A Replication-Limited Recombinant Mycobacterium bovis BCG Vaccine against Tuberculosis Designed for Human Immunodeficiency Virus-Positive Persons Is Safer and More Efficacious than BCG

Abstract

Tuberculosis is the leading cause of death in AIDS patients, yet the current tuberculosis vaccine, Mycobacterium bovis bacillus Calmette-Guérin (BCG), is contraindicated for immunocompromised individuals, including human immunodeficiency virus-positive persons, because it can cause disseminated disease; moreover, its efficacy is suboptimal. To address these problems, we have engineered BCG mutants that grow normally in vitro in the presence of a supplement, are preloadable with supplement to allow limited growth in vivo, and express the highly immunoprotective Mycobacterium tuberculosis 30-kDa major secretory protein. The limited replication in vivo renders these vaccines safer than BCG in SCID mice yet is sufficient to induce potent cell-mediated and protective immunity in the outbred guinea pig model of pulmonary tuberculosis. In the case of one vaccine, rBCG(mbtB)30, protection was superior to that with BCG (0.3-log fewer CFU of M. tuberculosis in the lung [P < 0.04] and 0.6-log fewer CFU in the spleen [P = 0.001] in aerosol-challenged animals [means for three experiments]); hence, rBCG(mbtB)30 is the first live mycobacterial vaccine that is both more attenuated than BCG in the SCID mouse and more potent than BCG in the guinea pig. Our study demonstrates the feasibility of developing safer and more potent vaccines against tuberculosis. The novel approach of engineering a replication-limited vaccine expressing a recombinant immunoprotective antigen and preloading it with a required nutrient, such as iron, that is capable of being stored should be generally applicable to other live vaccine vectors targeting intracellular pathogens.

The tuberculosis vaccine, Mycobacterium bovis bacillus Calmette-Guérin (BCG), one of the most widely used human vaccines, is administered to approximately 100 million infants annually (49). Unfortunately, as evidenced by the estimated 9 million new cases and 1.6 million deaths attributed to tuberculosis each year (52), mostly in BCG-vaccinated people, BCG is not highly efficacious against adult, pulmonary disease. However, BCG provides significant protection against disease and death due to childhood and disseminated forms of tuberculosis, supporting the vaccine's continued extensive use (1, 3, 7, 11).

As a live attenuated vaccine, BCG multiplies to a limited extent in immunocompetent hosts before cell-mediated immunity arrests its growth (22). This limited replication is likely necessary for optimal protective efficacy (31). Although it has a very good safety record in immunocompetent vaccinees, BCG, like all live vaccines, is contraindicated for immunocompromised individuals, such as persons with AIDS, in whom it can cause serious disseminated disease and even death (7, 36). Tragically, human immunodeficiency virus (HIV) is often highly prevalent in regions where tuberculosis is endemic; hence, many children immunized with BCG are potentially at risk of disseminated BCG infection. For many years, the World Health Organization (WHO) recommended that all healthy infants in areas with a high tuberculosis incidence, including asymptomatic HIV-positive infants, receive BCG because the risk of tuberculosis is so great. Recently, the WHO issued revised recommendations in which BCG is no longer advised for asymptomatic HIV-positive infants, as these individuals have a greatly increased risk of disseminated BCG disease (19, 53).

Because of the benefits of BCG and its continued worldwide use, we have sought to develop recombinant BCG vaccines with improved efficacy and safety. Previously, we demonstrated the first vaccine that is more potent than BCG in the stringent outbred guinea pig model of pulmonary tuberculosis, a model noteworthy for its relevance to human tuberculosis (26, 35). However, like BCG, this vaccine, a recombinant BCG strain (rBCG30) that overexpresses the Mycobacterium tuberculosis 30-kDa major secretory protein (r30; also called antigen 85B), is unsuitable for immunocompromised persons.

To render rBCG30 safe for immunocompromised persons while maintaining its protective efficacy, we generated BCG mutants with a restricted replication capacity that overexpress r30. One of these mutants, rBCG(mbtB)30, was specifically engineered such that it can undergo only limited replication in vivo through deletion of mbtB, which disrupts the synthesis of the siderophore mycobactin, making the strain mycobactin dependent and preventing normal iron acquisition. Although it is mycobactin dependent, rBCG(mbtB)30 is capable of significant yet limited residual replication in the absence of mycobactin, provided that the vaccine is preloaded with sufficient ferric mycobactin. Our second vaccine, rBCG(panCD)30, a pantothenate auxotroph capable only of minimal replication in the absence of pantothenate, was originally designed with the aim of supplementing the host diet with pantothenate to allow extended growth, although the success of this approach proved inconsistent in practice.

In this study, we examine the in vitro and in vivo growth characteristics of the rBCG(mbtB)30 and rBCG(panCD)30 vaccines, their ability to induce cell-mediated and protective immunity in guinea pigs, and their safety in immunocompromised SCID mice. We demonstrate that (i) the more replication-proficient vaccine, rBCG(mbtB)30, has greater protective efficacy than BCG; (ii) rBCG(panCD)30 is comparable in efficacy to BCG; and (iii) both vaccines are more attenuated than BCG in SCID mice. Both vaccines potentially provide a safe alternative and, in the case of rBCG(mbtB)30, a more effective alternative to BCG for protecting against tuberculosis in HIV-positive infants and adults.

MATERIALS AND METHODS

Bacterial strains and plasmids.

