The σ factors of Mycobacterium tuberculosis


  • Editor: Simon Cutting

Correspondence: Riccardo Manganelli, Dipartimento di Istologia, Microbiologia e Biotecnologie Mediche, Università di Padova, Via Gabelli 63, Padova 35121, Italy. Tel.: +39 498272366; fax: +39 498272355; e-mail:


Mycobacterium tuberculosis is a remarkable pathogen capable of adapting and surviving in various harsh conditions. Correct gene expression regulation is essential for the success of this process. The reversible association of different σ factors is a common mechanism for reprogramming bacterial RNA polymerase and modulating the transcription of numerous genes. Thirteen putative σ factors are encoded in the M. tuberculosis genome, several being important for virulence. Here, we analyse the latest information available on mycobacterial σ factors and discuss their roles in the physiology and virulence of M. tuberculosis.


On 24 March 1882, Robert Koch announced to the Physiological Society of Berlin that he had isolated and grown the tubercle bacillus, raising high hopes that a major plague would soon be eradicated (Kaufmann, 2003). Almost 125 years later, Mycobacterium tuberculosis remains one of the most widespread bacterial pathogens, infecting approximately one-third of the human population and killing about two million people every year (Raviglione, 2003). Several features contribute to the success of M. tuberculosis. It is capable of adapting to diverse harsh environments and blocking normal phagosome maturation. Mycobacterium tuberculosis can also enter a dormant state, allowing asymptomatic infections that persist for decades. Any weakening of the immune system response, possibly due to malnutrition, debilitating diseases or age, may result in the reactivation of latent bacilli (Smith, 2003).

The sophisticated infection and adaptation mechanisms used by M. tuberculosis are expected to require complex genetic programmes. Transcription initiation is the major step in the regulation of gene expression in prokaryotes (Browning & Busby, 2004). Although transcription is effected by RNA polymerase, an interchangeable σ factor is responsible for promoter recognition, thus directing the holoenzyme to specific genes (Gruber & Gross, 2003; Browning & Busby, 2004). Bacterial genomes encode a principal σ factor, devoted to the transcription of housekeeping genes, and may also contain a variable number of alternative σ factors (ASFs) (Gruber & Gross, 2003; Browning & Busby, 2004). As σ factors from a given organism generally have distinct promoter specificities, mixing and matching various σ factors with RNA polymerase represents a powerful way of modulating transcription profiles. Gene expression levels are also further modified by the action of transcription activators and repressors (Barnard et al., 2004).

The genome of M. tuberculosis is the largest of the obligate human pathogens and intracellular bacteria. It encodes c. 190 regulatory proteins, including 11 two-component systems, five unpaired response regulators, two unpaired histidine kinases, 11 protein kinases and over 140 other transcription regulators (Tekaia et al., 1999). Some of these gene products have been shown to be involved in the response to environmental stresses, such as cold shock (Shires & Steyn, 2001), heat shock (Manganelli et al., 1999; Stewart et al., 2002), hypoxia (Park et al., 2003), nitric oxide (Voskuil et al., 2003), iron or zinc starvation (Rodriguez & Smith, 2003; Canneva et al., 2005), carbon starvation (Betts et al., 2002), and surface and oxidative stresses (Manganelli et al., 2001b, 2002; Sala et al., 2003). Thirteen σ factor genes have also been predicted from the genome sequence of M. tuberculosis (Cole et al., 1998), and most of those investigated so far have been shown to be important for virulence (Smith, 2003; Manganelli et al., 2004b). In this paper, we summarize current knowledge regarding mycobacterial σ factors and review the latest advances on the role of σ factors in the physiology and virulence of M. tuberculosis.

Structure and classification of σ factors

σ factors can be categorized into two phylogenetically distinct families related to either σ54 or σ70 from Escherichia coli (Lonetto et al., 1992; Buck et al., 2000; Gruber & Gross, 2003). Members of the σ54 family of σ factors are relatively rare and are not found in mycobacteria. However, σ70-related σ factors are encoded in all bacterial genomes. These latter σ factors contain up to four conserved regions (Fig. 1a), which can further be divided into subregions (Lonetto et al., 1992; Gruber & Gross, 2003). Region 1 is located at the amino-terminus and contains subregions 1.1 and 1.2. Region 1.1 usually bears a negative charge and has been reported to inhibit the DNA binding of free σ factors (Dombroski et al., 1993). In some σ factors, a nonconserved region (NCR) connects region 1 to region 2. Region 2 is composed of four subregions. Subregion 2.4 is required in the recognition of the –10 promoter box, whereas subregion 2.3 is involved in the melting of the transcription bubble (Gardella et al., 1989; Siegele et al., 1989; Zuber et al., 1989; Panaghie et al., 2000). Region 3 comprises subregions 3.0, 3.1 and 3.2. Subregion 3.0 is implicated in the recognition of the extended −10 promoter elements (Barne et al., 1997; Sanderson et al., 2003). Finally, region 4 consists of subregions 4.1 and 4.2. This latter subregion is responsible for the recognition of the −35 promoter element and for the interaction with numerous transcription activators (Dove et al., 2003; Gruber & Gross, 2003; Paget & Helmann, 2003). All regions contribute, albeit to different extents, to the binding of the σ factor to RNA polymerase (Sharp et al., 1999; Gruber et al., 2001).

Figure 1.

 Schematic overview of conserved regions in σ factors. (a) Structural organization of σ factors of Groups 1, 2, 3 and 4. The broken lines link σ factor regions to the promoter element they recognize. The arrow indicates the transcription start site. (b) Architecture of conserved regions in Mycobacterium tuberculosis H37Rv σ factors. Numbers at the top and bottom of the coloured boxes indicate the residues at the beginning and end of each conserved region. The amino-terminus of each protein is represented by ‘1’ and the number of the last C-terminus residue is shown on the right side of each σ factor. σM* corresponds to the sequence from M. tuberculosis CDC1551.

