Identification of two Mycobacterium tuberculosis H37Rv ORFs involved in resistance to killing by human macrophages.

Background The ability of Mycobacterium tuberculosis to survive and replicate in macrophages is crucial for the mycobacterium's ability to infect the host and cause tuberculosis. To identify Mycobacterium tuberculosis genes involved in survival in macrophages, a library of non-pathogenic Mycobacterium smegmatis bacteria, each carrying an individual integrated cosmid containing M. tuberculosis H37Rv genomic DNA, was passed through THP-1 human macrophages three times. Results Two of the clones recovered from this enrichment process, sur2 and sur3, exhibited significantly increased survival relative to wild-type bacteria. In coinfection experiments, the ratio of sur2 colonies to wild-type colonies was 1:1 at 0 hours but increased to 20:1 at 24 hours post phagocytosis. The ratio of sur3 colonies to wild-type colonies was 1:1 at 0 hours and 5:1 at 24 hours. The M. tuberculosis ORFs responsible for increased survival were shown to be Rv0365c for the sur2 clone and Rv2235 for the sur3 clone. These ORFs encode proteins with as-of-yet unknown functions. Conclusions We identified two M. tuberculosis ORFs which may be involved in the ability of tubercle bacilli to survive in macrophages.


Background
Today, an estimated one-third of the world's population is infected with Mycobacterium tuberculosis, the causative agent of tuberculosis [1,2], and tuberculosis causes about 2 million deaths annually [1,2]. Many aspects of the interactions between M. tuberculosis and its human host remain unclear. Not only is this bacterium able to evade the defenses of the host's immune system, it is also able to persist in the body for years and may reactivate to cause disease decades after the initial infection. A better understanding of the interaction between M. tuberculo-sis and its human host is critical to developing new strategies to control the tuberculosis epidemic.
A key feature of the pathogenicity of M. tuberculosis is its ability to evade the antimicrobial processes of the macrophage and replicate intracellularly. Mycobacteria enter macrophages primarily by conventional receptor-mediated phagocytic pathways [3]. Following phagocytosis, phagosomes containing viable tubercle bacilli fail to acidify, apparently because of failure to insert a proton-ATPase pump into the phagosomal membrane [4][5][6].
This leads to altered vacuolar maturation such that phagosome-lysosome fusion is blocked and the mycobacteria-containing vacuoles end up with an internal pH of about 6 and markers of phagosomes and early and late endosomes [7][8][9][10][11]. The mycobacterial genes responsible for these processes are largely unknown.
One method for identifying bacterial genes involved in pathogenesis is to express these genes in a nonpathogenic host and isolate bacteria with increased virulence. This technique was first used to identify a gene that enables Yersinia pseudotuberculosis to invade HEp-2 cells [12]. In these studies, Escherichia coli bacteria expressing Y. pseudotuberculosis genes were used to infect HEp-2 cells. Only E. coli bacteria expressing the Y. pseudotuberculosis inv gene were able to invade the animal cells [12]. Using a similar approach, Arruda et al. [13] identified an M tuberculosis gene responsible for invasion of HeLa cells. We have previously used a similar technique to identify M. leprae genes involved in intracellular survival [14]. That is, by expressing M. leprae genes in E. coli, we were able to isolate recombinant bacteria that exhibited increased resistance to killing by murine bonemarrow derived macrophages [14].
A similar approach was used by Wei et al [15] to isolate 21 Mycobacterium smegmatis recombinant clones that displayed a greater than 2-fold enhancement in survival after 48 hours. M. smegmatis is a fast-growing, nonpathogenic species of Mycobacterium in which M. tuberculosis genes can be efficiently expressed [16][17][18]. An in depth analysis of one recombinant clone revealed that M. smegmatis recipients carrying the M. tuberculosis eis gene on an extrachromosomal multicopy plasmid displayed 2.4-to 5.3-fold greater survival in U937 macrophages than wild-type M. smegmatis bacteria at 24 to 48 hours post-infection [15]. The 42-kDa eis gene product has been shown to be associated with the mycobacterial cell surface and is released into extracellular medium, but its precise function is not yet known [19].
In the studies reported here, we used a similar enrichment scheme to identify two additional recombinant M. smegmatis clones, sur2 and sur3, that demonstrated enhanced survival during infections of THP-1 human monocyte-derived macrophages. The M. tuberculosis ORFs responsible for increased survival were shown to be Rv0365c for the sur2 clone and Rv2235 for the sur3 clone. These ORFs encode proteins with as-of-yet unknown functions.

