, yusL::pMUTIN, ywjF::pMUTIN, ykuF::pMUTIN, and yhfL::pMUTIN) are shown as representatives. The point of pMUTIN integration into each of the other genes is indicated by an arrow. Plasmid pMUTIN integration not only disrupts the target gene and replaces it with lacZ as a reporter of its expression but also places its downstream genes under the control of the spac promoter, which is repressed by E. coli LacI antagonized with IPTG. Amp r and Erm r indicate ampicillin resistance and erythromycin resistance genes, which are used as markers in cloning of the 5Ј-side inside portion of the target gene into pMUTIN in E. coli and in integration of the resulting plasmid into a target gene in B. subtilis, respectively. FEBRUARY 23, 2007 • VOLUME 282 • NUMBER 8 sion of the yisA-B-etfB-A, ykuF-G, yhfL, and yusL-K-J transcriptions were located just downstream of their transcription initiation bases, whereas the box for ywjF-acdA-rpoE one overlaps the "Ϫ35" region of its promoter (Fig. 6).

Fatty Acid Degradation Genes in B. subtilis
Identification of YsiA Boxes by in Vitro Analyses-To perform gel retardation analysis to determine whether the putative YsiA boxes function, the YsiA protein was produced in E. coli and purified by an anion exchange column chromatography to near homogeneity (Fig. 7). The purified YsiA protein exhibited a molecular mass of 40.2 kDa on gel filtration (data not shown), indicating that it is a dimer of two subunits (M r ϭ 22.0 kDa). On gel retardation analysis involving the purified YsiA protein (Fig.  8A), the ysiA fragment spanning bases Ϫ104 to ϩ143 containing a putative YsiA box for ysiA (YsiA box ysiA ) became increasingly retarded, forming a band corresponding to a protein-DNA complex, as the amount of YsiA protein in the assay mixture increased. Similarly, the ykuF, yhfL, yusL, and ywjF fragments comprising bases Ϫ46 to ϩ123, Ϫ65 to ϩ121, Ϫ39 to ϩ132, and Ϫ98 to ϩ73 with each of putative YsiA boxes became increasingly retarded, forming the bands for the protein-DNA complexed, as the amount of YsiA increased (Fig. 8A). DNase I footprinting analysis involving the purified YsiA (Fig. 9) verified that YsiA protein actually protected the regions containing the respective probable YsiA boxes against DNase I digestion, indicating that YsiA is able to bind to these YsiA boxes. To find more YsiA boxes, if any, we subjected the sequences of the five YsiA boxes identified to a webbased GRASP-DNA search (32), these sequences being scored as the best five among the 100 listed. However, there were no other candidate YsiA boxes than five in the regions of lcfA-ysiAB-etfBA, ykuFG, yusMLKJ, ywjF-acdA-rpoE, and yhfL.
The data obtained in gel retardation experiments (Fig. 8A) were used to calculate the equilibrium dissociation constants (K d ) that corresponded to the YsiA concentrations giving equal amounts of bound and unbound fractions for each fragment (Fig. 8B). The YsiA boxes, containing a consensus sequence (WTGAATGAMTANTCATTC-AN, where W, M, and N stand for A or T, A or C, and any base, respectively), can be classified into three groups according to the K d values for their affinity to YsiA; 20 and 21 nM for YsiA box ysiA and YsiA box ykuF , 37 and 43 nM for YsiA box yusL and YsiA box yhfL , and 65 nM for YsiA box ywjF , with was well correlated with their palindrome matching (Fig. 8B).

FIGURE 5. Primer extension analysis to map the 5 ends of the ysiA-B-etfB-A, ykuF-G, yusL-K-J, ywjF-acdA-rpoE transcripts.