The bacterial strains and plasmids used in this study are listed in Table 1. All BCG strains were derived from BCG Tice (Organon [now Schering-Plough], Kenilworth, NJ) and grown on Middlebrook 7H10 or 7H11 agar (BD, Sparks, MD) containing 10% (vol/vol) oleic acid-albumin-dextrose-catalase (BD) and 0.5% (vol/vol) glycerol or as unshaken cultures in Middlebrook 7H9 broth (BD) supplemented with 0.01% (wt/vol) tyloxapol and 0.2% (vol/vol) glycerol (7H9-TLX) at 37°C in an atmosphere of 5% CO₂-95% air, except where indicated. Hygromycin (50 μg/ml), kanamycin (10 or 50 μg/ml), apramycin (50 μg/ml), calcium d-pantothenate (50 μg/ml, except where indicated), and mycobactin J (0.1 μg/ml in plates and 0.01 μg/ml in broth, except where indicated) were included as appropriate. Ferric mycobactin J was obtained from Allied Monitor, Inc. (Fayette, MO), and is referred to simply as mycobactin J hereafter.

TABLE 1.

Bacterial strains and plasmids used in this study

Strain or plasmid	Description	Antibiotic resistanceb	Reference or source
Strains
E. coli strains
DH5α	General cloning strain	None	Gibco BRL
DY380	Recombineering strain	Tetr	33
M. bovis BCG strainsa
BCG	M. bovis BCG Tice, parental strain	None	Organon
rBCG30	rBCG30 Tice	Hygr	26
BCG mbtB	ΔmbtB::Kmr; mycobactin dependent	Kmr	This study
rBCG(mbtB)30	ΔmbtB::Kmr; mycobactin dependent, carries pNBV1-30 (Hygr)	Kmr Hygr	This study
BCG panCD	ΔpanCD::Aprr; pantothenate auxotroph	Aprr	This study
rBCG(panCD)30	ΔpanCD::Aprr; pantothenate auxotroph, carries pNBV1-30 (Hygr)	Aprr Hygr	This study
BCG trpD	ΔtrpD::Kmr; l-tryptophan auxotroph	Kmr	This study
rBCG(trpD)30	ΔtrpD::Kmr; l-tryptophan auxotroph, carries pMTB30 (Hygr)	Kmr Hygr	This study
Plasmids
pNBV1-30	Expression of 30-kDa antigen in mycobacteria, carries fbpB coding region and endogenous promoter	Hygr	17
pMTB30	Expression of 30-kDa antigen in mycobacteria, carries 4.5-kb genomic fragment	Hygr	15
pNBV1-mbtB	Expression of mbtB from endogenous mbtB promoter	Hygr	This study
pNBV1-panCD	Expression of panCD from hsp60 promoter	Hygr	This study
pNBV1-trpD	Expression of trpD from endogenous trpD promoter	Hygr	This study
pEX2	Mycobacterial allelic exchange vector derived from pPR27; ori ts sacB codBA	Hygr	18
phEX1	Derivative of phAE87 in which pZErO-2 replaces the pYUB328 cosmid	Kmr	This study
phAE87	Conditionally replicating mycobacteriophage	Ampr	2
pNBV1	E. coli-mycobacterial shuttle vector	Hygr	27
pZErO-2	Cloning vector	Kmr	Invitrogen
pUC19-Kmr	Carries kanamycin resistance cassette	Ampr Kmr	50
pPE207	Source of the apramycin resistance gene	Aprr	38

^aAll BCG strains were derived from M. bovis BCG Tice.

^bThe apramycin resistance gene [aacC4, also known as aac(3)IV] also confers resistance to kanamycin and gentamicin.

Escherichia coli strains DH5α (Gibco BRL [now Invitrogen], Carlsbad, CA) and DY380 (33) were used for cloning and were grown on Luria-Bertani agar, Luria-Bertani broth, or Terrific broth II (QBiogene, Carlsbad, CA) at 37°C (DH5α) or 32°C (DY380). Ampicillin (100 μg/ml), hygromycin (250 μg/ml), kanamycin (50 μg/ml), and apramycin (50 μg/ml) were included as appropriate.

Solubilization of mycobactin J.

Mycobactin J is highly soluble in ethanol but poorly soluble in aqueous solutions. Dilution of an ethanolic solution into medium results in immediate precipitation of mycobactin J, which is easily observed because ferric mycobactin is bright red. To dissolve mycobactin J at a high concentration in aqueous solution, 20 μl of 50-mg/ml mycobactin J in ethanol (1 mg) was mixed with 50 μl of 20% (wt/vol) tyloxapol in ethanol (10 mg), and the ethanol was evaporated under vacuum. The mycobactin J-tyloxapol solution was easily dissolved in 100 ml of 7H9 broth (or water), and the solution (10 μg/ml mycobactin J, 0.01% [wt/vol] tyloxapol) was sterilized by filtration (0.2 μm). Solubilization of 100 μg/ml mycobactin J required a corresponding increase in the tyloxapol concentration to 0.1% (wt/vol).

Southern hybridization.

