Virulence attenuation of two Mas-like polyketide synthase mutants of Mycobacterium tuberculosis

Tuberculosis (TB), a disease caused by Mycobacterium tuberculosis in humans, remains an important cause of mortality and morbidity worldwide (World Health Organization, 2002). Part of the success in combating this disease will involve the development of new antimycobacterial therapies enabling the control of multidrug-resistant tuberculosis. The complex cell envelope surrounding M. tuberculosis is one of the major causes of the resistance of this pathogen to host defence mechanisms (Daffé & Draper, 1998) and to therapeutic agents (Jarlier & Nikaido, 1994). It is also the site of action of many front-line anti-TB drugs (Chopra & Brennan, 1998). The identification of unique processes involved in the synthesis of cell envelope components of pathogenic mycobacteria may bring to light attractive drug targets for the development of new anti-TB drugs.

The existence of a large number of very long multiple methyl-branched fatty acids in pathogenic mycobacteria led to the suggestion that these molecules may play a significant role in virulence. In M. tuberculosis, lipids esterified with multiple methyl-branched fatty acyl substituents include sulfolipids (SL) (Goren & Brennan, 1979), di- and triacylated trehaloses (DAT and TAT) (Lemassu et al., 1991; Besra et al., 1992; Muñoz et al., 1997), polyacyltrehaloses (PAT) (Minnikin et al., 1985; Daffé et al., 1988) and phthiocerol dimycocerosates (DIM) (Daffé & Lanéelle, 1988) (Fig. 1). Although the last 50 years have seen considerable effort devoted to studying the possible roles of these molecules in pathogenesis (Goren & Brennan, 1979), we still know little about their biosynthesis. Lipids of the DIM family have received the most attention. In these lipids, 2,4,6,8-tetramethyl C₃₂ fatty acids and homologues (mycocerosic acids) are esterified to a long-chain β-diol, the phthiocerol (Fig. 1). The genetics and enzymology of the Mycobacterium bovis Bacille CalmetteGuérin (BCG) polyketide synthase Mas, which is involved in the synthesis of mycocerosic acids, have been studied in detail (for a review, see Kolattukudy et al., 1997). Moreover, DIM were recently shown to contribute to the impermeability of the cell envelope of M. tuberculosis (Camacho et al., 2001; Sirakova et al., 2001) and to the replication of this bacterium in the lungs of mice (Camacho et al., 1999; Cox et al., 1999). Analysis of the M. tuberculosis genome sequence has revealed the existence of several polyketide synthase genes (pks), seven of which are very similar to mas in terms of size, functional domains and relative positions of the domains (Sirakova et al., 2001). In an attempt to determine the function of mycobacterial pks genes, mutants deficient in the expression of the ppsB/ppsC, pks2, pks3/4, pks10 and pks15/1 genes were constructed by allelic replacement in M. tuberculosis or M. bovis BCG. Disruption of the pks2 gene in M. tuberculosis H37Rv resulted in a mutant that does not produce the hepta- and octamethyl-branched phthioceranic and hydroxyphthioceranic acids and is therefore SL-deficient (Sirakova et al., 2001) (Fig. 1). Disruption of the pks3/4 genes in M. tuberculosis H37Rv resulted in a mutant deficient in the production of the di- and trimethyl-branched fatty acids mycolipenates, mycolipodienates and mycosanoates required for the synthesis of some forms of PAT and DAT (Dubey et al., 2002; Rousseau et al., 2003) (Fig. 1). Disruption of ppsB and ppsC in M. bovis BCG led to a mutant unable to synthesize the phthiocerol and phenolphthiocerol derivatives required for the production of DIM and the M. bovis-specific phenolglycolipids (PGL) (Azad et al., 1997). Interestingly, disruption of the pks1 gene abolished the production of PGL but not that of DIM in M. bovis BCG (Constant et al., 2002), whereas in M. tuberculosis H37Rv (a strain naturally deficient in PGL production), a similar mutation in pks1 led to DIM deficiency (Sirakova et al., 2003). In M. bovis BCG, it was proposed that the fused pks15pks1 ORFs are involved in the elongation of p-hydroxybenzoic acid derivatives to form p-hydroxyphenylalkanoates which are in turn converted into phenolphthiocerol derivatives (Constant et al., 2002). In M. tuberculosis H37Rv, the two adjacent separate ORFs pks15 and pks1 are apparently unable to catalyse such an elongation reaction (Constant et al., 2002), and are instead thought to be involved in the synthesis of phthiocerol derivatives (Sirakova et al., 2003). Finally, the pks10 gene, disruption of which in M. tuberculosis H37Rv also caused DIM deficiency without affecting the ability of the mutant strain to synthesize mycocerosic acids, is thought to be involved in the production of phthiocerol derivatives (Sirakova et al., 2003). As for pks15pks1 in M. tuberculosis H37Rv, the precise function of pks10 is currently unclear.