Total RNA samples (45 g) from strains 168 (wild type) and BFS2426 (ysiA::pMUTIN) grown in LB medium to A 600 ϭ 0.5 were annealed with the respective primers (EysiA-R, EyhfL-R, EykuF-R, EyusL-R, and EywjF-R; Table 2) for the ysiA-B-etfB-A, yhfL, ykuF-G, yusL-K-J, and ywjF-acdA-rpoE transcripts, and then primer extension was performed as described in the text. Lanes 1 and 2 contained the run-off cDNAs from strains 168 and BFS2426, respectively. Lanes G, A, T, and C contained the products of the respective dideoxy sequencing reactions, performed as described in the text. The arrowheads indicate the run-off cDNAs resulting from primer extension analyses. For each analysis, part of the nucleotide sequence of the coding strand corresponding to the ladder is shown together with the transcription initiation base (bold type, ϩ1), and the deduced Ϫ10 and Ϫ35 regions (underlined). The current identification of the initiation base of the yusL-K-J transcription let us to shift the initiation codon (ATG) of the YusL protein to 78 bp downstream of that assigned previously (14). The results indicated that stearoyl-and oleoyl-CoAs with 18 carbon atoms inhibited this interaction most effectively and that the introduction of one double bond into stearoyl-CoA yielding oleoyl-CoA enhanced this inhibition significantly. Methylation of myristoyl-CoA (tetradecanoyl-CoA) at the 12th carbon also elevated it, 12-metyltetradecanoyl-CoA being one of the best inhibitors. Saturated acyl-CoAs with 3-8 carbon atoms including isovaleryl-CoA did not inhibit this interaction with YsiA box ywjF . Moreover, neither coenzyme A nor fatty acids inhibited the YsiA interaction with YsiA box ywjF (Fig. 10). Thus, long chain acyl-CoAs with 14 -20 carbon atoms, including oleoyl-and palmitoleoyl-CoAs and 12-metyletradecanoyl-and 13-metyltetradecanoyl-CoAs, are most likely the inducers of the YsiA regulon, interacting with the YsiA protein to prevent the binding to the YsiA boxes.
lacZ Expression under the Control of Each of the YsiA-regulated Promoters in pMUTIN Disruptants-The in vitro experiments suggested that YsiA was a repressor of the YsiA regulon members that interacted with their YsiA boxes to block their transcription and that this interaction was inhibited by long chain acyl-CoAs. Most of the gene products, the synthesis of which is highly repressed by YsiA, are enzymes involved in fatty acid ␤-oxidation.   (Fig. 5), a Shine-Dalgarno sequence for YusL was newly assigned. FEBRUARY 23, 2007 • VOLUME 282 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 5187
To determine whether the disruption of each of them affects their promoter activity, that is, YsiA repression is suppressed by it, we lined up pMutin disruptants of the genes regulated by YsiA (Fig. 4). As seen in gene organization of the vicinity of the pMUTIN-disrupted ysiA, yusL, ywjF, ykuF, and yhfL genes as representatives (Fig. 4), plasmid pMUTIN integration not only disrupts the target gene and replaces it with lacZ as reporter of its expression but also places its downstream genes under the control of the spac promoter which is repressed by E. coli LacI (29). Thus, the addition of IPTG to the medium induces the genes downstream of the disrupted gene. ␤-Gal synthesis in a series of pMUTIN disruptants was monitored during the logarithmic and stationary growth phases in the absence and presence of IPTG in the LB medium. Fig. 11A shows the monitoring of ␤-Gal synthesis in the pMUTIN disruptants of six genes (ysiB, etfB, etfA, yusL, yusK, and yusJ), which are supposed to be directly involved in fatty acid ␤-oxidation (Fig. 1). As shown in Fig.  11A, ␤-Gal synthesis in these disruptants started to drastically increase at the beginning of the stationary phase in the absence of IPTG. Even if IPTG was added to the medium, which induces the genes downstream of the disrupted genes, ␤-Gal synthesis in the disruptants of the yusJ and etfA, B genes, encoding acyl-CoA dehydrogenase, and the ␣and ␤-subunits of an electron transfer flavoprotein, respectively (Fig. 1), still greatly increased (Fig. 11A), suggesting that these disruptions might cause the accumulation of long chain acyl-CoAs, inducers for the YsiA regulon, in vivo (Fig. 10). We did not detect significant induction in the disruptants as to the yhfL, yukF, yukG, yusM, ywjF, and acdA genes but detected some increase in that of lcfA (Fig. 11B), although transcription of yusM and lcfA was unlikely to be directly regulated by YsiA because of a lack of any YsiA box sequences in their promoter regions (Fig. 6). These in vivo results support the above in vitro finding that long chain acyl-CoAs are inducers for the YsiA regulon.
