Molecular and Physiological Logics of the Pyruvate-Induced Response of a Novel Transporter in Bacillus subtilis

ABSTRACT At the heart of central carbon metabolism, pyruvate is a pivotal metabolite in all living cells. Bacillus subtilis is able to excrete pyruvate as well as to use it as the sole carbon source. We herein reveal that ysbAB (renamed pftAB), the only operon specifically induced in pyruvate-grown B. subtilis cells, encodes a hetero-oligomeric membrane complex which operates as a facilitated transport system specific for pyruvate, thereby defining a novel class of transporter. We demonstrate that the LytST two-component system is responsible for the induction of pftAB in the presence of pyruvate by binding of the LytT response regulator to a palindromic region upstream of pftAB. We show that both glucose and malate, the preferred carbon sources for B. subtilis, trigger the binding of CcpA upstream of pftAB, which results in its catabolite repression. However, an additional CcpA-independent mechanism represses pftAB in the presence of malate. Screening a genome-wide transposon mutant library, we find that an active malic enzyme replenishing the pyruvate pool is required for this repression. We next reveal that the higher the influx of pyruvate, the stronger the CcpA-independent repression of pftAB, which suggests that intracellular pyruvate retroinhibits pftAB induction via LytST. Such a retroinhibition challenges the rational design of novel nature-inspired sensors and synthetic switches but undoubtedly offers new possibilities for the development of integrated sensor/controller circuitry. Overall, we provide evidence for a complete system of sensors, feed-forward and feedback controllers that play a major role in environmental growth of B. subtilis.

the columns, and identical results were obtained (not shown). These results indicated that PftA and PftB form a hetero-oligomeric membrane protein complex. The PftAB complex operates as a pyruvate transporter. We asked whether pftAB encodes a pyruvate uptake system. During growth of C. glutamicum on pyruvate, pyruvate is taken up by the monocarboxylate transporter MctC, and as a result, a mctC mutant is unable to grow on pyruvate as the sole carbon source (16). We therefore tested for functional complementation of the C. glutamicum mctC mutant using an IPTG-inducible pftAB expression plasmid and appropriate controls. Since the mctC mutant cannot grow on pyruvate, we precultured the strains in rich medium, washed the cells, and inoculated a minimal medium plus pyruvate at an optical density of 600  (Fig. 2B), and the growth rates were between 0.04 and 0.06 h Ϫ1 (Fig. 2C and Fig. S2), which indicated that cell growth was IPTG dependent. Altogether, these results indicated that pftAB encodes a pyruvate import system. The PftAB complex is a pyruvate-specific facilitated transporter. We next asked whether PftAB operates as an active or passive pyruvate transport system. A straightforward approach to experimentally address this issue is to evaluate whether PftAB transports pyruvate along or against the concentration gradient of pyruvate. The concentration of intracellular pyruvate in B. subtilis cells grown in minimal medium containing glucose is about 1 mM (26). We therefore determined whether PftAB is specifically able to export intracellular pyruvate into minimal medium containing glucose (Fig. 3A). We quantified extracellular pyruvate during growth in M9G (glucose) of the WT, ΔpftAB, and P hs pftAB strains. The three strains exhibited an identical growth phenotype (Fig. 3B, top panel). The extracellular pyruvate measured in the growth medium of the WT and ΔpftAB strains steadily increased until the late exponential phase to about 0.03 g · liter Ϫ1 (~0.34 mM) with a specific production rate (q Pyr ) of 0.5 mmol · h Ϫ1 · g of cells Ϫ1 (dry weight). Afterward, pyruvate concentration dropped to zero, which suggests that cells rapidly used the pyruvate before entering stationary growth phase. This result is consistent with the well-known phenomenon of excretion of pyruvate by B. subtilis prior to its (re)assimilation when glucose is depleted (7,22). With the strain overexpressing pftAB (P hs pftAB), pyruvate culminated at~0.09 g · liter Ϫ1 (~1.02 mM) after 7 h of culture (Fig. 3B, bottom panel). The specific production rate of pyruvate (q Pyr ) peaked at~5 mmol · h Ϫ1 · g cells Ϫ1 (dry weight) after 4 h of culture and quickly decreased thereafter. The specific glucose consumption and specific acetate and citrate production rates did not show any differences between the three strains ( Fig. S3A to E). These results indicated that PftAB can specifically export pyruvate and also that there exists at least another pyruvate transport system allowing the import and export of pyruvate (consistent with the residual growth on pyruvate of the ΔpftAB mutant). Altogether, our findings prompted us to conclude that the gradient of pyruvate drove the PftAB-mediated transport of pyruvate.
