Comprehensive identification of conditionally essential genes in mycobacteria

All mycobacterial strains were grown in Middlebrook 7H9 broth (Difco) supplemented with 0.05% Tween 80 and AD (0.5% BSA, 0.2% dextrose, 0.085% NaCl) when grown in liquid culture. Mycobacterium smegmatis mc2–155 (13) and Mycobacterium fortuitum MBH1 (a clinical isolate provided by Gloria Brown, Massachusetts General Hospital, Boston, MA) were cultured on LB agar. Mycobacterium bovis BCG Pasteur and Mycobacterium tuberculosis H37Rv were cultured on Middlebrook 7H10 agar supplemented with AD or Middlebrook OADC enrichment (Becton Dickinson), respectively. For selection of mutants, solid media were supplemented with kanamycin at 20 (M. bovis BCG and M. tuberculosis), 50 (M. smegmatis), or 250 μg/ml (M. fortuitum). Selection of auxotrophs was performed on either minimal medium (7H10 agar supplemented with AD) or rich medium (7H11 agar supplemented with AD/0.32 mM thymine/5 mM adenine/0.3 mM guanine/50 μM thiamine/0.1 mM diaminopimelic acid).

PCR was used to amplify a fragment corresponding to each predicted ORF of M. tuberculosis H37Rv. To ensure that each primer pair would amplify only one product, each ORF in the genome was compared with all others by using blast (14). Sequences with an E value of <1e-5 were used to generate “mispriming libraries” for the next step. A set of 10 oligonucleotide primer pairs was generated for each gene with PRIMER3 (S. Rozen and H. J. Skaletsky (1997) at http://www-genome.wi.mit.edu/genome_software/other/primer3.html.), excluding primers that might prime on the mispriming library. Then, blast was used to eliminate primer pairs that generate products that could crosshybridize to other genes (>77% identity). Primer pairs that met these criteria were found for 3,855 of the 3,924 predicted ORFs. Many of the remaining ORFs represent transposons or phage-related genes. Thus, each predicted ORF of M. tuberculosis H37Rv was represented by a fragment that was predicted to have minimal cross-hybridization with other M. tuberculosis sequences. Fragments varied in length from 70 to 499 bp, with a median length of ≈350. Forward and reverse primers contained 5′ extensions: GGCATCTAGAG and CCGCACTAGTCCTC, respectively. PCRs with these gene-specific primers was performed by using 1.25 units of Taq and 0.15 units of Pfu polymerase (Stratagene) in 50 μl reactions containing 2.5 mM MgCl₂, 10% (vol/vol) DMSO and H37Rv genomic DNA as template. PCR conditions were 94°C for 2 min; 30 cycles of 94°C for 30 sec, 60°C for 30 sec, 72°C for 1 min; and 72°C for 5 min. These products were diluted 1:100, and 2.5 μl was used as a template for PCR with the following universal primers (both contain 5′ amino modifications including a three-carbon linker; Genosys, The Woodlands, TX): GAACCGATAGGCATCTAGAG and GAAATCCACCGCACTAGTCCTC. PCR was performed as follows: 94°C for 2 min; 3 cycles of 94°C for 30 sec, 40°C for 30 sec, 72°C for 1 min; 20 cycles of 94°C for 30 sec, 60°C for 30 sec, 72°C for 1 min; and 72°C for 5 min. PCR of all but 11 ORFs produced single products when analyzed by gel electrophoresis. Those reactions that did not produce products of appropriate size or produced multiple products were not included in further analysis. PCR products were purified by using multiscreen PCR plates (Millipore) and arrayed onto 3D-Link slides (Surmodics, Eden Prairie, MN) in duplicate, as recommended by the manufacturer.

The transposon donor phagemid, φMycoMarT7, was constructed in several steps starting with pMycoMar (15). φMycoMarT7 includes a transposon that encodes a kanamycin resistance gene and T7 promoters oriented so as to promote transcription into adjacent chromosomal DNA. The transposon is flanked by 29-bp inverted repeats. The entire sequence of the MycoMarT7 transposon has been deposited in GenBank (GenBank accession no. AF411123). The phagemid also contains the highly active C9 Himar1 transposase gene (16). A plasmid containing these elements and a bacteriophage λ cos site was cloned into PacI-digested φAE87, and phage stocks were generated as described (17). For transduction, mycobacterial cultures were grown to OD₆₀₀ of 0.8–1.0 in liquid culture. Cells were washed twice with MP buffer (50 mM Tris, pH 7.5/150 mM NaCl/10 mM MgSO₄/2 mM CaCl₂) and resuspended in 1/10 of the original culture volume in MP. Cells were infected with 10¹⁰ phage per ml of original culture for 3 h at 37°C. Transduced cells were plated directly onto solid media and cultured at 37°C. Libraries were collected by scraping colonies off plates. Aliquots were used either for genomic DNA isolation (see below) or frozen at −80°C.

