Multiple deletions in the polyketide synthase gene repertoire of Mycobacterium tuberculosis reveal functional overlap of cell envelope lipids in host–pathogen interactions

Authors


Summary

Several specific lipids of the cell envelope are implicated in the pathogenesis of M. tuberculosis (Mtb), including phthiocerol dimycocerosates (DIM) that have clearly been identified as virulence factors. Others, such as trehalose-derived lipids, sulfolipids (SL), diacyltrehaloses (DAT) and polyacyltrehaloses (PAT), are believed to be essential for Mtb virulence, but the details of their role remain unclear. We therefore investigated the respective contribution of DIM, DAT/PAT and SL to tuberculosis by studying a collection of mutants, each with impaired production of one or several lipids. We confirmed that among those with a single lipid deficiency, only strains lacking DIM were affected in their replication in lungs and spleen of mice in comparison to the WT Mtb strain. We found also that the additional loss of DAT/PAT, and to a lesser extent of SL, increased the attenuated phenotype of the DIM-less mutant. Importantly, the loss of DAT/PAT and SL in a DIM-less background also affected Mtb growth in human monocyte-derived macrophages (hMDMs). Fluorescence microscopy revealed that mutants lacking DIM or DAT/PAT were localized in an acid compartment and that bafilomycin A1, an inhibitor of phagosome acidification, rescued the growth defect of these mutants. These findings provide evidence for DIM being dominant virulence factors that mask the functions of lipids of other families, notably DAT/PAT and to a lesser extent of SL, which we showed for the first time to contribute to Mtb virulence.

Introduction

Mycobacterium tuberculosis (Mtb), the causative agent of tuberculosis, is a pathogen that infects one third of the world's population and kills more than 1.4 million people annually (WHO, 2012). Its pathogenicity is mainly due to its capacity to manage the hostile environment encountered within the host in order to survive and replicate. Mtb has evolved mechanisms of resistance to overcome the defence functions of macrophages; these involve the use of receptors for cell invasion, the modulation of normal progression of the phagosome into an acid and hydrolytically active phagolysosome, the regulation of local modulators of the immune response and the control of cell apoptosis. Despite the scale of the public health problem, our understanding of the survival strategies of Mtb remains elusive. In particular, the factors important for tuberculosis pathogenesis and their modes of action are far from being identified.

At the interface between the host and the pathogen, the mycobacterial cell envelope, and especially the outermost layer, constitutes its interface with the host. It contains unique cell-surface lipids, whose biosynthesis involves a family of enzymes known as polyketide synthases. These lipids can be classified into distinct categories according to their importance for Mtb biology (Neyrolles and Guilhot, 2011). On the one hand, some of them (e.g. mycolates) are major constituents of the mycobacterial outer membrane and their synthesis is essential for bacterial growth. On the other hand, non-covalently associated lipids located in the outermost part of the cell envelope, including mycocerosate-containing lipids [phenolic glycolipids (PGL) and phthiocerol dimycocerosates (DIM)] and trehalose-derived glycolipids [sulfolipids (SL) and di- and poly-acyl trehaloses (DAT, PAT)], do not seem to have major structural roles. Indeed, the loss of DIM is frequent after serial passages of Mtb in vitro, and is associated with a slight increase in the growth rate (Domenech and Reed, 2009; Kirksey et al., 2011), and clinical isolates deficient in PGL, SL or PAT synthesis have been isolated (Constant et al., 2002; Gonzalo Asensio et al., 2006). Nevertheless, the distribution of DIM, PGL, DAT/PAT and SL in the Mycobacterium genus is consistent with these lipids contributing to pathogenesis: DIM and PGL family lipids are synthesized by a limited number of mycobacterial species, including the major human pathogens Mycobacterium leprae, Mycobacterium ulcerans and Mtb; DAT/PAT have only been isolated from species of the Mtb complex; and SL have been isolated from Mtb and Mycobacterium canettii, but not from other members of the Mtb complex. While the role of DIM and PGL in virulence is well documented (for review, see Guilhot and Daffe, 2008), that of SL and DAT/PAT is still unclear. At the individual level, their role appears to be questionable based on the lack of attenuation in animal models displayed by the genetically engineered strains (impaired for the production of SL or DAT/PAT) in comparison to the parental strains (Rousseau et al., 2003a,b). Also, their production differs between clinical isolates (our unpublished results). Moreover, SL and DAT/ PAT together do not seem to contribute to bacterial virulence because the absence of both, in a Mtb H37Rv double mutant, does not significantly affect the growth of tubercle bacillus in a mouse model of infection (Chesne-Seck et al., 2008). Nonetheless, these glycolipids, when purified, exhibit biological activities relevant to the interaction between Mtb and host cells. For instance, purified SL have been described as modulating phagosome-lysosome fusion (Goren et al., 1976; Brodin et al., 2010) and the cytokine response of human phagocytes (Brozna et al., 1991). Similarly, DAT/PAT are potent inhibitors of leucocyte migration (Husseini and Elberg, 1952) and T cell proliferation (Saavedra et al., 2001), and participate in early interactions between Mtb and phagocytes (Rousseau et al., 2003a). Also, most clinical isolates of the various Mtb lineages produce these molecules (our unpublished data) suggesting that they contribute to the adaptation of Mtb to its pathogenic way of life. Therefore, the true importance of SL and DAT/PAT in the pathogenesis of tuberculosis remains unclear.

The discrepancy between data obtained in vivo and data obtained ex vivo can be reconciled if Mtb possesses compensatory virulence factors that complement the absence of trehalose-derived lipids. An alternative explanation is that trehalose-derived lipids have a subtle activity that is masked by the large effects of major virulence factors. We therefore re-examined the individual and collective contributions of SL, DAT/PAT and DIM to the pathogenesis of tuberculosis. We developed a genetic strategy involving sequential disruption of the biosynthesis pathways for these various lipids. We investigated the consequences of these single and multiple mutations for the interaction between Mtb and both an animal model and human cells. We demonstrated that there is functional overlap between DIM and DAT/PAT and SL in the virulence of Mtb, both families of molecules contributing to counteract the immune responses of human macrophages.

Results

Construction and characterization of lipid-deficient mutants

The strategy used for sequential disruption of SL, DAT/PAT and DIM biosynthesis is depicted in Fig. 1A. A set of polyketide synthases (PKSs) is required for the formation of the various and specific fatty acids incorporated into SL, PAT/DAT and DIM (for review see Mohanty et al., 2011). This includes: Pks2, a catalyser for the formation of phthioceranic and hydroxyphthioceranic acids found in SL; Pks3/4, an enzyme required for the formation of both phthienoic acids (also called mycolipenic acids) and their structurally saturated and dienoic analogues (mycolipanoic or mycolipodienic acids) present in DAT and PAT; and PpsA-E and Mas, proteins required for the synthesis of the structural units (the phthiocerol chain and mycocerosic acids respectively) constituting DIM. We also included Pks5 in this deletion strategy. Pks5 is highly similar to Pks2, Pks3/4 and Mas, and was found to be involved in the formation of the methyl-branched fatty acid incorporated into the lipooligosaccharide (LOS) of mycobacteria (Etienne et al., 2009). Although this compound has not been found in strains of the Mtb complex, with the exception of M. canettii, we suspected that minor forms of DAT may contain these specific methylated fatty acids. Indeed, mutations in the genomic locus dedicated to LOS synthesis, and containing pks5, alter the synthesis of some DAT forms (Brodin et al., 2010).

Figure 1.

Construction and lipid profile of polyketide-derived lipid-deficient mutants.

A. Strategy for sequential disruption of the biosynthetic pathways of SL, PAT/DAT and DIM and construction of lipid-deficient mutants.