Genomic DNA was isolated from BCG strains by phenol-chloroform extraction. Restriction fragments of genomic DNA were electrophoresed in agarose gels and transferred to positively charged nylon membranes (Hybond-XL; Amersham Biosciences, Piscataway, NJ) by capillary blotting in 0.5 M NaOH-1.5 M NaCl. Specific biotinylated oligonucleotide probes (for mbtB blot, 5′-biotin-CTGTCGGCTTAGCGCCGACGACGTCTATCTGG-3′; and for panCD blot, 5′-biotin-ACACAAAACGGCTCCCCGCCCATCTCAAGAAC-3′) were hybridized to the blots in modified Church and Gilbert hybridization buffer (0.5 M NaH₂PO₄-Na₂HPO₄, pH 7.8, 7% sodium dodecyl sulfate [SDS], 2 mM EDTA) (6), and detection was performed using a North2South chemiluminescence hybridization and detection kit (Pierce, Rockford, IL) according to the manufacturer's instructions.

Construction of BCG mbtB, BCG panCD, and BCG trpD mutants.

BCG mbtB and BCG trpD mutants were generated via an allelic exchange method that utilizes a temperature-sensitive, sacB-containing plasmid (40), while the BCG panCD mutant was generated via specialized transduction with a temperature-sensitive mycobacteriophage (2).

To generate the BCG mbtB mutant, we constructed an allelic exchange substrate by using a PCR strategy in which a BCG mbtB locus with a 3.9-kb deletion was created and a Km^r cassette from pUC19-Km^r (50) was inserted at the site of the deletion. The mutated allele was cloned into the allelic exchange vector pEX2 (18), and the plasmid was electroporated into BCG for allelic exchange as described previously (40). Previously, an M. tuberculosis H37Rv mbtB mutant was reported to grow in broth medium without mycobactin added to the medium (10), and initially, we were able to isolate a BCG mbtB mutant with the correct genotype, as determined by Southern blotting, in the absence of added mycobactin. However, after continued growth of the culture, a parental BCG genotype emerged, indicating that the original culture was likely a mixture of an mbtB mutant and the parental BCG strain, which outgrew the mutant. Therefore, we cultured the original strain in the presence of mycobactin and performed three rounds of colony purification, through which we were able to isolate a genetically pure and stable clone, which was verified by Southern blot analysis.

We constructed the BCG trpD mutant in the same manner as that for the BCG mbtB mutant. An allelic exchange substrate was constructed using a PCR strategy in which a BCG trpD locus with a 588-bp deletion was created and a Km^r cassette was inserted at the site of the deletion. This mutated allele was cloned into the allelic exchange vector pEX2, and the plasmid was electroporated into BCG for allelic exchange as described previously (40). A correctly constructed mutant was identified by its phenotype (Km^r Hyg^s l-Trp⁻), and its genotype was confirmed by Southern blot analysis.

To generate the BCG panCD mutant, we constructed an allelic exchange substrate by using a PCR strategy in which a BCG panCD locus with a 1,345-bp deletion was created and an apramycin resistance (Apr^r) gene from pPE207 (plasmid generously provided by Julian Davies [38]) was inserted at the site of the deletion {the apramycin resistance gene from pPE207 [aacC4, also known as aac(3)IV] also confers resistance to kanamycin and gentamicin (9)}. This mutated allele was cloned into the allelic exchange vector phEX1 (a derivative of the temperature-sensitive mycobacteriophage phAE87 [generously provided by William Jacobs {2}]), which was used to perform allelic exchange in BCG via specialized transduction (2). A correctly constructed mutant was identified by its phenotype (Apr^r Hyg^s Pan⁻), and its genotype was confirmed by Southern blot analysis.

Complementation of the mutants was achieved by electroporation of the strains with pNBV1 (27) plasmids carrying intact copies of the mutated gene or genes. Both mbtB and trpD were expressed from their endogenous promoters, while the panCD genes were expressed from the strong mycobacterial hsp60 promoter. Transformation with the complementing plasmids resulted in hygromycin-resistant clones that grew normally without an exogenous nutrient (mycobactin, pantothenate, or l-tryptophan).

Further details on the construction of the allelic exchange substrates and complementing plasmids are available in the supplemental material.

Exochelin analysis of BCG mbtB.

BCG and the BCG mbtB mutant were grown in 7H9 broth (without tyloxapol) containing 1 ng/ml mycobactin J for 8 weeks, from an initial calculated A₅₅₀ of 0.004 to a final A₅₅₀ of 0.53 (BCG) or 0.16 (BCG mbtB). In a second experiment in which 2 ng/ml mycobactin J was used, BCG mbtB reached the same final density as BCG (A₅₅₀ of 0.6). Exochelins were chloroform extracted from 270 ml of filtered culture medium, prepurified on a C₁₈ Sep-Pak cartridge (Waters, Milford, MA), and analyzed by high-performance liquid chromatography on a C₁₈ column (4.6 mm by 250 mm by 5.0 μm) (SunFire C₁₈; Waters, Milford, MA) as previously described (12).

Construction and analysis of BCG mutants overexpressing the M. tuberculosis 30-kDa major secretory protein (r30).