Results
Cosmids carrying M. tuberculosis genes [20] were electroporated into M. smegmatis LR222 to create a library of M. smegmatis transformants, each of which carries a pYUB178::H37Rv cosmid integrated into its chromosome. The library was generated from about 4000 independent transformants. This represents about 20 M. tuberculosis genome-equivalents given that ~225 cosmids contain one genome-equivalent of M. tuberculosis [20].
To enrich for clones with increased survival in human macrophages, the library was passed through THP-1 macrophages as shown schematically in Figure 1. THP-1 is a human monocyte-derived macrophage cell line [21] and wild-type M. smegmatis bacteria are rapidly killed by THP-1 macrophages [22]. After three rounds of enrichment, the resulting clones were analyzed individually by Southern blot to evaluate the number of different clones present as previously described [20]. Out of the 3000 colonies recovered, genomic DNAs from the bacteria of 90 randomly chosen colonies were digested with PstI and hybridized with the pYUB178 vector. Strains carrying different cosmids should exhibit a different pattern of hybridizing bands because of the presence of different M. tuberculosis H37Rv inserts. Two clones, designated sur2 and sur3, were each present three times and were chosen for further study. The remaining 84 clones each displayed unique patterns.
As a first step in the analysis of the sur2 clone, the time course of its survival in THP-1 macrophages was determined by infecting THP-1 macrophages, lysing the infected macrophages at various times after phagocytosis, and enumerating viable intracellular bacteria by plating on solid medium. Both parental and the sur2 bacteria were rapidly killed during the first few hours after phagocytosis ( Figure 2). The sur2 bacteria appeared to survive slightly better at the 9 hour time point, but the difference was not statistically significant.
To compare directly the relative ability of the sur2 and wild-type bacteria to survive in macrophages, THP-1 macrophages were infected with a mixture of a genetically marked control strain and the sur2 strain and the survival of each strain was followed independently as previously described [22]. The wild-type strain expresses the xylE gene product, catechol 2,3-dioxygenase, such that when its colonies are sprayed with catechol, they turn bright yellow, while wild-type colonies remain white. In essence, each well of the experiment contains an internal standard (the wild-type bacteria) to which to compare the survival of the recombinant bacteria. In experiments in which THP-1 cells were infected at a MOI of 50:1 (results in ~1 phagocytosed M. smegmatis bacterium per macrophage) with a mixture containing equal numbers of bacteria of the xylE-expressing control strain and a strain carrying the cosmid vector pYUB178, the ratio of recovered white colonies to yellow colonies was 1:1 at all time points (data not shown). This indicates that the survival of the xylE-expressing strain was the same as that of the wild-type and could be used an internal reference by which to measure the survival of other clones.
In coinfection experiments with xylE-expressing bacteria, both sur2 and sur3 bacteria exhibited increased survival ( Figure 3). Immediately after the 2 hour phagocytosis period (0 hr time point) the ratio of sur2 colonies to xylE-expressing colonies was 1:1 and by 12 hours it was about 7:1. By 24 hours, the ratio was approximately 20:1. The differences between the ratios at the zero time point and the subsequent time points were statistically significant (p < 0.005) for the 9, 12, and 24 hr time points. The ratio of sur3 colonies to control colonies increased from 1:1 at time 0 to 4:1 at 12 hours and to 5:1 at 24 hours (p < 0.005).