(19K): Fig. 1. Structures of some methyl-branched fatty acid-containing lipids of M. tuberculosis. A, The major sulfolipid SL-I; trehalose is sulfated at the 2' position and esterified with palmitic acid and the multimethyl-branched phthioceranic and hydroxyphthioceranic acids. B, Polyacyltrehalose (PAT); trehalose is esterified with stearic acid and the multimethyl-branched mycolipenic acids. C, 2,3-Di-O-acyltrehalose (DAT); trehalose is esterified with stearic acid and the multimethyl-branched mycosanoic acid. D, Phthiocerol dimycocerosates (DIM); the long-chain β-diol (phthiocerol moiety) is esterified with two mycocerosic acids; n=2022; n', n''=16, 18; p, p'=25; R=CH2CH3 or CH3. E, Phenolic glycolipid isolated from M. tuberculosis Canetti (PGL); m=1517.

In order to elucidate the role of other mas-like genes in M. tuberculosis, we constructed M. tuberculosis H37Rv mutants deficient in the expression of the mas-like genes pks5 and pks7 and undertook their characterization. Of all the M. tuberculosis pks genes, pks5 is the most similar to mas (66 % amino acid identity). pks7 is part of a four-pks gene cluster consisting of pks7, pks8, pks17 and pks9, flanked by two chalcone-synthase-like genes, pks10 and pks11. Pks7 shares 31 % amino acid identity with Mas. We report here the construction and biochemical characterization of these two mutants and the analysis of their growth characteristics in cellular and animal models. Bacterial strains and growth conditions.
Escherichia coli XL-1 Blue and DH5α, the strains used for cloning experiments, were routinely propagated in LuriaBertani (LB) broth and LB agar (Difco) at 37 °C. Mycobacterium smegmatis mc²155 was cultured at 37 °C in liquid LB medium supplemented with 0·05 % Tween 80 for competent cell preparation and in Middlebrook 7H9 broth (Difco) supplemented with 0·05 % Tween 80 for transduction. M. tuberculosis H37Rv (ATCC 25618) was cultured in Middlebrook 7H9 broth (Difco) supplemented with 10 % ADC enrichment (Difco) and 0·05 % Tween 80 or on agar Middlebrook 7H11 medium (Difco) supplemented with OADC (Difco). Kanamycin (20 µg ml^-1), hygromycin B (50 µg ml^-1), ampicillin (100 µg ml^-1) and sucrose (2 %) were added to the medium when required.

Electrotransformation, cloning and Southern blot analysis.
Molecular cloning, restriction endonuclease digestions and electrotransformation were performed by standard techniques. Southern blot analysis was performed as previously described (Pelicic et al., 1997).

Generation of M. tuberculosis pks5 and pks7 mutants by allelic replacement.
A pks5 mutant of M. tuberculosis H37Rv was constructed by allelic replacement using the Ts/sacB procedure described by Pelicic et al. (1997). The vector used for gene inactivation was constructed as follows. A M. tuberculosis H37Rv 5418 bp BamHI restriction fragment carrying part of the pks5 coding sequence was excised from the integrative cosmid I380 (kindly provided by Professor S. T. Cole, Institut Pasteur, Paris, France) and inserted into the BamHI site of the pBluescript II KS(-) vector. An internal 1746 bp SphI restriction fragment from the pks5 gene was then excised from the vector and replaced with the 1·2 kb kanamycin-resistance cassette (km) from pUC4K, such that km was flanked by 1·7 and 2·0 kb segments of pks5. The resulting disrupted pks5 : : km allele was then excised from the pBluescript II KS(-) vector and inserted together with the xylE gene into pPR27 (Pelicic et al., 1997), yielding p27P5KX, which was used in the allelic exchange experiment.

The method used to disrupt pks7 in M. tuberculosis H37Rv relies upon the use of a conditionally replicating mycobacteriophage (Bardarov et al., 1997). A DNA fragment containing part of the pks7 gene (bp 10444639 of the coding sequence) was amplified from genomic M. tuberculosis H37Rv DNA, using the following primers: sense primer, 5'-GGATCCCATTGAGGCGCAGGCG-3' (A), antisense primer, 5'-GGATCCGTCAAAACCCATCGCC-3' (B), introducing BamHI sites at the 5' and 3' ends of the sequence. The 3595 bp PCR product was inserted into the BamHI-digested pUC19 vector and a 1577 bp MluI fragment was replaced by a hygromycin-resistance gene cassette (hyg), such that hyg was flanked by 1·3 and 0·9 kb segments of pks7. The resulting disrupted pks7 gene and flanking regions were cloned into the cos vector pYUB572 and the recombinant cosmid was digested with PacI and ligated into PacI-digested phAE87 DNA. The ligation mixture was packaged using lambda in vitro packaging mix (Gigapack III, Stratagene), transduced into E. coli HB101 and plated on LB plates with hygromycin. DNA from several phasmid clones was isolated, confirmed by restriction digestion, electroporated into M. smegmatis strain mc²155 and plated for plaques at 30 °C. Individual plaques were tested for thermosensitivity, amplified and used to infect M. tuberculosis H37Rv as previously described (Sirakova et al., 2001).