Growth Defect with a Fatty Acid as the Sole Carbon Source Caused by Disruption of Several Members of the YsiA Regulon-E. coli fad genes regulated by FadR are required for cell growth on long chain fatty acids as carbon sources (34). However, no B. subtilis mutant unable to grow on long chain fatty acids as the sole carbon source has been reported, as far as we know. Thus, we examined whether disruptants of the YsiA regulon members grew on palmitic acid as the sole  The gel retardation analysis was performed as described in the legend to Fig. 8. The reaction mixtures (25 l) contained 0.02 pmol of the DNA probe containing YsiA box ywjF , which was the same probe as that used in the gel retardation analysis (Fig. 8), 220 nM YsiA protein (as dimer), 5 or 50 M acyl-CoA or fatty acid (A and B), or 100 M fatty acid (C), and 4% Me 2 SO, the concentrations given being final; acyl-CoAs and fatty acids were dissolved in Me 2 SO. The acyl-CoAs exhibiting a significant inhibitory effect on the formation of the DNA probe-YsiA complex are underlined. The K i values of long chain acyl-CoAs are shown in the upper part of A and were obtained in the other gel retardation experiments (data not shown). FEBRUARY 23, 2007 • VOLUME 282 • NUMBER 8 carbon source; we selected this fatty acid arbitrarily, because

DISCUSSION
DNA microarray analysis revealed 15 candidate target genes of YsiA, one of the TetR family of bacterial DNA-binding regulatory proteins. A BLASTP similarity search (17) of these products to the nonredundant protein sequence data base revealed that most of the target gene candidates exhibited high similarities to enzymes involved in fatty acid ␤-oxidation and closely related reactions, suggesting that YsiA might be a central regulator in fatty acid degradation (Fig. 1). Northern blotting indicated that these 15 genes are organized into five operons (lcfA-ysiA-B-etfB-A, ykuF-G, yhfL, yusM-L-K-J, and ywjF-acdA-rpoE) (Figs. 2 and 3). Among many transcripts resulting from these operons, the synthesis of the ysiA-B-etfB-A (5.1 kb), ykuF-G (3.5 kb), yhfL (1.7 kb), yusL-K-J (5.5 kb), and ywjF-acdA-rpoE (4.2 kb) transcripts was highly repressed by YsiA. Primer extension analysis revealed transcription initiation bases immediate upstream of ysiA, ykuF, yhfL, yusL, and ywjF that are preceded by Ϫ35 and Ϫ10 regions most likely recognized by A -associated RNA polymerase (Fig. 5). We found very similar palindromic sequences very close to the transcription initiation sites (Fig. 6), which were thought to be YsiA boxes.
To determine the regulatory function of the YsiA protein in vitro, it was produced in E. coli and purified to near homogeneity (Fig. 7), it being found to be a dimerized protein (M r ϭ 40.2 kDa). Gel retardation and footprinting analyses involving the purified YsiA protein revealed that the palindromic sequences associated with the ysiA, ykuF, yhfL, yusL, and ywjF promoters were YsiA boxes (Figs. 8 and 9). As shown in Fig. 8B, a consensus sequence deduced from the YsiA boxes was WTGAAT-GAMTANTCATTCAN, where W, M, and N stand for A or T, A or C, and any base, respectively. These sequences can be classified into three groups (ysiA and ykuF; yusL and yhfL; and ywjF) according to the K d values for their affinity to YsiA and to palindrome matching (Fig. 8B). As seen in comparison between the sequences of YsiA box ysiA and YsiA box ykuF , only A, A, and T bases at positions 5, 9, and 10 are conserved except bases common in all five of the YsiA boxes (Fig. 8B). Also, A, C, and T bases at the same positions are observed in the sequences of YsiA box yusL and YsiA box yhfL , although such conservation is not seen in that of YsiA box ywjF . Therefore, these base substitutions might be determinants as to the alteration of the K d values.
A web-based GRASP-DNA search (32) using the sequences of the five YsiA boxes indicated that these sequences were scored as the best five, and there were no more candidate YsiA boxes in the regions of lcfA-ysiA-B-etfB-A, ykuF-G, yusM-L-K-J, ywjF-acdA-rpoE, and yhfL. However, the synthesis of the 5.1-kb lcfA-ysiA-B-etfB-A and 7.0-kb yusM-L-K-J transcripts was under negative control of YsiA without any YsiA box-like sequence in the promoter regions upstream of lcfA and yusM. To explain this indirect transcriptional control through YsiA, we have to assume another transcription regulator whose synthesis and (or) operation is regulated by YsiA, but at present, we have no experimental evidence suggesting what this regulator is.
