Bacterial strains, plasmids and media

Bioinformatics

Homology searches were carried out using BLASTP ver 2.2.13 [25] The four homologs identified all had e-values <10^-3, and no other protein match approached significance. Prosite database information was obtained at http://www.expasy.ch/prosite/, using Release 20.56 dated November 4^th, 2009.

Construction of M. tuberculosis mutants

Targeted mutagenesis was carried out using a two-step strategy [26] in order to introduce an unmarked mutation without any potential polar effects.

ImpA

Primers tbimpA1 (CTCGACGTACAGGTTGAGCTATCC) and tbimpA2 (CTTCACCTGACCGATCGTCAGCTC) were used to amplify the impA gene and flanking regions (2,108 bp) from M. tuberculosis H37Rv using PCR. The resulting 2.1 kb fragment was cloned into the Eco RV site of pGEM5, producing pIMP50. A 200 bp Sph I fragment within impA was removed following partial digestion and religated to make pIMP51. The 2,348 bp Pvu II fragment of pIMP51 was cloned into p2NIL, producing pIMP57.

To create a deletion where the majority of impA was deleted (769 bp deleted from 813 bp), inverse PCR was performed on pIMP57. Primers tbimpAinv1 (TCGTGCCAGCTGACCAACGAATCCAAGTGCAT) and tbimpAinv2 (TCGTGCCAGCTGATAGGGGAACCAGAGGACTA) were used, simultaneously creating a deletion and introducing a Pvu II site in the deleted construct. Following the PCR reaction the DNA was digested with Dpn I for 1 h at 37C to destroy the template, then digested with Pvu II and religated to produce pFM74. Insertion of a Pac I gene cassette from pGOAL19 was cloned at the Pac I site of pFM74 producing the final delivery plasmid pFM75. The Pac I cassette carries lacZ and sacB, which can be used for positive and negative selection of unmarked mutant colonies, respectively.

SuhB

A 3,534 bp Xho I fragment of cosmid Y5ab was cloned into the Sal I site of plasmid p2NIL to produce pFM33. A fragment of 817 bp was deleted from the 874 suhB gene by inverse PCR on pFM33 using primers tbsuhBΔ1 (TCAGCATGCGTTCGTTGTCAGGTCGTGTC) & tbsuhBΔ2 (TCAGCATGCGATTCAACGGCCTAGAGC); this introduced a Sph I site in the deleted construct. Following treatment with Dpn I and Sph I, this was religated to produce pFM48. Insertion of the gene delivery cassette from pGOAL19 produced the final delivery plasmid pFM52.

ImpC

A 2,503 bp Stu I fragment of cosmid Y3A2 was cloned into the Pml I site of p2NIL, producing pFM31. A 731 bp deletion was generated in the 783 bp gene by inverse PCR on pFM31 using primers tbimpCΔ1 (TGCCAGCTGCATTAGATCGTCGTGGCTCA) & tbimpCΔ2 (TGCCAGCTGGAGGTGCTGACACGGCTC) to introduce a Pvu II site in the deleted construct. Following treatment with Dpn I and Pvu II, this was religated to produce pFM53. Insertion of the delivery gene cassette from pGOAL19 produced the final delivery plasmid pFM54.

CysQ

Primers tbcysQ1 (CCTGGTCGACCTGTTTCC) and tbcysQ2 (GCGGCTCTTTGACATCTTGT) were used to amplify the cysQ gene and flanking regions (2,748 bp) from M. tuberculosis H37Rv DNA. The product was cloned into the Pml I site of p2NIL, producing pFM145. Primers tbcysQΔ1 (AGTCAGGTCGTCCGTCAGATC) & tbcysQΔ2 (TACAACCAACTGGACCCCTAC) were used to generate a 666 bp deletion in the 804 cysQ gene by inverse PCR on pFM145. Following treatment with Klenow polymerase and T4 polynucleotide kinase (Promega), this product was religated to produce pFM148. Insertion of the gene delivery cassette from pGOAL19 produced the final delivery plasmid pFM151.