Biochemical analysis of cell envelope lipids in the wild-type and pks5- and pks7-disrupted mutants of M. tuberculosis.
[1-¹⁴C]Propionate (50 µCi; specific activity 55 Ci mol^-1, 2035 GBq mol^-1) and [1-¹⁴C]acetate (50 µCi; specific activity 56·7 Ci mol^-1, 2098 GBq mol^-1) (American Radiolabelled Chemicals) were added to 30 ml of 1012-day-old cultures (OD₆₀₀ 1·61·8) of M. tuberculosis H37Rv and its pks5- and pks7-disrupted mutants, and incubation was continued for a further 24 h at 37 °C in roller bottles. [1-¹⁴C]Propionate is a precursor molecule known to be incorporated into methyl-branched fatty acids. [1-¹⁴C]Acetate is incorporated into all classes of fatty acids, allowing the analysis of all lipid types. Cells were collected and total extractable lipids from bacterial pellets and culture supernatants were extracted as described by Sirakova et al. (2001). Total cellular and extracellular [1-¹⁴C]propionate- and [1-¹⁴C]acetate-labelled lipids from H37Rv wild-type and pks mutants were subjected to one- and two-dimensional thin-layer chromatography (TLC) on silica gel G plates in a variety of solvent systems to resolve lipids of various polarities (Table 1), and the lipids were detected by autoradiography. Labelled lipids recovered from TLC plates were eluted from the silica gel with chloroform/methanol (2 : 1, v/v). Fatty acid methyl esters were prepared from total lipids derived from [1-¹⁴C]propionate and from DIM, PAT, SL and DAT fractions isolated by TLC. Methyl esters of fatty acids and acetylated hydroxy-fatty acids were subjected to radio-GC analysis (Sirakova et al., 2001; Dubey et al., 2002). The methyl esters of fatty acids from the total [1-¹⁴C]propionate-derived lipids of both the wild-type strain and the pks mutants were also separated by argentation-TLC (5 % AgNO₃ in silica gel) using hexane/ethyl ether (9 : 1, v/v) as the developing solvent and each separated fatty acid methyl ester fraction was analysed by radio-GC using conditions similar to those previously described (Dubey et al., 2002). Mycolic acid methyl esters prepared from cells that had incorporated [1-¹⁴C]acetate were analysed as previously described (Kremer et al., 2002).

Table 1. Solvent systems for TLC analysis of lipids

Extraction of RNA from M. tuberculosis and RT-PCR experiments.
Cultures of M. tuberculosis were grown in 7H9 medium to OD₆₀₀ 0·6. For analysing the expression of the pks5 gene, bacteria were harvested and broken in 1 ml Trizol solution (Life Technologies) with mini glass beads using a Bead Beater apparatus (Polylabo) set at maximum speed. RNA was extracted with 200 µl chloroform/isoamyl alcohol (Ready Red, Appligene-Oncor). After 15 min centrifugation at 12 000 g, the aqueous phase was transferred to a tube containing 500 µl 2-propanol. Total RNA was precipitated for 10 min at room temperature, centrifuged for 10 min at 12 000 g and washed with 1 ml of a 75 % ethanol solution before resuspension in DEPC-treated water (Ambion). Contaminating DNA was removed by digestion with DNase I according to the manufacturer's instructions (Ambion). The same protocol was used to extract RNA from M. tuberculosis infecting resting mouse bone-marrow macrophages. Macrophages were infected at an m.o.i. of five bacteria per cell. Sixteen hours post-infection, cells were washed three times with fresh medium and then lysed with cell culture lysis reagent (Promega). RNA was extracted from the lysate as described above. Reverse transcription was performed using the Expand Reverse Transcriptase (Roche) with 2 µg total RNA and 50 pmol antisense primers specific for pks5 and sigA (Table 2). Quantitative PCR was performed on a Light Cycler apparatus (Roche), using the SYBR Green technique (Fast start DNA Master SYBR Green I, Roche) with specific primers for pks5 and sigA (Table 2). The number of cDNA copies present in each sample was determined as previously described (Manganelli et al., 1999). The number of pks5 mRNA copies in bacteria grown axenically and inside macrophages was compared to the number of copies of sigA mRNA in the same sample (Manganelli et al., 1999). RT-PCR experiments were performed on the RNA extracted from three independent 7H9 cultures of M. tuberculosis and from three independent preparations of macrophages infected with M. tuberculosis. Analysis of the pks7, pks8, pks17 and pks10 transcripts in the wild-type H37Rv and pks7 mutant strains was performed as follows. RNA was isolated from bacteria grown to mid-exponential phase. Chilled bacterial cells collected by centrifugation were resuspended in RNeasy lysis buffer (Qiagen), transferred to a 2 ml tube containing ceramic and silica beads (FastRNA blue) and disrupted using a FastPrep F120 instrument (Q.BIOgene). The extract collected by centrifugation was used to isolate total RNA using RNeasy kit (Qiagen) according to the protocol provided by the manufacturer. Reverse transcription was performed using random primers and SuperScript RNase H reverse transcriptase (Life Technologies). PCR on the cDNA was done with Platinum Taq DNA polymerase (InVitrogen) and the sets of specific primers listed in Table 2. A control without the reverse transcriptase verified the absence of DNA contamination.