To reinforce this conclusion, we cultivated the same strains in M9G supplemented with 0.15 g · liter Ϫ1 (~1.70 mM) pyruvate. In this experiment, the pyruvate gradient is reversed so that if PftAB operates as a facilitated transporter, we expect to monitor an import of pyruvate (Fig. 3C). As shown in Fig. 3D, the concentration of extracellular pyruvate measured for the WT strain showed a slight increase until 10 h of culture, followed by a drop as soon as cells entered stationary phase. Interestingly, this drop coincided with the induction of pftAB when glucose was depleted (see next paragraphs) (19). Consistently, in the ΔpftAB strain, the consumption of pyruvate was delayed and strongly reduced. Conversely, the extracellular concentration of pyruvate measured with the strain overexpressing pftAB dropped from the beginning of the culture to reach zero before entry into stationary phase. There were no differences in the specific glucose consumption and specific acetate and citrate production rates in the three strains ( Fig. S3F to J). Taken together, these results demonstrated that the gradient of pyruvate drove the PftAB-mediated transport of pyruvate across the cell membrane, which indicated that PftAB operates as a pyruvate-specific facilitated transporter. Making use of a simple model of facilitated transport (Text S1 and Table S2), we estimated that the PftAB complex had a maximum rate (J max ) of approximately 10.0 Ϯ 1.0 mmol · h Ϫ1 · g of cells Ϫ1 (dry weight) and an apparent affinity constant for pyruvate (K m ) of approximately 1.0 Ϯ 0.1 mmol · liter Ϫ1 .
Pyruvate induces expression of pftAB by binding of LytT upstream of the pftAB promoter. The lytST operon, which is located immediately upstream of pftAB, was recently shown to be involved in the induction of pftA upon the entry of Luria-Bertani (LB)-grown B. subtilis cells in stationary phase and essential during growth on minimal medium plus pyruvate (19). van den Esker et al. (19) proposed that LytST directly induces pftAB based on the chromosomal proximity of these two operons. We tested this assumption in vitro by performing electrophoretic mobility shift assay (EMSA) using a DNA fragment of 257 bp containing the intergenic region of the lytST and pftAB operons (Fig. 4A). As shown in Fig. 4B, increasing amounts of the response regulator His-LytT resulted in a shift in migration of the labeled DNA band, which indicated LytT DNA binding. The possibility of nonspecific binding of LytT was eliminated, as no shifted bands were observed when LytT was incubated with PCR fragments of~250 bp used as controls (not shown). EMSAs using either AfeI, AluI, or DdeI restriction product of the 257-bp fragment revealed that LytT binds within an 88-bp region bounded by the AfeI and AluI restriction sites and that DdeI cut inside the LytT binding site (Fig. 4B).
In silico comparative-genomic analysis recently predicted putative binding sites for a series of two-component systems (TCSs) of unknown functions (27). Hence, LytT may bind a DNA sequence within the 88-bp above-mentioned region, which is composed of two boxes of 13 nucleotides separated by 8 nucleotides (Box 1 and Box 2 [ Fig. 4A]). We therefore tested this hypothesis using synthetic sequences mutated for these two boxes either independently (P pftAB-ΔlytT1 and P pftAB-ΔlytT2 ) or together (P pftAB-ΔlytT1-2 ) (Fig. 4C). LytT shifted the nonmutated DNA fragment but none of the three synthetic fragments (Fig. 4D), which strongly suggested that each of the boxes is required for LytT binding.
