Comparison of the Construction of Unmarked Deletion Mutations in Mycobacterium smegmatis, Mycobacterium bovis Bacillus Calmette-Guérin, and Mycobacterium tuberculosis H37Rv by Allelic Exchange

Abstract

Until recently, genetic analysis of Mycobacterium tuberculosis, the causative agent of tuberculosis, was hindered by a lack of methods for gene disruptions and allelic exchange. Several groups have described different methods for disrupting genes marked with antibiotic resistance determinants in the slow-growing organisms Mycobacterium bovis bacillus Calmette-Guérin (BCG) and M. tuberculosis. In this study, we described the first report of using a mycobacterial suicidal plasmid bearing the counterselectable marker sacB for the allelic exchange of unmarked deletion mutations in the chromosomes of two substrains of M. bovis BCG and M. tuberculosis H37Rv. In addition, our comparison of the recombination frequencies in these two slow-growing species and that of the fast-growing organism Mycobacterium smegmatis suggests that the homologous recombination machinery of the three species is equally efficient. The mutants constructed here have deletions in the lysA gene, encoding meso-diaminopimelate decarboxylase, an enzyme catalyzing the last step in lysine biosynthesis. We observed striking differences in the lysine auxotrophic phenotypes of these three species of mycobacteria. The M. smegmatis mutant can grow on lysine-supplemented defined medium or complex rich medium, while the BCG mutants grow only on lysine-supplemented defined medium and are unable to form colonies on complex rich medium. The M. tuberculosis lysine auxotroph requires 25-fold more lysine on defined medium than do the other mutants and is dependent upon the detergent Tween 80. The mutants described in this work are potential vaccine candidates and can also be used for studies of cell wall biosynthesis and amino acid metabolism.

Mycobacterium tuberculosis, the agent of tuberculosis, is the leading cause of death in adults worldwide (13). The emergence of drug-resistant strains (47) and the problems associated with tuberculosis in human immunodeficiency virus-infected populations (17) have brought tuberculosis research to the forefront. The development of genetic techniques to study the biology of the organism is an important goal of mycobacterial research.

Considerable effort has gone into the development of allelic-exchange methods to selectively disrupt genes of various mycobacterial species. Several groups have used either small linear DNA fragments (3, 24, 41), long linear DNA fragments (4), or suicidal plasmids (8, 26, 36, 37, 39, 40, 42) to achieve allelic exchange in both fast- and slow-growing mycobacteria. Slow-growing mycobacteria such as M. tuberculosis and Mycobacterium bovis bacillus Calmette-Guérin (BCG) can integrate exogenous DNA into their chromosomes by both illegitimate and homologous recombination (1, 24). Allelic exchange in fast-growing mycobacteria such as Mycobacterium smegmatis is easier than in the slow-growing species; this has led to the idea that the homologous recombination machinery of slow-growing mycobacteria is rather inefficient (31).

Thus far, the only mutants constructed among the slow-growing mycobacterial species are those with genes disrupted with an antibiotic resistance marker. However, in many cases an antibiotic resistance marker may not be desirable. It may not be known whether a gene is essential, and targeted disruption does not let one ascertain essentiality. The failure to obtain a mutant might be due to the failure of the methodology and not to the essentiality of the gene. Furthermore, the possibility of polar effects from an inserted antibiotic resistance marker can prevent the disruption of a nonessential gene if that gene is located in an operon upstream of an essential gene. Also, there are a limited number of antibiotic resistance genes available for use in mycobacteria and making a marked mutation excludes one antibiotic from further consideration. In addition, mutants that are potential vaccine candidates should not contain antibiotic resistance determinants.

An ideal allelic-exchange system is one that can be used for the exchange of unmarked deletion alleles as well as alleles with point mutations. Constructing knockout mutants by in-frame deletions would negate the concerns with using a targeted disruption method. Such mutants are antibiotic sensitive and cannot revert, and the mutations should not be polar on the expression of downstream genes. By extension, the same technique could be used for allelic exchange of point mutations, allowing for a finer dissection of gene function. This type of unmarked allelic-exchange methodology, utilizing a plasmid unable to replicate in the organism of interest and selectable and counterselectable markers (14), has been successfully used for M. smegmatis (26, 39). We wanted to determine if such an allelic-exchange methodology would reproducibly work for the slow-growing mycobacteria M. bovis BCG and M. tuberculosis.

In this study, we describe a new mycobacterial suicide plasmid for allelic exchange of unmarked mutations with sacB sucrose counterselection. This counterselectable marker was previously reported to work for the allelic exchange of marked mutations in M. tuberculosis and M. bovis BCG (8, 38, 40). In this work, we demonstrate the reproducibility of allelic exchange of unmarked deletions in the chromosome of M. bovis BCG and M. tuberculosis. We chose to construct lysine auxotrophs of these two slow-growing mycobacteria and M. smegmatis, by allelic exchange of lysA, the gene encoding meso-diaminopimelate (DAP) decarboxylase, the last enzyme in the lysine biosynthetic pathway (51). We compared the kinetics of homologous recombination in these species and found that the frequency of allelic exchange at the lysA locus was remarkably similar among the three organisms. We also examined the nutritional requirements of the lysine auxotrophs and found striking differences among these mutants. To the best of our knowledge, this is the first report of the construction of unmarked deletion mutations in the genomes of slow-growing mycobacteria and the first direct comparison of the same allelic-exchange technique in both slow- and fast-growing mycobacteria. The results from this study suggest that the homologous recombination machinery in slow- and fast-growing mycobacteria may function with similar efficiency.

MATERIALS AND METHODS

Bacterial strains and culture methods.