Table 2. Primers used for RT-PCR experiments

Preparation and infection of murine bone-marrow macrophages.
Bone-marrow macrophages were prepared as previously described (Jackson et al., 1999) and infected at an m.o.i. of 0·5 bacterium per cell for 4 h at 37 °C in a 5 % CO₂ atmosphere. For experiments with activated macrophages, 100 U IFNγ ml^-1 and 10 ng TNFα ml^-1 were added to the medium 24 h before infection, and maintained throughout the infection process. On days 0 (4 h), 3, 5 and 7 the infected macrophage monolayers (3 wells per strain) were lysed in 200 µl cell culture lysis reagent (Promega) and the number of viable intracellular c.f.u. was evaluated by plating the lysed cell suspensions on 7H11, 7H11-Km or 7H11-Hyg plates. Infection experiments were carried out in duplicate using two independent batches of bacterial stocks and cell preparations.

M. tuberculosis growth and persistence in mice.
M. tuberculosis H37Rv wild-type and mutant strains were used to infect female BALB/c mice (68 weeks old) via the respiratory route. M. tuberculosis aerosols were generated from bacterial suspensions consisting of 10⁷ c.f.u. ml^-1 in phosphate-buffered saline (PBS) (pH 7·4) with 0·05 % Tween 80. Mice were exposed to the aerosols for 15 min. Four to six mice were used per experimental point and per strain. At various time points after infection (days 1, 7, 14, 20, 42 and 98 for the pks5 mutant; days 1, 7, 14, 22, 43, 64 and 100 for the pks7 mutant), the lungs and spleen were removed aseptically and homogenized in Sauton's medium diluted 1 : 4 and supplemented with Middlebrook OADC. The viable bacteria in the organs of infected animals were counted by plating serial dilutions of the organ homogenates on 7H11, 7H11-Km or 7H11-Hyg agar.

Disruption of the pks5 and pks7 genes by allelic replacement
A pks5 and a pks7 mutant of M. tuberculosis H37Rv were constructed by allelic replacement using the Ts/sacB method described by Pelicic et al. (1997) and the temperature-sensitive mycobacteriophage system developed by Bardarov et al. (1997), respectively. Disrupted alleles of the pks5 and pks7 genes were generated by replacing 1746 bp of the pks5 coding sequence and 1577 bp of the pks7 coding sequence with a kanamycin- and a hygromycin-resistance cassette, respectively. In pks5, this deletion resulted in the loss of the regions encoding the acyl transferase (AT), dehydratase (DH), enoyl reductase (ER), ketoreductase (KR) and acyl-carrier protein (ACP) domains. In pks7, the deletion resulted in the loss of the AT, DH and ER domains. Genomic DNA from two pks5 and three pks7 mutant candidates was analysed by Southern blotting and all mutants gave hybridization signals consistent with allelic replacement at the pks5 or pk7 locus (Fig. 2).