B. TLC analysis of the lipid content of H37Rv ATCC and H37Rv Pasteur WT strains and polyketide-derived lipid deficient mutants. Lipids were radiolabelled with [1-14C]-propionate. The TLC was run in petroleum ether/diethylether (90/10, v/v), chloroform/methanol (99/1, v/v) or chloroform/methanol/water (60/16/2, v/v/v).

To facilitate the sequential deletion strategy, we exploited an existing pair of H37Rv strains: the H37Rv ATCC strain, producing the three families of lipids analysed in this study; and the H37Rv Pasteur strain, containing a spontaneous single base substitution that introduces a stop codon into the pks3/4 gene (Cole, 1999). This mutation disrupts the polyketide synthase encoding gene pks3/4 and causes DAT/PAT deficiency (Fig. 1B). We used allelic exchange mutagenesis to delete an internal fragment from the targeted pks gene and insert a res-km-res cassette (see Experimental procedures for descriptions of the constructs). The antibiotic resistance marker was subsequently excised following expression of the γδ resolvase gene, leading to an unmarked mutation with a scar corresponding to one res sequence. Each mutant strain was tested by PCR using sets of primers designed to differentiate between strains with insertion of the disrupted allele by single cross over, illegitimate recombination or allelic exchange. A clone with the appropriate PCR profile was isolated and used for lipid analysis and subsequent gene disruption. The lipids produced by the various mutant strains were labelled with [1-14C] propionate and analysed by thin layer chromatography (Fig. 1B). The lipid profiles observed were consistent with the previously described functions of the various PKS: strains mutated for the pks3/4 gene were deficient in DAT/PAT synthesis; strains with a disruption of pks2 did not synthesize SL, and ppsE knock-out abolished DIM formation. Analysis of the H37Rv Pasteur-derived pks5 mutant failed to identify any lipid abnormality.

We evaluated the effect of cumulative lipid loss on the production of the remaining lipids of the outermost layer of the cell envelope. Exponentially growing strains producing either the three families of targeted lipids (DIM, SL, DAT/PAT), just two (DIM, SL) or none, were incubated with [1-14C] acetate or [1-14C] propionate for 24 h to label the total lipids or preferentially those containing methyl-branched fatty acids (notably DIM, SL and DAT/PAT). Lipids were then extracted and analysed by thin layer chromatography (Fig. S2). Spots corresponding to the major lipids of the outer layers of the mycobacterial cell envelope were quantified relative to total lipids and compared (see Fig. S2, table). No major change was observed in the lipid profile of strains deficient in the production of the different families of polyketide-derived lipids: there was no variation in the amount of DIM, SL, trehalose dimycolate (TDM) or trehalose monomycolate (TMM), associated with the loss of DAT/PAT (H37Rv ATCC versus Pasteur); nor in the amount of trehalose dimycolate (TDM) or trehalose monomycolate (TMM), in the strain deficient for DIM, SL and DAT/PAT synthesis (PMM127). However, the production of triacylglycerol (TAG) increased upon the loss of DIM, SL and DAT/PAT.

Furthermore, DIM-, DAT/PAT- and SL-deficient mutants were subjected to genome-wide transcriptomic analysis. As shown in Fig. 2A, only 17 genes were found to be differentially expressed in mutants deficient in either DIM (PMM135) or DIM and DAT/PAT (PMM174), and their wild-type (WT) counterpart (H37Rv ATCC); there was differential expression of only six genes for the pair PMM127 and WT H37Rv Pasteur. Impaired expression of eight genes, ppsE, drrA, drrB, drrC, pks3/4, mmpL8, papA1 and pks2 in the mutant strains was probably the direct result of the genetic disruption of ppsE, pks3/4 and pks2.

Figure 2.

Further characterization of lipid-deficient mutants.

A. Hierarchical clustering of expression profiles of differentially expressed genes in H37Rv ATCC, PMM135 (DIM-) and PMM174 (DIM-, DAT/PAT-) or H37Rv Pasteur (DAT/PAT-) and PMM127 (DIM-, DAT/PAT-, SL-). Red-blue display showing Mtb genes identified to be significantly and differentially expressed in the H37Rv ATCC, PMM135 and PMM174 or H37Rv Pasteur and PMM127 strains. Genes are ordered in rows, strains in columns (with four biological replicates per strain). Colouring indicates normalized expression values (Log2).

B. In vitro growth of H37Rv ATCC, H37Rv Pasteur (DAT/PAT-) and PMM127 (DIM-, DAT/PAT-, SL-) strains. Bacteria were grown in liquid Middlebrook 7H9 broth and bacterial growth was monitored by measuring the McFarland turbidity. Data are representative of two independent experiments.

Interestingly, two genes expressed at higher level in PMM135 and PMM174 mutants than in the WT strain, namely Rv1130 and Rv1131, encode homologues to the methylcitrate synthase (PrpC) and the methylcitrate dehydratase (PrpD), respectively, in other mycobacteria including M. smegmatis and M. marinum (Upton and McKinney, 2007). These enzymes belong to the methylcitrate cycle and are involved in succinate formation from propionyl-CoA. This suggests that impaired utilization in the mutant strains of propionyl-CoA for the synthesis of the methyl-branched fatty acids present in DIM and DAT/PAT results in the induction of another pathway for propionyl-CoA metabolism, as previously suggested (Upton and McKinney, 2007; Rhee et al., 2011). Another interesting result is the upregulation of the triacylglycerol synthase-encoding gene, Rv1760 (Daniel et al., 2004) in PMM135 and PMM174. This may account for the increased production of TAG by these mutant strains (Fig. 1B). The expression of the mbtD gene, involved in synthesis of the iron-chelating polyketide mycobactin, was downregulated in the mutant strains with reference to WT Mtb. The mycobactin synthesis pathway shares common substrates with the methyl-branched fatty acid synthesis pathway. Metabolic cross-talk between these two pathways has not previously been documented, but our results suggest that perturbation of polyketide-derived lipid synthesis may influence the formation of other polyketide-derived products, such as mycobactin. Finally, nine of the genes differentially expressed in PMM135, PMM174 and PMM127 strains are of unknown function.

Overall, the results of these transcriptome comparisons suggest that the loss of several polyketide-derived lipids does not cause substantial metabolic perturbations, at least in our in vitro growth conditions. This observation was confirmed by the comparison of the growth kinetics of mutants and control strains in standard growth medium. As depicted on Fig. 2B, there were no major differences in the growth behaviours of WT strains and the most-delipidated mutant, PMM127.

These results indicate that DAT/PAT, SL and DIM are individually and collectively dispensable for growth of Mtb. In addition, their loss is not associated with any major remodelling of the lipid composition of the mycobacterial cell envelope or of the cell metabolism.

