Mycobacterium tuberculosis Proteins Involved in Mycolic Acid Synthesis and Transport Localize Dynamically to the Old Growing Pole and Septum

Understanding the mechanism that controls space-time coordination of elongation and division of Mycobacterium tuberculosis (Mtb), the causative agent of tuberculosis (TB), is critical for fighting the tubercle bacillus. Most of the numerous enzymes involved in the synthesis of Mycolic acid - Arabinogalactan-Peptidoglycan complex (MAPc) in the cell wall are essential in vivo. Using a dynamic approach, we localized Mtb enzymes belonging to the fatty acid synthase-II (FAS-II) complexes and involved in mycolic acid (MA) biosynthesis in a mycobacterial model of Mtb: M. smegmatis. Results also showed that the MA transporter MmpL3 was present in the mycobacterial envelope and was specifically and dynamically accumulated at the poles and septa during bacterial growth. This localization was due to its C-terminal domain. Moreover, the FAS-II enzymes were co-localized at the poles and septum with Wag31, the protein responsible for the polar localization of mycobacterial peptidoglycan biosynthesis. The dynamic localization of FAS-II and of the MA transporter with Wag31, at the old-growing poles and at the septum suggests that the main components of the mycomembrane may potentially be synthesized at these precise foci. This finding highlights a major difference between mycobacteria and other rod-shaped bacteria studied to date. Based on the already known polar activities of envelope biosynthesis in mycobacteria, we propose the existence of complex polar machinery devoted to the biogenesis of the entire envelope. As a result, the mycobacterial pole would represent the Achilles' heel of the bacillus at all its growing stages.


Introduction
The mycobacterial cell envelope is of an exceptional complexity. One of the major challenges for Mycobacterium tuberculosis (Mtb) is probably achieving the synthesis of this thick multilayer barrier [1,2]. The Mtb envelope is partly responsible for its innate resistance to antibiotics and plays a major role in both its virulence and persistence. Tuberculosis (TB) reappeared in the 1990s and remains one of the main causes of lethality worldwide. The principal challenge is to fight the increasing number of multi-drug resistant (MDR) Mtb clinical isolates. The alternate periods of growth and dormancy of Mtb during infection [3] render the discovery of new effective antibiotics difficult. Mtb division, as well as the regulation of its cell cycle and the maintenance of cell wall homeostasis, are nonetheless weak points of the bacillus. Understanding the mechanisms that guarantee spatial and temporal coordination of Mtb envelope biogenesis during the cell cycle is a crucial step toward deciphering the role of this major player of Mtb pathogenesis [4,5].
In Mtb, two Fatty-Acid-Synthase systems (FAS), FAS-I [6] and FAS-II [7], are devoted to the synthesis of common fatty acids and of specific, long-chain, a-acylated, b-hydroxylated fatty acids, namely mycolic acids (MA) [8,9]. Mycolic acids constitute the major lipid component of the envelope and form the external mycomembrane [10,11]. They are either inserted in the mycomembrane as trehalose dimycolates and monomycolates (TDM and TMM) or linked covalently to the underlying arabinogalactan (AG), itself bound to peptidoglycan (PG) to form the MAPc [12,13]. Mycolic acids stem from the condensation of a medium chain-length fatty acid (C 24 -C 26 ) produced by FAS-I with a long meromycolic chain (up to C 60 ) produced by FAS-II ( Figure  S1) [14,15]. FAS-II is comprised of specialized and interconnected protein complexes [16,17,18] ( Figure S2). The MA biosynthetic pathway is assumed to be cytoplasmic, a notion recently reinforced by the discovery of a MA transporter in mycobacteria: the potential trans-membrane protein MmpL3 [19,20,21]. We hypothesized that at least part of the FAS-II multi-protein complexes may exist as stable entities and could be involved in the biogenesis of MAPc during bacterial growth.
The cell wall elongation process of actinobacteria [4] is different from that of other rod-shaped models such as Escherichia coli or Bacillus subtilis [22]. In the latters, nascent PG is made all along the side of the bacterium during elongation and at the septum during the division process [23] whereas in the former, the lateral PG derives from polar biosynthesis during elongation [24]. Among mycobacteria, the polar localization of PG synthesis is mainly due to the phospholipid-interacting protein Wag31, the ortholog of the B. subtilis DivIVA protein [25,26,27]. Wag31, by targeting negatively-curved membranes [28], is responsible for the polar localization and the septal re-localization of several PG biosynthesis proteins [29,30]. Wag31 is mainly located at the ''old pole'' (Figure 1), which is the active locus among mycobacteria and other actinobacteria where new molecular material of the lateral cell wall is added [31,32]. The ''septal pole'' (Figure 1), which represents the future pole and the developing septum, does not participate in lateral PG biosynthesis [33].
We demonstrate herein that reductases from the FAS-II core as well as the MA transporter MmpL3 are located at the active mycobacterial old poles and at the position of the septum prior to division. The condensing enzymes form polar foci but are also able to diffuse in the cytoplasm. FAS-II localization is correlated with the dynamics of cell division, with the localization of Wag31 and with the dynamics of the MA transporter MmpL3. These results describe the first observation of the localization of sophisticated machinery for fatty acid biosynthesis. In addition, they reveal the existence of a possible link between two central and essential processes: bacterial division and whole envelope biogenesis. Mycolic acids may be synthesized at the active poles and septa in order to be transported, probably by MmpL3, to the external mycomembrane.