Genomic DNA was isolated from mutant pools as described (18). A quantity (0.5 μg) of DNA/reaction was partially digested separately with HinPI and MspI (New England Biolabs). Fragments (500–2,000 bp) were purified from a 1% agarose gel with a Qiaquick gel purification kit (Qiagen, Chatsworth, CA). The resulting DNA was ligated to a 1,000-fold molar excess of the following adapter: CGACCACGACCA (includes 3′ C6-TFA-amino modification; Genosys) and AGTCTCGCAGATGATAAGGTGGTCGTGGT, by using T4 DNA ligase at 16°C. Ligated DNA (1 μl) was used as a template in two separate PCR reactions containing the adapter primer (GTCCAGTCTCGCAGATGATAAGG) and one of the transposon primers (primer 1-CCCGAAAAGTGCCACCTAAATTGTAAGCG or primer 2-CGCTTCCTCGTGCTTTACGGTATCG). PCR conditions were as follows: 94°C for 2 min; 5 cycles of 94°C for 30 sec, 72°C or 69°C for 30 sec, and 72°C for 1.5 min; 5 cycles of 94°C for 30 sec, 70°C or 67°C for 30 sec, and 70°C for 1.5 min; 15–25 cycles of 94°C for 30 sec, 68°C or 65°C for 30 sec, and 68°C for 1.5 min + 5 sec per cycle; and 72°C for 5 min. The low and high annealing temperatures were used for transposon primer 1 and 2, respectively. PCR products (250–500 bp) were purified from a 2% agarose gel and used as template for in vitro transcription reactions with T7 polymerase (Ambion, Austin, TX). After transcription, the template DNA was removed by using RQ1 DnaseI (Ambion), followed by extraction with phenol:chloroform and gel-filtration by using Centrisep-20 columns (Princeton Separations, Adelphia, NJ). RNA (1 μg) was labeled with either biotin or fluorescein by using ULS reagents (NEN, Boston, MA), purified with a Centrisep column, and concentrated by evaporation.

Microarrays were prehybridized in 5× SSC (1× SSC = 0.15 M sodium chloride/0.015 M sodium citrate, pH 7), 0.1% SDS, 100 μg/ml tRNA, and 1% BSA for 1 h, rinsed with water, and dried. Labeled RNAs were mixed in 50% (vol/vol) formamide, 5× SSC, 0.1% SDS, and 100 μg/ml tRNA, denatured at 65°C for 3 min, and hybridized to the microarray. Hybridizations were performed in a 40-μl volume under a coverslip at 42°C overnight. The arrays were washed in 5× SSC, 0.1% SDS at 42°C, then washed with 0.2× SSC, 0.1% SDS at room temperature, and finally washed in 0.2× SSC. Hybridized RNA was detected with the Micromax TSA system (NEN) using the manufacturer's protocol. Images were acquired with a Scanarray microarray scanner (GSI Lumonics, Watertown, MA) and analyzed with GENEPIX software (Axon Instruments, Foster City, CA).

We have used a Himar1-based mariner transposon to generate diverse libraries of transposon mutants in mycobacteria. To maximize the efficiency of mutagenesis, we used a mutant gene encoding a highly active Himar1 transposase (16) and transduction using a temperature-sensitive mycobacteriophage (17). As shown in Fig. 1, this transposon construct can be used to produce large numbers of mutants in a variety of mycobacterial species. Surprisingly, even species that are not lysed by the transducing phage when grown at the permissive temperature were transduced with the transposon. For example, this phage does not form plaques on Mycobacterium fortuitum lawns at 30°C. However, the transducing phage introduced transposons into this species with the same efficiency obtained in Mycobacterium smegmatis, a species in which the phage is lytic. Mutagenesis of the slowly growing pathogenic species, M. tuberculosis and M. bovis BCG, was more efficient than rapidly growing strains.

By using the transducing phage, we have produced libraries of greater than 10⁵ transposon mutants in both M. bovis BCG and M. tuberculosis. Thus far, the insertion sites from more than 40 mutants have been identified by DNA sequencing, and no evidence of multiple insertions in a single strain has been found. Himar1 transposons insert only into the dinucleotide TA, which is present >70,000 times in the genome of M. tuberculosis. A substantial number of these sites occur in essential regions that cannot sustain mutations, so the number of insertions that can be recovered is somewhat smaller. Therefore, in a library of 10⁵ mutants, an insertion will be present at nearly every TA. Assuming these sites are randomly distributed throughout the genome, an insertion will be present approximately every 60 bases. Therefore, a library of this size will contain an average of ≈17 insertions in a gene of 1 kb length.

To be a useful tool, transposon mutagenesis should produce a wide variety of different mutant phenotypes. To verify the complexity of our mutant libraries, we sought to select strains from our library that were resistant to antimycobacterial antibiotics.

Several drugs used to treat mycobacterial infections require activation by bacterial enzymes. In the absence of such an activity these agents are not converted to their active forms and exert no toxicity. Both isoniazid (INH) and ethionamide require such activation (19, 20), and mutations that alter the function of the activating enzymes are responsible for antibiotic resistance in many clinical M. tuberculosis isolates (20, 21). Because transposons often disrupt genes, we expected to find strains resistant to these antibiotics among the transposon mutants. Resistance to other antibiotics, however, does not result from gene inactivation. For example, M. tuberculosis strains that are resistant to streptomycin almost uniformly have point mutations in the rpsL gene, which alter affinity for the antibiotic (22). Because rpsL is essential for survival, we did not expect to find streptomycin resistance in the transposon mutant pool.