The bacterial strains used in this study are listed in Table 1. The genetic nomenclature for strains bearing an integrated suicide plasmid (DUP) was previously described (36). Escherichia coli cultures were grown in Luria-Bertani (LB) broth or on LB agar (Difco). Mycobacterial cultures were grown in Middlebrook 7H9 broth (Difco) with 0.05% Tween 80, on 7H9 medium solidified with 1.5% agar, or on Middlebrook 7H10 or 7H11 medium (Difco). All cultures were incubated at 37°C. All Middlebrook media were supplemented with 0.2% (vol/vol) glycerol and with 1× ADS (0.5% bovine serum albumin, fraction V [Boehringer Mannheim]; 0.2% dextrose; and 0.85% NaCl) for M. bovis BCG and M. tuberculosis cultures. Our basal media were 7H9 and 7H10 supplemented as described above. Sucrose was used in medium at a concentration of 2% (wt/vol), added after the medium was autoclaved and cooled to 55°C. Casamino Acids (acid-hydrolyzed casein; Difco) were used at a concentration of 0.2% (wt/vol). Individual amino acids were obtained from Sigma Chemical (St. Louis, Mo.) and used at a concentration of 40 μg/ml, unless indicated otherwise. The lysine analog S-(β-aminoethyl)-l-cysteine (AEC) was obtained from Sigma Chemical, dissolved in water, and used at a concentration of 3 mM. When required, the following antibiotics were used at the specified concentrations: carbenicillin (50 μg/ml, E. coli), kanamycin A monosulfate (25 μg/ml, E. coli, M. smegmatis, and M. bovis BCG), hygromycin B (50 μg/ml, E. coli, M. bovis BCG, and M. tuberculosis; 150 μg/ml, M. smegmatis). Hygromycin B was purchased from Boehringer Mannheim (50 mg/ml in phosphate-buffered saline), and all other antibiotics were purchased from Sigma Chemical. Note that we often found that pYUB412- and pYUB405-based plasmids were stable in E. coli only with use of both carbenicillin and hygromycin at 50 μg/ml in solid and liquid media. M. smegmatis plates were incubated for 3 to 5 days, while M. bovis BCG and M. tuberculosis plates were incubated for 3 to 4 weeks. M. bovis BCG and M. tuberculosis starter cultures were inoculated by using 1-ml frozen stocks in 10 ml of medium in 30-ml plastic medium bottles and incubated for 5 to 7 days on a shaker platform at 100 rpm. Larger cultures were inoculated from the starter cultures at a 1:50 dilution in 50 or 100 ml of medium within 490-cm² roller bottles (Corning) and incubated on a roller apparatus at 8 rpm for 5 to 7 days. For growth curves, mid- to late-exponential-phase cultures were centrifuged and washed with fresh medium lacking supplements, and the cells were resuspended appropriately and inoculated into test medium. Samples of M. tuberculosis and BCG cultures were mixed 1:1 with 10% phosphate-buffered formalin and fixed for at least 1 h prior to spectrophotometric measurement of optical density at 600 nm.

TABLE 1.

Strains used in this study

Strain	Description	Reference or source
E. coli K-12
HB101	F− Δ(gpt-proA)62 leuB1 glnV44 ara-14 lacY1 hsdS20 rpsL20 xyl-5 mtl-1 recA13	9
DH5α	F− [φ80dΔlacZM15]Δ(lacZYA-argF)U169 deoR recA1 endA1 hsdR17 glnV44 thi-1 gyrA96 relA1	18
M. smegmatis
mc2155	ept-1	46
mc21492	ept-1 DUP2 [(argS ΔlysA4 hdh′)*pYUB657*(argS lysA hdh)]	This work
mc21493	ept-1 ΔlysA4	This work
M. bovis BCG
Pasteur	Vaccine strain	Statens Seruminstitut
mc21601	Pasteur DUP3 [(argS lysA hdh thrC)*pYUB657*(argS ΔlysA5::res hdh thrC′)]	This work
mc21602	Pasteur DUP4 [(argS ΔlysA5::res hdh thrC")*pYUB657*(argS lysA hdh thrC)]	This work
mc21604	Pasteur ΔlysA5::res	This work
Connaught	Vaccine strain	AECOMa
mc21618	Connaught::pYUB668 homologous primary recombinant, clone 3	This work
mc22519	Connaught ΔlysA5::res	This work
M. tuberculosis
H37Rv	Virulent	AECOM
mc22998	H37Rv::pYUB668 homologous primary recombinant, clone 1	This work
mc22999	H37Rv::pYUB668 homologous primary recombinant, clone 2	This work
mc23026	H37RvΔlysA5::res	This work

^aAECOM, Albert Einstein College of Medicine. 

DNA methodologies.

DNA manipulations were done essentially as previously described (28). The plasmids used in this study are listed in Table 2. Plasmids were constructed in E. coli HB101 or DH5α cells and prepared by an alkaline lysis protocol (21). Plasmids used for recombination were purified with Qiagen columns as recommended by the manufacturer (Qiagen, Inc., Chatsworth, Calif.). DNA fragments used for plasmid construction were purified by agarose gel electrophoresis and recovered by absorption to a silica matrix (GeneClean; Bio 101, Vista, Calif.).

TABLE 2.

Plasmids used in this study

Name	Description	Reference or source
pKSI+	Apr; high-copy-number cloning vector	Stratagene
pMV261	Kmr; E. coli-mycobacterial shuttle vector	50
pET3d.lysA	M. tuberculosis Erdman lysA gene cloned into pET3d	15
pCVD442	AprsacB	14
pYUB328	Apr; PacI-excisable cosmid vector, ColE1	4
pYUB405	Apr Hygr; PacI-excisable cosmid vector, ColE1, does not replicate in mycobacteria	5
pYUB412	Apr Hygr; E. coli-mycobacterial shuttle, PacI-excisable cosmid vector, ColE1 origin, int attP, nonreplicative but integration proficient in mycobacteria	5
pYUB601	In vitro-repackaged pYUB412::lysA+ cosmid from mc2155 library	This work
pYUB604	4.4-kb EcoRI fragment from pYUB601 cloned in the EcoRI site of pMV261	This work
pYUB605	5.5-kb NotI self-ligated subclone of pYUB604	This work
pYUB607	3.4-kb NotI fragment from pYUB604 cloned into NotI site of pKSI+	This work
pYUB617	7.7-kb inverse XL-PCR product from pYUB604, containing a 1.2-kb deletion of lysA (ΔlysA4) marked with unique SnaBI site	This work
pYUB618	3.2-kb EcoRI fragment from pYUB617, bearing ΔlysA4, blunt end cloned into PacI sites of pYUB657	This work
pYUB631	2.5-kb PstI fragment from pCVD442, bearing sacB, cloned into same of pMV261	This work
pYUB635	1.3-kb XbaI-BamHI lysA gene from pET3d.lysA, cloned into same sites of pKSI+	This work
pYUB636	3-kb inverse XL-PCR product from pYUB635, containing 95-bp deletion of lysA marked with unique MluI site	This work
pYUB638	1.4-kb MluI res-aph-res cassette cloned into MluI site in pYUB636	This work
pYUB651	pYUB412 containing lysA+ of M. tuberculosis Erdman, under control of the M. bovis BCG groEL (Hsp60) promoter
pYUB657	3.5-kb NotI-NheI fragment from pYUB631, bearing groEL (Hsp60) promoter and sacB, cloned into the EcoRV site of pYUB405	This work
pYUB659	11-kb SnaBI fragment from cosY373 cloned into the EcoRV site of pKSI+	This work
pYUB665	1.7-kb NheI-BglII fragment from pYUB638 (ΔlysA::res-aph-res) replacing 300-bp NheI-BglII (lysA+) fragment in pYUB659	This work
pYUB667	pYUB665 with the aph gene resolved by passage in E. coli DH5α; Kms	This work
pYUB668	8.4-kb HpaI fragment from pYUB667 cloned into the PacI sites of pYUB657	This work
cosY373	pYUB382::M. tuberculosis H37Rv cosmid bearing the lysA operon	43