Polyketide-derived lipids collaborate to mediate the in vivo growth of Mtb

We next examined the effects of the loss of several polyketide-derived lipids on the virulence of Mtb. We compared the ability of the WT strains and lipid-deficient mutants (Table 1) to replicate and survive in lungs and spleen of BALB/c mice after intranasal infection. For H37Rv ATCC, the numbers of colony-forming units (cfu) increased to 4–5 logs both in lungs and spleen on day 28 post infection (Fig. 3A and C). The bacillary load then stabilized in both organs, and the cfu counts were maintained until the end of the experiment on day 56. The three mutant strains deficient for DIM synthesis (Table 1: PMM135, PMM56 and PMM127) grew less well than the DIM-containing controls (Table 1: H37Rv ATCC, H37Rv Pasteur and PMM57) in both lungs (Fig. 3B versus A) and spleen (Fig. 3D versus C). This is consistent with the role reported for DIM in the replication of Mtb in vivo (Camacho et al., 1999; Cox et al., 1999). In contrast, loss of DAT/PAT or DAT/PAT plus SL did not affect the capacity of Mtb to multiply and persist; over a period of 56 days, the bacterial loads in both lungs and spleen of H37Rv ATCC, H37Rv Pasteur (DAT/PAT-) and PMM57 (SL-, DAT/PAT-) were comparable (Fig. 3A and C). These results agree with previous reports that deficiency in SL and/or DAT/PAT has no significant consequence for the virulence of Mtb (Rousseau et al., 2003a,b; Chesne-Seck et al., 2008). However, when we tested the effect of the loss of DAT/PAT and SL combined with DIM deficiency, we observed a synergistic effect on the multiplication (Fig. 3B and D). Mice infected with PMM56 (DIM-, DAT/PAT-) had significantly lower bacterial loads in their lungs than those infected with strain PMM135 (DIM-) (Fig. 3B). After 28 days of infection, the loads were about 50-fold lower for PMM56 than for PMM135. The DAT/PAT deficiency also further decreased the capacity of the DIM-less strain to persist in the spleen although the difference did not reach statistical significance (Fig. 3D). Similarly, the cfu counts in lungs of PMM127 (DIM-, SL-, DAT/PAT-) infected mice were significantly lower than those from PMM135 (DIM-) infected mice on days 14, 28 and 56 post infection (Fig. 3B). The differences between cfu counts for PMM56 (DIM-, DAT/PAT-) and PMM127 (DIM-, SL-, DAT/PAT-) were significant only at day 56; this may have been due to the small difference in the sizes of the inocula (Fig. 3B). To confirm the effects of DAT/PAT deficiency in strains unable to synthesize DIM, we independently constructed a second strain (PMM174) containing mutations in pks3/4 and ppsE (Fig. 1A). We then compared the growth kinetics of PMM135 (DIM-) and PMM174 (DIM-, DAT/PAT-) in mice (Fig. 4). The cfu counts in both lungs and spleen were lower for the double DIM and DAT/PAT deficient mutant than the single DIM deficient mutant on day 14 and at subsequent time-points, confirming the findings for PMM135/PMM56 pair. Transformation of PMM174 with a plasmid carrying a functional pks3/4 gene (pWM228H) and resuming the production of DAT/PAT (Fig. 1B), restored growth in lungs and spleen of infected mice to a level comparable to that of PMM135 (Fig. 4).

Figure 3.

Polyketide-derived lipid deficiency decreases Mtb growth in BALB/c mice. A collection of single and multiple lipid-deficient mutants and the WT H37Rv were used to infect mice and data are presented as: (A, C) DIM-containing strains (H37Rv ATCC, H37Rv Pasteur, PMM57) or (B, D) DIM-less mutants (PMM135, PMM56, PMM127). Numbers of cfu in (A, B) lungs and (C, D) spleen were determined on days 1, 14, 28 and 56 by plating dilutions of homogenized tissues on Middlebrook 7H11 agar containing OADC. The dashed line corresponds to the detection limit. When counts in infected mice were below the detection limit, the number of cfu scored was 50 cfu per organ (1.7 log). Values are means ± SEM of cfu counts for four infected mice. The significance of differences between strains was evaluated: *P < 0.05, **P < 0.01, ***P < 0.005.

Figure 4.

DAT/PAT are required for Mtb growth in BALB/c mice. Mice were infected with either PMM135 (DIM-) (circles), PMM174 (DIM-, DAT/PAT-) (open squares) or the complemented strain PMM174 228H (DIM-) (solid squares). The numbers of cfu in (A) lungs and (B) spleen were determined on days 1, 14, 28 and 56 by plating dilutions of homogenized tissues on Middlebrook 7H11 agar containing OADC. The dashed line corresponds to the detection limit. When counts in infected mice were below the detection limit, the number of cfu scored was 50 cfu per organ (1.7 log). Values are means ± SEM of cfu counts for three or four infected mice. The significance of differences between strains was evaluated: *P < 0.05, **P < 0.01, ***P < 0.005.

Table 1. A collection of single and multiple lipid-deficient mutants and WT H37Rv strains
 DIMDAT/PATSL
H37Rv ATCC+++
H37Rv Pasteur++
PMM57+
PMM135++
PMM174 228H++
PMM56+
PMM174+
PMM127

These findings indicate that DIM are dominant virulence factors, which mask the functions of other families of extractable lipids in the cell envelope. This makes it difficult to analyse the roles of these lipids in vivo. While the construction of lipid-deficient mutants in a DIM-less background allowed us to demonstrate that DAT/PAT are also important for the pathogenesis of tuberculosis by contributing to the growth of bacilli in vivo, they also show that SL play a minimal role, if not any.

The loss of polyketide-derived lipids has no major impact on the functional integrity of the cell wall of Mtb

One of the factors that contribute to the success of Mtb as a pathogen is its ability to resist the bactericidal defences of the host. This is in part due to the structural properties of the cell envelope that forms a low permeability barrier, preventing the penetration of toxic metabolites. We previously reported that DIM deficiency affects cell wall function (Camacho et al., 2001), so we examined whether the most delipidated mutant, PMM127, is affected in its sensitivity to toxic compounds. When exposed to increasing concentrations of hydrogen peroxide (H2O2), PMM127 (DIM-, SL-, DAT/PAT-) as the WT H37Rv ATCC were killed in a dose-dependent manner between 8 and 24 mM H2O2, and there was no obvious difference between the relative sensitivities of the two strains (Fig. 5A). In addition, H37Rv ATCC and PMM127 (DIM-, SL-, DAT/PAT-) grew similarly when exposed to various pH in vitro (Fig. 5B), indicating that the lipid deficiency did not affect the acid tolerance of Mtb. H37Rv Pasteur and mutants defective for the production of a more limited set of lipids were similarly tested and results were comparable with those for H37Rv ATCC and PMM127 (data not shown).

Figure 5.

Polyketide-derived lipid deficiency does not substantially affect the functional integrity of the cell wall of Mtb or its sensitivity to propionate. H37Rv ATCC and lipid-deficient mutants, including PMM127 (DIM-, DAT/PAT-, SL-), were grown in liquid Middlebrook 7H9 medium. The toxicities of (A) H2O2, (B) pH, (C) ceftriaxone and (E) sodium propionate were evaluated at various doses and times of incubation, by the MTT assay (A, C) or by measuring McFarland turbidity (B, E). Data from two independent experiments are reported as means ± SEM of duplicate or triplicate samples. *P < 0.05, **P < 0.01. (D) Exponentially growing H37Rv ATCC and PMM127 were fixed and probed with non-relevant Ab or Ab directed against α-glucan, PIM or ESAT 6. Bacteria were examined using a JEOL 1200 EX transmission electron microscope. The results are expressed as the number of gold beads per bacterial surface and values are means ± SEM for ten bacteria on two grids. Images were acquired using a digital camera. Bar: 100 nm.

We also assessed whether polyketide-derived lipid loss impaired the bacterial susceptibility to the antibiotic, ceftriaxone, a β-lactam that targets enzymes involved in peptidoglycan assembly and located in the periplasm (Chambers et al., 1995). Ceftriaxone killed both H37Rv ATCC and PMM127 (DIM-, SL-, DAT/PAT-) in a dose-dependent manner, but the mutant exhibited a greater susceptibility to the antibiotic (Fig. 5C). A threefold higher concentration of antibiotic was required for a similar decrease in cell viability of strains containing DIM than for DIM-less mutants. In contrast, no difference in ceftriaxone sensitivity was specifically associated with absence of SL and DAT/PAT irrespective of the DIM context (Fig. 5C). Therefore, the absence of DIM increased the sensitivity to a β-lactam antibiotic, consistent with a previous report indicating that DIM contribute to the permeability barrier formed by the cell envelope (Camacho et al., 2001).