(43K): Fig. 2. Southern blot analysis of M. tuberculosis pks5 and pks7 knock-out strains. (a) Expected hybridization profile and Southern blot analysis of M. tuberculosis pks5 knock-out strains. Genomic DNA from pks5 mutant candidates was digested with BamHI. Lanes 12, pks5 mutants; lane 3, wild-type H37Rv. The specific probe P1 used to reveal the blot corresponds to the 5·42 kb BamHI restriction fragment carrying pks5 excised from cosmid I380. S, SphI; B, BamHI. (b) Expected hybridization profile and Southern blot analysis of a M. tuberculosis pks7 knock-out strain. Genomic DNA from the pks7 mutant candidate was digested with BamHI/EcoRI (lanes 1 and 3) and PstI (lanes 2 and 4). Lanes 12, wild-type H37Rv; lanes 34: pks7 mutant. The specific probe P2 used to reveal the blot corresponds to the 1·82 kb pks7 PstIEcoRV restriction fragment containing part of the gene segment deleted in making the construct and 801 bp DNA 3' of the deletion. Primer pair a/b was used to amplify the pks7 region chosen to generate the disruption construct. P, PstI; B, BamHI; M, MluI; EV, EcoRV; EI, EcoRI.

As the pks7 gene is part of a cluster of four pks genes pks7, pks8, pks17 and pks9 separated by short intergenic spaces (23 bp between pks7 and pks8, 3 bp between pks8 and pks17, 9 bp between pks17 and pks9) and flanked by two chalcone-synthase-like genes (pks10 and pks11), we could not exclude the possibility that the mutation in pks7 had a polar effect on the expression of neighbouring genes. To address this question, we examined the pks7 mutant for pks8, pks17 and pks10 transcripts using a gel-based RT-PCR assay. pks8, pks17 and pks10 transcripts were detected in the pks7 mutant (Fig. 3), indicating that the insertion of a hygromycin cassette within pks7 had not abolished the expression of neighbouring genes, although their levels of expression and regulation might differ from those in the wild-type H37Rv strain.

Table 2) and there were no products when reverse transcriptase was omitted. WT, wild-type H37Rv; M, pks7 mutant.

Biochemical characterization of the pks5 and pks7 mutants
We investigated the biochemical consequences of pks5 and pks7 gene disruption by labelling lipids with [1-¹⁴C]propionate, a precursor molecule known to be incorporated into multimethyl-branched fatty acids. The wild-type strain and the two pks mutants incorporated similar amounts (2025 %) of ¹⁴C into lipids. Radio-GC analysis of the methyl esters of fatty acids from total lipids showed that both pks mutants generated all the classes of fatty acids produced by the wild-type strain. No difference was observed in the fatty acid profiles of the wild-type strain and the pks5 mutant, but lower levels of mycocerosic acid labelling were observed in the pks7 mutant than in the wild-type strain (Fig. 4). Argentation-TLC of fatty acid methyl esters from total lipids, followed by radio-GC analysis of individual fatty acid fractions purified from the TLC plate, showed that both pks mutants produced all classes of fatty acids but that the labelling of the mycocerosic-acid-containing fraction was much less in the pks7 mutant than in the wild-type H37Rv and the pks5 mutant (data not shown). When total lipids were separated by TLC using 10 % ethyl ether in hexane as the solvent, about 30 % of the incorporated ¹⁴C was found in the dimycocerosyl phthiocerol (DIM) fraction in the wild-type strain, 40 % was found in the DIM fraction of the pks5 mutant, and the DIM fraction of the pks7 mutant was barely labelled (Fig. 5). The higher degree of labelling of DIM in the pks5 mutant affected both forms of DIM, the dimycocerosates of phthiocerol (DIM A) and the dimycocerosates of phthiodiolone (DIM B) (Fig. 5). The absence of DIM in the pks7 mutant was confirmed by treating the plate with dichromate/H₂SO₄ followed by heating. The charred band at R_F 0·6 for the wild-type strain was not observed for the mutant (data not shown). Similarly, DIM was absent from the culture filtrates of the pks7 mutant (data not shown). Thus, the pks7 mutant is deficient in the production of DIM but capable of generating mycocerosic acids.

(22K): Fig. 4. Radio-GC analysis of total fatty acid methyl esters derived from [1-14C]propionate in M. tuberculosis H37Rv (top), pks7 (middle) and pks5 (bottom) mutants. Retention time ranges for branched short-chain acids (A) mycolipanoic and mycolipenic acids (B), mycocerosic acids (C) and phthioceranic acids (D) are indicated.

(58K): Fig. 5. Autoradiograms of thin-layer chromatograms of lipids derived from [1-14C]propionate in M. tuberculosis H37Rv (WT) and two pks-disrupted mutants (Δpks5 and Δpks7). Total lipids were subjected to TLC on silica gel G with 10 % ethyl ether in n-hexane as the solvent. DIM A (dimycocerosates of phthiocerol) and DIM B (dimycocerosates of phthiodiolone) are indicated by arrows. The arrow at the top of the right-hand side shows the solvent front.