In order to in vivo validate the LytT binding site, gfp transcriptional fusions reporting the promoter activity of pftAB using either the wild-type sequence (P pftAB gfp) or the above-mentioned synthetic sequences (P pftAB-ΔlytT1 , P pftAB-ΔlytT2 , and P pftAB-ΔlytT1-2 ) were inserted at the pftAB locus in the WT strain and in a ΔlytST mutant strain ( Fig. 4C and Table 1). As expected, the growth of the ΔlytST strain was drastically reduced on pyruvate, while no growth defect was observed when replacing pyruvate by other substrates (Table S1). Because of this growth defect, cells were grown in M9SE with or without pyruvate. The green fluorescent protein (GFP) abundance was~4.5 units per OD 600 unit (U · OD 600 Ϫ1 ) for WT cells grown in M9SE with pyruvate (M9SEϩP) and barely detectable in M9SE (Table 2). In contrast, the GFP abundance was~0.5 U · OD 600 Ϫ1 for the ΔlytST strain under both conditions. The GFP abundance from the P pftAB-ΔlytT1 , P pftAB-ΔlytT2 , and P pftAB-ΔlytT1-2 synthetic reporter strains grown on M9SEϩP was similar to that in the ΔlytST strain (Table 2). Altogether, these results prompted us to conclude that each of the two predicted boxes is essential for the induction of pftAB by the LytST TCS. A genome-wide sequence homology search for other putative LytT binding motifs did not give any significant hits (not shown). Consistently, a comparative analysis performed using two recently published transcriptome data sets, M9P and M9SE (7), revealed that pftA and pftB were the only genes specifically induced in cells grown in M9P compared to cells grown in M9SE (Fig. S4). Altogether, these results established that pyruvate induced expression of only pftAB by binding the LytT response regulator upstream of the pftAB promoter.
Glucose represses transcription of pftAB by binding of CcpA to the ؊35 region of the promoter. Since PftAB can import and export pyruvate, its expression in cells grown on multiple carbon sources must be tightly regulated to ensure proper pyruvate homeostasis. The promoter of pftAB contains a putative cre site overlapping the Ϫ35 region, which suggests that pftAB expression may be under the control of CcpA, the master regulator of carbon catabolite repression (CCR). We tested this hypothesis in vitro by conducting EMSAs with increasing amounts of purified CcpA-His, serinephosphorylated HPr (P-Ser-HPr), and various restriction products of the previously described Cy5-labeled PCR fragment of 257 bp (Fig. 4A). As shown in Fig. 4E, increasing amounts of the CcpA/P-Ser-HPr complex resulted in a shift in migration of fragments of 170 bp for AluI, 116 bp for DdeI, and 189 bp for EcoRI. In contrast, no shift was observed after the fragment was treated by AfeI, which cuts in the putative cre box, the binding site for CcpA. These results prompted us to conclude that CcpA binds within a 48-bp region bounded by the DdeI and EcoRI restriction sites.
In order to validate in vivo that pftAB expression is under the control of the CcpA-dependent catabolite repression, we quantified the P pftAB gfp expression level in the WT, ΔccpA, ptsH1, Δcrh, and ptsH1 Δcrh strains under inductive or repressive conditions ( Table 3). The deletion of ccpA fully relieved the glucose-mediated repres- The binding region of LytT is detailed for the unmodified (P pftAB ), Box1-deleted (P pftAB-ΔlytT1 ), Box2-deleted (P pftAB-ΔlytT2 ), and Box1-Box2-deleted (P pftAB-ΔlytT1,2 ) reporter strains. Black bold letters stand for the two putative LytT DNA binding sequences (Box1, Box2). The red letters indicate the DNA sequence that replaced the deleted region in each strain. The violet shaded regions are similar to the binding sites identified in E. coli for the LytT-like YpdA response regulator (35). (D) EMSAs using purified His 6 -LytT (0, 0.5, 1, 1.5, 2, and 2.5 M) and the Cy5-labeled PCR fragment (represented in panel C) amplified from the P pftAB (267 bp), P pftAB-ΔlytT1 (259 bp), P pftAB-ΔlytT2 (263 bp), and P pftAB-ΔlytT1,2 (247 bp) reporter strains. (E) EMSAs using purified His 6 -CcpA sion of pftAB in M9P with glucose (M9PϩG). As expected, the glucose-mediated repression of pftAB was maintained in the ptsH1 and Δcrh single mutant strains (Table 3), which confirmed that the two cofactors can functionally replace each other (13). Consistently, the glucose-mediated pftAB repression was significantly reduced in the ptsH1 Δcrh double mutant. Hence, glucose repressed transcription of pftAB via CcpA and its cofactors by binding of the CcpA/P-Ser-HPr complex or the CcpA/P-Ser-Crh complex to the Ϫ35 region of the promoter.