B. subtilis YsiA is supposed to be a functional homolog of E. coli FadR, a global transcriptional regulator of fatty acid metabolism, the inducers of which are long chain acyl-CoAs (11). As for FadR, long chain acyl-CoAs with 14 -20 carbon atoms were considered to be inducers of the YsiA regulon, interacting with the YsiA protein to prevent the binding with the YsiA boxes (Fig. 10). The in vivo experiments involving pMUTIN disruptants of the target genes of YsiA (Fig. 11) revealed that the knock-out of acyl-CoA dehydrogenation through disruption of yusJ, etfB, or etfA presumably caused the accumulation of branched and straight long chain acyl-CoAs, which derepressed YsiA-regulated genes. In addition, the disruption of ysiB, yusJ, L, and K directly involved in fatty acid ␤-oxidation caused the ␤-Gal induction probably because of the accumulation of long chain acyl-CoAs, when their disruptants were incubated with palmitic acid (Fig. 12); only the ysiB disruptant among them grew on palmitic acid, a representative long chain fatty acid, as the sole carbon source. Fig. 12 also shows that the promoters directly repressed by YsiA are considerably derepressed during cell growth on palmitic acid, as observed in the disruptants of the yhfL, ykuF, ywjF, and acdA, which suggests that moderate inactivation of YsiA by long chain acyl-CoAs likely occurs in the cells growing on palmitic acid. These in vivo results support the in vitro finding that long chain acyl-CoAs are most likely the inducers of the YsiA regulon.
The lacZ fusion experiments also indicated that YsiA is considered to be able to repress its regulon members during the logarithmic and stationary growth phases in the LB medium, because ␤-Gal synthesis in the pMUTIN integrants as to the YsiA-regulated genes, except the yusJ, yusL, yusK, ysiB, etfA, and B genes involved in fatty acid ␤-oxidation, could not be induced during these growth phases (Fig. 11B). Moreover, the disruption of the ysiB, yusL, or yusK genes presumably might be suppressed by other paralogs even in the stationary growth phase in LB medium, as deduced from the result indicating that it did not induce ␤-Gal synthesis any more if IPTG was added to the medium.
The disruption of the etfA, B, yusL, K, and J genes, participating in the fatty acid ␤-oxidation cycle (Fig. 1), clearly affected the utilization of palmitic acid (Fig. 12). However, the growth defect of the disruptant of yusJ likely encoding acyl-CoA dehydrogenation was significantly more tolerant than were those of the disruption of yusL and yusK, which are likely to encode 3-hydroxyacyl-CoA dehydrogenase/enoyl-CoA hydratase complex and acetyl-CoA C-acyltransferase, respectively. These results implied that acyl-CoA dehydrogenation catalyzed by YusJ might be suppressed by its paralog(s), possibly AcdA, on growth with palmitic acid in minimal medium plates, which does not coincide with the suggestion that this dehydrogenation might be catalyzed only by YusJ in cells that have reached the stationary phase in LB medium (Fig. 11). On the contrary, the paralogs of YusL and YusK, including YsiB, might not support the growth on palmitic acid, which is different from the situation for the cells in the stationary growth phase in LB medium. Moreover, the knock-out of LcfA or YhfL, both being presumable long chain fatty acid-CoA ligases, did not affect palmitic acid utilization, suggesting that they compensate for each other or another paralog functions. On the other hand, it is most interesting that the deficiency of YkuG, presumably a cell wall protein with a peptidoglycan-binding domain, affected the utilization of palmitic acid. This implies that this protein might be involved in the transport of long chain fatty acids, which is worth investigating. Although the YwjF protein, a 4Fe-4S ferredoxin protein, was considered to be possibly associated with acyl-CoA hydrogenation, we obtained no evidence to support this prediction in the in vivo experiments involving its disruptant (Figs. 11 and 12). Furthermore, we have no idea at present how YusM and RpoE participate in fatty acid metabolism, because neither yusM knock-out nor rpoE expression freed from YsiA-regulation through pMUTIN integration into the genes upstream of rpoE affected palmitic acid utilization in the presence (Fig. 12) or absence (data not shown) of IPTG. Nevertheless, it is notable that a rpoE mutant strain has an altered morphology and shows a delay in exiting from the stationary phase (35).
