Here we report the high throughput visual screen of a high density M. tuberculosis mutant library for the search of microbial virulence genes involved in phagosome maturation arrest. The power of our approach lies both in: i/the analysis of a single labelling, i.e. “acidic compartments”, to quantify mycobacterial phagosome acidification, which allowed us to exclude any bias due to false co-localizations; ii/the possibility to distinguish between tens of thousands of mutant strains at the single strain level; iii/the possibility of a very stringent selection of the hits.
Many of the genes identified in our study are predicted to be located within operons (pstS3
, Rv1503c, Rv1506c, fadD28
, Rv3880c) 
. Using RT-PCR amplification of gene junctions we have confirmed this to be true for the genes Rv1503c-to-Rv1507c on which we focus our detailed analysis (data not shown). Although we could restore the lipid profile in the Rv1506c::Tn mutant by complementing with a plasmid encoding Rv1506c (data not shown), we have chosen as a general strategy to complement our mutants with integrating cosmids. These constructs integrate as single copy into the chromosome and encode the target gene as well as surrounding genes and regulatory elements. This strategy provides for a more realistic reproduction of wild-type gene expression levels of the target gene and has been successfully used in a number of previous studies (e.g. 
). However, cosmid complementation may complement polar effects of transposon insertion on neighboring genes and does not formally prove the involvement of a single mutated gene in an observed phenotype. The discussion below should be put into this context. The putative involvement of genomic loci is discussed rather than the exact function of given genes.
Here we found only one gene (fadD28
) that was already shown to play a part in phagosome maturation arrest and has been isolated in a previous similar global approach 
. In addition, Pethe et al.
have also identified Rv1522c/mmpL12
, encoding a putative lipid transporter, as involved in phagosome maturation arrest 
. Interestingly, this gene belongs to the same genomic locus as the Rv1503c and Rv1506c genes identified here. These three genes may be functionally related. The only modest overlap with datasets from other global screening studies examining related phenotypes 
may be due to the differences in bacterial strains, host cells and selection techniques employed in the different studies. Indeed, our study is the first to screen for intracellular phenotypes in the GC1237 W-Beijing clinical isolate of M. tuberculosis
and the selection used was highly stringent, isolating and sequencing only the 10 mutants that gave the strongest LysoTracker signals, i.e.
3 s.d. above the mean. It is likely that we have missed some borderline mutants in genes previously reported.
The fatty acid-CoA ligase-encoding gene fadD28
is required for the synthesis of phtiocerol dimycocerosates (PDIM). These complex lipids have already been shown to be involved in mycobacterial virulence 
, and a recent report has suggested that PDIM may directly prevent phagosome maturation through insertion into the phagosomal membrane 
. The lppM
gene was identified as involved in phagosome arrest and encodes a putative lipoprotein that may be anchored in the bacterial membrane. In favour of a membrane modification, the lppM
::Tn mutant grows in clumps relative to wild type (data not shown). The ppe54
gene encodes a protein belonging to the large PE/PPE family whose members have mostly been shown to be membrane bound and effective immunogens. Such features have not been reported so far for PPE54. Strikingly, its expression was found specifically upregulated in IFN-γ-activated macrophages 
suggesting a role in mycobacterial intracellular persistence in agreement with our findings. We also isolated a mutant in Rv3880c, recently named espL
as part of the ESX-1 secretion system 
. The exact function of EspL is unknown, however studies in M. marinum
suggest that proteins encoded by neighboring genes might interfere with phagosome maturation 
. Other components of the ESX-1 secretome might be involved in phagosome maturation arrest as well 
. More generally, the ESX-1 secretion system has been involved in mycobacterial virulence in various in vivo
and in vitro
systems, including in macrophages 
. Secreted or membrane-anchored products such as PDIM, lipoproteins, PE/PPE proteins and the ESX-1 secretome might interact with host cell endocytic machinery components yet to be identified.
Additional genes identified here include pstS3
and genes involved in molybdenum metabolism, namely moaC1
. The periplasmic phosphate binding lipoprotein-encoding gene pstS3
belongs to a larger genetic locus, the pho
regulon, involved in inorganic phosphate uptake, and previously reported to play a part in mycobacterial virulence 
. The pstS3
gene itself has been implicated in mycobacterial intracellular survival, possibly through phagosome maturation arrest 
. How phosphate uptake is linked to buffering phagosome acidification is unclear and requires further investigation.
The ability of some mutants for intracellular growth despite vacuole acidification is striking and may be explained by several reasons. The exact pH values of the mutants' phagosomes have not been precisely measured here, and it is possible that some mutants reside in more acidic vacuoles than others. As M. tuberculosis
is highly resistant to very low pH 
, phagosomes of mutants in which acidity is the only altered factor may still be permissive for mycobacterial growth. In addition, as can be seen in , a fraction of the phagocytosed mycobacteria does not localize in acidic compartments and thus can be the sub-population that indeed replicates at later time points. It is important to explicitly determine whether the bacterium being observed in a specific sub-cellular compartment is really the one that will replicate, and latest development of life time automated confocal imaging suggest that this may become foreseeable in a near future.