Malate represses transcription of pftAB by a malic enzyme-dependent but CcpA-independent mechanism. Malate was recently shown to hijack the usual CcpAmediated catabolite repression (13). Consistently, the expression level of P pftAB gfp was 0.4 U · OD 600 Ϫ1 for WT cells grown in M9P with malate (M9PϩM), which is about 9% of the expression level in M9P (Table 3). However, the malate-mediated repression of pftAB was not relieved in the ΔccpA mutant or in any of the ΔccpA, ptsH1, Δcrh, and ptsH1 Δcrh mutants (M9PϩM medium [ Table 3]). These results prompted us to conclude that malate repressed pftAB transcription by at least one CcpA-independent mechanism.
To identify the key players of this CcpA-independent mechanism, a mini-Tn10 insertion library was constructed from a B. subtilis pftAB reporter strain in a ΔccpA background. This library was then screened on plates for pftAB derepression in the presence of malate. Thirteen positive clones from six independent pools of transposants were isolated and further characterized (Fig. S5). Most of the clones with strong pftAB derepression were mutated in the TCS malK/malR and its regulon involved in malate transport (maeN) and utilization (maeA) (6,28). As maeA is the first gene of the maeA ywkB operon, a possible role of ywkB (of unknown function) could not be excluded. We thus analyzed the effect of the deletion of each gene. The inactivation of maeA fully relieved pftAB expression in M9PϩM, while the inactivation of ywkB had no significant effect (Table 3). These results showed that the CcpA-independent malatedependent repression requires maeA.
a More-complete relevant genotypes are given in Table 1. b The deleted LytT binding sites are shown in Fig. 4C. B. subtilis possesses four paralogous malic enzymes: YtsJ, a NADP-dependent malate dehydrogenase, plays the major role in malate utilization, whereas MaeA, MalS, and MleA are NAD dependent and are dispensable for growth on malate (29). The CcpAindependent repression of pftAB was dependent on the NAD-dependent activity, as the overexpression of malS in a ΔmaeA strain fully restored the CcpA-independent catabolite repression (Table 3). Altogether, our results indicated that the NAD-dependent malic enzyme activity but not the MaeA protein per se is essential to drive the malate-mediated CcpA-independent repression of pftAB.
The higher the pyruvate influx and/or the intracellular pyruvate concentration, the stronger the CcpA-independent repression of pftAB. The NAD-dependent malic enzyme MaeA catalyzes the transformation of malate into pyruvate during B. subtilis growth on malate (7). This prompted us to hypothesize that the end product of the reaction (i.e., intracellular pyruvate) is responsible for the CcpA-independent repression of pftAB. To test this hypothesis, we manipulated the concentration of intracellular pyruvate by two complementary means, making use of the kinetic properties of the PftAB transport system. P pftAB gfp reporter cells were grown in M9SE supplemented with different concentrations of pyruvate, and the rate of pyruvate uptake was controlled by modulating the level of PftAB (using the WT, ΔpftAB, and P hs pftAB strains) (Fig. 5A). Hence, if the intracellular pyruvate negatively affects pftAB expression, we will monitor in the presence of extracellular pyruvate a lower P pftAB activity in the WT and IPTGinduced P hs pftAB strains than in the ΔpftAB mutant, which is unable to take up pyruvate. GFP abundances in the WT and ΔpftAB strains were similar for extracellular pyruvate concentrations below 1 mM (note that this is the estimated K m of PftAB). For higher extracellular pyruvate concentrations, GFP abundance increased up to 14 U · OD 600 Ϫ1 in the ΔpftAB strain, while it slowly decreased down to 4 U · OD 600 Ϫ1 for the WT strain. Remarkably, GFP abundance in the IPTG-induced P hs pftAB strain peaked at~11 U · OD 600 Ϫ1 for an extracellular pyruvate concentration of about 1 mM and dropped afterward down to~1 U · OD 600 Ϫ1 (Fig. 5A). These results indicated that, although extracellular pyruvate triggered the induction of pftAB, this induction is strongly inhibited (14-fold between the ΔpftAB and P hs pftAB strains) by the activity of PftAB, i.e., by pyruvate uptake. It is worth noting that this feedback