In E. coli, the FadR protein is a transcription factor that plays a central role in the regulation of fatty acid metabolism, through its functioning as a switch that regulates the machinery required for fatty acid ␤-oxidation and the expression of a key enzyme in fatty acid biosynthesis coordinately (7,8). Fatty acid metabolism in B. subtilis is likely controlled by FapR (12) (controlling fatty acid biosynthesis and phospholipids) and YsiA (controlling fatty acid degradation). FapR presumably senses the in vivo concentration of malonyl-CoA, an increase of which inactivates this protein (12), whereas YsiA is most likely inactivated by an increase in the in vivo concentration of long chain acyl-CoAs (Figs. 10 and 11), as in the case of E. coli FadR. Interestingly, E. coli FadR is roughly 10 times more sensitive to long chain acyl-CoAs than B. subtilis YsiA (Ref. 11 and Fig. 10), which might reflect the dual regulatory roles of FadR in fatty FIGURE 13. High conservation among the YsiA sequences of various Gram-positive low GC bacteria. Alignment of the YsiA sequences of various Grampositive low GC bacteria is shown. A helix-turn-helix motif comprising 22 amino acids was predicted in the N-terminal region of B. subtilis YsiA with NPS@ (network protein sequence analysis) (36), which is well conserved among the aligned YsiA proteins. G., Genobacillus; O., Oceanobacillus. The eight ␣-helices were identified on x-ray crystallographic analysis of the YsiA protein (37), as indicated by boxed amino acid sequences. Helices 7 and 8 are supposed to be involved in the dimerization of YsiA subunits. The YsiA protein produced in E. coli happened to contain one lauroyl-CoA molecule per dimer, which allowed identification of the amino acids interacting with it, as indicated by amino acid numbers, where A and B stand for amino acids from the two subunits of the YsiA dimer.
acid synthesis and degradation, possibly requiring finer tuning of this coordinate regulation.
The BLASTP similarity search (17) also revealed that YsiA is highly conserved in many Gram-positive organisms including Bacillus, Clostridium, Streptomyces, and other related genera; alignment of the YsiA sequences of Bacillus and closely related genera is shown in Fig. 13. Moreover, its orthologs were unexpectedly found in far-diverse genera, such as Metanosarcina (Archaea), and Bordetella, Burkholderia, and Chromobacteria (betaproteobacteria). It is noteworthy that FapR, a global transcription factor for membrane lipid biosynthesis, is conserved in the Bacillus, Listeria, and Staphylococcus genera and also in Clostridium and other related genera (12), but YsiA is not conserved in the Listeria and Staphylococcus genera. This diverse but unique distribution of YsiA orthologs implies that horizontal gene transfer(s) might have occurred during its molecular evolution.
In the N-terminal region of B. subtilis YsiA (Fig. 13), a helixturn-helix motif covering amino acids 27-48 was predicted with NPS@ (network protein sequence analysis) (36). In the framework of a bacterial structural genomics project conducted at Structural GenomiX Inc. (37), the crystallographic structure of B. subtilis YsiA, one of the unique proteins whose function is unknown, was very recently determined and deposited under Protein Data Base identification code 1VI0 (38). As shown in Fig. 13, eight ␣-helices were identified; helices 7 and 8 are involved in establishment of the YsiA dimer; we also found in this study that YsiA is a dimer of two identical subunits. The YsiA protein was unexpectedly found to contain one molecule of lauroyl-CoA (38), leading to identification of the amino acids forming hydrogen bonds with H and O atoms of lauroyl-CoA. However, lauroyl-CoA hardly interfered with the complex formation between YsiA and YsiA box ywjF (Fig. 10), so that it was unlikely to cause its conformational change. As shown in Fig. 13, the amino acids of YsiA interacting with lauroyl-CoA are unexpectedly located in the C-terminal regions of both subunits, that is, the atoms of lauroyl-CoA interacted with amino acids derived from both subunits. In contrast, the atoms of acyl-CoA embedded in E. coli FadR interact exclusively with those of one subunit of the dimerized protein (39,40). As in the case of E. coli FadR protein, the adenosine 3Ј-phosphate and pyrophosphate moieties of acyl-CoA embedded in YsiA are also exposed to the solvent. From these points of view, crystallographic analysis of the unique interaction of YsiA with either acyl-CoA or YsiA box is a quite interesting subject worth investigating.