Genomic DNA was prepared either as previously described (22) or by a modified guanidinium protocol (33). Briefly, the cells from a 10-ml culture were lysed with 1.3 ml of a 3:1 mixture of chloroform-methanol. The lysate was then mixed with 1.3 ml of Tris-equilibrated phenol and 2 ml of GTC solution (4 M guanidinium thiocyanate, 0.1 M Tris [pH 7.5], 0.5% sarcosyl, with β-mercaptoethanol added to a final concentration of 1% prior to use). The upper phase was collected after centrifugation, and the genomic DNA was precipitated with isopropanol. Southern blotting and hybridization were done as previously described (36). Oligonucleotides for sequencing and PCR were synthesized by the Albert Einstein College of Medicine oligonucleotide synthesis facility.

Cloning and sequencing of the M. smegmatis lysA operon.

We used a library of genomic DNA from wild-type M. smegmatis mc²155 constructed in the cosmid vector pYUB412 to clone the lysA gene. The vector pYUB412 is an integration-proficient, PacI-excisable cosmid vector (5). This cosmid vector has the mycobacteriophage L5 attachment site (attP), the L5 integrase gene (int), and the hyg gene, conferring resistance to hygromycin. This vector efficiently integrates into the mycobacteriophage L5 attachment site (attB) of the mycobacterial chromosome and is stable (27). The pYUB412::mc²155 library was electroporated into the strain MCK3037, a lysine auxotrophic mutant of mc²155 generated by ethyl methanesulfonate mutagenesis (32). Transformants were selected on 7H10 medium lacking lysine, and Lys⁺ clones were screened for the hygromycin resistance marker carried on the cosmid vector backbone. One Lys⁺ Hyg^r clone was chosen for study, and the genomic DNA insert within the integrated cosmid was recovered by λ in vitro packaging (GigaPak III; Stratagene). The recovery procedure was as follows. The library insert DNA was flanked by PacI restriction endonuclease sites present in the cosmid vector, and since PacI sites do not exist in mycobacterial genomic DNA (25), PacI digestion of the genomic DNA releases the cosmid insert DNA. This DNA fragment was repackaged into PacI-digested arms of the cosmid vector pYUB412 by λ in vitro packaging, and a new cosmid (pYUB601) with the insert was recovered in E. coli. The cosmid pYUB601 insert DNA was subcloned into a 4.4-kb EcoRI fragment bearing the lysA gene in plasmid pYUB604. The plasmid pYUB604 and two subclones, pYUB605 and pYUB607, were templates for DNA sequencing with the Applied Biosystems Prism Dye Terminator Cycle Sequencing Core kit with AmpliTaq DNA polymerase (Perkin-Elmer) and an Applied Biosystems 377 automated DNA sequencer. Sequence data for both strands of the lysA operon of M. smegmatis were obtained from these subclones and by primer walking.

Construction of sacB suicide vector pYUB657.

A 2.5-kb PstI fragment from the E. coli sacB vector pCVD442 bearing sacB and its upstream regulatory region sacR were subcloned into the PstI site of the shuttle vector pMV261 downstream of the mycobacterial groEL (Hsp60) promoter, yielding the plasmid pYUB631. A 3.5-kb NotI-NheI fragment from pYUB631, bearing P_groEL-sacB, was cloned into the cosmid vector pYUB405, resulting in the final construct pYUB657 (see Fig. 1). The vector pYUB405 is a PacI-excisable cosmid vector unable to replicate in mycobacteria and encodes resistance to ampicillin and hygromycin (5).

FIG. 1.

Map of the suicide vector pYUB657. The vector pYUB657 cannot replicate in mycobacteria but has the ColE1 origin of replication for E. coli. The P_groEL-sacB cassette is indicated along with the sacR regulatory region (49). The vector has the bla gene, conferring resistance to ampicillin in E. coli, and the hyg gene, conferring resistance to hygromycin in mycobacteria. This vector is also a double-cos, PacI-excisable cosmid cloning vector (4). Useful cloning sites are indicated.

Construction of the M. smegmatis ΔlysA4 suicide plasmid pYUB618.

The plasmid pYUB604 was used as the template in an inverse PCR to produce a deletion within the lysA gene. Oligonucleotide primers Pv44 (5′-CCCGTCGTACGTACGAACCAGGTTGCGC-3′) and Pv45 (5′-CGAGTCGATACGTACTGCTGTGCCGCCC-3′) were used at 50 pmol each in an inverse XL-PCR in a Perkin-Elmer 9600 temperature cycler with the following program: 95°C for 5 min, 1 cycle; 93°C for 1 min and 68°C for 5 min, 16 cycles; 93°C for 1 min and 68°C for 5 min with the time increasing by 15 s for each cycle, 12 cycles; 72°C for 30 min. The reaction produced a 7.7-kb fragment with a 1.2-kb deletion within the lysA open reading frame (spanning nucleotide positions 2051 to 3251 of the sequence with GenBank accession no. AF126720) marked with a unique SnaBI site. The PCR product was gel purified, digested with SnaBI, and self-ligated to yield the plasmid pYUB617. A 3.2-kb EcoRI fragment from pYUB617 bearing the ΔlysA4 allele was cloned into the PacI sites of the mycobacterial sacB suicide vector pYUB657, resulting in the M. smegmatis ΔlysA4 suicide plasmid pYUB618.

Construction of the M. bovis BCG-M. tuberculosis ΔlysA5::res suicide plasmid pYUB668.

The lysA gene of M. tuberculosis was originally cloned and sequenced by Andersen and Hansen (2). The plasmid pET3d.lysA contains the lysA gene of M. tuberculosis Erdman cloned by PCR with primers designed on the basis of the previously published sequence (2, 15). A 1.3-kb XbaI-BamHI fragment bearing the lysA gene was cloned from pET3d.lysA into the same sites in pKSI⁺ to produce pYUB635. This plasmid was used as the template in an inverse PCR with the oligonucleotide primers Pv7 (5′-GATAGCGGTCACGCGTCTCGTGCGCGGTGGA-3′) and Pv8 (5′-TCCGTACGATACGCGTCAGCCACATCGGTTCG-3′) to generate a 95-bp deletion within the lysA gene marked with a unique MluI restriction endonuclease site. The inverse XL-PCR was done with a Perkin-Elmer 9600 temperature cycler and the program described above for plasmid pYUB617. The resulting 4.1-kb PCR product was gel purified, digested with MluI, and self-ligated to yield the plasmid pYUB636. The lysA deletion was marked with the aph gene, conferring kanamycin resistance, by insertion of a specialized aph cassette via the unique MluI site to yield pYUB638. This specialized cassette has an aph gene flanked by two γδ resolvase sites from the E. coli transposon γδ (Tn1000) (19). The presence of the resolvase sites made it possible to excise the antibiotic resistance marker by expressing the γδ resolvase in mycobacteria after the cassette had been inserted into the mycobacterial chromosome (7). For the purposes of this study, however, the res-aph-res marker was removed from pYUB638 by resolvase excision in E. coli DH5α prior to introduction into mycobacteria (see below).