To generate BCG mutants overexpressing the 30-kDa major secretory protein, we electroporated the plasmid pNBV1-30 (BCG mbtB and BCG panCD) or pMTB30 (BCG trpD) into the BCG mutant bacteria and selected hygromycin-resistant clones. The plasmid pNBV1-30 contains the coding region and endogenous promoter (a fragment of ∼500 bp upstream of the coding region) of fbpB cloned into the multicloning site of pNBV1, while pMTB30 contains a 4.5-kb genomic fragment which includes fbpB as well as several other genes (15, 17). Transformants were screened for expression and export of r30 by SDS-polyacrylamide gel electrophoresis (SDS-PAGE), and one clone was saved for all further analyses. To quantitate the overexpression of r30, we cultured BCG, rBCG30, rBCG(mbtB)30, and rBCG(panCD)30 in 7H9-TLX broth for 15 days, from an initial A₅₅₀ of 0.01 to a final A₅₅₀ of 1.9 to 2.3. Two cultures for each strain, inoculated from independent cultures and grown in parallel, were used to prepare concentrated culture filtrates. Cultures were centrifuged, and the supernatants were filtered (Acrodisc PF 0.8/0.2-μm filter; Pall Corporation, East Hills, NY) and concentrated to 100 to 200 μl with a Centricon Plus-20 concentrator (10,000-molecular-weight cutoff; Millipore, Billerica, MA). Concentrated culture filtrates equivalent to 1 ml of original culture volume were analyzed by SDS-PAGE followed by staining with Coomassie blue or immunoblotting. For immunoblotting, the proteins were transferred to a nitrocellulose membrane and probed with rabbit polyclonal antibodies specific for the M. tuberculosis 30-kDa antigen (diluted 1:25,000). The membranes were subsequently incubated with a 1:50,000 dilution of horseradish peroxidase-conjugated goat anti-rabbit antibodies (Bio-Rad, Hercules, CA), SuperSignal West Pico chemiluminescent substrate (Pierce, Rockford, IL) was added, and the 30-kDa antigen was visualized by exposure to X-ray film. Gels and blots were scanned, and quantitation of r30 was performed with TotalLab TL100 software (Nonlinear Dynamics, Newcastle upon Tyne, United Kingdom).

Growth of rBCG(mbtB)30 and rBCG(panCD)30 in broth.

Bacteria cultured to log phase in mycobactin- or pantothenate-supplemented broth were washed with 7H9-TLX to remove culture medium and inoculated at an initial A₅₅₀ of 0.01 into 30 ml 7H9-TLX containing no supplement or various concentrations of mycobactin J or pantothenate. The calculated carryover of mycobactin J and pantothenate from the original culture media was negligible. Duplicate cultures were monitored for growth by measuring the absorbance. When cultures of rBCG(mbtB)30 in the absence of mycobactin reached an A₅₅₀ of 1 (near the level at which parental BCG ceases growing under normal conditions), the cultures were subcultured again in 7H9-TLX broth lacking mycobactin to assess the maximum number of doublings possible in the absence of mycobactin. The number of bacterial doublings (generations) was determined by calculating the log₂ of the maximal absorbance achieved during the growth period and subtracting the log₂ of the initial absorbance, as follows: number of generations = log₂(maximal A₅₅₀) − log₂(initial A₅₅₀).

Intracellular growth of rBCG(mbtB)30 and rBCG(panCD)30 in human THP-1 macrophage monolayers.

THP-1 cells, a human monocytic cell line, were maintained in RPMI 1640 medium containing 10% heat-inactivated fetal bovine serum, 20 mM HEPES, and 2 mM l-glutamine at 37°C in an atmosphere of 5% CO₂-95% air. Cells were seeded at 3 × 10⁵ cells per well in 2-cm² 24-well tissue culture plates and differentiated with 100 nM phorbol 12-myristate 13-acetate for 4 days. The bacterial inocula were prepared by dilution of log-phase cultures grown in 7H9-TLX (plus the appropriate supplements) into tissue culture medium that included 10% human AB serum (Mediatech, Manassas, VA). When the bacterial cultures contained high concentrations of mycobactin J, the cells were washed with 7H9-TLX prior to dilution into tissue culture medium to prevent carryover of mycobactin J from the culture medium. The monolayers were infected with BCG strains at a multiplicity of infection of ∼1 bacterium per THP-1 cell for 2 h at 37°C in duplicate wells, after which the medium was removed and the monolayers were washed twice with medium. One milliliter of medium was added to the monolayers, and the plates were incubated at 37°C for 0 to 8 days. The medium was replaced with fresh medium on day 3 or 4 in the case of wells to be harvested at later time points. The mycobactin dependence of rBCG(mbtB)30 was assessed by supplementing the tissue culture medium with 0, 0.0002, 0.001, or 0.01 μg/ml of mycobactin J. The pantothenate dependence of rBCG(panCD)30 was assessed by supplementing the tissue culture medium with 0, 5, 50, or 100 μg/ml of pantothenate. CFU were enumerated at 0, 3, and 7 or 0, 4, and 8 days as follows. The culture medium (1 ml) was removed and added to 8 ml of dilution medium (7H9-10% oleic acid-albumin-dextrose-catalase-0.05% Tween 80). The monolayer was then lysed with 1 ml of 0.1% SDS in sterile distilled water, and the lysate was immediately added to the dilution tube. CFU were determined by plating bacteria on 7H10 plates containing the appropriate supplements. Colonies were enumerated after 2 to 3 weeks of incubation at 37°C.

Clearance of rBCG(mbtB)30 and rBCG(panCD)30 in guinea pigs.

Specific-pathogen-free 250- to 300-g outbred male Hartley strain guinea pigs from Charles River Breeding Laboratories (Wilmington, MA) were immunized in groups of 24 by intradermal administration in the right hindquarters of 10⁶ CFU of BCG, rBCG(mbtB)30, or rBCG(panCD)30. Three animals from each group were euthanized 1, 2, 3, 4, 6, 8, 10, and 15 weeks after immunization, and the right lung, spleen, and inguinal lymph nodes of each animal were cultured for CFU of BCG, rBCG(mbtB)30, or rBCG(panCD)30 on 7H11 plates containing pantothenate (50 μg/ml), mycobactin J (0.1 μg/ml), ampicillin (2.5 μg/ml), and amphotericin B (2.5 μg/ml). Colonies were enumerated after 2 to 3 weeks of incubation at 37°C.