regulation led to a repression of pftAB of~90%, which is similar to the level of the malate-dependent, CcpA-independent repression. Hence, the malate-dependent, CcpA-independent repression of pftAB most probably results from the strong increase in pyruvate influx and/or concentration generated by the high malic enzyme activity due to the induction of maeA in the presence of malate. The elevated pyruvate influx and/or concentration seems to alter the induction of pftAB by LytST. Although no DNA-binding regulatory protein other than MalR was identified by our transposon mutagenesis, the high malic enzyme flux in the presence of malate may activate a yet unidentified regulator responsible for the CcpAindependent repression of pftAB. Alternatively, the CcpA-independent repression of pftAB may be mediated by the LytST TCS itself, i.e., by reducing its activator activity. In order to test these assumptions, we constructed a synthetic fusion of the IPTGinducible P hs and the native P pftAB upstream of the gfp gene (P hs P pftAB gfp) (Fig. 5B). If the CcpA-independent repression of pftAB is mediated by an unknown DNA-binding repressor acting on or downstream of P pftAB , the expression driven by the upstream P hs will be altered by this repressor, somewhat acting as an artificial roadblock. In contrast, if the CcpA-independent repression of pftAB results from a lower activation of LytST, the expression from P hs will be fully relieved in the presence of IPTG. As shown in Fig. 5C, the malate-dependent repression was maintained on the synthetic P hs P pftAB gfp with or without IPTG. However, in the ΔccpA mutant, while the CcpA-independent repression by malate was maintained in the absence of IPTG, the repression was fully relieved in the presence of IPTG (Fig. 5C). This experiment revealed that CcpA repressed expression from the synthetic P hs P pftAB gfp in the presence of malate by binding on the Ϫ35 region of P pftAB (as in the presence of glucose) and by acting as a roadblock for the transcribing RNA polymerase recruited by P hs (Fig. 5B). Altogether, these results suggested that there is no other DNA-binding protein that repressed P pftAB in the presence of malate, which implies that the CcpA-independent pftAB repression actually resulted from a lower level of induction of pftAB by LytST.

DISCUSSION
In this study, we characterized the molecular and physiological logics of the pyruvate-induced response of a novel pyruvate transporter in B. subtilis. This novel bacterial transport system consists of a hetero-oligomeric complex of PftA and PftB which operates as a pyruvate-specific facilitated transporter. In the presence of extracellular pyruvate, the pftAB operon is induced by the TCS LytST. As for the transporters of alternative substrates, pftAB expression is repressed by the CcpA-dependent catabolite repression when a preferred carbon source, glucose or malate, is present in the medium. Unexpectedly, however, in the absence of preferred carbon source when the pyruvate influx is high, LytST activity is drastically retroinhibited (Fig. 6). Hence, LytST constitutes, together with the transporter PftAB, an original regulatory system ensuring proper adaptation to changing environments.
It was recently shown in B. subtilis forming biofilms that pftAB is induced in the presence of extracellular acetate (30). When we added acetate to the medium, we observed no induction of pftAB in exponentially growing cells. However, we observed a weak, dose-dependent induction when cells reached the stationary phase after growth in repressive conditions (data not shown). A reasonable hypothesis is that the reduction of the acetate export consequent to the presence of extracellular acetate led to an overflow metabolic shift toward an increase of pyruvate export. Indeed, it is well-known that when the intracellular pyruvate pool is high, the overflow metabolism, in particular pyruvate export, is stimulated (7). We showed that pyruvate export during exponential growth on glucose is independent from PftAB which is kept under the control of the catabolite repression (Fig. 6). Hence, at the onset of the stationary phase when the CcpA-dependent repression is relieved, the exported pyruvate triggered pftAB induction in a dose-dependent manner.