To include more DNA on both sides of the M. tuberculosis ΔlysA allele, we used cosmid cosY373 from the Sanger Centre M. tuberculosis H37Rv genome sequencing project (11). An 11-kb SnaBI fragment from cosY373, containing lysA situated in the middle, was subcloned into the EcoRV site of pKSI⁺ to yield plasmid pYUB659. To replace the wild-type lysA allele in pYUB659 with the ΔlysA::res-aph-res allele constructed above in pYUB638, we exchanged an internal NheI-BglII fragment of lysA encompassing the deletion region between these two plasmids. Because there is an additional NheI site at the 5′ end of the res-aph-res cassette, this exchange resulted in an additional deletion of 236 bp within the lysA gene. The resulting plasmid, pYUB665, contains a deletion within lysA totaling 331 bp and the res-aph-res cassette. We passaged the plasmid pYUB665 in E. coli DH5α (which has a γδ element capable of excising the aph gene from the ΔlysA::res-aph-res allele) and isolated a Kn^s derivative, plasmid pYUB667. DNA sequence analysis of pYUB667 showed that the aph cassette was absent and that a single res site that was in frame with respect to the lysA open reading frame remained. The mutant lysA allele in pYUB667 is designated ΔlysA5::res and has a total deletion of 331 bp of an internal portion of the lysA gene, but with the addition of the 136-bp res site, the net change in size of ΔlysA5::res compared to the wild type is a decrease of 195 bp. To produce the final suicidal plasmid for allelic exchange in M. bovis BCG and M. tuberculosis, an 8.4-kb HpaI fragment from pYUB667 was cloned into the PacI sites of the sacB suicidal vector pYUB657, resulting in plasmid pYUB668. This plasmid has approximately 4 kb of DNA flanking each side of the ΔlysA5::res allele.

Electroporation of mycobacteria.

M. smegmatis was electroporated as previously described (36). M. bovis BCG and M. tuberculosis were electroporated as described for M. smegmatis, except that all manipulations were done at room temperature instead of on ice and the expression step proceeded overnight for approximately 12 h prior to plating. For recombination experiments, 1 μg of covalently closed supercoiled plasmid DNA was used for each electroporation.

Nucleotide sequence accession number.

The DNA sequence of the 4,462-bp EcoRI fragment encoding the M. smegmatis lysA gene was submitted to GenBank and assigned the accession no. AF126720.

RESULTS

Allelic-exchange methodology.

Our basic procedure for making mutants with the sacB suicidal vector pYUB657 (Fig. 1) is a two-step allelic exchange (14, 36). A suicidal recombination plasmid is electroporated into cells, and primary recombinants are selected upon hygromycin medium. Since the plasmid cannot replicate, any hygromycin-resistant clones must have integrated the plasmid into the chromosome by a single-crossover event. Because of the presence of the sacB gene on the pYUB657 vector backbone, the Hyg^r clones are also sensitive to sucrose (Suc^s). Plasmid integration at the desired locus results in a tandem duplication (given the designation DUP) of the cloned region with the vector DNA in the middle. One such DUP clone is grown to saturation in supplemented medium, during which time individuals within the population undergo a second homologous recombination event between the duplicated regions. In this event, the plasmid vector is lost along with the hyg and sacB genes, leaving behind either the wild-type or the mutant allele, depending upon which side of the mutation the second recombination event occurred. This second recombination event occurs at a low frequency; thus, there must be a selection for the desired secondary recombinants. To select these clones, one takes advantage of the loss of the sacB gene; any clone losing the plasmid is now sucrose resistant (Suc^r). The culture is plated on supplemented medium containing sucrose to kill any clones that did not undergo a second recombination event. The sucrose-resistant clones are then screened for hygromycin sensitivity and the mutant phenotype.

Cloning of the mycobacterial lysA genes.

For this study, we chose to test our system by constructing lysine auxotrophs via deletion of the lysA gene, encoding meso-DAP decarboxylase, in M. smegmatis, M. bovis BCG, and M. tuberculosis. The lysA gene of M. tuberculosis was already available and could also be used for allelic exchange in M. bovis BCG due to the conservation of DNA sequences between the two species; however, the lysA gene of M. smegmatis was not available. We cloned the M. smegmatis lysA gene and resident operon as described in Materials and Methods. The lysA gene of M. smegmatis is 1,424 bp in length and has 77% homology with the lysA gene of M. tuberculosis, while the two LysA proteins have an 80% identity (16). The structure of the lysA operon is conserved among several mycobacteria and the related organism Corynebacterium glutamicum. In M. tuberculosis, the gene order is as follows: argS (arginyl-tRNA synthetase), lysA (meso-DAP decarboxylase), hdh (homoserine dehydrogenase), thrC (threonine synthase), PGRS-17 [poly(GC)-rich repeat 17], and thrB (threonine kinase) (43). Our sequence from M. smegmatis spans from upstream of argS through the hdh gene. A similar argS-lysA operon arrangement is seen for Mycobacterium leprae (36) and Brevibacterium glutamicum (renamed C. glutamicum) (34). The hdh gene product supplies homoserine, the precursor for Met and Thr biosynthesis (29), while the thrC and thrB genes are responsible for threonine synthesis (35).

Construction of an unmarked lysA deletion mutant of M. smegmatis.

We electroporated M. smegmatis mc²155 with the ΔlysA4 suicidal plasmid pYUB618 (see Materials and Methods for plasmid construction) and obtained an average of 15 Hyg^r clones per transformation, with primary recombination efficiencies of 10⁻⁵ (Table 3). Two cultures of one strain, mc²1492, were grown to saturation in 7H9-lysine medium, and dilutions were plated onto 7H10-lysine medium supplemented with sucrose. We obtained sucrose-resistant clones at a frequency of 10⁻⁴; screened 100 clones from each set for Suc^r, Hyg^s, and auxotrophy; and found three basic phenotypes: Suc^r Hyg^r prototrophic, Suc^r Hyg^s prototrophic, and Suc^r Hyg^s auxotrophic (Table 4, experiments 1 and 2). The largest group was the Suc^r Hyg^r prototrophic class, which likely resulted from inactivation of the sacB gene, since the clones were still resistant to hygromycin and did not appear to have arisen from a secondary recombination event. The other two Suc^r classes were Hyg^s and appeared to result from secondary recombination events; the first class retained the wild-type allele, while the second class retained the mutant allele and was auxotrophic for lysine. One mutant was given the designation mc²1493, and allelic exchange of lysA was confirmed by Southern blotting (Fig. 2A). The mutant grows equally well on defined 7H9 medium supplemented with lysine and on complex medium (7H9 supplemented with Casamino Acids or LB medium).