Immunogenicity and protective efficacy of BCG vaccines in guinea pigs.

Specific-pathogen-free 250- to 300-g outbred male Hartley strain guinea pigs from Charles River Breeding Laboratories were immunized in groups of 21 or 15 (sham group only) by intradermal administration in the right hindquarters of 10³, 10⁶, or 10⁸ CFU of BCG, rBCG30, BCG mbtB, rBCG(mbtB)30, rBCG(panCD)30, or rBCG(trpD)30 or were sham immunized with phosphate-buffered saline (PBS). In the first experiment, BCG mbtB and rBCG(mbtB)30 were cultured with either a low concentration of mycobactin J (0.01 μg/ml) or a high concentration of mycobactin J (100 μg/ml) prior to immunization. In later experiments, 10 μg/ml of mycobactin J was used. Five or 10 weeks after immunization, six animals from each group were shaved on the side over the ribs and administered 10 μg of highly purified r30 (16) intradermally, with the degree of induration assessed 24 h later. The remaining 15 or 9 (sham group only) guinea pigs in each group were challenged with an aerosol generated from a 7.5-ml single-cell suspension containing a total of 7.5 × 10⁴ CFU of M. tuberculosis Erdman, a dose that delivers ∼10 live bacteria to the lungs of each animal. Afterwards, guinea pigs were individually housed in stainless steel cages contained within a laminar-flow biohazard safety enclosure and allowed free access to standard laboratory food and water. Some groups of animals that were administered the rBCG(panCD)30 vaccine were fed a modified diet that incorporated a high dose of pantothenate in the food pellets (standard 5025 guinea pig diet supplemented with 9.5 g calcium d-pantothenate per kg of body weight, 500 times the usual amount in the 5025 guinea pig diet, prepared by TestDiet [Richmond, IN], a division of LabDiet and a Purina Mills, LLC/PMI Nutrition International company). Ten weeks after challenge, the animals were euthanized and CFU of M. tuberculosis in the lungs and spleen were assayed. CFU were determined by plating bacteria on 7H11 agar containing ampicillin (25 μg/ml) and amphotericin B (5 μg/ml). Colonies were enumerated after 2 to 3 weeks of incubation at 37°C.

Virulence of rBCG(mbtB)30 and rBCG(panCD)30 in SCID mice.

The virulence of rBCG(mbtB)30 and rBCG(panCD)30 was assessed in two parallel experiments, one examining survival of SCID mice and the other measuring bacterial burdens in SCID mouse organs. Nine to 10-week-old Fox Chase female SCID mice (CB-17/lcr-Prkdc^scid/Crl) from Charles River Breeding Laboratories were used for both experiments. In the experiment designed to assess survival, mice in groups of 20 were injected via the tail vein with 10⁶ or 10⁸ CFU of BCG, rBCG(mbtB)30, or rBCG(panCD)30 or were sham treated with PBS. Survival was assessed over a 40-week period. Animals were euthanized when they met preestablished humane end points. In the experiment designed to assess bacterial burdens in organs, mice in groups of 72 were injected with 10⁶ CFU of BCG, rBCG(mbtB)30, or rBCG(panCD)30 via the tail vein. At 0, 1, 2, 3, 4, 6, 8, 10, 16, 20, 30, and 40 weeks (except for BCG-infected animals, where no animals survived beyond 16 weeks), six mice per group were euthanized and CFU in their lungs, spleens, and livers were assayed. CFU were determined by plating bacteria on 7H11 containing pantothenate (50 μg/ml), mycobactin J (0.1 μg/ml), ampicillin (2.5 μg/ml), and amphotericin B (2.5 μg/ml). Colonies were enumerated after 2 to 3 weeks of incubation at 37°C. Several animals from the rBCG(mbtB)30 and rBCG(panCD)30 groups died over the course of the experiment, so for the 40-week final time point, animals from the survival experiment run in parallel were used.

Animal research was conducted in compliance with all relevant federal guidelines and UCLA policies.

Statistical analysis.

For guinea pig protective efficacy experiments, parametric analysis of variance (ANOVA) and nonparametric Kruskal-Wallis (K-W) methods were used to compare log CFU across immunization groups. For the mean comparisons by ANOVA, post hoc mean comparisons were judged to be statistically significant by use of the Fisher-Tukey least significant difference criterion. To combine CFU data across experiments, we first normalized individual animal log CFU values to the mean for the sham group in each experiment, thereby obtaining the proportional change from the sham mean, by using the following formula: standardized log CFU = (log CFU − mean log CFU for sham)/mean log CFU for sham. For the assessment of virulence in SCID mice, Kaplan-Meier survival analysis was performed and the log rank test was used to compare survival curves for different groups.

RESULTS

Construction of rBCG(mbtB)30 and rBCG(panCD)30.