σ70-related σ factors have been categorized into four groups on the basis of their structures and physiological roles (Gruber & Gross, 2003; Fig. 1a). Group 1 is composed of principal σ factors, which are essential genes comprising all four conserved regions. However, primary σ factors from Bacteroidetes are structurally closer to Group 2 σ factors and therefore constitute particular cases (Vingadassalom et al., 2005). Group 2 σ factors are only found in a limited number of bacterial species (Proteobacteria, Cyanobacteria and high-GC Gram-positive bacteria) and consist of primary-like σ factors. Although closely related to Group 1, these σ factors are normally not essential under laboratory growth conditions and lack most of region 1 (Fig. 1a). The best characterized Group 2 σ factors are involved in the transcription of general stress response and stationary phase survival genes. However, the function of Group 2 σ factors in high-GC Gram-positive bacteria is still poorly understood. Group 3 σ factors also contain conserved regions 2, 3 and 4, but are more distantly related to principal σ factors. Group 3 σ factors fall into clusters comprising evolutionarily related proteins with similar functions, such as heat shock, sporulation or flagellar biosynthesis. Group 4 is the largest and most heterogeneous collection of σ factors (Gruber & Gross, 2003). Also described as extracytoplasmic function (ECF) σ factors, members of this group contain conserved regions 2 and 4 only. Although the function of most Group 4 σ factors remains unknown, they are often involved in the response to stress conditions, such as iron limitation, oxidative stress and surface stress. Moreover, several σ factors from this category are important for virulence (Raivio & Silhavy, 2001; Bashyam & Hasnain, 2004). For example, AlgU is responsible for the regulation of alginate capsule biosynthesis in Pseudomonas aeruginosa (Helmann, 2002), BtrS regulates the expression of Type III secretion proteins in Bordetella (Mattoo et al., 2004) and σE is essential for intracellular multiplication of Salmonella typhimurium (Humphreys et al., 1999). Several M. tuberculosisσ factors are involved in virulence and are discussed below.

Structure and conserved regions in M. tuberculosisσ factors

Mycobacterium tuberculosis encodes 13 σ factors from all four groups of the σ70 family. Groups 1, 2 and 3 are represented by σA, σB and σF, respectively. The remaining 10 σ factors are part of Group 4 (Cole et al., 1998; Manganelli et al., 2004b). Figure 1b shows a schematic representation of all σ factors in which conserved regions are depicted. Interestingly, a carboxy-terminal extension of c. 120 amino acids is present in σG, σI and σJ, revealing an unusual σ factor architecture. This organization is conserved in the corresponding orthologues from other mycobacteria and in putative σ factors from other actinomycetes. In spite of a weak enrichment for hydrophobic residues, no clear sequence homology is detected between the corresponding C-terminus extension of σG, σI and σJ. No function can be readily associated to these σ factor extra domains, although part of a nuclear transport factor 2 (NTF2) domain is detected in the corresponding region of M. tuberculosisσG and in some mycobacterial orthologues. In eukaryotes, this domain interacts with Ras, a small GTPase protein. It is thus tempting to speculate that the long C-terminus extension found in σG, σI and σJ may constitute a surface for the interaction with proteins or molecules that would participate in the regulation of σ factor activity. It would also be interesting to see if this structure is processed or how it is accommodated in the context of a holoenzyme, provided that these proteins are genuine σ factors.

Comparative genomics of mycobacterial σ factors

The number of σ factors encoded in bacterial genomes is very variable. Data from the latest genome sequencing projects indicate that the number of σ factors varies from one in Mycoplasma species up to 65 in Streptomyces coelicolor (Kill et al., 2005). Although the number of σ factors generally increases with genome size, environmental bacteria and microorganisms that have developed differentiation programmes (i.e. sporulation) tend to have higher ASF/genome size (mega-base pairs, Mb) ratios than most obligate pathogens or commensal bacteria (Fig. 2). This is also in good agreement with the observation that the number of σ factor genes is usually correlated with the diversity of lifestyles encountered by a bacterium (Gruber & Gross, 2003). For example, 28 σ factor-encoding genes were detected in the genome of the saprophytic Mycobacterium smegmatis MC2 155, whereas only four σ factors are still functional in the genome of Mycobacterium leprae, an obligate pathogen. A notable exception is the intestinal symbiont Bacteroides thetaiotaomicron, which has the highest ASF/Mb ratio (7.8). Alternative σ factors in B. thetaiotaomicron could be involved in the regulation of components of its glycobiome, perhaps by imparting the ability to opportunistically retrieve and metabolize different types of glycans depending on their availability in the host niche (Xu et al., 2004). The genomes of nine mycobacterial strains have been fully sequenced to date and over 15 are underway. With the exception of M. leprae, which shows a very low ASF/Mb ratio (0.9), mycobacteria have relatively high ratios (2.7 for M. tuberculosis and Mycobacterium avium ssp. paratuberculosis, 3.6 for M. smegmatis). Interestingly, M. tuberculosis is the obligate pathogen with the highest ASF/Mb ratio (Fig. 2). A possible explanation for this could be that M. tuberculosis is a relatively young pathogen and that the high number of ASFs constitutes an evolutionary remnant from a previous lifestyle. Another possibility is that the events leading to a successful infection by M. tuberculosis are complex and require sophisticated regulatory mechanisms. This latter hypothesis is supported by the fact that all, but one, σ factor mutant strains analysed so far are attenuated for virulence, thus suggesting an important role of these genes in the biology of this bacterium.

Figure 2.

 Number of alternative σ factors per mega-base pairs (ASF/Mb) in various bacterial species.

Most M. tuberculosisσ factors have orthologues in the genomes of other mycobacteria (Fig. 3).

Figure 3.

 Phylogenetic tree comprising putative σ factor gene products from Mycobacterium tuberculosis (Mtb), Mycobacterium bovis AF2122/97 (Mb), Mycobacterium marinum (MM), Mycobacterium avium paratuberculosis K10 (MAP), Mycobacterium smegmatis MC2 155 (MSMEG) and Mycobacterium leprae (ML). Clusters were identified with respect to M. tuberculosisσ factors. The phylogenetic tree was produced using the drawgram application from the phylip 3.65 package (Felsenstein, 2005).

The genome of Mycobacterium bovis AF2122/97 is over 99.95% identical to the chromosome of M. tuberculosis (Garnier et al., 2003). All 13 σ factors are conserved, with very few amino acid substitutions. The genome sequence of M. bovis BCG Pasteur is currently being determined, and results obtained up to now reveal a striking homology with the σ factors of M. tuberculosis. However, the start codon of σK is mutated from a methionine to an isoleucine (M1I), resulting in the absence of the protein. Charlet et al. (2005) have reported that this mutation is found in several other M. bovis BCG strains sent from the Pasteur Institute in the 1927–1931 interval. A deletion of one gene has also occurred (nRD18) in M. bovis BCG Pasteur, leading to the fusion of σI to the next downstream gene, a putative alpha/beta hydrolase (Mostowy et al., 2003). The sigM and sigH genes have also been reported to be duplicated in M. bovis BCG Pasteur (Brosch et al., 2000; Manganelli et al., 2004b).