TABLE 3.

Electroporation efficiencies and primary recombination frequencies for lysA allelic exchange

Species and strain	Suicide plasmid	na	Avg no. of Hygr clonesb	Electroporation efficiencyc	Recombination frequencyd
M. smegmatis mc2155	pYUB618	2	15 ± 3	3 × 105	5 × 10−5
M. bovis BCG substrain Pasteur	pYUB668	10	5 ± 3	1 × 104	5 × 10−4
M. bovis BCG substrain Connaught	pYUB668	5	2 ± 1	1 × 103	2 × 10−3
M. tuberculosis H37Rv	pYUB668	10	3 ± 3	3 × 105	1 × 10−5

^an, number of electroporations for each species-plasmid combination. Each set was done with the same stock of electrocompetent cells. 

^bAverage number of hygromycin-resistant clones (± standard deviation) from each set of electroporations done with the suicide plasmids. 

^cElectroporation efficiency is the number of Hyg^r clones obtained from electroporations done with pYUB412, an attP/int Hyg^r vector that integrates into the attB site of the mycobacterial genome. The number of Hyg^r clones from pYUB412 electroporations is an indicator of the electroporation efficiency of the cells; the number of transformants obtained with an attP/int vector is equivalent to the number obtained with a replicating vector. We have never observed spontaneous resistance to hygromycin in the species studied in this paper. 

^dRecombination frequency is calculated by dividing the average number of Hyg^r clones obtained per electroporation with suicide plasmids by the electroporation efficiency obtained with the vector pYUB412. 

TABLE 4.

Recombination products from segregation of lysA DUP in different mycobacterial species

Species	Expt	Strain	Relevant genotypea	Mediumb	Sucr frequencyc	nd	Frequency of phenotypes in Sucr populatione
							sacB-inactivated Hygr prototrophs	Secondary recombinant
								Hygs prototrophs	Hygs auxotrophs
M. smegmatis	1	mc21492	DUP2	K	4	100	67	24	9
	2	mc21492		K	3	100	60	31	9
M. bovis BCG
Pasteur	3	mc21601	DUP3	KMT	4	48	2	63	35
	4	mc21602	DUP4	KMT	9	46	26	33	41
	5	mc21601	DUP3	Basal	0.2	92	9	91	0
	6	mc21601		K	0.9	86	15	73	12
	7	mc21601		KMT	3	90	11	61	28
	8	mc21601		CAA	6	78	8	92	0
Connaught	9	clone 3	Hom. pYUB668	K	NDg	47	15	51	34
	10	clone 9	Hom. pYUB668	K	ND	48	6	54	40
	11	clone 10	Hom. pYUB668	K	ND	47	10	77	13
	12	clone 2	Illeg. pYUB668	K	ND	48	100	0	0
	13	clone 4	Illeg. pYUB668	K	ND	48	96	4	0
	14	clone 8	Illeg. pYUB668	K	ND	47	98	2	0
	15	clone 11	Illeg. pYUB668	K	ND	95	100	0	0
M. tuberculosis H37Rv	16	mc22998	Hom. pYUB668	Basal	1	44	16	84	0
	17	mc22998		K	0.3	41	10	90	0
	18	mc22998		KMT	1	45	16	84	0
	19	mc22998		CAA	0.6	40	23	77	0
	20	mc22999	Hom. pYUB668	Basal	0.5	42	26	74	0
	21	mc22999		K	0.9	38	13	87	0
	22	mc22999		KMT	2	44	36	64	0
	23	mc22999		CAA	0.7	34	6	94	0
	24	mc22998	Hom. pYUB668	K200	2	39	44	56	0
	25	mc22998		K200/TW	10	287	20	80	0
	26	mc22998		K1	0.3	96	20	80	0
	27f	mc22998		K1/TW	1 L	96  L	17 L	83 L	0 L
					0.8 S	63  S	0 S	0 S	100 S

^aDUP designation is used for strains with pYUB668 integrated at lysA with known orientation (Table 1). “Illeg. pYUB668” refers to primary Hyg^r Suc^s clones in which pYUB668 integrated into the chromosome via illegitimate recombination. “Hom. pYUB668” refers to primary Hyg^r Suc^s clones in which pYUB688 integrated at lysA but the orientation of the duplication is unknown. 

^bType of medium used for outgrowth (Middlebrook 7H9) and sucrose selection (Middlebrook 7H10): basal (no supplementation), K (lysine at 40 μg/ml), KMT (lysine, methionine, and threonine each at 40 μg/ml), CAA (0.2% Casamino Acids), K200 (lysine at 200 μg/ml), K200/TW (lysine at 200 μg/ml plus 0.05% Tween 80), K1 (lysine at 1 mg/ml), and K1/TW (lysine at 1 mg/ml plus 0.05% Tween 80). 

^cNumber of Suc^r CFU per milliliter divided by the viable CFU per milliliter (expressed as n × 10⁻⁴). 

^dNumber of Suc^r clones screened. 

^eFrequencies of phenotypes expressed as percentages of the numbers of sucrose-resistant clones screened. Phenotypes are Hyg^r prototrophy (not secondary recombinant; sacB inactivated), Hyg^s prototrophy (secondary recombinant; wild-type lysA), and Hyg^s auxotrophy (secondary recombinant; ΔlysA). 

^fFor experiment 27, “L” refers to large colonies while “S” refers to small colonies seen on the sucrose selection medium. 

^gND, not determined. 

FIG. 2.