Detailed analysis of the more polar lipids labelled by [1-¹⁴C]propionate showed that no other lipid class was affected by the two pks gene disruptions. The distribution of ¹⁴C-radioactivity in the PAT, SL, and DAT fractions was very similar in the wild-type strain and in both pks mutants, except that the pks7 mutant incorporated 67 % more radioactivity into SL than did the wild-type strain (data not shown). Further detailed one- and two-dimensional TLC analyses of [1-¹⁴C]propionate- and [1-¹⁴C]acetate-derived lipids from cell pellets and culture filtrates of the H37Rv wild-type strain and the pks5 mutant collected at various growth stages (early and late exponential phase) were performed in a variety of solvent systems (Table 1). This allowed lipids of various polarities to be analysed, including all classes of phospholipids and phophatidylinositol mannosides, acyltrehaloses (trehalose mono- and di-mycolates, SL, PAT, DAT), DIM, glycosides, triacylglycerides, menaquinones and free fatty acids and mycolic acids. No difference between the two strains was found.

The fatty acids derived from [1-¹⁴C]propionate present in the individual TLC fractions, DIM, PAT, SL and DAT, were also analysed by argentation-TLC and radio-GC as previously described (Dubey et al., 2002), in order to define the methyl-branched fatty acid composition of each of these individual lipids. These analyses revealed a stronger ¹⁴C-labelling of hydroxyphthioceranic acids in the pks7 mutant than in the wild-type strain, probably accounting for the stronger labelling of SL in this mutant. No other significant difference in the fatty acid composition of individual ¹⁴C-propionate-labelled lipids between the wild-type strain and the pks7 mutant was observed (data not shown). The relative amounts of ¹⁴C incorporated from [1-¹⁴C]propionate into the different methyl-branched fatty acids present in PAT, SL and DAT in the pks5 mutant (including C₁₆-, C₁₈-branched fatty acids and mycolipenic, mycosanoic, mycolipodienic, mycolipanolic, phthioceranic, hydroxyphthioceranic and mycocerosic acids) were identical to those found in the wild-type strain.

The delipidated cells remaining after soluble lipid extraction were also saponified in 2-methoxyethanol containing 12 % H₂O and 5 % KOH and methylated as previously described (Sirakova et al., 2001) for analysis of potential cell-wall-linked methyl-branched fatty acids. TLC analysis of the fatty acid methyl esters prepared by this technique using hexane/ethyl ether (9 : 1, v/v) as the solvent system showed no difference between the wild-type H37Rv strain and the pks5 mutant (data not shown). Finally, to investigate differences in cell-wall-bound mycolates, the mycolic acid methyl esters from [1-¹⁴C]acetate-labelled delipidated cells were prepared and analysed as described elsewhere (Kremer et al., 2002). Again, no significant difference was observed between the wild-type strain and the two pks mutants (data not shown).

Thus, pks5 disruption did not seem to alter the fatty acid and lipid compositions of M. tuberculosis. The pks5 gene was also disrupted in M. bovis BCG and thorough analyses of soluble and cell-wall-bound lipids derived from [1-¹⁴C]propionate and [1-¹⁴C]acetate showed no detectable difference in labelling patterns (data not shown).

Growth kinetics of the pks5 and pks7 mutants in vitro and in mouse bone-marrow macrophages
The growth rates of the two mutants were similar to that of wild-type M. tuberculosis H37Rv in 7H9 broth (data not shown), suggesting that the changes affecting their cell envelope composition or structure had no effect on growth in this medium. We then analysed the ability of the two mutants to replicate and to persist within resting and activated mouse bone-marrow macrophages. The pks5 mutant displayed a similar rate of intracellular growth to the parental strain H37Rv in both resting and activated macrophages (Fig. 6). Therefore, pks5 is unlikely to be involved in the survival of M. tuberculosis within murine phagocytic cells, or in the resistance of mycobacteria to the respiratory burst that accompanies macrophage activation. In contrast, the pks7 mutant replicated 1·7-fold less than did the wild-type strain in resting macrophages over a 5 day period. It also replicated 1·6-fold less than did the wild-type strain in activated macrophages over a 7 day period (Fig. 6). However, this difference in growth between the two strains was not statistically significant in Student's t test (P<0·05).

(17K): Fig. 6. Multiplication and persistence of M. tuberculosis H37Rv wild-type (), pks5 () and pks7 (•) mutant strains in resting (a) and activated (b) murine bone-marrow macrophages. The reported values are the mean and standard deviation of data obtained from three independent wells in one typical experiment. This experiment was executed twice using independent cell and bacterial cultures.