The pftA and pftB genes were originally annotated as encoding homologs of the Staphylococcus aureus LrgA and LrgB membrane proteins (18). The molecular mechanisms controlling death and lysis during biofilm development of S. aureus are organized around the CidR-regulated cidABC and LytSR-regulated lrgAB operons. cidA and lrgA encode proteins that are believed to function as holin and antiholin, respectively, while cidB and lrgB are of unknown function (31). Although PftA and PftB share strong homologies with holin/antiholin systems, the finding that these proteins belong to a new class of effective bacterial transporters is consistent with the recent discovery of the chloroplastidic glycolate/glycerate transporter, PLGG1, that most likely evolved from a gene fusion of bacterial lrgA (pftA) and lrgB (pftB) homologs (32,33). Also consistent with our findings in B. subtilis, an impaired ability to utilize pyruvate was observed in the lytST-like lytSR mutant of Staphylococcus epidermidis (34). In addition, it was recently shown in the evolutionarily distant bacterium Escherichia coli that the LytST-like YpdAB TCS is weakly activated by pyruvate (35,36). Remarkably, the YpdB binding site is similar to the B. subtilis LytT binding site (Fig. 4C). However, it was proposed that in S. aureus, LytSR senses decrease in membrane potential and responds by inducing lrgAB transcription. It was also proposed that a primary intermediate of overflow metabolism, acetyl phosphate (acetyl-P), already shown to act as a small phosphodonor to response regulators (37), directly activates LytR in an alternative signaling pathway (31). Our data, however, demonstrated that the higher the pyruvate influx, the lower the induction level of pftAB (Fig. 5A). This finding argues in favor of a negative-feedback regulation by the level of intracellular pyruvate (Fig. 6). Besides, we observed no induction of pftAB in either the WT or ΔccpA strain under conditions that are known to give rise to elevated concentrations of the intermediates of overflow metabolism and high acetate excretion rates (7,22). Hence, if acetyl-P acts as a phosphate donor to LytT in B. subtilis, it does not result in its activation, as has been proposed for S. aureus LytR. Overall, there are undeniable similarities in gene sequences and signaling between the homologous staphylococcal LytSR, E. coli YpdAB, and B. subtilis LytST TCSs but also significant physiological and functional divergences, which are likely to be related to niche-specific evolutionary constraints.
Unpredictable changing environments necessitate appropriate responses for successful bacterial adaptation. Appropriate growth strategies rely on sensing systems that globally adjust gene expression via transcription factor-mediated feed-forward and feedback regulations. In particular, bacterial TCSs combine a sensor (i.e., the sensor kinase) with a feed-forward controller (i.e., the response regulator) to induce expression of target genes involved in the adaptation process. The pftAB expression levels in response to different pyruvate concentrations (Fig. 5A) revealed the existence of an additional feedback control acting on LytST. To gain insight into the molecular mechanism that governed the opposed, feed-forward and feedback LytST-mediated regulation of pftAB by extracellular and intracellular pyruvate, we developed a simple model of gene expression and explored competitive, noncompetitive, and uncompetitive inhibitions of LytT activation by intracellular pyruvate (see Text S1 in the supplemental material). We assumed that the LytT regulatory protein binds to the pftAB promoter as a homodimer upon phosphorylation by LytS, which can be well represented by a Hill equation (38), but phosphorylation of LytT by LytS has not yet been demonstrated. The model simulations perfectly mimicked the induction of the pftAB promoter in the ΔpftAB strain (Fig. S6). However, the model could not account for pftAB induction in the other genetic backgrounds, in which cells are still capable of utilizing extracellular pyruvate (Text S1). Other molecular mechanisms may explain the feedback regulation of LytST by intracellular pyruvate, such as inhibition of LytS autophosphorylation or hindered recruitment of the RNA polymerase (Fig. 6) or a phosphorylation statedependent proteolysis of LytT. To decide between these mechanisms and verify to which extent they apply to other TCSs, characterization of the transduction signal and of the pftAB regulation dependency on the LytT phosphorylation state should be performed. Indeed, such a retroinhibition challenges the rational design of novel nature-inspired sensors and synthetic switches but undoubtedly offers new possibilities for the development of integrated sensor/controller circuitry.
In conclusion, the influx and efflux of pyruvate from exponential to stationary growth phases result from the timely (re)routing of the central carbon metabolic fluxes and of the dynamics of the intermediate concentrations. Hence, the intracellular and extracellular sensor systems and the feed-forward and feedback regulations of pyruvate uptake ensured with respect to cell adaptation a tight management of pyruvate homeostasis.