Because the recombinant clones contain integrated pYUB178::H37Rv cosmids, the following strategy was used to isolate cosmids corresponding to those in the sur2 and sur3 clones. First, genomic DNA from the sur2 clone was digested with PstI, and genomic DNA from the sur3 clone was digested with BamHI to generate frag-ments of each integrated cosmid carrying oriE, aph, and a portion of the M. tuberculosis genomic DNA insert. The presence of oriE allows the recombinant to replicate as a plasmid in E. coli. The digestion products were treated with T4 DNA ligase and transformed into E. coli XL 1-Blue. Plasmid DNA from the resulting kanamycin-resistant colonies were analyzed by restriction site mapping and partial sequencing of the M. tuberculosis genomic DNA insert as described in Materials and Methods. For the sur2 recombinant, PCR primers were designed to amplify a 0.8 kb region of the M. tuberculosis insert from M. tuberculosis genomic DNA for use as a probe in colony blot experiments. Probing colony blots of the E. coli (pYUB178::H37Rv) library with the 0.8 kb PCR fragment led to the isolation of a 4 kb plasmid. This plasmid contained an intact oriE and aph gene, a portion of the integrase gene, and 1.1 kb of M. tuberculosis genomic DNA ( Figure 4). PCR, DNA sequence, and Southern blot data indicated that the recovered plasmid was the same as the cosmid integrated in the sur2 genome (data not shown). For example, sequencing of PCR amplicons of the junctions between the mycobacterial sequences and vector sequences revealed that the junctions in the recovered plasmid were identical to those in the sur2 genomic DNA.
Comparison of the sequence of the cloned 1.1 kb fragment with the M. tuberculosis H37Rv genome sequence [23] revealed that it contains portions of the Rv0366c and Rv0365c genes ( Figure 4). In the sur2 clone, the  To determine if the Rv0365c gene was responsible for the increased survival of sur2 bacteria, the full-length M. tuberculosis Rv0365c ORF as well as a truncated ORF similar to that present in the insert in the sur2 clone were subcloned into the pHIP vector downstream of the M. tuberculosis hsp65 promoter. The truncated ORF contained the 969 bp of Rv0365c present in the sur2 insert followed by a stop codon but did not contain the 5 amino acids corresponding to the vector contribution to the ORF. In coinfection experiments, the ratio of M. smegmatis bacteria expressing the full-length Rv0365c ORF to xylE-expressing wild-type bacteria was 1:1 at time zero and increased to approximately 10:1 at 12 hours and to 11:1 at 24 hours ( Figure 5). The clone expressing the truncated Rv0365c ORF exhibited slightly less, but not statistically significantly different, increased survival (1:1 at time zero, 8.6:1 at 24 hrs).
To investigate the difference in survival at 24 hours of the full-length Rv0365c ORF expressing bacteria and sur2 bacteria, a strain expressing Rv0365c and xylE was constructed and used in THP-1 coinfections with the original sur2 clone. In this coinfection, the ratio of white (sur2) to yellow (xylE and Rv0365c expressing) colonies remained 1:1 through 12 hours and then increased to ~3:1 at 24 hrs, consistent with the above-described observations.
A cosmid corresponding to the one in the sur3 clone was isolated from the E. coli (pYUB178::H37Rv) library by probing colony blots with the 4.5 kb BamHI/EcoRl fragment of the M. tuberculosis genomic DNA insert in the plasmid recovered from the sur3 clone. PCR, DNA sequence, and Southern blot data indicated that the recovered cosmid was the same as the cosmid integrated in the sur3 genome (data not shown). The ~10.7 kb cosmid in the sur3 chromosome contains a 5.76 kb fragment of M. tuberculosis H37Rv genomic DNA ( Figure 6) [23]. This region encodes eight potential ORFs designated Rv2233-Rv2240c as well as the tRNA for valine.