Growth kinetics of the pks5 and pks7 mutants in an in vivo mouse model
We then assessed the virulence of the pks5 and pks7 mutants in mice infected via the respiratory route. Consistent with the demonstrated role of DIM in the in vivo replication of M. tuberculosis (Camacho et al., 1999; Cox et al., 1999), the DIM-deficient pks7 mutant showed attenuated growth during the acute phase of infection in both organs examined (Fig. 7a). One-twentieth the number of c.f.u. was recovered from the lungs of mice infected with the pks7 mutant than from those of H37Rv-infected mice 21 days post-infection. The difference between the mutant and the wild-type strain was even more marked in the spleen, with 180 times more wild-type than mutant c.f.u. recovered at the same time point. These differences in c.f.u. counts were statistically significant in Student's t test (P<0·05) on days 7, 14 and 22 in the lungs and on day 21 in the spleen. At later stages of the infection, differences tended to lessen, suggesting that pks7 plays a less important role in virulence during the persistent phase of infection.

(19K): Fig. 7. Multiplication and persistence of M. tuberculosis H37Rv wild-type, pks5 and pks7 mutant strains in the lungs and spleen of mice infected via the respiratory route. (a) M. tuberculosis H37Rv wild-type () and pks7 mutant (•). (b) M. tuberculosis H37Rv wild-type () and pks5 mutant (). The multiplication and persistence of the wild-type and mutant strains in the lungs (left panels) and spleens (right panels) of infected animals was followed by counting bacteria in these organs over a 100-day period. Results are expressed as means and standard deviations of c.f.u. counts for four to six infected mice.

The pks5 mutant also multiplied much less extensively than did the parental strain in the lungs and in the spleen during the acute phase of infection (Fig. 7b). This difference between the two strains was especially obvious between days 21 and 42. During these 3 weeks, the rates of multiplication of the pks5 mutant were about one-third and one-fifth of those of the wild-type strain in the lungs and in the spleen, respectively. There were no statistical differences between c.f.u. counts on day 1. However, differences in c.f.u. counts were statistically significant on days 7, 20, 42 and 98 in the lungs and on days 42 and 98 in the spleen (P<0·05).

Expression of the pks5 gene in 7H9-grown and intracellular-grown M. tuberculosis
The virulence attenuation of the pks5 mutant in mice despite its apparent lack of biochemical phenotype when grown under axenic conditions led us to question the regulation of the expression of pks5 during in vitro growth and host infection. We used a quantitative real-time RT-PCR assay to compare the level of transcription of pks5 when M. tuberculosis is grown axenically or inside mouse bone-marrow macrophages. The amount of cDNA produced, which is proportional to the amount of the specific transcript present in the original RNA sample, was measured. In all experiments, sigA was used as an internal standard. sigA is an essential housekeeping sigma factor in M. tuberculosis, and the amount of sigA mRNA remains constant in different growth conditions (Manganelli et al., 1999) and during macrophage infection (Manganelli et al., 2001). This experiment, performed on three RNA preparations for each culture condition, indicated that pks5 was expressed both in 7H9 medium and inside macrophages. The amount of pks5 transcripts was found to be 0·355±0·048 that detected for sigA in M. tuberculosis grown in 7H9 medium, and 0·221±0·056 that detected for sigA in M. tuberculosis grown intracellularly. This difference in expression of the pks5 gene under axenic and intracellular conditions was not statistically significant in Student's t test (P<0·05).

In an effort to define the functions associated with the polyketide synthase genes (pks) of M. tuberculosis we followed a global genetic approach involving the construction and biochemical characterization of pks-deficient mutants (Sirakova et al., 2001; Dubey et al., 2002; Sirakova et al., 2003). In this study, we disrupted the mas-like genes pks5 and pks7 and analysed the effects of their inactivation on fatty acid composition and virulence in M. tuberculosis.

Biochemical analyses of the pks7 mutant indicated that this strain is deficient in DIM synthesis although it produces mycocerosic acids. Therefore, pks7, like the ppsAE, pks10 and pks15/1 genes, is probably involved in the synthesis of phthiocerol derivatives (Azad et al., 1997; Sirakova et al., 2003). As the pks7 gene is part of a cluster of four pks genes pks7, pks8, pks17 and pks9 separated by short intergenic spaces, we checked whether the mutation in pks7 had a polar effect on the expression of neighbouring genes using a gel-based RT-PCR assay. Transcripts of the pks8, pks17 and pks10 genes were detected in the pks7 knock-out strain, indicating that the insertion of a hygromycin cassette within pks7 had not abolished the expression of adjacent genes. The two adjacent ORFs pks8 and pks17 together contain all the domains required for the synthesis of methyl-branched fatty acids (Sirakova et al., 2001), and Pks8/17 displays 32 % amino acid identity to Mas. Pks9 (1017 amino acids in length) is 42 % identical to Pks8 (1603 amino acids in length) on a 809 amino acid overlap. Therefore, Pks7, Pks8/17 and Pks9 appear to belong to the same polyketide synthase family, and could be involved in the same biosynthetic pathway leading to the production of DIM. Although the reason of the existence of such a high number of polyketide synthases dedicated to the production of phthiocerol derivatives and DIM is unclear, the fact that the pks7, pks10, pks15/1 and ppsAE mutants fail to synthesize DIM suggests that the polyketide synthase enzymes encoded by these genes serve non-redundant catalytic functions. Like Pks15/1 and Pks10, Pks7 may act upstream from PpsAE in the pathway, catalysing the formation of precursors that are further elongated by PpsAE to give the final phthiocerol products. Careful enzymic studies are required to define precisely the reactions catalysed by each of these enzymes.