The genome of Mycobacterium microti, which is part of the M. tuberculosis complex, is also being sequenced. The only major difference in σ factor-encoding genes between this species and M. tuberculosis is a nonsense mutation (E122X) in σF. Considering that an M. tuberculosis sigF mutant strain is attenuated for virulence (Geiman et al., 2004), it is possible that this mutation contributes to the lower pathogenicity of M. microti with respect to the other members of the M. tuberculosis complex.

Some variation exists in σM, with most of the σM region 4.2 altered in M. tuberculosis H37Rv with regard to the other mycobacteria (Fig. 1b). The introduction of a cytosine at position 476 in the H37Rv strain causes a frameshift that modifies residues in region 4.2 and displaces the stop codon 78 bp downstream of its original location. Alignments with various sigM paralogues indicate that this mutation is restricted to the M. tuberculosis H37Rv strain. However, another mutation is observed in the same region of M. bovis AF2122/97 sigM, in which a substitution of the CG dinucleotide at position 477–478 also causes a frameshift in region 4.2. As the M. tuberculosis H37Rv and M. bovis AF2122/97 strains are fully virulent, it is likely that these mutations do not affect the activity of σM, or that σM is not required for virulence.

All M. tuberculosisσ factors are represented in the genome of M. smegmatis MC2 155, with the exception of σC and σK (Waagmeester et al., 2005). Of 28 putative M. smegmatis MC2 155 σ factor-encoding genes, seven are related to σH, five to σJ, two to σL, three are not significantly homologous to any σ factor of M. tuberculosis and the remaining 11 have single orthologues in M. tuberculosis (Fig. 3). It is likely that the relatively large number of predicted ASFs in M. smegmatis reflects the saprophytic lifestyle of this organism.

Of a total of 19 predicted σ factor-encoding genes, M. avium paratuberculosis K10 (Li et al., 2005) lacks sigK but encodes five gene products showing some homology to σJ and two that are similar to σF (Fig. 3).

We have also been able to identify 16 putative σ factors from the unfinished genome sequence of Mycobacterium marinum (Fig. 3). Interestingly, M. marinum is apparently the only mycobacterial species not included in the M. tuberculosis complex to have a functional σK, suggesting that this protein may regulate cellular functions restricted to some disease-causing mycobacteria. Three genes were also found to be related to σJ and two to σC. However, no orthologue of σI has yet been detected in this bacterium. It is interesting to note that many predicted σ factors appear to belong to few subfamilies, such as σH- or σJ-related proteins in M. smegmatis and σJ-like proteins in M. avium. The reasons for this are still unknown. It is possible that the genes encoding these σ factors are located in close proximity to recombination hot spots, favouring their duplication. It would be extremely interesting to see whether the functions of related σ factors from a given organism are distinct or overlapping.

As mentioned above, only four σ factor-encoding genes are still functional in M. leprae (Cole et al., 2001). Recently, it has been proposed that the loss of different σ factors may have triggered the accumulation of pseudogenes in this organism (Madan Babu, 2003). Indeed, 1604 ORFs and 1116 pseudogenes are found in the genome of M. leprae (for a comparison, M. tuberculosis has 3959 ORFs and only six pseudogenes). Orthologues of M. tuberculosisσA, σB, σC and σE are nonetheless conserved in spite of this massive gene decay (Fig. 3). sigL is not detected, even as a pseudogene, in the chromosome of M. leprae.

Promoters recognized by mycobacterial σ factors

Relatively few mycobacterial promoters reported in the literature can be unambiguously associated to a σ factor (Jacques et al., 2005). A σA consensus promoter sequence has been proposed based on the similarity of some mycobacterial promoters to the σ70 consensus in E. coli (Unniraman et al., 2002). As regions 2.4 and 4.2, involved in the recognition of the −10 and −35 promoter elements, respectively, are very well conserved between E. coli and M. tuberculosis, it is likely that most σA-dependent promoters resemble the σ70 consensus. As σB is highly homologous to σA, it would not be surprising if their respective regulons at least partially overlapped. In support of this hypothesis, σA and σB have been reported to initiate transcription at the same nucleotide at the Bacillus subtilis sinP3 promoter (Predich et al., 1995; Jacques et al., 2006).

Genes affected by the deletion of sigC have also been identified by comparing the gene expression profiles of a sigC mutant with the parental M. tuberculosis CDC1551 strain (Sun et al., 2004). Approximately 200 genes have been reported to be differentially expressed at different points of the growth curve. The upstream regions of 82 sigC mutant early log phase down-regulated genes were used to search for promoter sequences potentially recognized by σC. A degenerated σC consensus sequence [SSSAAT-N(16–20)-CGTSSS] was proposed upstream of 18 of these genes (Sun et al., 2004). A logo representing the position specific weight of nucleotides in the −35 and −10 promoter boxes of the 18 predicted σC-dependent promoters is shown in Fig. 4. According to this logo, only three nucleotides from the −10 and −35 promoter elements significantly contribute to promoter discrimination by σC. As the authors have not identified any transcriptional start sites to support their predictions, further work will be needed to assess the validity of these proposed σC-dependent promoters.

Figure 4.

 Logos representing promoter sequences proposed to be recognized by various Mycobacterium tuberculosisσ factors. The y-axis of each logo corresponds to the bits of information. Allowed spacing ranges are indicated by the numbers in brackets. Appropriate references are shown for each σ factor. Logos were generated by submitting all MtbRegList database (Jacques et al., 2005) σ factor-associated sequences or suggested promoter sequences from the cited references to WebLogo (Crooks et al., 2004).

Two groups have generated sigD mutants and have measured the differences in gene expression relative to the corresponding wild-type strain using DNA microarrays. However, the results reported by the two laboratories are in strong contrast, making it difficult to define the σD regulon and the prototypic σD-dependent promoter. Calamita et al. (2005) reported 61 down-regulated genes in their sigD mutant, whereas 51 were reported by Raman et al. (2004). Of these genes, only three (Rv1815, Rv0440 and Rv1883c) were affected in the two sigD mutants. Moreover, two genes (Rv1297 and Rv0704) were reported to be down-regulated by Calamita et al., but were up-regulated in the other study. These data were obtained using two different genetic backgrounds, M. tuberculosis H37Rv and CDC1551. Nonetheless, the important lack of overlap between the two studies is still difficult to justify. The proposed σD promoter consensus sequences are also hard to reconcile (Fig. 4).