Figure 3
Survival of recombinants relative to wild-type M. smegmatis. THP-1 macrophages were infected with an equal mixture of xylE-expressing bacteria and sur2 bacteria (hatched) or sur3 bacteria (horizontal stripe). The ratio of the recovered white-to-yellow colonies is shown for -2 hours (initial inoculum), 0 hours (immediately after phagocytosis interval), and at 3, 6, 9, 12, and 24 hours after phagocytosis. The ratios represent the average of the results of three independent experiments. Error bars represent the standard deviation in the ratio of white-to-yellow colonies between experiments. 5DWLR RI ZKLWH WR \HOORZ FRORQLHV

Figure 4
Map of the M. tuberculosis H37Rv DNA integrated in the sur2 genome. The integrated 4 kb cosmid is between the two attachment sites, attL and attR.

PstI PstI
To determine which ORF(s) was responsible for the increased survival exhibited by sur3 bacteria, each ORF was subcloned individually into the expression vector pHIP. The two potential operons, Rv2233-Rv2235 and Rv2238c-Rv2240c, were also subcloned into pHIP. The recombinant bacteria were examined for survival in the macrophage in coinfection experiments at 0, 6, and 12 hours post phagocytosis ( Figure 7). Each of the recombinants tested exhibited a 1:1 ratio of white-to-yellow colonies at 0 hours. Two of the recombinants demonstrated an increase in the ratio of white-to-yellow colonies over time, while the ratio remained at 1:1 for the other eight recombinants. One of the two recombinants contained the potential operon of Rv2233-Rv2235, and the other contained Rv2235 only. The ratio of white-to-yellow colonies for bacteria expressing ORFs Rv2233-Rv2235 was 2:1 at 6 hours and ~6:1 at 12 hours. For bacteria expressing Rv2235, the ratio of white-to-yellow colonies was ~2:1 at 6 hours and >3:1 at 12 hours. The survival of recombinant bacteria expressing the Rv2235 ORF relative to wild-type was more directly compared to the relative survival of sur3 by doing the coinfections in parallel. The ratios of white-to-yellow for both recombinant bacteria and sur3 bacteria were 1:1 at 0 hours, ~5:1 at 12 hours, and >5:1 at 24 hours. The differences in the relative survival of the three strains were not statistically significant.
A PCR fragment containing the M. tuberculosis ORFs Rv2233, Rv2234, and Rv2235 without the upstream promoter region, was cloned into the pBPhin vector, which does not contain a promoter to express the inserted DNA. In coinfections with the xylE-expressing bacteria, the recombinant bacteria containing Rv2233, Rv2234, and Rv2235 did not exhibit the same increase in macrophage survival as the sur3 clone (data not shown), suggesting that the Rv2235 ORF is expressed using signals upstream of the Rv2233 ORF. Database searches revealed that Rv0365c encodes añ 41-kDa protein which displays significant homology only to a truncated Corynebacterium glutamicum hypothetical protein of unknown function (42% identity in a 296 aa overlap) located upstream of a gene encoding a fructose-bisphosphate aldolase. Rv2235 encodes a conserved hypothetical membrane protein of about 30-kDa and shares a motif with the SURF-1 family of proteins. The other two ORFs in the putative Rv2233-Rv2235 operon, Rv2233 and Rv2234, share homology with several proteins in data base searches. Rv2233 may encode a putative phosphatase, and Rv2234 may encode a low molecular weight protein tyrosine phosphatase [23].