FAS-II complex components are located at one bacterial pole
Mtb genes encoding the FAS-II reductases MabA and InhA, and the condensing enzymes KasA and KasB, were merged to the C-terminal end of gfp and expressed under the control of the P amiE inducible promoter in the non-pathogenic bacterium model M. smegmatis mc 2 155 (Msm). In a first series of experiments, large field fluorescence microscopy images, hereafter called static images, were acquired six hours after induction. GFP alone emitted a bright and diffuse fluorescence throughout the cytoplasm (Figure 2A). Conversely, the reductase fusions yielded intense foci located mainly at one bacterial pole (Figure 2A). In the strains expressing either GFP-KasB or GFP-KasA, polar foci were also detectable although together with a diffuse fluorescence in the cytoplasm (Figure 2A). To quantify foci formation, a polar index (PI) was calculated and defined as the ratio of the number of bacteria containing at least one polar focus reported to the total number of fluorescent bacteria (N). Over one thousand bacteria expressing each fusion were analyzed on several replicates. Two groups were found ( Figure 2B). The first group, e.g. MabA (89.47%61.20, N = 1100) and InhA (63.86%61.56, N = 2000), had a high PI while the second group, e.g. KasA (13.34%61.40, N = 900) and KasB (21.02%62.40, N = 1700), had a lower PI. Due to the inaccuracy of quantifying and comparing static images from different snapshots, the total amount of fluorescence in live bacterial culture was assessed by fluorometry as a function of induction time. This enabled to verify whether PI was correlated or not with fusion expression levels. Eight hours after induction ( Figure 2C), the strain containing GFP alone (PI = 0%) emitted the maximum fluorescence (280 RFU). MabA (PI = 89%) and KasB (PI = 21%) yielded intermediate fluorescence levels (45 RFU and 35 RFU,respectively) whereas KasA (PI = 13%) and InhA (PI = 63%) yielded very low levels (below 10 RFU). The absence of correlation between fluorescence levels and PI values (as compared between Figure 2B and Figure 2C) was further confirmed by Western blotting experiments performed on the same cultures ( Figure 2D). Western blotting results were perfectly correlated with fluorometry results (see comparison between Figure 2D and Figure 2C) and thus were not correlated with PI values, indicating that foci formation was not due to protein overexpression but rather due to preferential localization of FAS-II at the pole.

Dynamics of polar localization of FAS-II proteins
In order to analyze foci formation and localization and avoid potential problems due to protein accumulation, a dynamic analysis was implemented. The goal was to observe and follow the very early stages of foci formation during bacterial division, starting just after the beginning of the induction of GFP-fusion expression with a low amount of inducer (0.02% acetamide). Time-lapse microscopy experiments were designed using a recent and efficient agar pad microscopy technique [34]. In the GFPproducing strain (Movie S1), the signal was bright and diffuse in which no foci were observed. In the GFP-MabA producing strain (Movie S2), at the beginning of induction, the first foci ( Figure 3, foci a, 3 h; and b, 5 h), were observed at the septal poles of the mother bacterium. These foci appeared at mid-cell at the position of the next division septum. These types of foci were often apparently split in two after the end of the division process ( Figure 3, foci a 1-2 , 5 h; b 1-2 , 9 h). Thereafter, they became established at the new poles of the first pair of daughter bacteria ( Figure 3, foci a, b, c, 9 h to 13 h). At the end of induction the images resembled the static images from the first experiments ( Figure 2A) where this protein was present at only one pole, leading to conclude that these foci correspond to dynamic protein complexes. The same results were obtained with the InhA fusions (Movie S3), displaying foci which were first visible at the septal poles and thereafter establishing at one preferential pole. Because of the lower level of expression of the KasA fusion (Movie S4), it was more difficult to follow foci dynamics at the beginning of induction although in the micro-colonies, GFP-KasA foci were also first visible at the septal poles and then appeared very rapidly at the poles. Finally, the KasB fusion (Movie S5) was diffuse in the cytoplasm at the early stage of induction and the foci were only visible at one pole at the end of the experiment. There was no typical envelope labeling by FAS-II fusions observed at any time, indicating that the FAS-II protein tested are probably cytoplasmic. We thus hypothesized that these proteins ultimately participate in pole structure in order to ensure elongation of the lateral mycomembrane.

FAS-II proteins are localized with the old pole marker protein Wag31
Since FAS-II localization appeared to follow the reported localization of the polar protein Wag31 [29,35], a recognized ''old pole marker'', Wag31-FAS-II colocalization was therefore studied. For purposes of clarity, the word colocalization is used in this study to describe the localization of two proteins at the same focus but without implying the existence of any molecular interaction. A strain was constructed bearing a single but ectopic chromosomal copy of a mCherry-Wag31 fusion under the control of the constitutive promoter pHSP60 of the integrative vector pMV361. This type of construct, using the pHSP60 promoter, has been successfully used previously to study Wag31 localization [36]. When this strain was observed in fluorescence microscopy (Figure S3), the mCherry fusion was more visible at one pole as also commonly observed by others [36,37]. This pole has been shown to represent the old pole ( Figure 1) [29]. The fusions were introduced into the mCherry-Wag31 producing strain after which the fluorescence of both GFP and mCherry was analyzed as performed above on static images. As observed in the absence of mCherry-Wag31 (Figure 2A), GFP remained diffuse whereas GFP-MabA and GFP-InhA yielded intense polar foci, mainly at one pole ( Figure 4A). GFP-KasA and GFP-KasB still appeared diffuse with a lower extend of polar spotting ( Figure 4A). The overall localization pattern of the FAS-II fusions was comparable to that of the previous fusions ( Figure 2A). The statistical relevance of the localization ( Figure 4B) was analyzed as above on individual bacteria expressing each fusion (N) stemming from several individual slides (n.6). The polar indices of MabA (93.9360.59 N = 800), KasA (23.8764.56 N = 500) and KasB (13.4961.43 N = 500) were not statistically different from the previously observed indices (see comparison between Figure 2B and 4B) whereas the PI of GFP-InhA (35.9962.14 N = 700) was slightly reduced. While there was no dramatic effect of the presence of the mCherry-Wag31 fusion on the percentage of localization of the FAS-II proteins, there was the presence of colocalization of both Wag31 and FAS-II fusions foci. To quantify this phenomenon on a large number of bacteria a colocalization index was defined as being the percentage of bacteria exhibiting Wag31-GFP colocalization relative to the number of GFP-mCherry-positive bacteria. Colocalization indices ( Figure 4C) were very high for all FAS-II proteins: MabA (83.31%63.27, N = 800), InhA (93.93%62.60, N = 700), KasA (89.71%63.37, N = 500) and KasB (89.52%63.60, N = 500). In order to evaluate the distribution of foci within individual bacteria, a more in-depth analysis was performed on a subset of bacteria by scanning more than 100 organisms of each type. On each scan ( Figure 4D), the maximum fluorescence was plotted against the normalized length of the bacteria. A dot represented on a graph can refer to a focus (at the pole) or simply to the maximum of diffuse fluorescence along the cell. Localization of the reductases and colocalization with Wag31 at the old poles were clearly confirmed ( Figure 4D). Furthermore, the maximum florescence emitted by KasA and KasB, which displayed the lowest polar indices, were clearly preferentially observed in proximity of the old poles along with Wag31, thus suggesting an actual preference for this localization ( Figure 4D). The distribution of the polar foci clouds represented only about 20% of total cell size, i.e. less than 1 mm, probably due to light diffusion. This distribution was similar for FAS-II proteins and for Wag31, already known to be localized close to the polar membranes. In contrast, the maximum accumulation of GFP alone was never found at the old pole suggesting a polar exclusion of GFP alone. These analyses clearly demonstrate that the FAS-II proteins were preferentially located at the old pole with Wag31. Since Wag31 is acknowledged to be an old pole marker, we can conclude that the polar FAS-II proteins were located at the active growing ''old pole'' of the bacteria.