Mutagenesis

Deletion plasmids were constructed as described above. The delivery plasmids were introduced into M. tuberculosis H37Rv or M. tuberculosis H37Rv ino1, and single crossovers (SCOs) isolated by selection for blue hyg^R kan^R colonies. One SCO colony was plated onto 2% (wt/vol) sucrose-50 μg ml^-1 X-Gal to isolate bacteria with a second crossover; this will lead to mutant or wild-type cells depending on the location of the recombination event. In order to screen for impC mutant, DNA was extracted from sucrose^S kan^S white colonies (obtained from plating M. tuberculosis FAME9 onto sucrose medium) and analysed by PCR using primers that flank the impC gene (TBC1: GGACCGCGATCAGTATGAGT and TBC2: TCGACACAGAATCCGCTAGA). Strains carrying the impC wild-type allele would produce a band of 1148 bp whereas strains carrying an impC mutation would carry the deletion band of 417 bp. Mutant candidates and a wild-type control were digested with Pvu II and subjected to Southern blot analysis using a 2.5 kb impC probe (impC plus flanking region). The wild-type strain showed a 4 kb band whilst the mutant showed a 3.2 kb deletion band along with a 2.5 kb band for the integrated impC copy

Complementation

A construct expressing the impC gene was made by PCR amplification of the impC gene, together with 288 bp of upstream sequence using chromosomal M. tuberculosis H37Rv as template DNA. The primers tbimpCBamP (CGCGGATCCGGCGATGGTGACAT) and tbimpCBam (CGCGGATCCTTACCCGGCGTTGAGC) were used. The product was digested with Bam HI and cloned into the Bam HI site of pBluescript-SK+ to produce pFM94. The Hin dIII cassette of pUC-Gm-int, carrying the int and gm genes was cloned into the Hin dIII site of pFM94 to produce pFM96.

A construct expressing the cysQ gene was made by PCR amplification of the cysQ gene including 352 bp of upstream sequence using M. tuberculosis H37Rv; chromosomal template DNA; primers tbcysup (GCATAGAGCAGGAGGTTTGC) and tbcysend (GCGCCACGCGTCGGCGAT) were used. The PCR product was treated with T4 polynucleotide kinase and cloned into the Sma I site of pBluescript-SK+ to produce pFM160. The Hin dIII cassette of pUC-Gm-int, carrying the int and gm genes was cloned into the Hin dIII site of pFM160 to produce pFM164.

Site-directed mutagenesis

Site-directed mutagenesis was carried out using the non-PCR-based Quickchange kit (Stratagene). Oligonucleotides D86N-forward (GGATCGTAGACCCGATCAACGGCACCAAAAACTTTGTGC) & D86N-reverse (GCACAAAGTTTTTGGTGCCGTTGATCGGGTCTACGATCC) were used to prime DNA synthesis with pFM96. Sequencing confirmed the presence of the required mutation.

Real-time quantitative PCR

Cell wall analysis

Bioinformatics analysis

There are four genes in the M. tuberculosis genome that encode proteins with significant homology to IMPases. All four M. tuberculosis proteins are equally distant from the human IMPase (PDB structure 1IMA; 22-30% identity, 37-46% similarity) [34] and the aligned amino acid sequences are shown in Figure 1A. The four proteins are only as similar to each other, as to the human protein (27-32% identity, 36-44% similarity).

These four genes are generally conserved in other actinomycete genomes, with for example, apparent orthologs in Mycobacterium avium, Mycobacterium smegmatis, and Corynebacterium glutamicum (data not shown). M. leprae, which has many pseudogenes, has no functional impA. Other genomes do also have a small number of other IMPase genes (thus, M. avium has a fifth paralog that is similar to cysQ). While levels of homology between the different M. tuberculosis IMPase paralogs are moderate (22-30% amino acid identity), similarities between orthologs are much higher (for example, 75-79% identity between M. tuberculosis and M. leprae, and 51-67% identity between M. tuberculosis and M. smegmatis).

The genomic contexts of these genes are shown in Figure 2. As with M. smegmatis [24], the impA gene (Rv1604) lies in the middle of the main his operon between hisA and hisF. The stop codon of hisA overlaps with the putative start codon of impA, and the stop codon of impA overlaps with the putative start codon of hisF. These impA genes are 70% identical.

The suhB gene (Rv2701c) was named in the original genome annotation [35], because it is the gene most similar to the Escherichia coli suhB gene. The E. coli suhB gene was so-named because deletion of the gene resulted in a cold-sensitive phenotype, and suppression of a thermosensitive rpoH mutation [36]. It has also been shown to suppress secY [37], dnaB [38], and era [39] mutations. However, these phenotypes are not related to the enzymatic properties of the protein, as they are unaffected by a null point mutation in the active site [40] (Figure 1B). Furthermore, inositol production is not believed to occur in E. coli, so the biological context is very different from that in mycobacteria. Recombinant SuhB from M. tuberculosis has been confirmed to have IMPase activity [41]. SuhB is monocistronic in M. tuberculosis (Figure 2).