Discussion
There are several potential limitations of isolating M. tuberculosis genes involved in intracellular survival using an enrichment procedure. One limitation is that this protocol is biased towards the recovery of clones with the greatest increase in survival relative to wild-type [14]. Thus, the recovered clones are not a random collection of genes involved in intracellular survival and hence the number of genes involved for survival cannot be calculated. This approach is also biased towards identifying genes expressing proteins that directly interfere with the antimicrobial processes of the macrophage. Some types of genes involved in resistance to killing, such as genes that are part of a multi-enzyme pathway, may not be isolated. Also, genes necessary for intracellular survival as opposed to resistance to killing are not likely to be isolated.
Two general classes of clones might be recovered following enrichment of the M. smegmatis recombinant library for clones with increased intracellular survival. One type might be clones that carry M. tuberculosis genes that confer enhanced resistance to the antimicrobial processes of the macrophage. Another type might be clones whose M. tuberculosis gene products are involved in attachment or invasion or increase phagocytosis. The two genes that were isolated in the studies reported here confer enhanced resistance rather than increased uptake. That is, the ratios of bacteria expressing either Rv0365c or Rv2235 to wild-type bacteria were 1:1 in both the initial mixture and inside the macrophages at the end of the phagocytosis period. Differences in survival compared to wild-type did not become apparent until about 9 hours post-phagocytosis for bacteria expressing Rv0365c and about 6 hours post-phagocytosis for those expressing Rv2235.
In this study, the enrichment process resulted in the isolation of two strains containing small, integrated plasmids rather than the expected 30-50 kb cosmids [20]. Small plasmids corresponding to the integrated ones were present in the original E. coli (pYUB178:H37Rv) cosmid library. PCR, sequencing, and Southern blot analysis demonstrated that the plasmid integrated in the sur2 clone was approximately 4 kb containing 1.1 kb of M. tuberculosis DNA and the plasmid integrated in the sur3 clone was approximately 10.7 kb containing about 5.7 kb of M. tuberculosis DNA. Cosmid libraries frequently contain clones without DNA inserts [24], so it is not too surprising that clones with small DNA fragments were present in the E. coli library after infection with λ phage. The small plasmids may have a growth or replication advantage causing them to be over-represented in the library following the various amplification steps. The ~2 kb deletion of vector DNA in the sur2 clone results in a fusion of the ORF of the L5 integrase with that of the M. tuberculosis insert such that the fused ORF encodes a protein that contains only the amino-terminal 73 aa of the 344 aa L5 integrase. The observation that the sur2 clone contains an integrated plasmid corresponding to the plasmid isolated from the cosmid library suggests that either a) the hybrid protein retains integrase activity, b) the plasmid inserted into the attachment site by homologous recombination between attP and attB, or c) a functional integrase was provided in trans by a second cosmid transiently present in the original transformant or by a gene present in the M. smegmatis LR222 genome. However, repeated attempts to electroporate the 4 kb plasmid into M. smegmatis LR222 did not generate any stable kanamycin-resistant transformants (data not shown). These results suggests that the hybrid protein does not retain integrase activity, that integration by homologous recombination into attP is unlikely, and that provision of integrase activity by an M. smegmatis gene chromosomal does not occur. The simplest explanation is that integrase was provided in trans by a second cosmid transiently present in the original transformant, and indeed, M. smegmatis bacteria carrying an integrated copy of the 4-kb plasmid can be readily isolated following electroporation of a mixture of the 4-kb plasmid and a plasmid that expresses integrase (unpublished results).
Data base searches did not reveal any homologies that could be used to predict functions for the gene products of ORFs Rv0365c or Rv2235. Rv0365c encodes a hypothetical protein of 376 amino acids which displays significant homology only to a Corynebacterium glutamicum hypothetical protein of unknown function (42% identity in a 296 aa overlap) [23]. Rv2235 encodes a hypothetical protein of 271 aa with three putative transmembrane domains and which displays significant homology only to hypothetical protein MLCB1243.32c in M. leprae, to which it is 74% identical [23,25]. This M. leprae homologue was evident in Southern blot experiments done under low stringency conditions. Rv2235 also contains the SURF-1 signature sequence [26] and modest homology to members of the SURF-1 family such as the SURF-1 protein of Caulobacter crescentus (25% identity; 40% similarity) [27]. SURF-1 proteins are ~33-kDa, integral membrane proteins whose precise function is not known. In eukaryotic cells, SURF-1 proteins are involved in the assembly and maintenance of mitochondrial respiratory chain complexes including cytochrome oxidase [26,28]. The homology to SURF-1 proteins raises the possibility that Rv2235 could play a role in resisting the antimicrobial activities of macrophages by helping to maintain the stability or function of an important cellular process, akin to the stabilizing role of chaperonins during a heat shock.
In addition to the genes described in this report, other studies with M. smegmatis recombinants and/or mutants have implicated 11 other M. tuberculosis genes in intracellular survival. Genes identified using enrichment or screening protocols include Rv2962c and Rv2958c (probable glucuronsyl transferases), Rv2220 (glutamine synthetase A1), Rv3913-Rv3914 (thioredoxin, thioredoxin reductase) and Rv2416c (eis, unknown function) [14,15,22,29]. By screening insertional mutants of M. smegmatis, Lagier et al [30] isolated 8 mutants with impaired ability to survive in human peripheral blood monocyte-derived macrophages and identified the M. tuberculosis gene corresponding the mutated M. smegmatis gene for five of them. The genes included: Rv3052c (probable nrdI) which is postulated to be involved in deoxynucleotide production under stressed conditions; Rv0101 which is a nonribosomal peptide synthetase that displays strong homology with a Pseudomonas nonribosomal peptide synthetase required for the synthesis of the pyoveridine, a siderophore involved in iron uptake; Rv3420c which displays homology with the S18 ribosomal protein acetyltransferase which behaves as a heat shock protein in Chlamydia trachomatis; and Rv0497 and Rv3604c which are hypothetical conserved membrane proteins of unknown function.
It should be noted that these M. tuberculosis genes generally confer only a limited enhancement of the survival of M. smegmatis bacteria in the human macrophages. That is, usually only a few per cent of the recipients are viable 24 to 48 hours post-infection. This reinforces the concept that the intracellular survival of a pathogenic mycobacteria is a complex multifactoral process. The precise role(s) of any of the identified genes in the intracellular survival of mycobacteria is not yet known, although several of the identified genes have features of stress response genes. This is not unexpected given the relatively small enhancements of the survival of the M. smegmatis recombinants and the numerous environmental stresses encountered in the macrophage. Additional studies, such as the construction and characterization of targeted knock-out mutants, will be needed to determine the roles of the proteins encoded by these genes in the survival of M. tuberculosis in human macrophages.