FAS-II localization dynamics follow Wag31 dynamics with no molecular interaction
In order to compare FAS-II and Wag31 dynamics, time-lapse experiments were performed on bacteria expressing both types of fusions. The expression of mCherry-Wag31 being constitutive, polar Wag31 foci were visible in the mother cells at the beginning of the experiments. As observed on static images ( Figure 4A), GFP alone was diffuse in the cytoplasm (Movie S6). In this movie, the presence of Wag31 at the old pole and its re-localization to the septum were clearly visible (Movie S6). In the GFP-FAS-II producing strains (Movies S7-S10), bacterial division was observed together with Wag31 and FAS-II dynamics. The dynamics of GFP foci were analyzed in detail for the GFP-MabA fusion (Movie S7 and Figure 5). Prior to the detection of GFP fluorescence, mCherry-Wag31 mainly observed at the old poles then, as the elongation progressed (1 h, bacteria 1 and 2), Wag31 was relocated to the septa as previously observed [29,30]. The GFP fusions were first detected at these new poles and slightly at the old poles (2 h, bacteria 1 1 and 1 2 ). Progressively thereafter, establishment was observed at the old poles (see after 5 h, bacteria 1 1.1 ). At the end of the experiments, GFP fluorescence was primarily observed at the old poles together with the brighter Wag31 fluorescence. Bacterium no. 2 presented here ( Figure 5) displayed exactly the same dynamics with a delayed appearance at the beginning of the induction. Because of the lower expression of the InhA fusion, the dynamics of this fusion was more difficult to follow although the majority of GFP foci were established with the brighter Wag31 foci (Movie S8). Finally, the KasA fusion (Movie S9) and the KasB fusion (Movie S10) were mainly diffuse in the cytoplasm as already observed on static images and foci were observed only at the end of the experiments. When GFP fusion expression was induced, new proteins entered the system via the pole in construction (septal pole) and then established progressively to the old pole with Wag31.
To ascertain whether dynamic FAS-II-Wag31 colocalization was due, or not, to molecular interactions, in vitro Co-immunoprecipitation (Co-IP) experiments were performed with Wag31 against FAS-II proteins. In vitro 35 S-labeled HA-tagged FAS-II proteins (h-proteins) bound to magnetic beads coated with anti-HA antibodies were used to trap labeled c-Myc-tagged Wag31 protein (c-Wag31). No Co-IP bands were observed, except with Wag31 itself ( Figure S4A). When the Co-IP experiments were performed in the reverse direction ( Figure S4B), with h-Wag31 against c-FAS-II proteins, only non-specific binding was observed with FAS-II proteins ( Figure S4B, IPØ). This non-specific binding was not observed when known negative controls were used ( Figure  S4C) including human c-Lamin (c-Lam) or mycobacterial Nat protein (Arylamine N acetyltransferase, [38]). This indicates that FAS-II proteins localize to the old pole and follow Wag31 dynamics without any direct molecular interaction with Wag31.
In vivo activity of GFP-fusions FAS-II proteins are organized in specialized complexes ( Figure  S2) interconnected with one another [16,17,18]. We showed above that representative members of theses complexes were located at the old poles during elongation and at the septa prior to division. It was thus tempting to postulate that the biosynthesis of mycolic acids may occur at these precise foci. To date, there is no known biochemical clue for locating MA biosynthesis in vivo. Thus, to address this question, we chose to analyze GFP-fusion activity in vivo. The kasB gene is the only FAS-II gene that is not essential in both M. marinum [39] and Mtb [40]. In the absence of KasB, which is part of the E2-FAS-II complex [17] and is responsible for the final step of meromycolate elongation, the MA are shortened by two to four carbons [40]. For the present purposes, we constructed a kasB-KO mutant of Msm ( Figure S5) that was viable, indicating that kasB is also ''non-essential'' in Msm. In this mutant, the alphaand epoxy-MA were shortened by four carbons ( Figure 6A, 6B) whereas the alpha'-MA, which appeared as KasB independent, were not affected ( Figure S6). The mutant phenotype was reverted by complementation with the wild type kasB gene and, more importantly, by the GFP-KasB fusion ( Figure 6A). The localization mCherry-Wag31 fusion-expressing bacteria. GFP fluorescence images (GFP; leftmost column) together with mCherry fluorescence images (Che, middle column) of Msm::mCherry-wag31 expressing GFP fusions with MabA, InhA, KasA, KasB, or GFP alone (Ø) were acquired after 6 hours of induction. The merged images (right columns) allowed visualizing the polar colocalization of the fusions. Magnification and scale bars are identical to those indicated in Figureô   pattern of GFP-KasB in the complemented mutant was identical to that observed with the wild type (data not shown). The GFP-KasB fusion was active in vivo.
Given that the other FAS-II genes (mabA, inhA, kasA) are essential [41,42] and part of operons, it was not easily achievable, in our system, to test the GFP-fusion complementation of interrupted genes. MabA and kasA genes exist within operons whereas inhA lies at the end of an operon. Nevertheless, we had some inkling with regard to InhA. InhA is the primary INH (isoniazid, also known as isonicotinylhydrazine) target and its overproduction induces an increase in the level of INH resistance by trapping the INH-NAD adducts in its catalytic pocket, thus reducing INH concentration within the bacteria [43]. We first verified that the overproduction of the wild type inhA gene in our strain indeed increased INH resistance (18.75 fold; 75 mg/ml, Figure 6C) whereas the production of GFP alone did not affect this level (4 mg/ml). A clear increase in resistance threshold, (12.5 fold; 50 mg/ml), was also observed when GFP-InhA was produced. The GFP-InhA fusion was likely able to bind the INH-NAD adducts in vivo thereby inducing an increase in cell resistance to INH. Finally, to demonstrate that the GFP-InhA fusion was active, the complementation of an inhA thermosensitive mutant Msm strain (mc 2 2359; inhA ts ; [44]) was tested ( Figure 6D). At 42uC, in the presence of acetamide, the inhA ts strain containing the void vector pLAM12 was unable to grow. In contrast, when either the wild type InhA protein or, more interestingly, the GFP-InhA fusion was expressed, the complemented strains were thermo-resistant indicating that the fusion was also able to complement the InhA deficiency. These results demonstrate that the GFP-InhA as well as the GFP-KasB fusion may be active in vivo while forming foci.