The outermost layer of the mycobacterial cell envelope is a capsular layer that contains polysaccharides, proteins and small amounts of lipids, some of which contribute to pathogenesis (for review, see Forrellad et al., 2013). The suggestion that the absence of DAT/PAT affects the capsule attachment (Dubey et al., 2002) prompted us to examine whether the absence of polyketide-derived lipids affected exposure at the bacterial surface of capsular molecules contributing to mycobacterial pathogenicity. We examined the surface expression of phosphatidyl-myo-inositol mannosides (PIM), α-glucan and the ESX-secreted protein (ESAT-6), which are all involved in pathogenesis (Forrellad et al., 2013), by labelling intact bacteria with antisera and examination by immuno-gold-EM (Sani et al., 2010). The bacterial surfaces of H37Rv ATCC and PMM127 (DIM-, DAT/PAT-, SL-) were labelled homogeneously with the monoclonal antibody specific for α-glucan (Geurtsen et al., 2009), and there was no quantitative difference in labelling between the two strains (Fig. 5D). Likewise, the lipid deficiency did not affect labelling by antibodies directed against an epitope of the glycolipid PIM6 or the protein ESAT-6 (Fig. 5D). Importantly, the presence of the detergent Tween-80 in the bacterial culture medium, previously shown to remove the mycobacterial capsule (Sani et al., 2010), decreased cell surface labelling by these antibodies (data not shown). Therefore, the absence of DIM or SL or DAT/PAT did not significantly affect PIM, α-glucan, and ESAT-6 exposure on the mycobacterial cell surface. In all likelihood, these lipid deficiencies do not have major effects on the structure of the cell envelope.

Given that Mtb can adapt to metabolic stresses associated with the accumulation of potentially toxic intermediates of propionate by incorporating them into methyl-branched lipids of the cell wall (Lee et al., 2013), we next examined whether the lipid-defective mutant PMM127 grew on propionate as the sole carbon source. Increasing the propionate concentration in the medium from 10 to 20 mM severely impaired the growth of H37Rv ATCC, as expected, and also of PMM127 (DIM-, SL- and DAT/PAT-) (Fig. 5E). However, no difference in sensitivity to propionate between the two strains was detected (Fig. 5E).

Taken together, these data argue against a model in which the in vivo attenuation of these mutants, deficient in lipids and notably in trehalose-derived lipids, is due to substantial abnormalities in the cell envelope structure or to greater sensitivity to propionate.

Polyketide-derived lipids contribute to the outcome of Mtb in human macrophages

The attenuation in vivo of the mutants deficient for multiple polyketide-derived lipid synthesis may reflect an altered interaction with human macrophages (hMDMs), the main target cell of Mtb. We therefore evaluated the consequences of polyketide-derived lipid deficiency on the capacity of Mtb to infect hMDMs. As expected, the percentage of infected cells after 1 h with the DIM-less mutant PMM56 [multiplicity of infection (moi) 10:1] was 25% lower than the WT strain H37Rv Pasteur (33 ± 5 versus 43 ± 4 %, P < 0.03, n = 3) (Astarie-Dequeker et al., 2009). By contrast, there was no difference between the three mutants PMM135 (DIM-), PMM56 (DIM-, DAT/PAT-) and PMM127 (DIM-, DAT/PAT-, SL-) in the ability to infect hMDMs, suggesting that SL and DAT/PAT do not contribute to macrophage infection (data not shown).

We then examined the consequences of lipid deficiency on the capacity of Mtb to grow within hMDMs. We previously reported that, at an moi of 10:1, there is no difference in bacterial loads between WT strain and DIM-less mutant after 96 h of infection (Astarie-Dequeker et al., 2009). A recent study shows that hMDMs restrict Mtb growth more effectively at an moi of 1:1 than 10:1 (Welin et al., 2011). For this reason, we revisited the role for DIM using a lower moi of 2:1, and followed the intracellular bacterial growth for 24 h and 168 h by counting the number of fluorescent bacteria per cell. Under these conditions, the bacterial count per macrophage at 168 h was significantly lower for DIM-less mutants (PMM135 and PMM56) than WT controls (H37Rv ATCC and H37Rv Pasteur) (Fig. 6A). Accordingly, we then tested the intracellular load for the PMM135 (DIM-), PMM56 (DIM-, DAT/PAT-) and PMM127 (DIM-, DAT/PAT-, SL-) mutant strains over time. Interestingly, while the three mutants grew at a similar rate for the first 48 h, the PMM56 (DIM-, DAT/PAT-) and PMM127 (DIM-, DAT/PAT-, SL-) replicated more slowly than PMM135 (DIM-) at a later time point (Fig. 6B). Of note, the replication of PMM127 (DIM-, SL-, DAT/PAT-) in macrophages was consistently slightly slower, albeit non-significantly, than that of PMM56 (DIM-, DAT/PAT-) (Fig. 6B). To confirm the phenotype associated with DAT/PAT deficiency, we performed additional experiments with PMM174 (DIM-, DAT/PAT-) and the pks3/4-complemented strain called PMM174 228H. Similar to PMM56 (DIM-, DAT/PAT-), the growth of PMM174 was slower than that of PMM135 (DIM-) (Fig. 6C), and complementation with a functional copy of pks3/4 restored both the production of DAT/PAT (Fig. 1B) and the intracellular growth capacity approaching that of PMM135 (Fig. 6C). The loss of DIM (Fig. 6D) and DAT/PAT (Fig. 6E–F) decreased simultaneously the percentage of infected macrophages at 168 or 144 h, strongly suggesting that lipid deficiency affects also the capacity of Mtb to propagate infection into new macrophages. Of note, H37Rv ATCC and the PMM135 and PMM174 mutants induced a weak cell death, which averaged 2–4% of total macrophages (Fig. S3B), as evaluated by using 7-actinomycin D (7-AAD) to assess cell membrane permeability (Fig. S3A). However, when data are expressed as the percentage of 7-AAD-positive cells among GFP-positive macrophages (i.e. infected cells), H37Rv ATCC induced significantly higher cell death than PMM135 and PMM174 (Fig. S3C). This makes it likely that the observed differences in percentage of infected macrophages were due to variability in cell death.

Figure 6.

The loss of polyketide-derived lipids impairs Mtb growth in human macrophages. hMDMs were put in contact for 1 h with (A, D) H37Rv ATCC or H37Rv Pasteur and the corresponding DIM-less mutants, (B, E) PMM135 (DIM-), PMM56 (DIM-, DAT/PAT-) and PMM127 (DIM-, DAT/PAT-, SL-), or (C, F) PMM135, PMM174 (DIM-, DAT/PAT-) and the complemented strain PMM174 228H (DIM-) at an moi of 2:1. Once infected, the cells were washed and further incubated in the presence of serum. At various times thereafter, cells were fixed and processed to quantify the number of bacteria per macrophage (A–C) and the percentage of infected cells (D–F). For each set of conditions, duplicate experiments were performed, and at least 100 cells were counted per slide. The values reported are means ± SEM of three to eight independent experiments. The significance of differences between strains was evaluated. *P < 0.05, **P < 0.01, ***P < 0.005.

This set of data demonstrates that DIM and DAT/PAT are important for the outcome of Mtb in hMDMs. A similar trend was observed for SL but the effect was very modest and not statistically significant. However, this observation is consistent with the slight attenuation observed in mice for the PMM127 (DIM-, SL-, DAT/PAT-) mutant relative to PMM56 (DIM-, DAT/PAT-) (Fig. 3B).