Conclusions
Using an enrichment and screening procedure, two M. tuberculosis genes, Rv0365c and Rv2235, were identified that could confer an enhanced ability to survive in human macrophages to normally susceptible M. smegmatis recipients. The functions of these two proteins are not known. This study brings the number of M. tuberculosis genetic loci that have been implicated in enhancing the intracellular survival of M. smegmatis cells to 13. The precise role(s) of any of the identified genes in the intracellular survival of mycobacteria remain to be elucidated.

Bacterial strains, plasmids, and growth conditions
The bacterial strains and plasmids used in this study are listed in Table 1. The E. coli strain XL 1-Blue was obtained from Stratagene (La Jolla, Calif). The M. smegmatis strain LR222 was obtained from Dr. Jack Crawford, Tuberculosis and Mycobacteriology Branch, Centers for Disease Control and Prevention (CDC), Atlanta, GA. The pYUB178 plasmid and the λ phage library of pYUB178::H37Rv cosmids were generously provided by Dr. William Jacobs, Albert Einstein University, New York, NY [20]. The cosmid library contains 30-50 kb fragments of M. tuberculosis genomic DNA generated by partial Sau3A digestion cloned into BclI-digested pYUB178. In the λ phage library, ~225 cosmids represent one genome-equivalent of M. tuberculosis [20]. An E. coli (pYUB178::H37Rv) library was created by infecting E. coli XL1-Blue with the λ phage library. Bacteria from ~4000 colonies were recovered and pooled, and cosmid DNA was isolated. The pooled pYUB178::H37Rv cosmid DNAs were electroporated into M. smegmatis LR222, and kanamycin-resistant colonies were isolated. Bacteria from ~4,000 colonies (representing ~20 genome equivalents) were recovered and pooled to generate the M. smegmatis (pYUB178::H37Rv) library. Because the cosmids integrate into the mycobacteriophage L5 attachment site in the M. smegmatis genome, a single copy of the M. tuberculosis DNA is maintained in the M. smegmatis transformants [20].

DNA manipulations
All enzyme reactions were performed as recommended by the manufacturers (GIBCO Bethesda Research Laboratories, Inc., Gaithersburg, Md., New England Biolabs, Beverly, Mass.). Cosmid DNA from E. coli was prepared using the Qiagen Plasmid Maxi Kit (Qiagen, Valencia, Calif.) according to manufacturer's instructions for lowcopy number plasmids. Wizard Plus Minipreps Kits (Promega, Madison, Wis.) were used to isolate plasmid DNA from E. coli strains. Mycobacterial genomic DNA was isolated as previously described [33].

PCR
The primers used in this study and their sequences are listed in Table 2 [21]. To differentiate the THP-1 cells into macrophage-like cells, the cells were treated with 10 µM phorbol myristate acetate (PMA) (Sigma) as follows [21]. All PMA manipulations were done under low light conditions. THP-1 cells were harvested by centrifugation for 10 minutes at 228 × g, and the pellet was resuspended in RPMI 1640/10% FCS/ 10 µM PMA to give a cell density of approximately 1 × 10 6 THP-1 cells/mL. Three milliliters of the suspension was added to each well of a six-well tissue culture plate (Costar, Corning, NY). The plates were incubated for 48 hours at 37°C in 5% CO 2 . The medium was removed from each well, the attached cells were washed once with RPMI 1640/10% FCS, and 3 ml of fresh RPMI 1640/10% FCS was added. The cultures were incubated at 37°C in 5% CO 2 for an additional 48 hours. Immediately prior to infection, cells in each well were washed once with fresh RPMI 1640/10% FCS.