Localization of the Mycolic acid export machinery
Recently, the potential membrane protein MmpL3 has been identified in Mtb as an MA transporter [19,45]. It is involved in the transport of TMM to the external mycomembrane. Because of the localization of major components of FAS-II observed herein, we hypothesized that MA may be synthesized at these loci to be subsequently transported at the actively growing envelope regions. GFP was therefore merged to the MmpL3 C-terminal given that this domain was predicted to be intra-cytoplasmic (Toppred2, [46], http://bioweb.pasteur.fr/seqanal/interfaces/toppred.html). MmpL3, a RND family protein [47], was predicted to be organized in twelve transmembrane helices. The localization of this fusion was analyzed as above. Upon performing time-lapse experiments with MmpL3-GFP alone (Movie S11), or in the presence of mCherry-Wag31 (Figure 7, extracted from Movie S12), a very clear envelope staining was observed together with preferential labeling of one pole and a very bright and flashing labeling of the septum, likely due to the presence of the double membrane at this position. The first conclusion was that MmpL3 might indeed be in the membrane or at least, in the envelope. Its C-terminal domain, fused with the GFP, could be intracytoplasmic because GFP is not fluorescent when it is located in another compartment [48]. As a localization control, a GFP fusion of the first ten transmembrane helices of MmpL3 (MmpL3-N10) was constructed and its localization analyzed in vivo as described above. Upon performing time-lapse experiments with MmpL3-N10-GFP in the presence of mCherry-Wag31 (Figure 8, extracted from Movie S13), the preferential MmpL3 polar localization was not observed. MmpL3-N10, which remained in the membrane, or at least in the envelope, was no more accumulated at the pole. The polar accumulation (P) of MmpL3 was quantified as described in Materials and Methods and compared to that of MmpL3-N10 ( Figure 8B). MmpL3-N10 localization was not polar (P = 1.16 6 0.15 N = 113) whereas MmpL3 was enriched by more than 5 fold at the pole (P = 5.3560. 38 N = 192).
The aspect of the images of the polar localization of MmpL3 differed significantly from that observed with the FAS-II components. The former was ''crescent-shaped'', suggesting a membrane localization of MmpL3. MmpL3 was located in the envelope, probably in the cytoplasmic membrane, and was preferentially located at the growing bacterial tips containing the larger amounts of Wag31. In addition, MmpL3 was visible very early in the septa, likely before the appearance of Wag31 and membrane curvature. This polar localization of the MA transporter appeared to be due to a MmpL3 domain located between the end of helix 10 and the C-terminal end of the protein.