BCG mutants were generated by allelic exchange, and their construction was confirmed by Southern blotting (Fig. 1A and B). The stability of the mutations was confirmed by repeated subculturing (>50 generations) in the absence of antibiotic selection (data not shown). To generate derivatives of these mutants similar to rBCG30 (26), in that M. tuberculosis r30 is overexpressed, we electroporated the strains with plasmid pNBV1-30, and clones that stably expressed r30 were selected. Even in the absence of selective pressure, we found that expression was stable for at least 40 generations of growth in broth (data not shown). rBCG(mbtB)30 and rBCG(panCD)30 expressed and secreted 8-fold and 12-fold more r30, respectively, than BCG did and 1.4-fold and 2-fold more r30, respectively, than rBCG30 did (Fig. 1C and D). Deletion of mbtB from M. tuberculosis results in a strain that cannot synthesize mycobactin or the structurally similar, but more hydrophilic, exochelin family of iron-binding siderophores (10, 12). Consistent with this, we detected no exochelins by high-performance liquid chromatography analysis of chloroform extracts of BCG mbtB culture filtrates under conditions in which the parental BCG strain produced abundant exochelins (Fig. 2).

FIG. 1.

Construction of rBCG(mbtB)30 and rBCG(panCD)30. (A) Maps of the parental BCG mbt locus and the disrupted allele of BCG mbtB, which contains a Km^r cassette (kan) inserted into a 3.9-kb deletion, removing nearly the entire mbtB coding region. Genomic DNA from the BCG parental strain and the BCG mbtB mutant was digested to completion with BamHI and probed with a biotinylated oligonucleotide that hybridizes to mbtA. (B) Maps of the parental BCG panCD locus and the disrupted allele of BCG panCD, which contains an Apr^r gene (apr) inserted into a 1.3-kb deletion, removing the entire panCD coding region. Genomic DNA from the BCG parental strain and the BCG panCD mutant was digested to completion with PshAI and probed with a biotinylated oligonucleotide that hybridizes to BCG3667c. The genomic organization and gene nomenclature are from the annotated BCG Pasteur genome (4; http://genolist.pasteur.fr/BCGList/). The sequence of the BCG genome is nearly identical to that of the M. tuberculosis H37Rv genome (8; http://genolist.pasteur.fr/TubercuList/) in the regions shown, with the only differences being two single nucleotide polymorphisms in the 11.5-kb region of the mbt locus and one single nucleotide polymorphism in the 5-kb panCD region. SDS-PAGE (C) and immunoblot analysis (D) of culture filtrates (equivalent to 1 ml of culture) of BCG, rBCG30, rBCG(mbtB)30, and rBCG(panCD)30 were performed. The arrows indicate the overexpression and secretion of the recombinant 30-kDa major secretory protein by rBCG30, rBCG(mbtB)30, and rBCG(panCD)30 compared with that by the BCG parental strain. Results for one representative experiment of two are shown. M, molecular mass markers in kb (A and B) or in kDa (C); B, BamHI; P, PshAI.

FIG. 2.

The BCG mbtB mutant is defective in exochelin synthesis. The chloroform extracts from 54-ml BCG and BCG mbtB culture filtrates were loaded onto a C₁₈ column (4.6 mm by 250 mm by 5.0 μm) (SunFire C₁₈; Waters) equilibrated with buffer A (0.1% trifluoroacetic acid) and eluted with the following gradient: 0 to 50% buffer B, 0 to 10 min; 50 to 100% buffer B, 10 to 60 min; and 100% buffer B, 60 to 70 min (buffer B is 0.1% trifluoroacetic acid and 50% acetonitrile). The eluate was monitored at both 220 nm and 450 nm, a wavelength at which iron-binding exochelins have a strong absorbance; the 450-nm absorbance values are shown. Results for one representative experiment of two are shown. Extracts from cultures grown in the presence of 2 ng/ml mycobactin J are indicated (+ mycobactin J). The low level of mycobactin J supplementation had no apparent effect on exochelin production by the parental BCG strain. Mycobactin J was strongly retained on the C₁₈ column, requiring 95% acetonitrile for elution, and so was not detected under the gradient conditions used.

Mycobactin dependence of rBCG(mbtB)30.

rBCG(mbtB)30 was dependent on exogenous mycobactin when it was grown in broth culture containing 40 μg/ml ferric ammonium citrate (∼130 μM Fe), and maximal growth was achieved at ≥10 ng/ml mycobactin (Fig. 3A). Subsequent to culture in the presence of mycobactin, rBCG(mbtB)30 was capable of limited growth in the absence of exogenous mycobactin, which we hypothesized was due to mycobactin incorporated into the bacterial cell wall, as carryover of culture medium was negligible. To test this hypothesis, we cultured rBCG(mbtB)30 with various concentrations of mycobactin, washed the cells, and measured growth in the absence of exogenous mycobactin (Fig. 3B). The number of generations of growth in the absence of exogenous mycobactin was dependent upon the concentration of mycobactin in the original inoculum culture, increasing linearly with the logarithm of the mycobactin concentration (Fig. 3C). For each doubling of the mycobactin concentration in the original culture, there was almost one doubling of growth in the absence of mycobactin, suggesting that the ferric mycobactin complex was efficiently stored and/or absorbed by the bacteria. In support of ferric mycobactin storage, bacterial cell pellets obtained after centrifugation of cultures grown in the presence of 10 μg/ml mycobactin were a reddish color, which was not removed by washing (data not shown).

FIG. 3.