Enrichment protocol
About 10 6 bacteria of the M. smegmatis (pYUB178::H37Rv) library were inoculated into Middlebrook 7H9 media containing 10 µg kanamycin/mL and then were grown to midlog phase (OD 600 ~ 0.3). The bacteria were harvested by centrifugation for 1 minute at 16,000 × g and washed twice with RPMI 1640/10% FCS. The bacteria were suspended in RPMI 1640/10% FCS at 5 × 10 8 bacteria/mL. The enrichment process was as follows ( Figure 1): differentiated THP-1 macrophages were infected by adding 3 mL of the bacterial suspension to each well of a six-well plate. The multiplicity of infection (MOI) was ~500 bacteria per THP-1 cell. The culture was left at 37°C in 5% CO 2 for 2 hours, which resulted in the phagocytosis of about 10 bacteria per macrophage. After the phagocytosis period, each well was washed twice with RPMI 1640/10% FCS to remove free bacteria. To kill any remaining extracellular bacteria, 3 mL of fresh RPMI 1640/10% FCS/200 µg amikacin/mL (Sigma) was added to each well. The infected THP-1 cultures were incubated at 37°C for an additional 5 hours. Each well was then washed twice with 3 mL of RPMI 1640/10% FCS, and then 1 mL of 0.1% (v/v) Triton X-100 (Sigma) was added to each well to lyse the macrophages. The wells were scraped with a rubber policeman, and the lysates were removed and diluted for plating on TSA containing 10 µg kanamycin/mL. After 3 days of incubation at 37°C, bacteria from the resulting colonies were harvested, suspended in Middlebrook 7H9 media containing 10 µg kanamycin/mL and 0.05% (v/v) Tween 80, and pooled for reinfection of THP-1 macrophages. The enrichment process was done for a total of three cycles. After the third round of enrichment, individual clones were isolated and analyzed.

Southern blots
PstI-digested genomic DNA was electrophoresed through a 1.0% agarose gel, denatured, neutralized, and transferred by capillary blotting to a Hybond -N + membrane (Amersham, Arlington Heights, IL). The blots were hybridized to pYUB178 plasmid DNA labeled using the ECL Direct Nucleic Acid Labeling and Detection System (Amersham). All hybridization and washing steps were done at 42°C under either stringent (0.1 M NaCl) or non-stringent conditions (0.5 M NaCl) according to kit instructions.

Coinfection assay to measure survival
Coinfection assays were done as previously described [22]. Briefly, separate cultures of recombinant and of xylE-expressing bacteria were grown to midlog phase (OD 600 = ~0.3). The bacteria from each culture were harvested by centrifugation for 1 minute at 16,000 × g, washed twice with RPMI 1640/10% FCS, and resuspended in RPMI 1640/10% FCS at a concentration of 1.5 × 10 8 bacteria/mL. Equal volumes of the two bacterial suspensions were mixed to produce a suspension containing a 1:1 ratio of recombinant-to-control bacteria. A portion of the combined mixture was plated onto TSA plates to determine the number of colony forming units (CFUs) of each strain in the initial inoculum (the -2 hour time point in figures). The bacterial suspension was diluted with RPMI 1640/10% FCS to give approximately 5 × 10 7 bacteria/mL, and 3 mL was added to each well containing 1 × 10 6 THP-1 macrophages (MOI of 50 bacteria/macrophage). The cultures were incubated for 2 hours at 37°C in 5% CO 2 to allow phagocytosis to occur, and then each well was washed twice with RPMI 1640/10% FCS to remove unphagocytosed bacteria. Typically, this results in one phagocytosed bacterium per macrophage. To kill extracellular bacteria, 3 mL of RPMI 1640/10% FCS containing 200 µg amikacin/mL was added to each well.
Cultures were incubated at 37°C in 5% CO 2 . At various times, the medium was removed from each of three wells, and 1 mL of 0.1% (v/v) Triton X-100 in H 2 0 was added to each well to lyse the macrophages. Each lysate was diluted as necessary, and portions were plated on TSA plates. The cultures which were assayed immediately after the addition of the media with amikacin serve to measure of the number of phagocytosed viable bacteria; the time at which these cultures were assayed was considered time zero (t 0 ).
After a 3-day incubation at 37°C, the TSA plates from each time point were stored overnight at 4°C. The following day, the plates were sprayed with 0.5 M catechol (Sig-ma) in 50 mM potassium phosphate (pH 7.5) to distinguish the xylE-expressing colonies (yellow) from the recombinant colonies (white). Storing the plates overnight at 4°C results in a stronger yellow color. To determine the percent survival of a particular clone at time point X, the number of CFUs at time × was divided by the number of CFUs at to and multiplied by 100.