The third homologous gene is Rv3137, which we have called impC. It appears to be the first gene in a two-gene operon; a 457 bp intergenic gap upstream of impC suggests it has its own promoter., and a second gene, pflA, is predicted to start only 14 bp downstream, so is probably co-transcribed. PflA shows homology to pyruvate formate lyase-activating proteins. Beyond this is a cluster of fad genes (fadE24-fadE23-fadB4), but the gap beyond pflA and fadE24 is 79 bp, so is less likely to be part of the same operon.

The fourth homologous gene is cysQ (Rv2131c), so-named because it is most similar to the E. coli cysQ gene. E. coli cysQ mutants are cysteine auxotrophs during aerobic growth [42]. Interestingly M. smegmatis contains two paralogs of this gene.

Two sequence motifs have been described for IMPases in the Prosite database [43] (see legend to Figure 1B). One motif, near the N-terminus contains the metal-binding aspartate residues of the active site, and the other lies near the C-terminus. All of the gene products except SuhB had small differences from at least one of the two IMPase motifs (Figure 1B). However, they all contain the important metal-binding residues in both motifs.

The M. tuberculosis impA and suhB genes are dispensable

The impA gene has previously been shown to play an indirect role in inositol synthesis in M. smegmatis [24], and a knockout plasmid construct was therefore prepared to isolate an M. tuberculosis impA mutant. As the gene lies within the his operon (Figure 2), this plasmid carried an unmarked deletion that would not have polar effects. The mutant was generated using a two-step method [26], and grew well on solid medium. Unlike the M. smegmatis impA mutant which had altered colony morphology, there were no obvious differences in colony morphology between the wild-type and mutant strains.

We carried out a similar experiment to determine whether suhB plays a role in inositol metabolism. Again, a deletion construct was prepared, and an unmarked mutant isolated, with no obvious differences in colony morphology.

Inactivation of CysQ

We constructed a plasmid to delete the cysQ gene. Initially, we were unable to obtain a mutant; of 97 double crossovers (DCOs) screened in the presence of inositol, all were wild-type. We therefore made a merodiploid strain by integrating a second copy of cysQ into the single crossover (SCO) strain, and repeating the selection for DCOs on sucrose. Using this method, 24 out of 30 colonies were found to be mutants. The ability to isolate a mutant only in the presence of a functional copy of the gene indicates that this gene was essential under the conditions tested. It could be inferred that cysQ synthesizes all the inositol in the cell, or all the inositol for a specific essential molecule. However, this hypothesis is improbable, as, if true, we would predict thatmutants would be inositol auxotrophs, yet no mutants were isolated even in the presence of high levels of inositol. One possibility is that inositol does not penetrate the cell wall, which is known to be highly impermeable. However, as we had successfully isolated a mutant lacking inositol-1-phosphate synthase (an inositol auxotroph), only when the media was supplemented with extremely high levels of exogenous inositol (50-77 mM) [23], it seems that inositol does enter the cell in sufficient quantities but permeability to this molecule is poor. This suggests that even a slight increase in the requirement for inositol might make mutant isolation impossible, since we had reached the limits of inositol solubility. We reasoned that an increase in the availability of inositol by introduction of a porin might allow a mutant to be isolated. We therefore electroporated an integrating plasmid (pMN013) carrying the M. smegmatis porin gene mspA [44, 45] (for which M. tuberculosis has no orthologue) into the SCO strains, and repeated the sucrose selection. Using this method, we successfully isolated a cysQ mutant in the presence of 77 mM inositol. We screened 16 DCO colonies and two were mutants.

We then plated the mutant on inositol-free medium, and were surprised to observe normal growth, indicating that once the mutant has been isolated, it does not require inositol. To determine whether the mutant could survive in the absence of the mspA gene, the integrating plasmid containing this gene was switched with an empty integrative plasmid carrying the hygromycin resistance gene (pUC-Hyg-int). This would result in the replacement of the cysQ-carrying plasmid, leaving a stain with no functional cysQ. Surprisingly, we were able to obtain cysQ mutants using this approach although we had failed to isolate a mutant by our standard mutagenesis procedure. We therefore conclude that cysQ is also dispensible, and a cysQ mutant does not require inositol for growth.