Gene deletion. Deletions of pftA (ysbA), pftB (ysbB), pftAB (ysbAB), and lytST were performed by sequence replacement with a chloramphenicol resistance cassette (Cm r ) expressed under the control of a constitutive promoter. The Cm r cassette was PCR amplified from plasmid pDG1661 using primers containing either a ClaI, SmaI, or XbaI restriction site (see Table S3 in the supplemental material). PCR products were purified using the Wizard SV gel and PCR clean-up system (Promega, Madison, WI) and then digested by the proper restriction enzymes. Regions containing pftA, pftB, pftAB, and lytST were PCR amplified from genomic DNA using primers listed in Table S3. PCR products were purified and subcloned into the pGEM-T Easy vector according to the supplier's protocol. Reverse PCR were performed on the resulting plasmids using primers containing either the ClaI, SmaI, or XbaI restriction site (Table S3). After purification and digestion, PCR products were ligated to the corresponding Cm r -containing PCR product.
Inducible gene expression. Isopropyl-␤-D-thiogalactopyranoside (IPTG)-inducible expression of lytST was performed by fusing lytST to the P spac promoter from pMUTIN4 using BamHI and SacI restriction sites, followed by the single crossover integration of the resulting plasmid upstream pftAB in the ΔlytST strain. IPTG-inducible expression of malS and pftAB were carried out by fusing these genes to the P hs promoter using plasmids pDR111 (kind gift of David Rudner), pDG1661, and pDG1664 (42). Briefly, PCR-amplified fragments (Table S3) of pftAB and malS were cloned into the pDR111 plasmid by SalI/NheI digestion leading to pDR111-P hs pftAB and by SalI/SphI digestion leading to pDR111-P hs malS. A PCR on pDG1664 was performed to introduce a NotI restriction site using appropriate primers (Table S3). The P hs pftAB from pDR111-P hs pftAB was then cloned into pDG1664-NotI using BamHI and NotI restriction sites. The spectinomycin resistance cassette (Spec r ) of pDR111-P hs malS was replaced by Cm r from pDG1661 using EcoRI and SacI. The pDR111, pDG1661, and pDG1664 derivative plasmids were inserted by double crossover at the amyE and thrB loci, respectively. IPTG-inducible expression of pftAB for complementation of the C. glutamicum mctC mutant was performed by inserting the PCR-amplified sequence of pftAB (Table S3) in the pXmj19 plasmid (kind gift of Gerd Seibold) using SmaI and SalI restriction sites.
Promoter reporter fusions. Fusion of the P pftAB promoter with lacZ (for transposon mutagenesis) was constructed using the pDG1661 plasmid and the PCR-amplified P pftAB using appropriate primers (Table S3). The PCR fragment was inserted by HindIII/BamHI restriction/ligation. Fusion of P pftAB with the gfpmut3 gene was carried out by ligation-independent cloning using the pBSBII plasmid (Table S3) as described previously (43). Reverse PCR on the pBSBII-P pftAB plasmid using proper primers (Table S3) was used to substitute parts or all of the LytT binding site from the promoter region of pftAB by BamHI. The pBSBII derivative plasmids were inserted at the pftAB locus by single crossover. Fusion of the P hs and P pftAB promoters upstream of gfpmut3 was performed by PCR amplification of the P pftAB gfp-containing region from the pBSBII-P pftAB using appropriate primers (Table S3) and subsequent insertion downstream of P hs into pDR111. The resulting plasmid was integrated by double crossover at the amyE locus, which led to the P hs P pftAB gfp transcriptional fusion.
Live-cell array and fluorescence analysis. Experiments and analyses were performed as previously described (22,43). Cells were grown in 100 l of medium in 96-well plates (Cellstar; Greiner Bio-One) and incubated at 37°C under constant shaking in a Synergy II microplate reader (BioTek). The optical density at 600 nm (OD 600 ) and fluorescence were measured every 10 min. Each culture was performed in at least three technical replicates by two biological replicates (more than six values). The mean green fluorescent protein (GFP) concentration in exponentially growing cells was expressed as unit per OD 600 , with 1 unit being equivalent to 1 pM fluorescein (Fig. S6).
Transposon mutagenesis. The mini-Tn10 delivery vector pIC333 was used for transposon mutagenesis as previously described (44,45). This plasmid was introduced into B. subtilis strain TC64 (Table 1) at 25°C using erythromycin selection. Single transformants were used to inoculate independent cultures at 25°C in LB plus spectinomycin. In early exponential growth, the temperature was shifted to 40°C, and cultures were allowed to grow for 4 more hours. Appropriate culture dilutions were spread on solid M9 medium with pyruvate and malate (M9PϩM) plus spectinomycin and 0.04% X-Gal (5-bromo-4-chloroindolyl-␤-D-galactopyranoside) and incubated at 25°C to screen for clones with derepressed P pftAB lacZ expression (blue colonies). Thirteen clones from independent pools of transposants were isolated in which the lacZ gene was expressed on M9PϩM (Fig. S4). These clones were selected for backcross experiments, and genes inactivated by the transposon were identified by target rescue and sequencing.