Transcriptome profiling of a sigE mutant relative to the wild-type strain has also been used to identify genes that could be regulated by σE. Thirty-eight genes required σE for their full expression during exponential growth and 23 genes were not appropriately induced in the absence of sigE during sodium dodecylsulphate (SDS)-mediated surface stress (Manganelli et al., 2001b). As σE shows a significant homology to S. coelicolorσR, we have searched for the proposed σR consensus sequence (Paget et al., 2001) upstream of genes identified in the screening. On the basis of these data, a σE-dependent promoter consensus has been proposed (Fig. 4).

A σF-dependent promoter was first identified using in vitro transcription assays at the usfXP1 promoter (Beaucher et al., 2002). In a more recent study, a sigF deletion mutant was constructed and transcription profiles were obtained at various points of the growth curve (Geiman et al., 2004). The authors used the sequence of the usfXP1 promoter boxes to conduct a DNA motif search upstream of genes that were down-regulated in the sigF mutant. By restricting the number of mismatches to a total of three in the −10 and −35 promoter boxes, 14 genes were proposed to be directly regulated by σF and a consensus promoter sequence was proposed (Fig. 4). However, it is unknown whether the inhibition of σF activity by its cognate anti-σ factor UsfX (see below) was relieved under these experimental conditions. Recently, the sigB promoter has been reported to be recognized by a σF-containing holoenzyme in in vitro transcription assays. The sequence of this σF-dependent promoter was significantly similar to the usfXP1 sequence (Dainese et al., 2006). However, the expression of sigB was not significantly affected in the absence of σF under the growth conditions used by Geiman et al. (2004).

sigH mutant strains have been generated in three different laboratories (Raman et al., 2001; Kaushal et al., 2002; Manganelli et al., 2002). The sigH promoter was first shown to be autoregulated in M. smegmatis, and σH-dependent promoters were next identified in M. tuberculosis (Raman et al., 2001). Microarray experiments also revealed genes differentially regulated in the absence of sigH at different points of the growth curve (Kaushal et al., 2002) and after diamide stress (a thiol-specific oxidizing agent) (Manganelli et al., 2002). All three studies proposed an almost identical σH promoter consensus sequence, which was also related to the sequence recognized by S. coelicolorσR (Paget et al., 2001; Fig. 4).

σL-dependent promoters have recently been identified by two groups (Hahn et al., 2005; Dainese et al., 2006). In one approach, sigL was expressed from an acetamide-inducible promoter, whereas, in the other, the anti-σL anti-σ factor gene (see below) was disrupted to mimic σL-inducing conditions. The transcription start sites of genes up-regulated under these conditions were determined and an identical σL consensus sequence was proposed (Fig. 4). In vitro transcription assays performed by both groups supported these findings. Interestingly, the promoters recognized by σE, σH or σL can be very similar. In fact, these three σ factors use the same promoter to initiate the transcription of sigB (Dainese et al., 2006). However, no other genes appear to be regulated by more than one of these σ factors, suggesting that subtle differences may allow the discrimination of promoters by σE, σH and σL.

Posttranslational regulation of mycobacterial σ factors

The modulation of σ factor levels is a simple mechanism to regulate gene transcription profiles. However, the activity of some σ factors can be further regulated by antagonist proteins. Anti-σ factors directly interact with specific σ factors. The resulting sequestering of σ factor hinders their association to RNA polymerase, thus preventing a potential effect on gene expression profiles until an appropriate stimulus is sensed by the bacterium. To date, four anti-σ factors have been identified in the genome of M. tuberculosis (Fig. 5). Of these, RseA, RshA and RslA are part of the zinc-associated anti-σ factor (ZAS) family (Paget et al., 2001). These latter proteins contain an HXXXCXXC motif found in many anti-σ factors and are located downstream of their cognate σ factor-encoding gene (Gehring et al., 2001; Newman et al., 2001; Paget et al., 2001).

Figure 5.

 Regulatory network of σ factors. Posttranslational regulators and transcriptional relationships among σ factors are shown. →, transcriptional regulation; T, posttranslational regulation.

The rseA gene (Rv1222) encodes a σE-specific anti-σ factor (S. Rodrigue and L. Gaudreau, manuscript in preparation). It is preceded by sigE in a hypothesized operon also comprising htrA, coding for a putative membrane serine protease, and tatB, coding for a putative protein belonging to the Tween Arginine Translocator (Tat) secretion system. Similarly, the E. coli surface stress-responsive σ factor RpoE is part of an operon containing a transmembrane anti-σ factor and a membrane-bound serine protease that regulate its activity in response to the presence of misfolded proteins in the periplasmic space (Raivio & Silhavy, 2001). However, several experiments designed to detect interactions between RseA, HtrA and TatB were unsuccessful (A. Cascioferro, E. Dainese, R. Manganelli and R. Provvedi, unpublished results). The signal and the potential effectors involved in the release of σE by RseA remain unknown.

RshA is a σH-specific antagonist that is capable of directly sensing redox potential through specific cysteine residues (Song et al., 2003). In a reducing environment, RshA binds σH, preventing its interaction with the RNA polymerase. In contrast, disulphide bonds can form between RshA cysteine residues under oxidizing conditions, resulting in the release of σH that becomes free to interact with RNA polymerase (Song et al., 2003). Heat treatment has also been reported to disrupt the RshA–σH interaction (Song et al., 2003).

RslA is a transmembrane protein also bearing the ZAS anti-σ factor family HXXXCXXC motif. RslA has recently been shown to bind to σL (Hahn et al., 2005). This binding has been shown to be specific and is able to inhibit σL-dependent transcription in vitro (Dainese et al., 2006). The gene encoding σL is transcribed from two different promoters. One is located inside the coding region of mapA, the sigL upstream gene, and is expressed constitutively during all growth phases. The second promoter is σL dependent (Hahn et al., 2005; Dainese et al., 2006). σL would thus be expressed but kept inactive by RslA until an appropriate signal was detected by the bacterium. As RslA is a transmembrane protein, it is tempting to speculate that the signal leading to the release of σL by RslA originates from the cell envelope or from the extracellular environment.

sigF is cotranscribed with its upstream gene, usfX (Rv3285c, also named rsbW), which encodes a σF-specific anti-σ factor. UsfX has been shown to directly interact with σF and to inhibit the σF-dependent transcription from the usfXP1 promoter (Beaucher et al., 2002). Interestingly, UsfX is posttranslationally regulated by two anti-anti-σ factors: RsfA and RsfB (Beaucher et al., 2002). Both anti-σ factor antagonists are able to bind to UsfX and would disrupt the UsfX–σF complex, thus ensuring the release of σF and transcription from σF-dependent promoters. RsfA interacts with UsfX under reducing conditions. However, a disulphide bridge formed between two cysteine residues would prevent this interaction under oxidative conditions (Beaucher et al., 2002). The activity of RsfB has been proposed to be controlled by phosphorylation, but the putative kinase and the stimulus leading to this modification are still unknown (Beaucher et al., 2002).