Southern blots of genomic DNA from four mycobacterial lysA deletion mutants. (A) Genomic DNA from wild-type M. smegmatis mc²155 (lane 1) or the M. smegmatis auxotroph mc²1493 (lane 2), digested with EcoRI and probed with a 3.3-kb EcoRI fragment from plasmid pYUB617, encompassing the ΔlysA4 allele. The wild-type fragment is the expected 4.4-kb size, while the genomic DNA from the mutant has the expected 3.2-kb fragment. (B) Genomic DNA from wild-type BCG substrain Pasteur (lane 1), BCG substrain Pasteur auxotroph mc²1604 (lane 2), wild-type BCG substrain Connaught (lane 3), BCG substrain Connaught auxotroph mc²2519 (lane 4), wild-type M. tuberculosis H37Rv (lane 5), and M. tuberculosis H37Rv auxotroph mc²3026 (lane 6), digested with BssHII and probed with a lysA PCR product obtained from BCG substrain Pasteur wild-type genomic DNA. Digestion of wild-type genomic DNA with BssHII splits the lysA gene over two fragments, one of which is 1.1 kb in size and the other of which is 1.2 kb. Digestion of genomic DNA from the deletion mutants yields the same 1.2-kb fragment seen in the wild type with a 0.9-kb fragment, corresponding to the deletion site, replacing the 1.1-kb fragment. The blot in panel B shows the expected shift in size of the 1.1-kb fragment down to 0.9 kb in all three mutants (lanes 2, 4, and 6). The invariant 1.2-kb fragment shows a lower intensity in the blot due to a lower percentage of homology to the probe, relative to the 1.1- and 0.9-kb fragments.

Construction of an unmarked lysA deletion mutant of M. bovis BCG substrain Pasteur.

We used the suicide plasmid pYUB668 (see Materials and Methods) to construct an unmarked, in-frame deletion of lysA (ΔlysA5::res) in the genome of M. bovis BCG substrain Pasteur. After electroporation of BCG substrain Pasteur with the suicide plasmid, we obtained an average of five Hyg^r clones per transformation with a primary recombination efficiency of 10⁻⁴ (Table 3). We screened several Hyg^r Suc^s clones by PCR to determine which of the primary clones were homologous recombinants. The PCR screen used an oligonucleotide primer specific for the res site at the deletion site and primers specific for the chromosomal DNA sequences flanking the insert DNA cloned into the suicide plasmid. Three of four clones examined had incorporated the suicide plasmid pYUB668 at the lysA locus, while the fourth appeared to be the result of an illegitimate recombination event (data not shown). We chose two clones for further study, mc²1601(DUP3) and mc²1602(DUP4), both of which had integrated pYUB668 at lysA but had differed in the orientation of the duplication (Table 1). The two strains were grown to saturation in 7H9 medium supplemented with lysine, methionine, and threonine and then plated upon the same type of medium containing sucrose. We used this combination of amino acids to ensure that any unforeseen polar effect of the ΔlysA5::res allele on the downstream Met and Thr biosynthetic genes would not prevent the isolation of mutants. The results of the sucrose selection are shown in Table 4, experiments 3 and 4. We obtained Suc^r clones at a frequency of 10⁻⁴ and observed the same three classes of secondary recombinants that we saw in the M. smegmatis experiments. Allelic exchange was confirmed in strain mc²1604, a mutant derived from DUP3 strain mc²1601 (see Southern blot in Fig. 2B). The auxotroph mc²1604 does not revert, and no suppression was observed in two independent cultures of 5 × 10⁹ CFU each.

The kinetics of allelic exchange of lysA in M. bovis BCG substrain Pasteur was surprisingly similar to that of M. smegmatis, prompting us to examine the reproducibility of this system. We repeated the sucrose selection with M. bovis BCG substrain Pasteur DUP3 strain mc²1601 with cultures grown in basal medium or medium supplemented with Lys, Met-Thr-Lys, or Casamino Acids (acid-hydrolyzed casein). We obtained Suc^r clones from each of the respective cultures at a frequency similar to those seen in the previous experiment with mc²1601 (Table 4, experiments 5 through 8; compare to experiment 3). The distribution of the three phenotypic classes in the Suc^r population was also similar, except that we did not obtain any lysine auxotrophs from cultures grown in basal medium lacking lysine (as expected) or, surprisingly, Casamino Acids medium (Table 4, experiments 5 and 8).

Use of allelic exchange to distinguish homologous from illegitimate primary recombinants.

When using the two-step allelic-exchange methodology with the slow-growing mycobacteria, it is important to identify primary recombinants that resulted from illegitimate recombination and those which resulted from homologous recombination. This can be done by PCR screening (as we did for the above-described experiment) or Southern blotting, although these screening methods are difficult when using large recombination substrates. We reasoned that it should be possible to distinguish between the two types of recombinants by observing the phenotypic frequencies in the pool of Suc^r secondary clones. Presumably, any primary recombinant resulting from a homologous integration of the plasmid at lysA would be able to undergo a second recombination event and lose the plasmid, while a recombinant that had integrated the plasmid via illegitimate recombination would be unable to do the same. Any Suc^r clones arising from an illegitimate recombinant would result from inactivation of the sacB gene as seen above, and all these clones should also be Hyg^r.

We tested this idea in a series of lysA allelic-exchange experiments with M. bovis BCG substrain Connaught. Electroporation of BCG substrain Connaught with the suicide plasmid pYUB668 yielded an average of two Hyg^r clones per electroporation for a primary recombination efficiency of 10⁻³ (Table 4). We chose seven Hyg^r Suc^s BCG substrain Connaught::pYUB668 primary recombinants, grew them in medium supplemented with lysine, plated for sucrose-resistant clones, and then screened the Suc^r clones for hygromycin sensitivity and auxotrophy (Table 4, experiments 9 through 15). Three of the seven primary recombinants (clones 3, 9, and 10) gave rise to phenotypic populations similar to that seen for M. bovis BCG substrain Pasteur DUP strains mc²1601 and mc²1602 (compare results in Table 4). Therefore, these three primary clones (3, 9, and 10) were homologous primary recombinants. Two clones (2 and 11) yielded only Suc^r Hyg^r prototrophs, while the remaining clones (4 and 8) yielded a majority of Suc^r Hyg^r prototrophs and a small number of Suc^r Hyg^s prototrophs (Table 4). These four primary clones (2, 4, 8, and 11) were therefore classified as illegitimate recombinants. One BCG substrain Connaught lysine auxotroph, derived from clone 3, was designated strain mc²2519, and allelic exchange was confirmed by Southern blotting (Fig. 2B).

Construction of an unmarked, in-frame lysA deletion mutant of M. tuberculosis H37Rv.