Discussion
The main objective of the present study was to determine whether the biogenesis of mycolic acids is coordinated or not with bacterial envelope elongation. Through the use of a mycobacterial model of Mtb, M. smegmatis, we were able to localize MA biosynthesis FAS-II proteins in vivo and compare their localization dynamics to that of the polar protein Wag31. Wag31 is a main component of mycobacterial elongation and division [29,37] and has been found to be associated with the curved membranes of old poles as well as of the division septum [49]. Herein; proteins of the FAS-II complex core [16,17,18] were observed to form foci associated with both the old poles and septal poles. While the FAS-II condensing enzymes were mainly diffuse, they were also able to form foci at the same locations. In dynamic studies, newlysynthesized ectopic FAS-II proteins were first located at the septal region suggesting that they may participate very early in pole construction. Thereafter, they were established rapidly at the old poles together with polar Wag31 foci [29]. Such polar localization of FAS-II proteins as well as their dynamic localization to the old poles along with Wag31 suggest the existence of a spatiotemporal coordination between MA biosynthesis and Wag31-controlled polar PG synthesis.
A key question was to determine whether foci, which were observed herein in bacteria expressing ectopic GFP fusions, correspond to actual complexes or to protein aggregates or inclusion bodies (IB). In rod-shaped bacterium E.coli, which does not display the same polar cell-wall elongation process as mycobacteria, the poles are not the active sites for PG synthesis [22]. In E.coli, the poles and the regions between the chromosomes are the areas where protein aggregates accumulate [50]. Furthermore, these foci do not separate after division and are preferentially accumulated in the mother cell during cellular aging [51]. In contrast, in mycobacteria, poles are very active loci where lateral PG synthesis [22] and most likely free MA synthesis [52], occur during bacterial elongation. In addition, poles are also involved in protein export [53], DNA transfer [32], chromosomal partitioning and DNA replication [54,55] as well as phosphorylation by Ser/Thr protein kinases [56]. Mycobacterial poles are thus certainly not suitable for detrimental protein accumulation. A second argument against polar IB formation is the fact that asymmetric FAS-II polar localization was observed in the present model. Actinobacteria exhibit apical (polar) growth [24] and the distribution of polar proteins involved in PG biosynthesis is asymmetric with a preference for old poles [29]. Even though old pole-containing mycobacteria do not grow faster [57], polar growth itself appears to be asymmetric with the old pole being more active than the new pole [31]. Indeed, we observed herein a localization preference of FAS-II proteins to the old pole containing Wag31 whereas GFP alone seemed excluded from these poles. The foci clouds of FAS-II and Wag31 were located at the same position in the bacteria, at the very end of the organism. As commonly observed in E.coli, inclusion bodies are otherwise more likely to be found further inside the bacterial tips. Finally, three additional observations in the present study advocate against IB formation. First, there was no quantitative correlation between foci formation and GFP-fusion expression levels. Secondly, two of the four fusion proteins tested (KasB and InhA) were active in vivo while forming foci. Thirdly, during the division process, we observed foci splitting which can be compared to the active polarisome splitting in Streptomyces [58]. These finding strongly suggest that the FAS-II foci revealed the presence of FAS-II complexes mainly at the old poles and septal poles. Mycolic acids are believed to be synthesized in the cytoplasm and transported to the periplasm as TMM by the trans-membrane RND-family transporter MmpL3 [19]. As expected, in view of its secondary structure and function, MmpL3 was observed at the level of the mycobacterial envelope. It is likely inserted in the plasma membrane similarly to that proposed in the original MmpL3 study [19]. Moreover, RND family proteins such as gram negative efflux pumps for example are inserted in the inner membrane and driven by the proton motive force [59] solely present at this level of the envelope. In addition, the MmpL3 Cterminal GFP fusion used herein was fluorescent, indicating that the C-terminus domain of MmpL3 is probably located in the cytoplasm. GFP functions as an efficient reporter for analysis of protein topology since it is fluorescent when located in the cytoplasm and non-fluorescent when transported outside the plasma membrane [48]. MmpL3-GFP was clearly more abundant in the membrane of Wag31-containing old poles while displaying a crescent-shaped fluorescence. This typical shape is similar to that observed for DivIVA, an ortholog of Wag31 in B.subtilis [28]. Moreover, this shape was very different from the shape of the GFP-FAS-II foci, suggesting that FAS-II was not inserted in the membrane as in the case for MmpL3 or occupying the inside of the membrane as in the case of Wag31. Finally, the most impressive images were obtained when the bacteria formed a septum: MmpL3 yielded a very intense labeling at the position of the septum probably due to its insertion into the double membrane of the septum. This MmpL3 localization pattern strongly resembled that reported for membrane Ser/Thr protein kinase PknB [56]. Indeed, Wag31 has been shown to be regulated by PknA and PknB [29]. PknA and B have been involved in the inhibition of many MA biosynthesis proteins such as KasA [60], mtFabH [61], MabA [62], InhA [63,64] and the methyltransferase PcaA [65]. This suggests the existence of a common regulatory network between FAS-II, Wag31 and the cell wall elongation process. From the present data and in view of the literature, we can propose that mycolic acids are probably primarily exported at the poles and septa in proximity to Wag31 and FAS-II foci and at the locus of PG biosynthesis. The disappearance of the polar accumulation of MmpL3 upon deletion of its C-terminal end suggests that this region is probably involved in its localization. It also suggests that MmpL3 may be localized through a diffusioncapture mechanism using its C-terminal end as signal [66].
A puzzling question was to ascertain whether the polar localizations of FAS-II and MmpL3 reflect MA biosynthesis and export localization. Some reports can indeed be interpreted in this manner when compared with our results. Mtb polar lysis has been observed in very early EM images of isoniazid-treated wild-type bacilli [67]. Isoniazid targets InhA and if we assume that synthesis and export of mycolic acids are primarily polar, InhA inhibition will leave ''holes'' in the polar mycomembrane and ultimatly induce specific unipolar lysis as previously reported [67]. More recently, polar lysis was also observed when arabinan synthesis was inhibited with benzothiazinone [31,52]. Finally, a link with MA synthesis was also attributed following MmpL3 localization. Because MmpL3 is accumulated at the pole, MA export (and probably MA synthesis) can be polar. By using unnatural trehalose analogs, two separate reports have shown that Ag85 activity, the mycoloyl-transferase that links mycolic acids to arabinogalactan, is concentrated in the envelope and at the poles of mycobacteria with a pattern strongly resembling that of MmpL3 herein [68,69]. The linking of mycolic acids to the cell wall may occur mainly at these positions. As suggested by the present results, the C-terminal domain of MmpL3 may be involved in this localization.
In conclusion, the polar localization of external membrane lipid biosynthesis enzymes observed in the present study constitutes a novel observation in bacteria and a fortiori in mycobacteria. Moreover, the observation of the dynamic localization of the MA transporter MmpL3 strongly reinforces the possibility of a positioning of MA biosynthesis in these regions. In keeping with the existing literature, the present data further suggest the existence of the polar localization of complete MAPc synthesis and its coordination with bacterial cell elongation. Peptidoglycan is already known to be synthesized at these positions [70] as is likely arabinan [52]. Moreover, mycolic acids are linked to arabinogalactan at these positions [68,69]. Present findings also show that FAS-II proteins, MmpL3 and Wag31 are present in the same region, at the same time in the bacteria; furthermore, it has recently been demonstrated that INH can affect the coupling of division and elongation [57]. Altogether, all the above data further place the pieces of the puzzle together and support the proposed model of localized envelope biogenesis machinery coordinated with the process of cell elongation and division. This coordinated process may ultimately represent the Achilles' heel of the pathogenic bacterium Mycobacterium tuberculosis.