Polyketide-derived lipids are required to drive Mtb in a poorly acidic phagosome

DIM contribute to the control of phagosomal pH (Fig. S4) consistent with our previous report (Astarie-Dequeker et al., 2009) and with a mutant library screening approach, which previously identified mutations in DIM biosynthesis genes as being associated with failure to block phagosomal acidification (Pethe et al., 2004; Stewart et al., 2005; Brodin et al., 2010). Interestingly, none of these genetic screens identified mutations in DAT/PAT or SL biosynthetic genes. However, Brodin et al. reported that mutations affecting the balance of SL and DAT synthesis altered the capacity of Mtb to remodel its phagosome (Brodin et al., 2010). To determine if the advantage conferred by polyketide-derived lipids to bacterial growth is associated with control of phagosomal pH, hMDMs were infected with our single and multiple mutants. The numbers of bacilli colocalizing with the acidotropic dye, LysoTracker, were counted. A large proportion of phagosomes containing PMM135 (DIM-) colocalized with LysoTracker during the early stage of infection (Fig. 7A), consistent with previous finding (Astarie-Dequeker et al., 2009); this proportion then decreased gradually over the following 144 h. In comparison, the two double DIM and DAT/PAT-deficient mutants, PMM56 and PMM174, accumulated in phagosomes that were significantly more frequently labelled with LysoTracker (Fig. 7A and B), and this phenotype was lost when PMM174 was complemented with a functional pks3/4 gene (Fig. 7B). The triple lipid-deficient mutant PMM127 (DIM-, DAT/PAT-, SL-) also accumulated in phagosomes that were more acid than those containing PMM135 (DIM-) (Fig. 7A). A larger, but non-significant, proportion of PMM127 than PMM56 (DIM-, DAT/PAT-) was also found in LysoTracker-positive phagosomes.

Figure 7.

Polyketide-derived lipid deficiency induces acidification of Mtb-containing phagosomes. hMDMs were infected for 1 h with GFP-expressing DIM-less mutants deficient in one or more lipids at an moi of 2:1: (A) PMM135 (DIM-), PMM56 (DIM-, DAT/PAT-) and PMM127 (DIM-, DAT/PAT-, SL-) or (B–D) PMM135 (DIM-), PMM174 (DIM-, DAT/PAT-) and the complemented strain PMM174 228H (DIM-). At various times after infection, (A, B) cells were incubated with LysoTracker for 1 h, fixed and processed for microscopy or (C, D) were fixed, permeabilized, and immunostained with polyclonal anti-serum against H+-ATPase (C) or MAb against CD63 (D) and processed for fluorescence microscopy. (A) A representative micrograph of infected cells collected at 144 h and analysed by confocal microscopy. Bar: 11 μm. For each marker, 100 phagosomes from at least 10 different fields were counted in duplicate. The values reported are means ± SEM of three experiments. The significance of differences between strains was evaluated. *P < 0.05, **P < 0.01.

Infected hMDMs were immunostained for the vacuolar type proton-transporting ATPase (H+-ATPase), an enzyme contributing to phagosomal acidification (Lukacs et al., 1990), and for CD63, a marker of fusion with late endosomes/lysosomes (Astarie-Dequeker et al., 2002). The greater accumulation of PMM174 (DIM-, DAT/PAT-) in acidified phagosomes than PMM135 (DIM-) (Fig. 7B) correlated with more H+-ATPase in the phagosomal membrane (Fig. 7C); again, this phenotype was reversed in the complemented strain PMM174 228H (Fig. 7C). In contrast, PMM135 and PMM174 similarly colocalized with CD63 (Fig. 7D).

Our data strongly suggest that DAT/PAT share with DIM the capacity to prevent the recruitment of the H+-ATPase to the phagosomal membrane and thereby inhibit the acidification of the bacterial environment.

Polyketide-derived lipids contribute to the intracellular growth of Mtb by preventing phagosome acidification

We had established that mutants deficient in DIM or DAT/PAT synthesis exhibited a growth defect and a higher association with LysoTracker. We therefore investigated whether the absence of lipid-dependent manipulation of phagosomal pH was responsible for the intracellular growth defect. We tested if the prevention of phagosomal acidification by the H+-ATPase inhibitor bafilomycin A1 (Drose and Altendorf, 1997) rescued the growth defect associated with DIM or DAT/PAT deficiency. Pretreatment with bafilomycin A1 reduced the early acidification of phagosomes containing H37Rv Pasteur and PMM56 (DIM-), and abolished the difference between their percentage of lysotracker-positive phagosomes at 24 and 168 h post infection (Fig. 8A). Interestingly, the bacterial growth at 168 h was greater in the presence than absence of bafilomycine A1 for both strains; furthermore, the growth defect in untreated macrophages of PMM56 (DIM-) with reference to the DIM-producing strain H37Rv Pasteur was abolished by bafilomycine A1 (Fig. 8B). Similar results were obtained for PMM135 (DIM-) and PMM174 (DIM-, DAT/PAT-) (Fig. 8C and D).

Figure 8.

Inhibition of the vacuolar H+-ATPase by bafilomycin A1 prevented phagosomal acidification and rescued the growth defect associated with polyketide-derived lipid deficiency. hMDMs were pretreated for 1 h with 100 nM bafilomycin A1 or vehicle control (DMSO) and for the first 24 h after infection. Cell were then infected with GFP-expressing (A, B) H37Rv Pasteur and PMM56, or (C, D) H37Rv ATCC, PMM135 and PMM174, at an moi of 10:1 for 24 h infection or 2:1 for 168 h. Cells collected after 24 or 168 h were incubated, or not with LysoTracker for 1 h, fixed and processed to quantify (A, C) the number of LysoTracker-positive phagosomes, or (B, D) the number of bacteria per macrophage. 100 phagosomes from at least 10 different fields were counted in duplicate (A, C) and at least 100 cells were counted per slide (B, D). The values reported are means ± SEM of three to four experiments performed in duplicate. The significance of the effect was assessed by comparing data for lipid-deficient mutants to those for parental control strains: *P < 0.05, **P < 0.01, ***P < 0.005.

Therefore, early inhibition of H+-ATPase activity prevents acidification of phagosomes and rescues the growth defects of DIM and DAT/PAT-deficient mutants. This strongly suggests that DAT/PAT and DIM contribute to the intracellular growth of Mtb by provoking an exclusion of the H+-ATPase from the phagosome membrane, thereby providing a non-acidified compartment for bacterial proliferation.

Discussion

For years, the discrepancies between in vitro and in vivo data have long interfered with our understanding of the role of polyketide-derived lipids in Mtb pathogenicity. Indeed, the biological activities described for polyketide-derived lipids were consistent with their contribution to the interaction between Mtb and its host and to its virulence: for example, SL can modulate host cell responses, including phagosome–lysosome fusion and cytokine production (Goren et al., 1976; Pabst et al., 1988; Zhang et al., 1988; Brozna et al., 1991), and DAT/PAT can strongly inhibit the proliferation of T cells (Saavedra et al., 2001), and the production of pro-inflammatory cytokines (Lee et al., 2007). However, investigations in vivo showing no phenotype for Mtb mutants deficient in the synthesis of SL or DAT/PAT, or both (Converse et al., 2003; Rousseau et al., 2003a,b; Chesne-Seck et al., 2008) did not support a role in pathogenesis of Mtb. In this study, we established that: (i) polyketide-derived lipids from Mtb (including, SL, DAT/PAT and DIM) have overlapping functions during the interaction between the bacteria and the host, thereby explaining, at least in part, the absence of phenotype of SL and/or DAT/PAT deficient mutants; (ii) these molecules contribute to Mtb fitness during the acute phase of the infection, probably through phagosome remodelling and blockade of H+-ATPase recruitment; and (iii) the effects of DIM are dominant over those of SL and DAT/PAT. Therefore, our findings reconcile previously in vitro and in vivo data for these molecules and substantially advance our understanding of the role of polyketide-derived lipids as virulence factors of Mtb.