Although no experimental data support these hypotheses, Raman et al. (2004) have proposed that σD could be regulated by an anti-σ factor, and the existence of a putative σK anti-σ factor has also been postulated (Charlet et al., 2005). Posttranslational modifications of σ factors, such as phosphorylation and proteolytic maturation of pro-σ factors, have also been reported in other bacteria (Stragier et al., 1988; Hughes & Mathee, 1998; Klein et al., 2003). No such modifications have been reported to date for mycobacterial σ factors.

Physiological roles of M. tuberculosisσ factors

σA (Group 1)

Both σA and its Mycobacterium smegmatis orthologue are essential for bacterial viability (Gomez et al., 1998) (I. Smith, The Public Health Research Institute, Newark, NJ; pers. commun.). The sigA structural gene is constantly expressed in H37Rv and has often been used as an invariant internal standard in quantitative reverse transcriptase-polymerase chain reaction (RT-PCR) experiments (Manganelli et al., 1999, 2001a; Dubnau et al., 2002). Hu & Coates (1999) reported that the sigA mRNA half-life is unusually long (more than 40 min) when compared with 2.4 min for the sigB mRNA. Interestingly, it has recently been shown that, in some clinical strains, but not in H37Rv, expression of sigA increases after phagocytosis. This overexpression appears to be responsible for an enhanced intracellular growth and an increased resistance to superoxides, suggesting that σA modulates the expression of virulence genes (Wu et al., 2004). In support of this hypothesis, an arginine to histidine substitution at residue 515 (R515 H) of σA in M. bovis ATCC35723 was shown to cause an attenuation of virulence in a guinea pig model of infection (Steyn et al., 2002). However, the same mutation did not confer attenuation in a different animal model (Australian brushtail possum) (Collins et al., 2003). The R515H mutation is located at the C-terminus of the protein, in subregion 4.2, which is known to interact with transcriptional activators in other bacteria (Dove et al., 2003). Using yeast two-hybrid assays, Steyn et al. (2002) showed that this domain interacts with WhiB3, a putative transcription regulator, and that this interaction is prevented by the R515 H substitution. Inactivation of the WhiB3 structural gene conferred the same phenotype to a wild-type M. bovis strain relative to the sigA R515H mutant strain, suggesting that the σA–WhiB3 interaction was the cause of the observed attenuation. Interestingly, whiB3 inactivation in M. tuberculosis resulted in only a partial attenuation of virulence, implying that different genetic backgrounds play an important role in the expression of this phenotype, or that additional transcription activator(s) may contact this surface of σA in M. tuberculosis. Recently, the R515H substitution was associated to a new phenotype: the inability to grow at low phosphate concentrations (Collins et al., 2003). Taken together, these findings suggest that, in addition to its primary σ factor role, σA is involved in the modulation of specialized functions that are essential for virulence.

σB (Group 2)

σB is the sole Group 2 σ factor of M. tuberculosis. It is dispensable for growth in both M. smegmatis and M. tuberculosis (Mukherjee et al., 2005) (I. Smith, The Public Health Research Institute, Newark, NJ; pers. commun.). The sigB gene is located about 3 kb downstream of sigA in all mycobacterial species tested so far (Doukhan et al., 1995). σB is very similar to the C-terminal portion of σA that contains conserved regions 2, 3 and 4, and includes the subregions involved in promoter recognition. It is interesting to note that this is also the case with σ70 and σ38, the Group 1 and Group 2 σ factors, respectively, in E. coli (Tanaka et al., 1993). An M. tuberculosis sigB knockout mutant is more sensitive to various stresses, such as SDS-induced surface stress, heat shock, oxidative stress (I. Smith, The Public Health Research Institute, Newark, NJ; pers. commun.) and exposure to vancomycin (R. Provvedi, manuscript in preparation), suggesting an important role in stress responses. However, a σB mutant was still able to grow normally in human macrophages (I. Smith, The Public Health Research Institute, Newark, NJ; pers. commun.). Experiments to determine the role of σB in an animal model of infection are in progress.

Mukherjee et al. (2005) have recently demonstrated that overexpression of M. tuberculosis sigB in M. smegmatis results in a prolonged generation time and a marked change in colony morphology. This phenotype was linked to the constitutive production of surface hyperglycosylated polar glycopeptidolipids that are usually produced in response to carbon starvation, suggesting an additional role of σB in the adaptation to stationary phase or to growth in nutritionally poor environments.

The transcriptional regulation of sigB is extremely complex. In vitro transcription experiments have recently shown the presence of two different promoters upstream of sigB. One of them is recognized by σF-containing RNA polymerase, and the other is recognized by RNA polymerase containing either σE, σH or σL, suggesting that σB is required for adaptation to multiple environmental conditions (Dainese et al., 2006). σ factors often self-regulate their own expression. However, σB-containing RNA polymerase does not initiate transcription upstream of sigB (Dainese et al., 2006). Under standard physiological growth conditions, sigB transcription is almost completely σE dependent (Manganelli et al., 2001b). sigB transcription is induced after exposure to heat shock, SDS-mediated surface stress and oxidative stress mediated by diamide. The induction following heat shock or oxidative stress is mediated by σH (Manganelli et al., 2002), whereas σE is responsible for sigB induction in response to surface stress (Manganelli et al., 2001b). The conditions leading to sigB transcription by σL or σF are currently unknown. Recently, the response regulator MprA has been shown to bind upstream of sigB, with a small effect on transcription during growth under physiological conditions (He et al., 2006).

In sum, although the precise physiological role of σB is still unclear, as for the other Group 2 σ factors of high-GC Gram-positive bacteria, it seems reasonable to predict a role for σB in the general stress response and in adaptation to stationary phase and carbon starvation.