Plasmid constructions
The four Mtb FAS-II genes (inhA, rv1484; mabA, rv1483; kasA, rv2245; kasB, rv2246) were isolated from the previously described pGAD-T7::fas-II and pGBK-T7::fas-II derivatives [17] as either Nde1-BamH1 (kasA and mabA), Nde1-EcoR1 (inhA), or Nde1-Msc1 fragments (kasB). They were cloned under the control of the acetamidase promoter (P amiE ) at the corresponding sites of the inducible vector pLAM12 [71] to produce the pLAM12::fas-II derivatives (Table S1). For the construction of the fluorescence control vectors, the egfp and the mCherry genes were amplified respectively from pEGFPN1 (Clontech) and from pmCherry-N1 (Clontech) using specific pairs of primers creating a Nde1 site at the ATG position of each gene and a second site (egfp, HindIII; mcherry, BamH1) at the end of each gene (Table S2). These genes were then cloned at the same sites in pLAM12 to yield pLAM12::gfp and pLAM12::mCherry (Table S1). For construction of N-terminal GFP fusions with the FAS-II proteins, the egfp gene was amplified from the pEGFPN1 vector using a second set of primers creating inphase Nde1 sites both at the ATG and STOP positions (Table S2). The egfp gene was inserted at the Nde1 sites of each of the four FAS-II genes in the pLAM12 derivatives. For construction of the Wag31 fusion, the wag31 gene (rv2145c) was amplified by PCR from Mtb H37Rv genomic DNA using specific primers (Table S2) creating a Nde1 site at the ATG position and cloned as an Nde1-BamH1 fragment under the control of the pHSP60 promoter of the replicative vector pMV261-Nde [17], a derivative of pMV261 [72]. Using the same strategy as above, the mCherry gene was cloned, as an intermediate, at the Nde1 site of pMV261::wag31 to yield a N-terminal fusion with Wag31. The fusion was further transported at an equivalent position of the integrative vector pMV361 [72]. The same wag31 Nde1-BamH1 DNA fragment was also cloned at the equivalent positions of the Matchmaker III yeast two-hybrid cloning vectors pGAD-T7 and pGBK-T7 (Clontech) to produce pGAD-T7::wag31 and pGBK-T7::wag31. In the same manner, the Mtb nat gene (rv3566) was cloned in the Matchmaker III vectors as Nde1-BamH1 fragments. Finally, the mmpL3 gene (rv0206c) and its N-terminal end containing only the first ten transmembrane helices (Mp3-N10) were cloned at the Nde1 site of pLAM12::gfp to construct the C-terminal MmpL3-GFP and MmpL3-N10-GFP fusion plasmids pLAM12::mmpL3::gfp and pLAM12::mp3-N10::gfp. All constructs were verified by restriction analysis and sequencing (Millegen).

Analysis of GFP-fusion expression in vivo
Msm liquid cultures were induced with 0.2% acetamide in early log phase (OD 590nm = 0.5) and samples were collected every two hours. Fluorescence at 525 nm (excitation at 485 nm) and OD 590nm were measured in micro-titration plates (Optiplate-96 well) with a FLx800 fluorescence reader (Biotek Instruments). Acquisition of fluorescence in the sample (F(s)) was performed in triplicate on serial dilutions of each culture. The data, expressed in relative fluorescence units (RLU), were corrected for the variation in fluorescence of the medium (F(m)) and the background fluorescence of the Msm/pLAM12 control strain (F(c)). Fluorescence (F) was also reported to the OD of the sample (OD(s)) and of the control strain (OD(c)) using the following formula: F = [(F(s) -F(m))/OD(s))]/[(F(c) -F(m))/(OD(c)]. The means 6 SD (standard deviation) of the data were plotted against the time of induction using Graphpad Prism 5 software (GraphPad Software Inc.). The same culture samples were analyzed by Western blotting during the course of induction. Aliquots of known and equivalent OD were lysed, fractionated on 10% SDS-PAGE and appropriate amounts of total protein were blotted using mouse monoclonal anti-GFP (Roche Applied science ref: 11814460001) as primary antibodies. The blots were revealed by bioluminescence (ECL detection Kit, Millipore) and visualized by fluoroimaging.

In vitro Co-immunoprecipitation (Co-IP)
In vitro transcription/translation of the genes of interest with the TnT Quick Coupled Transcription/Translation System (Promega) was performed as described previously [18] using supercoiled DNA from pGAD-T7 or pGBK-T7 derivatives expressing either the FAS-II genes [16,17] or the wag31 and nat genes as substrates (Table S1). This system allowed generating L-[ 35 S]methionine-labeled proteins tagged with either the HA (hemagglutinin) epitope (pGAD-T7 derivatives) or the c-Myc epitope (pGBK-T7). The products were named h-proteins (HA-tagged) or c-proteins (c-Myc-tagged). For Co-IP experiments, the h-proteins and the c-proteins were mixed together with goat anti-mouse IgG magnetic beads (Dynabeads M-450) coated with monoclonal anti-HA antibodies (Sigma). Proteins ratios were adjusted to 1:1 as described by assessing their specific activities (phosphorimaging) and correcting for their differences in the number of methionine residues. The co-IP reactions were performed as previously described [18] and analyses performed by separation on SDS-PAGE (4-20%) and phosphorimaging.