It is emerging that functional overlap, redundancy and buffering relationships are major features of biological processes where they may confer robustness to systems (Hartman et al., 2001). The importance of these phenomena has been illustrated in host/pathogen interactions (Kvitko et al., 2009; O'Connor et al., 2011; 2012). For instance, the intracellular bacterial pathogen, Legionella pneumophila, may tolerate deletion of 31% of its effectors, known to be translocated into host cell during infection, without being significantly affected for its capacity to invade and replicate within mouse macrophages (O'Connor et al., 2011). Functional overlap of this type may reinforce the robustness of pathogens and thereby allows them to deal with genetic variation of the putative host and to expand the repertoire of targetable host cells (Kvitko et al., 2009; O'Connor et al., 2012). Our results with Mtb establish that there is such functional overlap among polyketide-derived lipids exhibiting related (SL and DAT/PAT) or unrelated (DIM versus SL or DAT/PAT) structures. We found that all these compounds contribute to remodelling the Mtb-containing phagosome and to prevent its acidification. The lack of attenuation phenotype associated with the loss of SL or DAT/PAT in animal and cellular models was initially interpreted as indicating that these molecules had no role in pathogenicity or that their activity was restricted to human cells (Converse et al., 2003; Rousseau et al., 2003a,b; Chesne-Seck et al., 2008). However, we believe that the analysis of the roles of DAT/PAT and SL in virulence using lipid-deficient mutants was uninformative because of the presence of DIM, which exert a functionally dominant effect. Our study in mice confirmed that DIM-deficiency caused a reduction of about 2-log in bacterial titres in lungs and spleen; the loss of DAT/PAT and/or SL alone had no effect. However, the absence of DIM and both DAT/PAT and SL led to a more severely attenuated phenotype. The effect could be mostly attributed to DAT/PAT because, although SL deficiency tended to impair bacterial growth in DIM-less background in all cellular and animal models used, the differences were not statistically significant. These observations contradict those described recently by Gillmore et al., whose study reported that the loss of SL-1 in an stf0-deleted mutant slightly enhanced Mtb survival in human macrophages (Gilmore et al., 2012). The authors ascribe this effect to a decrease in mutant sensitivity to a human cationic antimicrobial peptide (Gilmore et al., 2012). At this point, we do not have an explanation for the discrepancy between these results.

Cell envelope lipids could contribute to the pathogenesis of Mtb in several ways: they may have a structural role in the bacterial cell envelope; a ‘sink’ activity important for sequestrating toxic compounds produced during intracellular growth; or induce modulatory effects of the host immunity. These various possibilities are not mutually exclusive. We confirm here that DIM have a role in the permeability of the envelope (Camacho et al., 2001), and DIM are also partners in the interaction with the host (Astarie-Dequeker et al., 2009). A pks3/4 mutant has been reported to have a severely affected morphology (Dubey et al., 2002) and to be more efficient than the WT strain for infecting phagocytic and non-phagocytic cells (Rousseau et al., 2003a). This implicates trehalose-containing lipids in maintaining the integrity of the cell envelope structure. However, our data do not support this conclusion and provide compelling evidence that deficiency in trehalose-containing lipids does not cause major modifications of the cell envelope structure. Indeed, mutants defective in SL or DAT/PAT alone or both in combination displayed the same sensitivity to toxic compounds including H2O2 and the antibiotic, ceftriaxone. The absence of DIM increased the sensitivity to ceftriaxone, but additional removal of DAT/PAT and SL had no additional effect. The absence of DIM, DAT/PAT and SL did not affect either the arrangement of components exposed on the cell envelope surface, such as PIM6, ESAT-6 and α-glucan, all involved in Mtb–host interactions (Forrellad et al., 2013).

Methyl-branched fatty acid anabolism was proposed to relieve the toxicity of propionate, which accumulates during growth on odd-chain fatty acids as a carbon source (Jain et al., 2007; Upton and McKinney, 2007). Our data strongly suggest that the alternative mechanisms for alleviating this metabolic pressure are sufficiently effective for the loss of methyl-branched fatty acid biosynthetic pathways to be tolerated. Indeed, the (DIM, DAT/PAT, SL)-deficient mutant and the control strain, H37Rv ATCC, grew similarly on various propionate concentrations. Consistent with this idea, our transcriptome analysis revealed that the genes prpC and prpD, encoding the first two specific steps of methylcitrate cycle, are upregulated in DIM and DIM, DAT/PAT mutants. Lee et al. demonstrated that shutting down the methylcitrate cycle makes Mtb dependent on either methyl-branched fatty acid synthesis or the methylmalonyl pathway for propionyl-CoA detoxification (Lee et al., 2013). Our results do not contradict this conclusion but indicate that the methylcitrate cycle may compensate for the loss of the main PKSs utilizing propionyl-CoA. These findings illustrate a buffering relationship between these pathways for a key process of Mtb biology.

Our data are consistent with a role of polyketide-derived lipids as effectors of Mtb virulence. We establish that these molecules are essential for Mtb to control the intra-phagosomal pH and as a consequence to survive and multiply in human macrophages. The absence of DIM drove bacteria into phagosomes that accumulated an acidotropic dye early in infection (Astarie-Dequeker et al., 2009) and impaired growth (present study). These phenotypes were exacerbated by the additional removal of DAT/PAT and SL. The difference in intracellular survival between lipid-deficient mutants and control strains cannot be attributed to different intrinsic sensitivities to the pH encountered in phagosomes because they were found to be equally acid-sensitive in vitro. Consistent with this, searches for Mtb mutants more sensitive to acidic pH failed to identify any mutation in DIM, SL or DAT/PAT biosynthetic pathways (Vandal et al., 2008). We suggest that polyketide-derived lipids participate in the intracellular growth of Mtb by blocking phagosome acidification. Indeed, blocking the vacuolar H+-ATPase with bafilomycin A1 prior to and throughout the first 24 h of infection, dramatically impairs the early and/or late acidification of phagosomes containing DIM- or DAT/PAT-less mutants and simultaneously rescued their intracellular growth defect.

How these lipids control the phagosomal pH remains to be established. Following the contact between Mtb and phagocytes, the biophysical properties of the host cell membrane change, probably because of the insertion of DIM into this structure (Astarie-Dequeker et al., 2009). We are currently investigating whether this mechanism is responsible for H+-ATPase exclusion and whether other polyketide-derived lipids, such as SL and DAT/PAT, act in the same way. Interestingly, our collection of multiple deletion mutants and our results provide the microbiological tools and conceptual framework for deciphering, in greater detail, the biological activities of the polyketide-derived lipids in the Mtb cell envelope.

Experimental procedures

Antibodies, fluorescent probes and reagents

The mouse antibody (MAb) anti-PIM6 and anti-capsule monoclonal antibodies were generous gifts from B. Appelmelk (Amsterdam, The Netherlands). MAb against ESAT-6 was from R. Brosch (Institut Pasteur, Paris, France). Rabbit polyclonal anti-serum against the human H+-ATPase proton pump and mouse monoclonal anti-body against human CD63 were purchased from Synaptic Systems (Göttingen, Germany) and Caltag Laboratories (Burlingame, USA) respectively. Anti-mycobacteria rabbit serum (1:50) was obtained as previously described (N'Diaye et al., 1998). Secondary antibodies and Lyso-Tracker Red DND-99 were purchased from Molecular Probes (Eugene, USA). Goat anti-MAb labelled with gold beads was purchased from Aurion (Wageningen, The Netherlands).