σF (Group 3)

The sigF gene was first thought to be present only in the genomes of some slow-growing pathogenic mycobacteria, prompting research on this σ factor (DeMaio et al., 1996). However, it is now clear that sigF is found in all sequenced mycobacterial genomes. When introduced in M. bovis BCG, the expression of the M. tuberculosis sigF gene is induced on exposure to several antibiotics (rifampin, ethambutol, cycloserine and streptomycin), anaerobiosis, cold shock, oxidative stress, nutrient depletion and after entry into stationary phase. The strongest induction was observed under anaerobic conditions with the addition of metronidazole (DeMaio et al., 1996; Michele et al., 1999). Less information is available regarding the expression of sigF in M. tuberculosis. In this latter bacterium, sigF transcription levels were measured after exposure to oxidative stress, nutrient depletion, cold shock, anaerobiosis, and after entry into stationary phase. An increase in sigF transcription was only detected in response to nutrient depletion (Manganelli et al., 1999; Betts et al., 2002), suggesting a different regulation of this gene in M. tuberculosis and M. bovis BCG despite the close similarity of these organisms (Graham & Clark-Curtiss, 1999; Manganelli et al., 1999).

The gene encoding σF was disrupted in M. tuberculosis, and the resulting mutant was able to grow to a threefold higher density in stationary phase than the wild-type strain. Moreover, the sigF mutant did not exhibit any lag phase after being diluted from a dense culture into fresh medium (Chen et al., 2000). The sigF mutant showed the same sensitivity to heat shock, cold shock, hypoxia and long-term stationary phase relative to the parental strain. However, the mutant strain was shown to be more permeable to a hydrophobic solute, suggesting that σF regulates components of the mycobacterial envelope. In support of this hypothesis, the sigF mutation was shown to confer negative neutral red staining, suggesting a reduced synthesis of cell wall-associated sulpholipids (Chen et al., 2000). These findings may, in turn, explain the eightfold increased sensitivity of the sigF mutant to rifampin.

When used to infect human monocytes, the sigF mutant did not show any difference from the wild-type (Chen et al., 2000). Nevertheless, it was attenuated for virulence in a mouse infection model. In the absence of sigF, the mutant strain was able to persist in lung tissues, but the bacterial load in the organs was lower than that observed with the parental strain. Moreover, histopathological analyses of lungs and spleen confirmed minor disease progression in mice infected with the mutant strain (Geiman et al., 2004).

In order to identify genes requiring σF for their expression, total RNA from the sigF mutant and from the wild-type strain was compared using DNA microarrays. Thirty-eight genes were down-regulated in the sigF mutant in the exponential phase, 187 in early stationary phase and 277 in late stationary phase, suggesting a major role of σF in the adaptation to stationary phase (Geiman et al., 2004). About 50% of the down-regulated genes encoded hypothetical proteins or proteins of unknown function. Additional genes were involved in the biosynthesis and structure of the cell envelope, and others encoded transcriptional regulators of the MarR, GntR and TetR families. Finally, the σF structural gene and the gene encoding the Group 4 σ factor σC were down-regulated in the mutant, suggesting a complex regulatory cascade (Geiman et al., 2004).

σC (Group 4)

σC and σE are the only Group 4 (ECF) σ factors encoded by the M. leprae genome (Cole et al., 2001), suggesting an important role in mycobacterial physiology and virulence. In a previous study, we have shown that sigC is expressed during exponential phase, and that the level of its mRNA strongly decreases during stationary phase and after exposure to different stress conditions, such as heat shock, cold shock, oxidative stress and surface stress (Manganelli et al., 1999). An M. tuberculosis mutant lacking a functional sigC has been described recently (Sun et al., 2004). The survival of the resulting strain was not affected in activated mouse macrophages relative to the wild-type strain. However, the sigC mutant was significantly attenuated in mouse aerosol infection time-to-death assays. Interestingly, the sigC mutant strain was able to reach the same bacterial load as the wild-type strain in mouse lung (Sun et al., 2004). However, histopathological analyses revealed that mutant-infected mice had reduced inflammatory infiltrates when compared with mice infected with the wild-type or complemented strains. This could be explained by a lower toxicity of the sigC mutant strain, rather than by a hampered capacity to proliferate in the host tissues. The authors also showed that the mutant was more sensitive than the wild-type to mechanical killing caused by the process of nebulization, suggesting a difference in the cell envelopes of the two strains (Sun et al., 2004).

Using DNA microarrays, about 200 genes were reported to be down-regulated in the sigC mutant at different points of the growth curve. In addition to sigC, some virulence-associated genes, such as hspX, senX3 and mtrA, encoding the α-crystalline homologue, a two-component sensor kinase and a two-component response regulator, respectively, were also down-regulated in the mutant strain. It is unclear whether these genes are directly regulated by σC.

σD (Group 4)

Hyperphosphorylated guanidine (p)ppGpp is used as a signalling molecule in the stringent response, which modulates bacterial gene expression for long-term survival under starving conditions (Magnusson et al., 2005). In M. tuberculosis, (p)ppGpp is produced by a gene required for virulence, RelM. tuberculosis (Dahl et al., 2003). sigD has recently been suggested to be part of the RelM. tuberculosis regulon (Dahl et al., 2003), and is induced during starvation (Betts et al., 2002). A sigD mutant strain of M. tuberculosis was independently constructed and characterized by two laboratories in the H37Rv and CDC1551 strains (Raman et al., 2004; Calamita et al., 2005). The CDC1551-derived mutant strain was still able to survive and grow in resting and activated macrophages, but induced a lower level of tumour necrosis factor-α (TNF-α) production by macrophages relative to the wild-type strain (Calamita et al., 2005). When used to infect mice, both sigD mutant strains were able to grow at the same level as the wild-type in the organs. However, the histopathology and ability to kill the animals were strongly reduced in the absence of sigD (Raman et al., 2004; Calamita et al., 2005).