Recovery of the cloned M. tuberculosis genomic DNA fragment
One µg of sur2 genomic DNA was digested with the restriction enzyme PstI then ethanol precipitated. The precipitated DNA was resuspended in 20 µL T4 ligase buffer (GIBCO BRL) and then ligated for 2 hours at room temperature with 0.5 units of T4 ligase (GIBCO BRL). The ligation mixture was electroporated into electrocompetent E. coli, and kanamycin-resistant transformants were isolated. After sequencing the M. tuberculosis H37Rv insert in the recovered plasmid, PCR primers were designed to amplify a product containing a portion of the cloned M. tuberculosis H37Rv genomic DNA. The PCR product was used as a probe for colony blots to find cosmids in the E. coli (pYUB178::H37Rv) library carrying at least a portion of the cosmid integrated in the sur2 clone.
A plasmid carrying a portion of the M. tuberculosis H37Rv insert in the sur3 clone was isolated in a similar manner from a BamHI digest of sur3 genomic DNA. The BamHI/EcoRI fragment of the M. tuberculosis H37Rv insert of the recovered plasmid was used as a probe in colony blots to find cosmids in the E. coli (pYUB178::H37Rv) library.

Colony blots
Portions of the E. coli (pYUB178::H37Rv) cosmid library were plated on LB agar containing 50 µg kanamycin/mL, and colony blots were performed with the ECL Direct Nucleic Acid Labeling & Detection System (Amersham) according to manufacturer's instructions. Colonies hybridizing with the probe of interest were removed as plugs and incubated in LB containing 50 µg kanamycin/ mL for approximately 30 minutes at 37°C. Dilutions of this culture were plated on LB agar containing 50 µg kanamycin/mL to give well-separated colonies. Colony blots were performed and positive clones were selected for further study.

Subcloning ORFs into pHIP and pBPhin
Unless otherwise stated, all M. tuberculosis open-reading frames (ORFs) were generated by PCR from M. tuberculosis H37Rv genomic DNA as full-length ORFs without their natural promoters. The PCR primers (Table 2) were designed to contain restriction enzyme sites for cloning of the amplicon into the pHIP vector downstream of the hsp65 promoter as well as the eight base-pairs (bp) upstream of the start codon of the gene being cloned. Because the hsp65 promoter in pHIP contains a ribosome binding site (rbs), the spacing between the rbs and the start codon of the cloned gene is about the same in these constructs as the spacing between the rbs and the start codon of the hsp65 gene in wild-type M. tuberculosis.
A truncated form of the Rv0365c gene was generated by PCR to contain 969 bp of the 970 bp of the truncated ORF present in the sur2 clone followed immediately by a stop codon. This fragment was BamHI-digested and ligated to BamHI-digested pHIP to form pHIP3. An amplicon containing the hsp65 promoter and full-length Rv0365c gene was generated by PCR amplification of the pHIP2 plasmid. This amplicon was NotI-digested and cloned into the NotI site of the pHIP1 plasmid containing the xylE gene to generate pHIP4. Plasmid pBPhin1 was constructed by ligating a BamHI-digested PCR fragment containing the Rv0365c gene plus 24 bp located upstream of it to BamHI-digested pBPhin. Plasmid pHIP13 was created by cloning a BglII-digested PCR fragment containing Rv2233,Rv2234, and Rv2235 into the Bam-HI site downstream of the hsp65 promoter of pHIP. Plasmid pHIP14 was created by cloning a BamHI-digested PCR fragment containing Rv2238c, Rv2239c, and Rv2240c into the BamHI site downstream of the hsp65 promoter of pHIP such that Rv2240c was proximal to the hsp65 promoter. Plasmid pBPhin2 was created by ligating the BglII-digested Rv2233-Rv2235 fragment into the BamHI site of pBPhin.

Electroporation
All electroporations were conducted using a Bio-Rad Pulse Controller (Bio-Rad, Hercules, Calif). Preparation and electroporation of competent E. coli XL 1-Blue cells were done according to Bio-Rad instructions. Competent M. smegmatis LR222 cells were prepared and electroporated as described by Jacobs et al. [34].

DNA sequencing
All sequencing reactions were prepared with the Applied Biosystems, Inc. (ABI) PRISM Dye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems) according to the manufacturer. All sequencing was conducted using an ABI 373 DNA Sequencing System (Applied Biosystems).

Statistical analysis
Results were analyzed by the two-sample T test.

DNA and protein homology analyses
DNA and protein database searches were performed using the BLAST services (blastn, blastp, and psi-blast) at the National Center for Biotechnology Information (NC-BI) [35,36] and the genomes site at The Institute for Genomic Research [http://www.tigr.org] . Protein domain/motif searches were performed using PSI-BLAST [37], and CD-Search (RPS-BLAST) [36] at NCBI and the SMART utility at the European Molecular Biology Laboratories [38,39].