Except when specified in the text, other chemicals were obtained from Sigma Chemical (USA).

Bacterial strains and growth conditions

Mtb H37Rv ATCC (ATCC), Mtb H37Rv Pasteur (the sequenced strain from Institut Pasteur) and lipid-deficient mutants (Table S1 and Fig. 1A) were cultured at 37°C in liquid Middlebrook 7H9 medium (Invitrogen, Cergy-Pontoise, France) supplemented with ADC (Becton Dickinson, Sparks, USA) or on Middlebrook 7H11 agar containing OADC (ADC and 0.005% oleic acid) (Becton Dickinson). Kanamycin, hygromycin and Tween-80 were added as required at concentrations of 40 μg ml−1, 50 μg ml−1 and 0.05% (v/v) respectively.

Construction of M. tuberculosis mutant strains

The Mtb mutants were constructed either using the thermosensitive counterselectable (ts/sacB) plasmid pPR27 (Pelicic et al., 1997) or the thermosensitive bacteriophage phAE87 as described by Bardarov et al. (2002). The main features of the strains and plasmids used are summarized in Table S1. Briefly, an approximately 2 kb DNA fragment containing part of the targeted gene was amplified by PCR from genomic DNA using specific oligonucleotides (Table S2) and inserted into a cloning vector. An internal fragment was excised using restriction enzymes and substituted by the res-Ωkm-res cassette (Malaga et al., 2003), generating an allelic exchange substrate (AES) composed of the Km cassette flanked by two arms of approximately 1 kb specific of the targeted gene. This AES was inserted either into the ts/sacB plasmid pPR27 or into the bacteriophage phAE87. The resulting plasmid or bacteriophage was transferred into the recipient M. tuberculosis strain, and allelic exchange mutants were selected on 7H11 agar plates as described previously (Pelicic et al., 1997; Bardarov et al., 2002).

Clones were analysed with primers res1 and res2, specific for the Km cassette, and primers located within the small internal DNA fragment substituted by the Km cassette in the AES and oligonucleotides hybridizing to regions located outside the AES. An example of the mutation strategy and genetic characterization of the mutants is provided in Fig. S1. One clone giving the pattern corresponding to allelic exchange was isolated and retained for further analysis.

The res-Ωkm-res cassette was recovered by transferring pWM19, containing the resolvase gene of transposon γδ, into the Km-resistant mutant and by screening for Kms clones among transformants (as described in Malaga et al., 2003). Kms clones were tested by PCR, as illustrated in Fig. S1, and for each construct, one clone displaying a PCR profile consistent with excision of the res-Ωkm-res cassette was isolated and retained for further analysis.

Complementation of M. tuberculosispks3/4 mutants

The pks3/4 gene is organized as an operon with the downstream genes papA3 and mmpL10. Therefore, we cloned these three genes to complement the pks3/4 gene disruption. A 11 kb DNA region covering these three genes was assembled from four smaller DNA fragments obtained by PCR amplification (using primers 3A-3B, 3C-3D, 3E-10B and 10A-10C) with H37Rv ATCC genomic DNA as the template. This piece of DNA was inserted into a derivative of pMV361, from which the original phsp60 promoter had been deleted (Stover et al., 1991), and a hygromycin resistance gene was inserted. The resulting plasmid was named pWM228H.

Lipid extraction and analysis

Each Mtb strain was grown to exponential growth phase in 10 ml 7H9 liquid medium supplemented with ADC and 0.05% Tween, and labelled by incubation with 0.4 μCi ml−1 [1-14C] propionate (specific activity of 54 Ci mol−1) or 0.4 μCi ml−1 [1-14C] acetate (specific activity of 56 Ci mol−1) for 24 h. Lipids were extracted as described elsewhere (Constant et al., 2002). The lipid profiles were compared by spotting equivalent amounts of crude extracts (resuspended in CHCl3 at a final concentration of 20 mg ml−1) on TLC plates, which were then run in various solvent systems [petroleum ether/diethyl ether 9:1 (v/v) for DIM; CHCl3/CH3OH/H2O 60:16:2 (v/v/v) for DAT and SL; CHCl3/CH3OH 99:1 (v/v) for PAT]. Labelled lipids were visualized and analysed with a Typhoon PhosphorImager (Amersham Biosciences). Counts per minute corresponding to each lipid spot were quantified and divided by the total amount of radioactivity in the corresponding lane to give the relative amount of the analysed lipid.

Microarray hybridization and data analysis

Mycobacteria were grown to mid-log phase in liquid Middlebrook 7H9-ADC medium containing 0.02% Tween-80. Samples were centrifuged for 5 min at 2500 g, and the bacteria resuspended in lysis buffer (RNeasy mini kit, Qiagen) containing 10% β-mercaptoethanol, and lysed using a bead beater. Total RNA was extracted using the RNeasy mini kit, and DNA contaminants were removed using DNase 1 (Ambion, Austin, TX, USA). The amount and purity of RNA was measured using a NanoDrop ND-1000 (Thermo Scientific); RNA integrity was assessed using a 2100 BioAnalyzer (Agilent Technologies, Les Ulis, France) and the RNA Nano 6000 kit (Agilent Technologies). Only samples showing RIN-values ≥ 9 were used.

Double-stranded cDNA was synthesized from 2 μg of total RNA using random primers and the cDNA SuperScript One-Cycle cDNA synthesis kit (Invitrogen), according to the manufacturer's protocol. cDNA was labelled with Cy3 using Cy3-dCTP and the One-Color DNA labelling kit (Roche NimbleGen, Madison, WI, USA).

Gene expression was analysed using a 12 × 135K custom-designed microarray (Roche NimbleGen) covering 3951 open reading frames of Mtb H37Rv with six 60-mer oligonucleotide probes per gene. Each hybridization mixture contained 2 μg Cy3-cDNA and the hybridization cocktail of the Hybridization kit (Roche NimbleGen). Samples were incubated for 5 min at 65°C and 5 min at 42°C prior to loading. Hybridization was performed for 17 h at 42°C using the Roche NimbleGen Hybridization System 4. Microarrays were washed according to the Roche NimbleGen Arrays User's Guide and scanned using an MS200 microarray scanner (Roche NimbleGen). The resulting xys files, which were extracted using Deva Software (Roche NimbleGen), were imported into R (Bioconductor). The microarray datasets were corrected for background and normalized by summarizing the intensity values of the probes in a probe set using the robust multi-array average (RMA) algorithm. Genes were considered to be significantly differentially expressed if they displayed at least a 1.5 Log2-fold difference in expression ratio between conditions and an adjusted P-value of ≤ 0.05.

Immunogold-electron microscopy

Bacteria were grown and fixed in paraformaldehyde (PFA) and glutaraldehyde (Delta Microscopies, Ayguevives, France) as described previously (Sani et al., 2010), and incubated for 15 min with nickel coated formvar/carbon grids (Delta Microscopies). Grids were then floated on drops of PBS to wash off the overflow, and aldehyde groups were neutralized by incubating twice for 15 min with NH4Cl. Grids were then blocked for 15 min with PBS containing 1% bovine serum albumin (BSA) (Euromedex, Souffelweyersheim, France), and incubated first with the primary MAb directed against PIM6 (1:150), capsular α-glucan (1:5) or ESAT6 (1:500) and next with the secondary goat anti-MAb labelled with gold beads (1/50). Grids were examined with a JEOL 1200 EX transmission electron microscope (JEOL, Peabody, MA, USA) at 80 kV and the number of gold beads per bacterial surface was determined. Images were acquired using a digital camera (AMT, Woburn, MA, USA) at 40–300 000× magnification.