σE (Group 4)

The gene encoding σE is induced after exposure to various environmental stress conditions, such as heat shock, SDS detergent or vancomycin-mediated cell surface stress (Manganelli et al., 1999; R. Provvedi, manuscript in preparation) and diamide oxidative stress (Raman et al., 2001; Manganelli et al., 2002), as well as during the growth of M. tuberculosis in human macrophages (Graham & Clark-Curtiss, 1999; Schnappinger et al., 2003). In accordance with these observations, disruption of sigE in M. tuberculosis H37Rv leads to an increased sensitivity to heat shock, SDS, oxidative stress (Manganelli et al., 2001b) and vancomycin (R. Provvedi, manuscript in preparation). The disruption of sigE in M. tuberculosis H37Rv impaired the growth of the resulting strain in resting THP-1-derived macrophages (Manganelli et al., 2001b) and in human dendritic cells (Giacomini et al., 2006). Interestingly, when used to infect dendritic cells, the mutant strain was still able to promote their differentiation, but stimulated the production of about tenfold more interleukin-10 (IL-10) than the wild-type parental strain, suggesting a different interaction with the immune system (Giacomini et al., 2006). The M. tuberculosis H37Rv sigE mutant was also shown to be attenuated in BALB/C mice (using bacterial load in the organs and histopathology as criteria) and in severe combined immunodeficient (SCID) mice (using death as a criterion) (Manganelli et al., 2004a). Attenuation in mice of a sigE mutant was also studied in the M. tuberculosis CDC1551 genetic background with similar results (Ando et al., 2003).

sigE is transcribed by multiple promoters (Wu et al., 1997). One of these has a clear σH consensus sequence and has been shown to be responsible for the σH-dependent induction of sigE after heat shock and exposure to diamide (Raman et al., 2001). Recently, the induction of sigE after SDS-mediated surface stress (as well as after treatment with Triton X-100 or alkaline pH) has been shown to depend on the MprA/MprB two-component system. Interestingly, these two genes are part of the σE regulon as described below (He et al., 2006).

Genes regulated by σE on surface stress have been identified by DNA microarray analyses (Manganelli et al., 2001b) and encode proteins involved in fatty acid degradation, such as the isocitrate lyase icl1, whose role in virulence has been clearly established (Munoz-Elias & McKinney, 2005), fadE23 and fadE24. These last two genes are induced by isoniazide (Wilson et al., 1999) and may function as components of a shunt pathway that degrades/detoxifies the fatty acids that accumulate as a result of the inhibition of mycolic acid biosynthesis. Their σE-dependent induction following surface stress supports the hypothesis of their role in the biology of the mycobacterial cell wall. Other σE-controlled genes encode transcriptional regulators, including σB, the two-component regulatory system composed of MprA and MprB (Zahrt & Deretic, 2000), heat shock proteins, such as HtpX and Hsp, and surface exposed proteins with unknown function. The induction of a putative operon containing Rv2744c (known as the 35-kDa antigen) was also observed. The product of this gene has some homology to PspA, a protein involved in the extracytoplasmic stress response in Proteobacteria, and may play a role in maintaining cytoplasmic membrane integrity and/or the proton-motive force (Darwin, 2005).

σH (Group 4)

In M. tuberculosis, the gene encoding σH is induced after heat shock (Manganelli et al., 1999; Raman et al., 2001), after exposure to the thiol-specific oxidizing agent diamide (Raman et al., 2001; Manganelli et al., 2002) and during macrophage infection (Graham & Clark-Curtiss, 1999). The σH-encoding gene was inactivated in two different genetic backgrounds: H37Rv (Raman et al., 2001; Manganelli et al., 2002) and CDC1551 (Kaushal et al., 2002). sigH was shown to be essential for survival at high temperature and in oxidative environments (Raman et al., 2001; Manganelli et al., 2002). However, it was dispensable for growth in THP-1-derived macrophages and for resistance to the bactericidal activity of activated murine macrophages (Manganelli et al., 2002). When used to infect mice, the sigH mutant strains were able to grow in the mouse organs at the same bacterial load as the parental strains (Kaushal et al., 2002; Manganelli et al., 2002). However, the animals infected with the mutant strain showed a reduced histopathology, with smaller and less abundant granulomas (Kaushal et al., 2002). By studying the cellular immune response of infected mice, the same authors demonstrated that, despite the high bacterial burden, the mutant produced a blunted, delayed pulmonary inflammatory response, and recruited fewer CD4+ and CD8+ T cells to the lung in the early stages of infection. Finally, the mutant strain killed the animal more slowly than the wild-type strain (Kaushal et al., 2002).

Genes regulated by σH include its own structural gene and genes encoding σE, σB, DNA repair proteins, general stress response proteins, enzymes involved in thiol metabolism, such as thioredoxin and thioredoxin reductase, enzymes involved in cysteine and molybdopterin biosynthesis, and Rv2466c, a protein of unknown function with a glutaredoxin active site (Manganelli et al., 2002). Glutaredoxin is an enzyme involved in the glutathione cycle, an alternative pathway to reduce intracellular disulphide bonds. However, actinomycetes do not synthesize glutathione but utilize mycothiol instead (Fahey, 2001). It is consequently reasonable to hypothesize that Rv2466c is involved in the mycothiol cycle. In support of the possible involvement of σH in the mycothiol cycle, the S. coelicolor mutant lacking the σH orthologue produces less mycothiol than the corresponding parental strain (Paget et al., 1998).

σJ (Group 4)

The gene encoding σJ has been shown to be strongly induced in stationary phase cultures (Hu & Coates, 2001) and during growth in human macrophages (Cappelli et al., 2006). Recently, a sigJ mutant was obtained in H37Rv (Hu et al., 2004). Unexpectedly, the ability to survive in prolonged stationary phase cultures was not affected by the inactivation of sigJ. Moreover, when the mutant strain was used to infect intravenously BALB/c mice, its growth rate in lung and spleen was the same as the wild-type strain. The two strains were also able to reach the same bacterial burden in the host tissues, suggesting that σJ is not involved in virulence, at least in this mouse model of infection. However, the mutant was more sensitive than the wild-type strain to H2O2, suggesting a role of σJ in the oxidative stress response (Hu et al., 2004).

σK (Group 4)

In a recent paper, Charlet et al. (2005) have demonstrated that the decreased expression of two antigenic proteins (MPB70 and MPB83) observed in some strains of M. bovis BCG is due to a sigK polymorphism causing a mutation in the translation start codon. By complementing the mutated strains with a wild-type copy of sigK and performing DNA microarray experiments, the authors identified two chromosomal loci including six and seven genes regulated by σK, respectively. The first locus contained the genes encoding MPB70 and MPB83 and other surface-associated or secreted proteins of unknown function. The second locus included the genes encoding σK, a putative amine oxidase, a cyclopropane-fatty-acyl-phospholipid synthase and four other proteins of unknown function. Interestingly, sigK was part of a locus lost in an attenuated M. bovis strain obtained by signature-tagged mutagenesis (Collins et al., 2005). The sigK-encoding gene is present in all M. tuberculosis complex species and in M. marinum, but not in other mycobacteria, such as M. avium ssp. paratuberculosis and M. smegmatis.