The impC gene is essential

We attempted to construct an unmarked impC deletion mutant. The first step of the mutagenesis to produce SCOs worked well, however, when cells carrying a second crossover were isolated, only wild-type bacteria were obtained. In theory, the second crossover could take place on either side of the deletion, resulting in either mutant or wild-type cells. The fact that we obtained only wild-type cells (n = 48) suggested that the mutants are not viable.

These initial mutagenesis experiments were carried out in the absence of exogenous inositol. We therefore repeated the mutagenesis, including different levels of inositol in the media at all times. Again, only wild-type bacteria were isolated following the second cross-over (n= 97; 16 on 15 mM inositol, 8 on 30 mM, 16 on 46 mMl, and 57 on 77 mM).

The inability to obtain a mutant may be due to other factors, such as a low frequency of recombination on one side of the gene, even though the length of flanking DNA should be sufficient (847 and 874 bp). Therefore we constructed a merodiploid strain by inserting a second functional copy of impC into the SCO strain. This extra copy was present on an L5-based integrating vector, and contained 288 bp upstream of impC, which was likely to carry its promoter. When this strain (FAME9) was plated onto sucrose to isolate DCOs, three out of eight colonies isolated had lost the original copy of impC. The fact that this gene could only be lost when a second copy of the gene is present suggests that impC is essential for survival, even in the presence of high levels of exogenous inositol (Fisher's exact test, p < 0.01, comparing only the experiments with 77 mM inositol and the complemented strain). To further investigate the essentiality of the impC gene, and in view of what was observed with cysQ, we introduced the mspA gene into the impC SCO strain; this time we were not successful in obtaining a mutant, indicating that the difficulty we encountered making an impC mutant differed from cysQ.

A difference between an IMPase mutant and an ino1 mutant may be that inositol-1-phosphate accumulates in the IMPase mutant, which might somehow be detrimental to the cell. We therefore carried out the essentiality experiment in an ino1 mutant background. The impC mutant construct was introduced into M. tuberculosis ino1, and a SCO strain isolated. On plating for DCOs, only wild-type colonies were isolated (n = 38), again suggesting that this was not the explanation for the essentiality.

Site directed mutagenesis of impC

Our results suggest that impC does not have a critical role in inositol production and hence our inability to obtain an impC mutant may indicate that impC has a different or secondary function that prevents isolation of a mutant. For example, the enzyme might form part of an enzyme complex, and play a vital structural role in maintaining the integrity of that complex. Deletion of the gene would then have both enzymatic and structural effects. An analogous situation was found with the E. coli SuhB protein; where phenotypes in suhB mutants were not related to IMPase activity, as a point mutation in the active site did not produce the suppressing phenotype [40]. We therefore used the same approach to try to separate enzymatic activity from a structural role.

A D93N change in E. coli SuhB and an equivalent D90N change in the human IMPase suppress activity [40, 46] (Figure 1B). Site-directed mutagenesis was used to introduce a corresponding mutation (D86N) in the M. tuberculosis impC gene using the integrating plasmid pFM96 previously used for complementation. This plasmid (pFM123) was introduced into the SCO strain FAME7, and the resultant strain (FAME11) was streaked onto sucrose/inositol plates. DCO colonies were analysed, and, in contrast to the situation with pFM96, all were shown to be wild-type (n = 52). The fact that the functional impC gene could not be replaced by this mutated gene, even in the presence of inositol (p < 0.01), shows that the mutation did inactivate enzymatic activity, and (assuming that the structure was not affected) that it is this enzymatic activity that is essential, rather than an additional structural role.

Enzyme activities

PIM, LAM and mycothiol levels are normal in the impA, suhB and cysQ mutants

Cell extracts of the mutant strains were prepared for the assay of inositol-containing molecules (cell envelope glycolipids and mycothiol). TLC analyses showed that PIMs were normal in the mutant strains (Figure 3A), whilst polyacrylamide gel electrophoresis (Figure 3B) and sugar compositional analysis (not shown) demonstrated normal levels of LAM and LM. Mycothiol levels were assayed by HPLC analysis; levels in the impA, suhB and cysQ mutants were similar to wild-type (see Figure 4).

Gene expression levels of imp genes in M. tuberculosis