We selected an M. bovis BCG mutant library containing 10⁵ independent mutants on streptomycin, INH, and ethionamide. As shown in Fig. 1, we did not find any streptomycin-resistant strains. However, resistance to INH and ethionamide was present at much higher than spontaneous rates. By using PCR, we mapped the sites of transposon insertions in the INH-resistant mutants. All twenty INH-resistant strains had insertions in the katG gene that encodes the enzyme required for INH activation. As seen in Fig. 1, independent insertions were found throughout the INH gene. Similarly, 6 of 12 ethionamide-resistant mutants contained insertions in the gene encoding the known activator, etaA. In addition, multiple insertions in the ndh gene were found (data not shown). ndh mutants also are known to be resistant to ethionamide (23). The rates of INH and ethionamide resistance were consistent with the rates expected if transposition had occurred randomly throughout the chromosome.

TraSH uses a DNA microarray to map simultaneously all transposon insertions in a pool of mutants. A microarray was constructed that contains a fragment of DNA derived from each predicted ORF in the genome. To identify the location of insertions in a mutant pool, we produced labeled RNA that hybridized to the chromosomal sequences immediately adjacent to each transposon. As shown in Fig. 2B, genomic DNA was isolated from a pool of mutants and fragmented by using restriction enzymes with sites that appear frequently in the M. tuberculosis genome. Adapters were ligated to the resulting DNA fragments, and PCR was performed by using adapter- and transposon-specific primers to amplify the regions adjacent to the transposon-insertion sites. PCR products were size-selected by agarose gel electrophoresis to produce fragments ≈250–500 bp in size (containing ≈100–350 bp of chromosomal sequence) to minimize overlap into adjacent ORFs. In vitro transcription then was performed by using T7 RNA polymerase that recognizes outward-facing T7 promoters located in the transposon. This RNA was labeled and hybridized to the DNA microarray. RNA derived from mutant pools, which have been grown under different selective conditions, can be labeled with fluorophores that emit at different wavelengths. By mixing these labeled RNA's before hybridization, differential gene requirements are identified (see Fig. 2A).

TraSH only identifies mutants that lie close to the region represented by the ORF fragment immobilized on the DNA microarray. By convention, these fragments are referred to as probes (24). An average probe is 330 bp in length, whereas an average TraSH target (the RNA derived from in vitro transcription) hybridizes to ≈225 bp of chromosomal sequence. Assuming that 50 bp of overlap between the DNA probe and the RNA target are required for hybridization, an average probe can be identified by targets generated from ≈500 bp of chromosome. In a random transposon library of 10⁵ mutants, an average probe would identify ≈10 different mutants.

We compared the growth of libraries containing ≈10⁵ M. bovis BCG mutants on two different defined media. After mutagenesis, bacteria were grown on rich medium (Middlebrook 7H11 agar, which contains all amino acids with added supplements, as described in Materials and Methods) or minimal medium (Middlebrook 7H10 agar). Genomic DNA was prepared from the surviving colonies under both growth conditions. TraSH targets were synthesized from each pool and labeled with different dyes. The targets then were mixed and hybridized to the DNA microarray.

For each DNA probe on the microarray, we determined the ratio of intensity produced by each dye. Probes that produced only a low-intensity signal (less than two times the background intensity) from the dye representing the population grown on rich medium were omitted from the analysis (≈30% of genes). Genes that produced ratios of rich/minimal medium in the top 5% in two independent experiments are reported in Table 1. Fig. 3 depicts the data obtained from the chromosomal regions surrounding several representative auxotrophic mutations.

We found 21 genes that met the above criteria. Of these 21 genes, 12 have annotated functions based on homology, and 10 are known to be involved in biosynthesis. The other two genes, gmhA and fadB, are similar to genes involved in heptose and fatty-acid metabolism, respectively. It is unclear why these two genes should be required for growth on minimal medium. The remaining identified genes do not have clear homology to genes of known function.

The 10 genes that are predicted to be involved in biosynthesis can be grouped into four different metabolic pathways. First, leuB, leuD, and ilvA are all involved in leucine synthesis (25). Second, mutations in subI, a sulfate transporter, result in methionine auxotrophy (12). Similarly, cysH and cysD are involved in the reduction of sulfate, which is necessary for the synthesis of both cysteine and methionine (26). Third, glnE, gltB, and amiB2 are involved in the metabolism of glutamine, glutamate, and nitrogen. Finally, thiC is necessary for thiamin synthesis (25). The identification of multiple genes in a pathway provides significant validation for this method of functional analysis. In addition, our results suggest that only a limited number of pathways can be exploited for the generation of auxotrophic mutants in pathogenic mycobacteria (see Discussion).