PftA and PftB localization and quantitative pulldown assays. Fusion of the SPA (sequential peptide affinity) sequence at the 3= end of either pftA or pftB was performed using the pMUTIN-LICSPA plasmid, previously adapted from the pMUTIN-SPA plasmid for ligation-independent cloning to facilitate high-throughput (HTP) cloning (46) and appropriate primers (Table S3). Each SPA-tagged strain was precultured overnight in the presence of erythromycin at 37°C. The culture was initiated by a 500-fold dilution (to an OD 600 of 0.001) in 2 liters of M9 medium with glutamate and succinate plus pyruvate (M9SEϩP). When the OD 600 reached 0.3, cultures were centrifuged at 3,000 ϫ g for 10 min at 4°C. The pellet was resuspended in 50 ml of cold buffer (10 mM Tris-Cl [pH 8], 150 mM NaCl) and instantly frozen in liquid nitrogen. The cells were disrupted using a French press. Membrane and cytosolic fractions were separated by ultracentrifugation at 100,000 ϫ g for 1 h. The membrane pellet was solubilized using buffer A (10 mM Tris-Cl [pH 7.5], 150 mM NaCl, 1 mM EDTA) supplemented with 1% DDM (n-dodecyl-D-maltoside). Protein concentrations were measured by the Bradford method, and identical protein amounts of the two fractions were loaded on 10% SDS-polyacrylamide gels for Western blotting using anti-FLAG antibodies. In addition, protein complexes from the membrane fraction were pulled down using a SPA purification method (47), then separated by native polyacrylamide gel electrophoresis (PAGE), and analyzed by nanoscale liquid chromatography coupled to tandem mass spectrometry (nanoLC-MS/MS) (Text S1).
Affinity capture of the SPA-PftA PftB-His complex. The pftAB operon was PCR amplified from the B. subtilis BSB168 chromosome with primers OOD141 and OOD142 (Table S3). The OOD142 primer allowed the PftB C-terminal fusion of six His residues. After digestion by XhoI and NotI, the PCR fragment was ligated to the pSG-SPA-NTER plasmid (48), enabling the PftA N-terminal fusion of SPA and placing the SPA-pftA pftB-His 6 synthetic operon under the control of a xylose-dependent promoter P xyl . The plasmid was used to generate strain TC199 by double crossover at the amyE locus. Transformants were selected for resistance to spectinomycin and lack of amylase activity. Expression of SPA-pftA pftB-His was induced by 1% xylose (vol/vol) at mid-log growth and harvested 3 h later. Cell membranes were prepared after cell disruption by sonication in buffer A and centrifugation at 100,000 ϫ g for 1 h. Membranes were solubilized using buffer A supplemented with 1% n-dodecyl-␤-D-maltoside (DDM). The solution was either mixed with nickel-nitriloacetic acid (Ni-NTA) agarose resin (Invitrogen) after the addition of 10 mM imidazole or mixed with the CaM Sepharose-4B resin (GE Healthcare) after the addition of 2 mM CaCl 2 and left overnight at 4°C on a rotating wheel. For the Ni-NTA agarose column, washes were performed with buffer A supplemented with 20 mM imidazole, and the elution was performed with 250 mM imidazole. For the CaM Sepharose-4B column, washes were performed by buffer A supplemented with 0.1 mM CaCl 2 , and the elution was performed with buffer A supplemented with 3 mM EGTA. Both columns were also used sequentially, in both ways. Prior to the analyses by Western blotting, proteins were precipitated by the addition of 10% trichloroacetic acid (TCA).
Electrophoretic mobility shift assay. The lytT coding sequence was PCR amplified using primers enabling insertion of a His 6 tag between the start codon and the coding sequence and cloning into the E. coli pJ411 expression vector (DNA2.0, Newark, CA), and His 6 -LytT proteins were expressed in strain