Growth of rBCG(mbtB)30 in broth culture and in human THP-1 macrophages. (A) rBCG(mbtB)30 was cultured in medium containing 0.01 μg/ml mycobactin J and washed before inoculation into broth containing 0 to 100 ng/ml mycobactin J (as indicated to the right of the graph). Growth of the parental BCG strain (grown without mycobactin) is shown for comparison. (B) rBCG(mbtB)30 was cultured in medium containing 0.01, 0.1, 1, or 10 μg/ml mycobactin J (as indicated to the right of the graph) and washed before inoculation into broth lacking mycobactin J. (C) Plot of the number of generations of growth of rBCG(mbtB)30 in the absence of mycobactin versus the log₂ of the concentration of mycobactin J in the medium of the inoculating culture. (D) rBCG(mbtB)30 was cultured in medium containing 0.01 μg/ml mycobactin J and washed before addition to THP-1 monolayers containing 0 to 10 ng/ml mycobactin J in the tissue culture medium (as indicated to the right of the graph). Growth of the parental BCG strain (grown without mycobactin) is shown for comparison. (E) rBCG(mbtB)30 was cultured in medium containing 0.01, 1, or 10 μg/ml mycobactin J (as indicated to the right of the graph) and washed before addition to THP-1 monolayers. The number of bacterial generations over the 7-day course of infection is indicated to the far right of the growth curves (data are means for two independent experiments). For broth cultures (A and B), the cultures were inoculated to an initial calculated A₅₅₀ of 0.01, and growth was monitored by measuring absorbance. Data are the means ± SE for duplicate cultures. For infection of THP-1 macrophages (D and E), CFU were enumerated 0, 3, and 7 days after infection. Data are the mean log CFU ± SE for duplicate wells. In many instances, the error bars are smaller than the symbols. Each experiment was repeated at least once, with similar results.

rBCG(mbtB)30 also required exogenous mycobactin for sustained intracellular growth in human THP-1 macrophages, with maximal growth occurring with ≥1 ng/ml mycobactin (Fig. 3D). Growth of rBCG(mbtB)30 over the initial 3 days showed little dependence on exogenous mycobactin, but sustained growth over the entire 7-day infection required mycobactin in the medium. As in broth culture, after the strain was preloaded with a high concentration of mycobactin, rBCG(mbtB)30 exhibited limited intracellular growth (∼3 generations with ≥1 μg/ml mycobactin) in the absence of mycobactin (Fig. 3E).

Complementation analysis was performed by transforming the parental BCG mbtB mutant with plasmid pNBV1-mbtB. The transformed strain was fully complemented, growing normally in the absence of mycobactin (see Fig. S1A in the supplemental material).

Pantothenate dependence of rBCG(panCD)30.

Like a previously characterized M. tuberculosis panCD mutant (45), rBCG(panCD)30 is a pantothenate auxotroph, and growth in broth culture was dependent upon exogenous pantothenate (Fig. 4A). Maximal growth of rBCG(panCD)30 was achieved at ≥10 μg/ml pantothenate (Fig. 4A). rBCG(panCD)30 was capable of slight growth (∼2 doublings) in the absence of exogenous pantothenate. However, growth could not be increased by increasing the pantothenate concentration in the original inoculum culture from 50 to 500 μg/ml (Fig. 4B), unlike the situation with rBCG(mbtB)30 and mycobactin (Fig. 3B). Limited replication in the absence of pantothenate was also observed during intracellular growth in human macrophages (Fig. 4C). Growth of rBCG(panCD)30 over the initial 4 days was similar to that of BCG, but sustained growth over the 8-day infection required pantothenate supplementation of the tissue culture medium. Depleting rBCG(panCD)30 of its internal stores of pantothenate before infection of THP-1 macrophages by growth in medium lacking pantothenate reduced its ability to replicate subsequently without supplementation (0.9 versus 2.4 bacterial generations) (compare Fig. 4D with Fig. 4C).

FIG. 4.

Growth of rBCG(panCD)30 in broth culture and in human THP-1 macrophages. (A) rBCG(panCD)30 was cultured in medium containing 50 μg/ml pantothenate and washed before inoculation into broth containing 0 to 50 μg/ml pantothenate (as indicated to the right of the graph). Growth of the parental BCG strain (grown without pantothenate) is shown for comparison. (B) rBCG(panCD)30 was cultured in medium containing 50 or 500 μg/ml pantothenate (as indicated to the right of the graph) and washed before inoculation into broth lacking pantothenate. (C and D) rBCG(panCD)30 was cultured for 3 days in medium containing 50 μg/ml pantothenate (C) or in medium lacking pantothenate (to deplete the strain of its internal pantothenate supply) (D) and then added to THP-1 monolayers. The tissue culture medium was supplemented (filled symbols) or not supplemented (open symbols) with 100 μg/ml pantothenate, as indicated to the right of the graph (unsupplemented medium contains 0.25 μg/ml pantothenate). In a second experiment (not shown), 5 or 50 μg/ml of pantothenate was added to the medium, with similar results. The number of bacterial generations over the 8-day course of infection is indicated to the far right of the growth curves (data are means for two independent experiments). Growth of the parental BCG strain (grown without pantothenate) is shown for comparison. For broth cultures (A and B), the cultures were inoculated to an initial calculated A₅₅₀ of 0.01, and growth was monitored by measuring absorbance. Data are means ± SE for duplicate cultures. For infection of THP-1 macrophages (C and D), CFU were enumerated 0, 4, and 8 days after infection. Data are mean log CFU ± SE for duplicate wells. In many instances, the error bars are smaller than the symbols. Each experiment was repeated at least once, with similar results.

Complementation analysis was performed by transforming the parental BCG panCD mutant with plasmid pNBV1-panCD. The transformed strain was fully complemented, growing normally in the absence of pantothenate (see Fig. S1B in the supplemental material).

Attenuated growth in guinea pigs.