One of the most striking findings of our study is the independent isolation of two pairs of mutants in genes located within the same genetic loci, namely Rv1503c/Rv1506c, and moaC1
. Given the size and coverage of our library, it is unlikely that such isolations happened by chance. The adjacent genes moaC1
are part of a chromosomal region (Rv3108-20) presumably involved in molybdopterin biosynthesis. We have recently shown that this gene cluster actually forms a genomic island that arose in the ancestor of the tubercle bacilli through horizontal transfer from environmental proteobacteria 
. Molybdopterin is a precursor of the so-called molybdenum cofactor (MoCo), a coenzyme for various oxidoreductases, including nitrate reductase and sulfite oxidase for instance. Our results are reminiscent of a recent study in which lesion in moeB1
, another gene possibly involved in MoCo biosynthesis and distant from the Rv3108-20 locus, was shown to affect mycobacterial phagosome biogenesis 
. The genome of the tubercle bacillus contains several loci potentially involved in MoCo biosynthesis 
, and a few gene products in M. tuberculosis
harbour molybdopterin/MoCo-binding sites and might use MoCo as a cofactor. This is the case of the nitrate reductase subunit-encoding genes, narG
, of the formate and aldehyde dehydrogenase-encoding genes fdhF
, respectively, as well as of two other gene products, Rv0197, an uncharacterized oxidoreductase, and Rv0218, a possible sulfite oxidase. Although we can only speculate at this stage, we can anticipate that MoCo-dependent redox reactions play a crucial part in early intracellular fate of the tubercle bacillus, which should deserve further attention in future studies.
We also isolated two independent mutants carrying genetic disruptions in Rv1503c and Rv1506c. The homologues of these genes in M. marinum
belong to a locus involved in the synthesis of LOS 
. Here we showed that, like other M. tuberculosis
, our W-Beijing mother strain does not synthesize LOS. However we showed that the Rv1503c::Tn and Rv1506c::Tn mutants are impaired in the synthesis of other acyltrehalose-containing lipids, namely DAT, and that they overproduce sulfoglycolipids (SGL), Ac3
SGL and more importantly Ac4
SGL. DAT and SGL are glycolipids based on trehaloses 
. Although some of the genes involved in DAT 
and SGL 
synthesis have been identified, knowledge of the enzymes, transporter and regulators involved in these pathways is still incomplete. Transcriptional regulation of the genes involved in synthesis of these molecules has been shown to be coordinated by the PhoP regulator 
. This suggests that Ac4
SGL overproduction in the mutant strains may be due to a compensation mechanism associated with impairment of DAT production. However, the Rv1503c and Rv1506c proteins do not carry the signature of enzymes such as acyltransferases and polyketide synthases (Pks), and it is thus unclear what their exact involvement in DAT synthesis could be. The Rv1505c gene encodes a putative acyltransferase and the Rv1505c and Rv1506c genes are predicted to form an operon; the possibility thus remains that impairment of DAT synthesis in the Rv1506c mutants, and eventually in the Rv1503c mutants, is actually due to inactivation of Rv1505c because of operon disruption and/or of polar effects. Interestingly, transcription of Rv1505c, as well as that of other genes located in the same genomic region, namely Rv1517, Rv1518, Rv1522c/mmpL12
, has been proposed to be regulated by PhoP 
. Whether and how these genes are involved in DAT synthesis will require further investigation and is beyond the scope of the present study.
Most importantly, our results suggest that Ac4
SGL, but not DAT, increase phagosome acidification. This is at variance with conclusions from an early study suggesting that mycobacterial sulfolipids (SL) tend to impair phagosome maturation 
. However these authors incubated purified sulfolipids directly with the cells, and the model systems used in this study and in ours are barely comparable. Nevertheless, our results obtained with beads suggest that SL increase phagosome maturation, which may account, at least in part, for the increased trafficking of the Rv1503c and Rv1506c mutants, although the exact mechanism of increased acidification of these two mutants has not been formally identified here. Furthermore, analysis of the production of cytokines and chemokines using protein arrays upon infection with Rv1503c and Rv1506c mutants did not reveal any significant differences compared to the wild type strain (data not shown). This suggests that the mutants altered trafficking phenotypes were not caused by changes in their ability to induce different inflammatory cytokine production 
and indicates that other mechanisms might account for the observed trafficking behaviour. For instance, other possible explanations include an overall altered cell wall, or a more generally reduced intracellular fitness of the mutants, which both may impact intracellular trafficking. In this regard, it is interesting to notice that the phoP
mutant shows a significant, though partial, relocation to an acidic phagosome (). The PhoP transcriptional regulator controls the expression of many genes, some of which have been involved in phagosome remodelling; this is the case of fadD28
for instance 
. It also regulates the synthesis of important lipids, such as diacyltrehaloses and sulfolipids 
, which may be involved in intracellular trafficking according to our data. The partial inability of the phoP
mutant to arrest phagosome maturation may thus be due to a defect in production of these factors. It may also be due to a broader defect in fitness 
. Further study of our mutants and other mutants isolated in previous screens will help understand the general relationship that may exist between mycobacterial fitness and intracellular trafficking.
In summary, the unbiased approach developed here allowed us to identify novel mycobacterial genes involved in M. tuberculosis
intracellular parasitism, and suggests the involvement of important M. tuberculosis
-specific glycolipids in this process. This approach can easily be adapted for the comprehensive quantitative analysis of mycobacterial vacuole sorting, as well as for the study of intracellular parasitism by pathogenic microorganisms, and for phenotypic drug screening