The same methodology and suicide plasmid, pYUB668, described above were used to construct a lysA deletion mutant of M. tuberculosis H37Rv. We observed primary recombination efficiencies that were similar to those observed in experiments with BCG substrain Pasteur (Table 3). We chose six Hyg^r Suc^s primary recombinants, grew them in lysine-supplemented medium, and plated for sucrose-resistant recombinants. All six primary recombinants gave rise to phenotypic populations, similar to the results seen with the BCG substrain Pasteur mc²1601 DUP3 strain grown in basal medium and shown in Table 4, experiment 5 (data not shown). We concluded that these primary clones were all likely homologous recombinants but that something was wrong with the system, since we did not isolate any auxotrophs. The sucrose selection was repeated with two of these primary recombinant strains, mc²2998 and mc²2999, grown in several types of medium: basal, Lys, Met-Thr-Lys, and Casamino Acids (Table 4, experiments 16 through 23). In the subsequent experiments, the frequency of sucrose resistance was in the range of 10⁻⁵ to 10⁻⁴ (Table 4). Again, we failed to obtain auxotrophs and confirmed that the phenotypic frequencies within the Suc^r population were similar to those in the experiment that failed to yield Lys⁻ BCG mutants on basal medium (compare experiments 17 and 18 with experiment 5 in Table 4). Furthermore, the results from the M. tuberculosis primary recombinants were unlike the results obtained with the M. bovis BCG substrain Connaught illegitimate primary recombinants. These results suggested to us that all our primary recombinants were indeed homologous but that for some reason any auxotrophs resulting from a secondary recombination event were nonviable. Apparently, the medium could not support the growth of an M. tuberculosis lysine auxotroph. We decided to determine if our inability to isolate a lysine auxotroph of M. tuberculosis was due to the inability of the organism to transport lysine.

Transport of lysine in mycobacteria.

To investigate lysine transport in M. tuberculosis, we used the toxic lysine analog AEC. AEC is transported via lysine importers; the lysine permeases of E. coli (LysP) and C. glutamicum (LysI) were identified by using AEC-resistant mutants (45, 48). AEC inhibits aspartokinase, the enzyme catalyzing the first step of the aspartate amino acid family pathway responsible for the synthesis of Met, Thr, Ile, Lys, and meso-DAP, the last being a component of the cell envelope peptidoglycan and the precursor to lysine (23, 44). AEC alone is capable of inhibiting the growth of E. coli but requires the addition of threonine to inhibit the growth of corynebacteria (44). Presumably, full AEC sensitivity in corynebacteria requires repression of the threonine branch of the pathway by threonine.

The growth curves of M. tuberculosis H37Rv and M. bovis BCG substrain Pasteur in media with or without AEC and Thr are shown in Fig. 3. We used a molar concentration of 3 mM for AEC and Thr, a concentration that is close to the 40 μg/ml used for amino acid supplementation in our studies. As seen in Fig. 3, neither AEC nor Thr alone has an inhibitory effect upon the growth of the two species; however, the combination of the two does inhibit growth, with M. bovis BCG experiencing the greatest inhibition compared to that of M. tuberculosis. One interpretation of the results of this experiment is that lysine uptake is not as efficient in M. tuberculosis as in M. bovis BCG. The BCG lysine auxotrophic mutant mc²1604 does not grow well in medium supplemented with lysine at concentrations below our standard concentration of 40 μg/ml (data not shown). This suggests that a decrease in transport efficiency of M. tuberculosis compared to that of M. bovis BCG might preclude us from isolating an M. tuberculosis lysine auxotroph. Since our inability to isolate a lysine auxotroph of M. tuberculosis might be due to inefficient lysine transport by the organism, we made another attempt by using medium with increased amounts of lysine.

FIG. 3.

Effect of AEC on the growth of wild-type M. bovis BCG and M. tuberculosis H37Rv. Growth curve data were obtained as described in Materials and Methods. (A) M. bovis BCG substrain Pasteur; (B) M. tuberculosis H37Rv. Basal, 7H9 medium; AEC, basal medium with 3 mM AEC; Thr, basal medium with 3 mM threonine; AEC/Thr, basal medium with AEC and threonine at 3 mM each.

Identification of media that support the growth of an M. tuberculosis H37Rv lysine auxotroph.

Allelic exchange with the M. tuberculosis pYUB668-carrying homologous primary recombinant strain mc²2998 was repeated with modified media with increased amounts of lysine. Experiments utilizing media containing lysine at 200 μg/ml, at 200 μg/ml with 0.05% Tween 80, or at 1 mg/ml did not yield any auxotrophs (Table 4, experiments 24 to 26). However, auxotrophic mutants were isolated when medium containing lysine at 1 mg/ml with 0.05% Tween 80 was used (Table 4, experiment 27). The mutant colonies were much smaller than were the wild-type colonies and were easily identified on the sucrose selection plates (Table 4, experiment 27).

One mutant was designated mc²3026, and allelic exchange of lysA was confirmed by Southern blotting (Fig. 2B). No reversion or suppression was seen in 3 × 10⁹ CFU. The mutant grows slowly, requiring approximately 4 to 5 weeks to form a large colony on solid medium, and has an approximate doubling time of 48 h in liquid medium (data not shown). Surprisingly, the mutant can grow on 7H10 solid medium supplemented with Casamino Acids and also on 7H11 (supplemented with Casitone, a pancreatic digest of casein) but requires high concentrations of lysine if lysine is the sole supplement. It has an absolute dependency upon Tween 80 regardless of the type of solid medium.

DISCUSSION

Several groups have demonstrated the use of suicide plasmids for allelic exchange in fast- and slow-growing mycobacteria. The most efficient are those systems using a counterselectable marker; for mycobacteria, workers have successfully used rpsL (36, 42), pyrF (26), and sacB (40). The most promising counterselectable system for the slow-growing mycobacteria is sacB, which confers sensitivity to sucrose. Methods using sacB were used for the targeted disruptions of ureC in M. bovis BCG (40) and M. tuberculosis (37) and the erp gene of M. bovis BCG and M. tuberculosis (8).

We decided to construct a new sacB suicidal vector, pYUB657, and test it for the construction of unmarked, in-frame deletion mutants in the slow-growing mycobacteria. In these studies, we saw an opportunity to examine homologous recombination in the mycobacteria from a practical standpoint. The bane of allelic exchange in slow-growing mycobacteria has been the propensity with which these organisms incorporate exogenous DNA into their genome via illegitimate recombination (1, 24, 31). Allelic exchange in M. smegmatis is relatively easy, and this species does not appear to integrate DNA via illegitimate recombination. Several workers have suggested that the homologous recombination machinery is rather inefficient in the slow-growing mycobacteria. It is generally believed that illegitimate recombination occurs at a higher frequency than does homologous recombination in the slow-growing mycobacteria, but this does not necessarily mean that homologous recombination is defective in these organisms (31).