Bacterial sensitivity to oxygen radicals and ceftriaxone

Bacterial sensitivity was assessed using the dimethylthiazolyl diphenyl tetrazolium bromide (MTT) assay as described previously (Mosmann, 1983). A suspension of exponentially growing bacteria was adjusted to a final optical density at 600 nm (OD600) of 0.05 and aliquot was added to wells in microtitre plates and incubated for 6 days at 37°C each of a series of concentrations of hydrogen peroxide (up to 24 mM) or ceftriaxone (up to 32 μg ml−1). Triplicate wells were used for each experimental condition. MTT (Sigma, St Louis, MO) was then added to each well and incubation continued at 37°C for 24 h. A lysis buffer containing 20% sodium dodecyl sulfate in 50% N,N-dimethylformamide (pH 4.7) was then added, and the plates were incubated overnight. Cell viability was determined by measuring absorbance at a wavelength of 600 nm using a microplate reader (Asys Expert Plus, Biochrom, Cambridge, UK). Results are reported as percentages of viable cells for treated and untreated bacteria.

Bacterial growth assays under various conditions of pH or various propionate concentrations

For each assay, bacteria were grown in 7H9 liquid medium supplemented with ADC to an OD600 of 0.5. To test the effect of pH, the cultures were diluted 1 in 10 in 5 ml 7H9 medium containing ADC, previously adjusted to a pH in the range 6.5 to 4.5 with 37% hydrochloric acid, and incubated at 37°C for up to 4 days. To test for propionate toxicity, cultures were diluted (1 in 100) in 5 ml 7H9 medium supplemented with 0.5% albumin, 0.085% NaCl, 0.05% Tween 80, and 10 mM or 20 mM sodium propionate as the carbon source, and incubated at 37°C for up to 15 days. Bacterial growth was monitored by measuring the McFarland turbidity. Duplicate flasks were used for each experimental condition.

Mouse infection

Animal studies were conducted following the CNRS guidelines for housing and care of laboratory animals. All protocols were reviewed and approved by the relevant ethics committee (Comité d'Ethique Midi-Pyrénées; reference MP/10/12/02/11). Female BALB/c mice, 7 to 8 weeks old, were purchased from Centre d'Elevage Janvier (Le Genest St Isle, France) and housed in the IPBS ASB3 animal facilities. Each mouse was infected with approximately 2 × 102 bacteria in PBS via the intranasal route. At the indicated times post infection, bacterial load was evaluated in spleen and lungs as follows. Aseptically removed lungs and spleen were homogenized in 5 ml of PBS containing 0.05% Tween 80 using a gentleMACS dissociator (Miltenyi Biotec, Mercoeur, France) and serial dilutions of the organ homogenates were plated onto solid 7H11 medium. The numbers of cfu were enumerated 21 days later.

Human cell cultures and infection

Human blood samples, procured by the Etablissement Français du Sang of Toulouse (France), were collected from fully anonymous non-tuberculous control donors. Peripheral blood monocytes were isolated as previously described (Astarie-Dequeker et al., 1999), and cultured for 7 days on sterile glass coverslips in 24-well tissue culture plates (5 × 105 cells per well) containing RPMI 1640 (Gibco, Cergy Pontoise, France) supplemented with 2 mM glutamine (Gibco) and 7% heat-inactivated human AB serum. The culture medium was renewed on the third day. Human macrophages derived from monocytes (hMDMs) were washed twice with fresh RPMI medium before use.

The infection assay was performed as previously described (Astarie-Dequeker et al., 2009). Exponentially growing GFP-expressing mycobacteria were dispersed using glass beads. Macrophages were then infected with bacteria at the appropriate moi in RPMI 1640 and infection was allowed to proceed for 1 h at 37°C under an atmosphere containing 5% CO2. Extracellular bacteria were then removed by three successive washes with fresh medium.

Phagocytosis and bacterial load assays

Phagocytosis was assessed as previously described (Astarie-Dequeker et al., 2009). After being infected, hMDMs were fixed with 3.7% PFA and extracellular mycobacteria were labelled with rabbit anti-mycobacteria Ab (1:50), which was detected by a Rhodamine Red-conjugated goat anti-rabbit secondary Ab (1/100). Preparations were visualized under a Leica DM-RB epifluorescence microscope and the number of intracellular bacteria per macrophage and the percentage of cells having ingested at least one bacterium were determined. For each set of conditions, duplicate experiments were performed, and at least 100 cells were counted per slide.

Colocalization experiments with LysoTracker, H+-ATPase and CD63

Colocalization of markers with bacteria-containing phagosomes was evaluated as previously described (Astarie-Dequeker et al., 2009). At the end of the infection time, hMDMs were thoroughly washed and incubated with fresh RPMI 1640 supplemented with 7% heat-inactivated human serum at 37°C under an atmosphere containing 5% CO2. After various times, hMDMs were washed and incubated with the acidotropic dye LysoTracker Red (1:2000) in RPMI 1640 for 1 h, fixed with 3.7% PFA for 1 h, washed and mounted on slides with DAKO mounting medium. For H+-ATPase or CD63 analysis, fixed hMDMs were permeabilized by incubation with 0.3% Triton X-100 for 10 min at room temperature (RT), blocked by incubation with 0.3% BSA and incubated with rabbit polyclonal antiserum against H+-ATPase (1/100) or mouse anti-CD63 Ab (1:100) for 1 h at RT, and then with Rhodamine-Red-conjugated goat anti-rabbit or anti-mouse Ab (1/1000), washed and mounted on slides. Coverslips were viewed with a Leica DM-RB fluorescence microscope or a Leica TCS-SP2 confocal scanning microscope. All images were processed with Adobe Photoshop software. Colocalization was determined as the fraction of phagosomes with GFP fluorescence that was associated with markers. For each marker, 100 phagosomes from at least 10 different fields, in duplicate, in at least three independent experiments were counted for each time.

Statistics

Results are expressed as means ± standard error of the mean (SEM) of the indicated number of experiments (n) performed at least in triplicate. The data were analysed by Student's t-test using the GRAPHPAD PRISM 5·00 for Windows (GraphPad Software, San Diego, CA, USA). P < 0·05 was used as the limit of statistical significance.

Acknowledgements

We are grateful to B. Appelmelk (VU University Medical Center, Amsterdam) and R. Brosch (Institut Pasteur, Paris) for the gift of antibodies, and to F. Laval and M. Daffé (IPBS) for the help with the mass spectrometry analyses. We thank the Genotoul plateforms TRI (IPBS) for epifluorescence and confocal microscopy imaging, TRI (IBCG, Toulouse), notably S. Balor, for the electron microscopy imaging, and ANEXPLO (IPBS) for the animal experiments. We thank F. Viala (Photographist, IPBS, Toulouse) for help with graphical work. We also thank I. Vergne and G. Lugo-Villarino for critical reading of the manuscript. C.P. is recipient of a fellowship from the Fondation pour la Recherche Médicale. This work was supported by the Centre National de la Recherche Scientifique (CNRS), the Fondation pour la Recherche Médicale (‘Equipe FRM’ DEQ20090515399), the ANR (2010-PATH-007-01/GeMoA) in the frame of ERA-Net PathoGenoMics and by European Structural Funds (FEDER) and the Région Midi-Pyrénées (CPER 2007–2013) for the scientific project and for the ASB3 animal facility.

Ancillary