Construction of the Mycobacterium smegmatis kasB knockout strain
The Msm kasB gene (MSMEG_4328) was disrupted by using the ''recombineering method'' developed by van Kessel and Hatfull [71]. The linear allelic exchange substrate (AES; Figure S5) was constructed using a three fragments fusion-PCR [73] with the indicated primers (Table S2). The AES was constituted, from 59 to 39, by the fusion of the 800 bps of DNA upstream from the initiator codon of kasB, followed by the Zeocin (InvivoGen) resistance gene (Sh ble,) from the pZeo vector [74] and ended by the 800 bps of DNA downstream from the stop codon of kasB. The substrate was subsequently introduced by electro-transformation into a recipient Msm strain already containing a hygromycinresistant derivative of the recombineering plasmid pJV53 encoding for phage recombinases [71]. The construction and the selection of the mutant were essentially performed as previously described [71]. The chromosomal structure of the mutant ( Figure S5) was verified by PCR with suitable primers (Table S2 and Figure. S5). After curing of the recombineering plasmid, the mutant strain was transformed by appropriate plasmid to produce the control strain (Msm-DkasB/pLAM12) and the complemented strains (Msm-DkasB/pLAM12::kasB and Msm DkasB/pLAM12::gfp-kasB)).

Mycolic acid analysis
MA was performed essentially as described [75]. Bacterial residues obtained after lipid extraction with organic solvents were saponified at 110uC for 3 h in a mixture of KOH (40%) and methoxyethanol (C 3 H 8 O 2 ) (1:7, v/v). After acidification, fatty acids were extracted with diethyl ether and methylated with an ethereal solution of diazomethane. The mycolate patterns of the strains were determined by analytical high performance thin-layer chromatography (HPTLC) on Silica gel 60 (Merck) using dichloromethane 100% as eluent and revealed with rhodamine B (Sigma). The different classes of mycolates were separated by preparative TLC (Macherey-Nagel Silica Gel 60), using the same eluent. The lengths of the various types of purified mycolates were determined by MALDI-TOF mass spectrometry as previously described. MALDI mass spectra were acquired in reflectron mode using an Applied Biosystems 4700 Analyzer equipped with a Nd:YAG laser (355 nm wavelength, ,500-ps pulse and 200Hz repetition rate). The shots (2500 total) were accumulated in positive ion mode and MS data were acquired using the instrument default calibration. Mycolate samples were dissolved in chloroform at a concentration of 1 mM, and were directly spotted onto the target plate as 0.5 ml droplets, followed by the addition of 0.5 ml matrix solution. Samples were allowed to crystallize at room temperature. The matrix used was 2,5dihydroxybenzoic acid (10 mg ml21) in CHCl 3 /CH 3 OH (1 : 1).

Microscopy and video-microscopy
Msm liquid cultures expressing either the GFP protein fusions alone, or the GFP fusions together with the mCherry-Wag31 fusion were induced in log phase (OD 590nm 0.5) with 0.2% acetamide during 6 hours. Bacteria were pelleted, concentrated (20X) in PBS, and spread (10 mL) on poly-L-lysine coated glass slides (PolyScience). Slides were mounted with Dako mounting medium and observed with a large field Leica DMIRB fluorescence microscope at 63X magnification (Leica HCX Plan Apo x63/1.40-Oil). GFP (488 nm/509 nm) was revealed with a GFP-ET filter cube (excitation: BP 470/40; dichroic mirror, 495; emission: LP 525/50) whereas mCherry fluorescence (587 nm/ 610 nm) was visualized with a N2.1 Leica filter cube (excitation: BP 515-560; dichroic mirror, 580; emission LP 590). Images were captured with a CoolSnap HQ2 CCD camera (1 pixel = 6.45 mM), acquired with MetaMorph 7.1.7 software and processed with ImageJ software (National Institutes of Health, Bethesda, MD). All different images were acquired with the same exposure time (200 ms). Image processing consisted in equivalent adjustments of brightness and contrast on complete images. Gamma and LUT (Look Up Table) values were not modified and were left as linear on each channel. For statistical analysis and quantification of the polar indices, multiple slides (n) of independent experiments were used. An average of 1500 bacteria (N; see text) were counted from a minimum of three different slides and a maximum of 16 slides (3,n,16). Student t-tests were performed on raw data using the GraphPad Prism 5.0 software. Polar indices were expressed as percentages of polar foci in the fluorescent population and presented as mean with standard deviation (SD). For precise analysis of colocalization, each channel (GFP and mCherry) on microscopy images from individual bacteria producing both types of fusion (greater than 100) were scanned using the ROI manager of ImageJ software (NIH-USA). Thereafter, the GFP and the mCherry fluorescence profiles of each bacterium were obtained with the Plot Profile plugin. The maximum fluorescence levels of each scan of a given strain were normalized to the highest fluorescence value in this strain and plotted against their position along the long axis of the bacterium. The total length of each bacterium was reported to 1.
In order to quantify the polar localization of MmpL3, images from wide field fluorescence microscopy of Msm expressing either MmpL3-GFP or MmpL3-N10-GFP were analyzed with ImageJ software (NIH-USA) using the Plot Profile function. On each individual bacterium, three plots along 0.7 mm width lines were obtained: the first inside the cytoplasm (Background Fluorescence, F BG ), the second along the long axis of the bacterium and passing through one pole (Polar Fluorescence, F PO ), and the third along the short axis of the bacterium and crossing the membranes (Membrane fluorescence, F MB ). On each bacterium, the maximum fluorescence (FMax) of each plot was collected. The analysis was performed on more than 100 individual bacteria. The polar localization fold ratio was calculated with the formula, [|F PO Max-F BG Max|]/[|F MB Max-F BG Max|], yielding a value of 1 when there was no polar localization and a value superior to 1 when the fluorescent protein was accumulated at the membrane pole. Polarization was expressed as mean 6 SEM (Standard Error to the Mean), with N represent here the number of bacteria analyzed.
Time-lapse video microscopy was performed using an agar pad method essentially as previously described [34]. The bacteria were grown to mid-log phase, and aliquots (400 mL) were applied onto the coverslip of 35 mm glass bottom dishes (MatTek). After 5 minutes, the liquid was removed by aspiration and the bacteria were covered by 7H9 medium (10% ADC) at 37uC containing 0.6% Noble agar (Sigma), along with the appropriate antibiotics and the inducer (acetamide, 0.02%). After 1 hour solidification, the plates were pre-incubated for three hours at 37uC. The fluorescence was detectable in control GFP-producing bacteria. This time was chosen as the starting point (t = 0) for all time lapse experiments. Thereafter, the plates were observed as above using the Leica DMIRB florescence microscope equipped with a 37uC humid chamber. GFP and mCherry fluorescence was visualized with a GFP/mCherry-ET filter cube (Chroma Technology; excitation, no selection; dichroic filter, SP490-BP570/40; emission BP522/44-BP633/54). Individual frames (100 ms exposure) were acquired at regular time intervals (from 10 min to 2 hours) with MetaMorph 7.1.7 software during 14 to 20 hours. The representative movies were generated using ImageJ software. The required channels were extracted from the time-lapse image stacks and combined as described in the supplemental movie legends, after which they were converted to AVI (Audio Video Interleave) format at 0.5 to 1.5 frames per second using JPEG compression. Minimal image processing was performed as described above for static images. Figure S1 The mycolic acid biosynthesis pathway. The substrates and products of the Fatty Acid Synthase of type I (FAS-I) and II (FAS-II) are indicated together with the biochemical reactions (black arrows) necessary to achieve the biosynthesis of mature mycolic acid (see [8] and [9] for review). The enzymes responsible for each reaction are identified in colored circles. The cofactors necessary for reduction reactions to occur (NADPH, NADH) were omitted for reasons of clarity. The proximal and distal positions of the meromycolic chain modifications by MA-Mtfs are indicated as P and D respectively. Intermediate reactions not detailed (acyl-chains activation) are symbolized by interrupted arrows. The synthesis of the meromycolic chain is initiated by the condensation by the mtFabH protein of the acyl-CoA products of FAS-I with malonyl-ACP by MtFabD to give a keto-acyl-ACP. After reduction by the keto-acyl-ACP reductase MabA, followed by dehydration by the hydroxyl-acyl-dehydratases HadAB or HadBC and reduction by the enoyl-ACP reductase InhA, the acylchain enters into a new cycle of elongation through condensation by the keto synthase KasA or KasB with a new malonyl-ACP unit. After its synthesis, the meromycolic chain is adenylated and ligated by FadD32 onto Pks13, which is the terminal condensing enzyme that links the meromycolic chain to a carboxylated alpha chain produced by FAS-I. The remaining keto function of the generated mycolic motif is then reduced by CmrA to yield the mycolic acid.  Figure   S3 Localization of mCherry-Wag31 in Msm::mcherry-wag31. Enlarged images of wide-field microscopy experiments on mCherry-Wag31 fusion-expressing bacteria as presented in Figure 2 and Figure 4. Brightfield images (BF; leftmost column) together with mCherry fluorescence images (Che, middle column) of Msm mc 2 155::pMV361::mCherry (Ø) and Msm mc 2 155::pMV361::mCherry-wag31 (Wag31) were acquired 6 hours after dilution of the main culture and in the presence of anhydro-tetracyclin (50 ng/mL) for the latter. The merged images (right columns) allowed visualizing the localization of Che-Wag31 mainly at one pole of the bacteria (the old pole). Total optical magnification: 630 X. Scale bar: 5 mm.   Movie S1 Localization of GFP. Growth of Msm derivatives expressing GFP alone under the control of the p amiE promoter was followed by fluorescence time lapse video-microscopy. The movie (left panel) obtained by merging the bright field channel (in red) with the GFP channel (in green) is presented with the movie (right panel) displaying only the GFP channel. AVI file (0.8 frames per second) was obtained as described in Materials and Methods at 0.8 frames per second with one frame per hour. Time intervals are indicated in hours (h). Scale bar: 5 mm.