The results of this work suggest that homologous recombination in M. bovis BCG and M. tuberculosis is as efficient as that in M. smegmatis. First, the frequency of integration of suicidal plasmids into the chromosomes of the fast and slow growers is similar, within the range of 10⁻⁴ to 10⁻⁵ (except for BCG substrain Connaught, the frequency of which was 10⁻³; this might be an inflated value, however, due to an unusually low electroporation efficiency with the control vector pYUB412). While the number of primary recombinants obtained in M. bovis BCG and M. tuberculosis is lower than that obtained in M. smegmatis, the differences in the numbers of primary recombinants and recombination frequencies are small, and the electroporation frequencies are, at best, only an approximation. We suspect that any significant differences in primary recombination frequencies between slow-growing mycobacteria and M. smegmatis likely reflect a difference in DNA entry into the cells, since it is generally agreed that higher electroporation efficiencies are possible with M. smegmatis than with the slow growers.

The primary recombination frequencies for the slow-growing mycobacteria include both homologous and illegitimate recombinants; thus, a direct comparison between the frequencies of primary recombination in fast- and in slow-growing mycobacteria may not be valid. However, we think that more illegitimate recombination occurs with linear DNA than with plasmid DNA; thus, the contribution of illegitimate recombination to the primary recombination frequencies is likely to be small. The recombination experiments described in this work used covalently closed, supercoiled plasmid DNA. In preliminary work (data not shown), we found that electroporation of linear insert DNA from our recombination plasmids into BCG yielded 10-fold more clones than did electroporation with the covalently closed circular plasmids, but all the clones obtained with linear DNA were illegitimate recombinants. In addition, we rarely obtained hygromycin-resistant clones when we electroporated the sacB suicide vector pYUB657 lacking a DNA insert for recombination into M. bovis BCG or M. tuberculosis (data not shown). The difference in recombination results with linear substrates and covalently closed circular DNA substrates may be due to linear DNA being more recombinogenic than circular DNA, since it has free ends available for strand invasion (31). The illegitimate recombination mechanism in slow-growing mycobacteria is not characterized in any detail, but we hypothesize that the illegitimate recombination machinery may be relatively more sensitive to linear DNA than is the homologous recombination machinery. In this view, linear DNA might stimulate illegitimate recombination to a much higher degree than homologous recombination.

Comparing homologous recombination frequencies among these three species is more straightforward when one examines the frequencies of secondary recombination events. When we subjected cultures to sucrose selection, we obtained sucrose-resistant clones in the range of 10⁻⁴ to 10⁻⁵ for all three species, the same as the frequency seen for the primary recombination of the plasmid into the chromosome. In the sucrose-resistant population, we observed three phenotypic classes, two of which resulted from a recombination event and one that we believe did not. The latter class, the Suc^r Hyg^r prototrophs, was designated “sacB-inactivated” clones, since they were still hygromycin resistant. Inactivation of sacB in BCG, at a frequency similar to that observed in this study, has been noted previously (40). Counterscreenable markers can be inactivated at an approximate frequency of 10⁻⁵ in M. smegmatis by the action of mobile insertion elements (10). We have also seen a similar phenomenon, at a lower frequency, with use of the rpsL system for allelic exchange in M. smegmatis (36).

In this study, we sought to construct mutants with a deletion in lysA, conferring a lysine auxotrophic phenotype. Unexpectedly, the lysine auxotrophs that we obtained in this study have different lysine requirements. The M. smegmatis mutant is the most flexible in its requirements, growing on chemically defined medium supplemented with lysine as well as on medium supplemented with Casamino Acids. In contrast, we could not isolate auxotrophs of BCG substrain Pasteur by using Casamino Acids-containing medium. The compositional analysis of the Casamino Acids used in this study showed that our medium should have a lysine concentration that is threefold greater than the amount required for the BCG lysine auxotrophs (12). Neither the BCG substrain Pasteur nor the BCG substrain Connaught lysine auxotrophs are able to grow on solid medium if Casamino Acids or Casitone (a pancreatic digest of casein) is used as the source of lysine. Previously studied Met and Leu auxotrophic mutants of BCG can grow on casein medium, unlike the BCG lysine auxotrophs described in this study (24, 30). In more recent work with transposon mutagenesis of BCG, there were attempts to assay the efficiency of mutagenesis by screening for amino acid auxotrophy (6). The only mutants that were obtained were Leu auxotrophs, as isolated previously. This led to some concern that the transposition mechanism might not be random, which would be detrimental to a mutagenesis system (5). However, all of these attempts utilized medium containing casein preparations. Under such conditions, lysine auxotrophs would not be isolated. It is possible that the casein phenomenon described here is more widespread and could explain the dearth of auxotrophs in the above-described experiments. We are currently investigating why the BCG lysine auxotrophs fail to grow on medium containing casein.

We were unable to isolate lysine auxotrophs of M. tuberculosis H37Rv until we used medium with a high concentration of lysine and 0.05% Tween 80. As was the case for M. bovis BCG, we could not isolate M. tuberculosis mutants by using Casamino Acids; however, once we obtained a mutant, we found that it could grow on Casamino Acids medium or Casitone, as long as there was Tween 80 in the medium. Since the M. tuberculosis mutant is dependent upon the presence of Tween 80, we assume that our failure to obtain a mutant by using Casamino Acids medium was due to the absence of Tween in the selection medium. It is important to note that Tween 80 does not allow the BCG auxotrophs to form colonies on Casamino Acids medium. Based upon the AEC toxicity data, we can conclude that M. tuberculosis H37Rv does not transport lysine as effectively as does M. bovis BCG. Alternatively, since AEC toxicity requires transport of threonine as well, the AEC results could be explained by inefficient threonine transport. However, the high lysine requirement of the mutant and the dependency upon Tween 80 would support the former conclusion, since Tween 80 is believed to increase the permeability of the mycobacterial cell envelope (20). The primary phenotypic difference between the M. bovis BCG and M. tuberculosis mutants is that the M. bovis BCG mutants require lysine supplementation alone, while the M. tuberculosis mutant requires Tween 80 along with either lysine at a high concentration or Casamino Acids.

The auxotrophic mutants that we obtained in this study will be useful in a variety of applications. We hope to use the M. bovis BCG and M. tuberculosis lysine mutants for the construction of DAP auxotrophs (peptidoglycan mutants), as we have done for M. smegmatis (36). We are also developing a series of vectors bearing the lysA gene that could be used for the expression of foreign antigens in the BCG auxotrophs; the presence of the lysA gene would maintain the plasmids in vivo in the absence of antibiotic selection. We are also testing the behavior of the BCG mutants in animals in the hope that the mutants could be used in human immunodeficiency virus-infected populations as a safer alternative to live, wild-type BCG vaccine. One major goal of mycobacterial research is the development of attenuated strains of M. tuberculosis that could be used as potential vaccine strains. Such mutant strains would be unable to grow in a host, or would grow for only a short time, lasting long enough to prime the immune system. To this end, we are currently examining the growth kinetics of the M. tuberculosis auxotroph in animal models.