σL (Group 4)

A sigL mutant of M. tuberculosis H37Rv was independently obtained in two different laboratories. When the sigL mutants were used to infect mice, they were shown to grow in the organs at the same rate as the wild-type parental strain, but the lethality of the infection was strongly reduced (Hahn et al., 2005; Dainese et al., 2006). sigL mutation also resulted in higher sensitivity to the detergent SDS and to the superoxide generator plumbagine. However, exposure to these two compounds did not result in the induction of sigL transcription (Dainese et al., 2006).

In order to characterize the σL regulon, a mutant strain lacking both sigL and rslA was complemented by integrating a single wild-type copy of sigL into its chromosome. In the resulting strain, the RslA anti-σ factor is absent and σL is continuously available to associate with RNA polymerase and constitutively drive the transcription of its target genes. The transcription profile of the partially complemented strain was compared with that of the sigL-rslA mutant on DNA microarrays, and 28 induced genes, included in 13 putative transcriptional units, under σL control were identified. Several of these transcriptional units contained genes encoding proteins involved in cell envelope-related processes, such as an operon including genes (pks10, pks7, pks8, pks17, pks9 and pks11) probably involved in the biosynthesis of dimycocerosyl phthiocerol (a component of the cell envelope that is essential for virulence) (Sirakova et al., 2003), and another operon containing three genes, two of which are probably involved in fatty acid transport (mmpL13a and mmpL13b) (Sonden et al., 2005). Other genes under σL control included its own structural gene, the gene encoding the Group 2 σ factor σB, and two proteins probably involved in the control of the redox state of exported proteins (Dainese et al., 2006). Hahn et al. (2005) obtained similar results using a bacterial strain overexpressing sigL from an acetamide-inducible promoter.

σM, σG and σI (Group 4)

Practically no information is available in the literature for these three σ factors. Recently, a strong induction of sigG was reported during the infection of human macrophages (Cappelli et al., 2006).

Concluding remarks

During the past few years, mycobacterial σ factors have become the subject of great attention. Their critical role in the physiology and pathogenicity of M. tuberculosis is widely accepted. Nonetheless, the understanding of σ factor physiology in M. tuberculosis is just beginning. Defined in vitro-inducing conditions are known for only three M. tuberculosisσ factors (σE, σH and σB) (Manganelli et al., 2004b). This represents a serious obstacle for the identification of σ factor regulons by differential gene expression profiling, which, in turn, impairs our ability to infer the physiological function(s) regulated by each σ factor. Investigators have attempted to circumvent this problem by comparing the transcriptomes of σ factor mutant strains with the corresponding parental strains in the absence of specific inducing conditions. However, the relevance of these data is difficult to assess, as only a relatively small fraction of the differentially expressed genes is likely to be directly regulated by the σ factor of interest. Alternative methodologies, such as overexpression of a specific σ factor, may constitute an interesting option. However, novel approaches, such as chromatin immunoprecipitation, which allows the detection of direct transcription factor binding sites in vivo, would be highly desirable (Lee et al., 2005). In this regard, we have begun to determine the genome-wide location of σ factor binding sites in M. bovis BCG, an attenuated strain closely related to M. tuberculosis (S. Rodrigue, J. Brodeur, A. Gervais, P-É Jacques, R. Brzezinski & L. Gaudreau, manuscript in preparation).

One of the most interesting and challenging aspects of σ factor biology is the great complexity of regulatory networks, which is well represented in M. tuberculosis by the regulation of the sigB gene (Fig. 5). sigB can be transcribed by RNA polymerase holoenzymes containing any of four different σ factors, three recognizing the same promoter (Dainese et al., 2006) σE is responsible for the basal expression of sigB and for its induction on surface stress (Manganelli et al., 2001b); σH is responsible for sigB, sigE and sigH induction following oxidative stress and heat shock (Manganelli et al., 2002). Both σL and σF can lead to the transcription of their own structural genes and of sigB under yet unknown conditions (Dainese et al., 2006). σE, σH, σL and σF are also posttranslationally regulated by specific anti-σ factors, and the σF-specific anti-σ factor is further regulated by two specific anti-anti-σ factors (Beaucher et al., 2002; Dainese et al., 2006). Moreover, the response regulator MprA has recently been shown to bind upstream of sigB and sigE, hence regulating their basal level of expression and sigE induction following surface stress (He et al., 2006). Interestingly, at least four of the five σ factors included in this network are essential for virulence.

Several questions should be addressed by future work. For example, the stimulus and molecular mechanisms leading to σ factor release by their cognate anti-σ factors have only been described for RshA (Song et al., 2003). Moreover, relatively few genes have been unequivocally reported to be directly regulated by a σ factor. As this number grows, it will be interesting to investigate the importance of functional redundancy between σ factors, considering that consensus promoter sequences of at least σE, σH and σL can overlap. From this perspective, it would also be interesting to delete multiple σ factor genes from the genome of M. tuberculosis (Raman et al., 2001) to create novel attenuated strains that would express most antigens but would be incapable of surviving under mild stress conditions. Such strains could constitute attractive candidates as novel live attenuated vaccines for tuberculosis. It will also be important to assign a function to more genes in the M. tuberculosis genome in order to confidently extrapolate the physiological functions regulated by each σ factor. Studies addressing these issues will be of great value to better understand the role of σ factors in the physiology and virulence of M. tuberculosis.


The authors would like to thank Issar Smith and Joëlle Brodeur for critical reading of the manuscript.

This work was supported by a grant from the Natural Sciences and Engineering Research Council of Canada (NSERC) genomics projects awarded to L.G., R. Brzezinski and J. Goulet. L.G. holds a Canada Research Chair on the mechanisms of gene transcription. S.R. is the recipient of studentships from NSERC and Fonds de recherche en santé du Québec (FRSQ). P.É.J. was supported by the Fonds québécois de la recherche sur la nature et les technologies (FQRNT). R.M.'s laboratory is funded by the Istituto Superiore di Sanità, P.N. AIDS 50F.24, Ministero della Istruzione, della Università e della Ricerca (MIUR), Progetti di Interesse Nazionale (PRIN) 2003 n. 2003059340 and the University of Padova, P.A. CPDA047993.