(AVI)
Movie S2 Localization of GFP-MabA. Growth of Msm derivatives expressing the GFP-MabA fusion under the control of the p amiE promoter was followed by fluorescence time lapse videomicroscopy. Movie treatment is identical to that of Movie S1 (AVI) Movie S3 Localization of GFP-InhA. Growth of Msm derivatives expressing the GFP-InhA fusion under the control of the p amiE promoter was followed by fluorescence time lapse videomicroscopy. Movie treatment is identical to that of Movie S1 (AVI) Movie S4 Localization of GFP-KasA. Growth of Msm derivatives expressing the GFP-KasA fusion under the control of the p amiE promoter was followed by fluorescence time lapse videomicroscopy. Movie treatment is identical to that of Movie S1 except that the AVI file was obtained at 1.5 frames per second with one frame obtained every 15 min. (AVI) Movie S5 Localization of GFP-KasB. Growth of Msm derivatives expressing GFP-KasB alone under the control of the p amiE promoter was followed by fluorescence time lapse videomicroscopy. Movie treatment is identical to that of Movie S1 (AVI) Movie S6 Localization of mCherry-Wag31 and GFP. Growth of Msm::pMV361 mCherry-Wag31 derivatives expressing GFP alone under the control of the p amiE promoter was followed by fluorescence time lapse video-microscopy. The movie (left panel) obtained by merging together the bright field channel (in blue), the GFP channel (in green) and the mCherry channel (in red) is presented with the movie displaying only the merged GFP and the mCherry channels (right panel). The AVI file was obtained as described in Materials and Methods at 0.5 frames per second. With one frame obtained every hour. The time intervals are indicated in hours (h). Scale bar: 5 mm.

(AVI)
Movie S7 Localization of mCherry-Wag31 and GFP-MabA. Growth of Msm::pMV361 mCherry-Wag31 derivatives expressing the GFP-MabA fusion under the control of the p amiE promoter was followed by fluorescence time lapse video-microscopy. Movie treatment is identical to that of Movie S6 except that the AVI file was obtained at 1.5 frames per second with one frame obtained every 10 min. (AVI) Movie S8 Localization of mCherry-Wag31 and GFP-InhA. Growth of Msm::pMV361 mCherry-Wag31 derivatives expressing the GFP-InhA fusion under the control of the p amiE promoter was followed by fluorescence time lapse video-micros-