Wild-type H37Rv and the pks7 mutant replicated equally well in 7H9 broth, but the mutant grew slightly less well in activated and non-activated mouse bone-marrow macrophages. However, this difference in intracellular growth between the two strains was not statistically significant. This suggests that, although involved in the permeability of the cell envelope (Camacho et al., 2001), DIM do not contribute in a major way to the resistance of the tubercle bacillus to the intracellular environment. We also cannot rule out the possibility that increased amounts of other lipids such as SL compensate for the absence of DIM in the pks7 mutant. The pks7 mutant showed reduced virulence during the acute phase of infection in the lungs and spleen of mice infected via the respiratory route. This result is consistent with previous reports that mutations in other genes that caused DIM deficiency resulted in attenuation (Camacho et al., 1999; Cox et al., 1999). Interestingly, 3 weeks post-infection, fadD26 and fadD28 DIM-deficient mutants of M. tuberculosis Erdman and Mt103 were found to be five- to tenfold more attenuated for growth in the lungs of BALB/c and C57BL/6 mice infected intravenously (Camacho et al., 1999; Cox et al., 1999), than the pks7 mutant in our mouse model of infection. As the pks7 mutant produces mycocerosic acids whereas the fadD28 mutant does not (Fitzmaurice & Kolattukudy, 1998), it is tempting to speculate that the presence of mycocerosic acids in the pks7 mutant accounts for its higher virulence. As mentioned above, it is also possible that increased incorporation of the methyl-branched precursors into other lipids, such as SL, in the pks7 mutant partially restores some virulence in this strain.

In contrast to the situation encountered in all msl-deficient mutants examined so far, and despite the high level of sequence similarity between pks5 and mas, the disruption of pks5 caused no major change in the fatty acid and lipid contents of the mutant strain. The mutant produced all the major methyl-branched fatty acid containing lipids, including DIM, in similar amounts to the wild-type strain. Interestingly, the replication of the pks5 mutant was unaffected in mouse bone-marrow macrophages, but this mutant replicated much less efficiently than the wild-type strain in the lungs and spleen of BALB/c mice infected via the aerosol route, suggesting an important role for Pks5 or for its polyketide product in the course of infection. These findings provide support for the belief that key cell envelope lipids or polyketides might only be produced under certain culture conditions or during host infection (Minnikin et al., 2002) and raise questions about the expression of pks5 in M. tuberculosis grown under axenic conditions or in vivo. A search for pks5 transcripts in M. tuberculosis H37Rv grown in 7H9 medium by quantitative real-time RT-PCR indicated that the pks5 gene was expressed and that its level of expression was about one-third that of sigA. Therefore, pks5 is expressed in vitro. The absence of detectable changes in the polyketide composition of the pks5 mutant may then be due to the too low level of expression of the pks5 gene or to the fact that the transcript of this gene is not translated into an active product. Interestingly, the level of expression of pks5 inside macrophages was found to be similar to that detected in vitro. It is possible that the very low level of polyketide produced by pks5 has an important role in mouse infection but not in macrophage infection. Alternatively, the expression of pks5 within the host may cause the production of a polyketide that is not produced at detectable levels in culture or inside macrophages, and this polyketide product may be important for virulence. Clearly, more genetic and biochemical analyses are required to determine the level of expression of pks5 in vivo and the effects of the inactivation of this gene on the cell envelope and virulence of M. tuberculosis.

In conclusion, our study provides two more examples of mas-like polyketide synthase genes, the expression of which is required for the full virulence of M. tuberculosis during host infection. Due to their role during host infection and their restricted distribution to pathogenic mycobacterial species, the enzymes of this family constitute interesting drug targets for the development of new anti-TB drugs that could be used in addition to classical treatments.

We thank Eddy Maranghi and Professor Gilles Marchal for their help in infecting the mice. This work was supported by Ministère de la Recherche et de la Technologie grant no. 1412146.00, Institut Pasteur, and grants AI46582 and AI35272 from the National Institute of Health. C. R. holds a Ministère de la Recherche fellowship.