BCG is capable of limited replication in immunocompetent guinea pigs before cell- mediated immunity arrests its growth (22). Since this limited replication may be necessary for optimal protective efficacy, we assessed the ability of ferric mycobactin-loaded rBCG(mbtB)30 and of rBCG(panCD)30 to replicate in guinea pigs. Both mutants had a greatly reduced ability to replicate in immunocompetent guinea pigs compared with BCG (Fig. 5). While no organisms were detected for rBCG(panCD)30 at any time point, some growth of ferric mycobactin-loaded rBCG(mbtB)30 was detected in the spleen and lymph nodes 3 weeks after inoculation, consistent with the greater residual replication of ferric mycobactin-loaded rBCG(mbtB)30 in vitro.

FIG. 5.

Limited replication of rBCG(panCD)30 and rBCG(mbtB)30 in guinea pigs. Guinea pigs in groups of 24 were immunized by intradermal administration of 10⁶ CFU of BCG, rBCG(mbtB)30, or rBCG(panCD)30. rBCG(mbtB)30 was cultured with a high concentration of mycobactin J (10 μg/ml) prior to immunization. One, 2, 3, 4, 6, 8, 10, and 15 weeks after immunization, three animals per group were euthanized, and CFU of BCG, rBCG(mbtB)30, and rBCG(panCD)30 in the lungs, spleen, and inguinal lymph nodes were assayed. Data are mean log CFU ± SE. The limit of detection was 1 log CFU.

Immunogenicity and protective efficacy in guinea pigs.

We immunized guinea pigs with BCG, rBCG30, and attenuated BCG mutants in a series of four experiments. BCG and rBCG30 were administered at our standard dose of 10³ CFU, and the BCG mutants were administered at 10³, 10⁶, or 10⁸ CFU. Five or 10 weeks after immunization, we tested six animals in each group for cutaneous delayed-type hypersensitivity (DTH) to purified r30 (Fig. 6), and the remaining animals were subjected to a low-dose M. tuberculosis Erdman aerosol challenge (Fig. 7). The four strains that express high levels of r30—rBCG30, rBCG(mbtB)30, rBCG(panCD)30, and rBCG(trpD)30—elicited a strong cutaneous DTH response to r30, while BCG and BCG mbtB, which express low levels of the endogenous 30-kDa protein, did not (Fig. 6). Increasing the dose of BCG to 10⁶ CFU did not result in a cutaneous DTH response to r30 (Fig. 6B and D). These results for BCG are consistent with previous studies from this laboratory showing that immunization with BCG fails to elicit a strong DTH response to the 30-kDa protein (23, 25, 26), which all commonly used BCG strains produce in relatively small amounts (Fig. 1C and D) (22). Both rBCG(mbtB)30 and rBCG(panCD)30 elicited strong DTH responses only at doses of 10⁶ and 10⁸ CFU, not at a dose of 10³ CFU, whereas rBCG30 elicited a strong response at a dose of 10³ CFU.

FIG. 6.

Immunogenicity of rBCG(mbtB)30 and rBCG(panCD)30 in guinea pigs. In a series of four experiments (A to D), guinea pigs in groups of six were immunized by intradermal administration of 10³ CFU of BCG or rBCG30 or 10³, 10⁶, or 10⁸ CFU of BCG mutants or were sham immunized with PBS. The BCG mutants were administered at 10⁶ CFU except where indicated. In two experiments, BCG was administered at 10⁶ CFU in addition to 10³ CFU. BCG mbtB and rBCG(mbtB)30 were cultured with 10 μg/ml mycobactin J prior to immunization, except for panel A, where the strains were cultured with 0.01 or 100 μg/ml mycobactin J. Some of the groups that received the rBCG(panCD)30 vaccine were fed a diet supplemented with a high level of pantothenate (+pan). Five (C) or 10 (A, B, and D) weeks after immunization, the animals were skin tested by intradermal administration of highly purified M. tuberculosis 30-kDa major secretory protein, and the degree of induration was assessed 24 h later. Data are mean diameters of induration ± SE. †, P = 0.06; *, P = 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P = 1 × 10⁻⁸ (ANOVA; compared with BCG-immunized guinea pigs).

FIG. 7.

Protective efficacy of rBCG(mbtB)30 and rBCG(panCD)30 in guinea pigs. In a series of four experiments (A to D) paralleling those described in the legend to Fig. 6, guinea pigs in groups of 15 (except for the sham-immunized group of 9 animals) were immunized by intradermal administration of BCG, rBCG30, or BCG mutants as described in the legend to Fig. 6. Ten weeks after immunization, the animals were challenged with a low-dose aerosol of M. tuberculosis Erdman. Ten weeks after challenge, the animals were euthanized and CFU of M. tuberculosis in the lungs and spleen were assayed. Data are mean log CFU ± SE. Open symbols indicate groups immunized with BCG strains not overexpressing the 30-kDa antigen and sham-immunized animals. Closed symbols indicate groups immunized with BCG strains overexpressing the 30-kDa antigen.

When vaccinated guinea pigs were subjected to aerosol challenge with M. tuberculosis, rBCG(mbtB)30, administered at a dose of 10⁶ CFU, provided protection superior to that of BCG, as well as that of BCG mbtB, in two experiments (Fig. 7A and B; Table 2). In a third experiment, rBCG(mbtB)30 provided protection equivalent to that of BCG (Fig. 7D; Table 2). For the three experiments combined, guinea pigs immunized with rBCG(mbtB)30 showed a high level of protection compared with that of sham-immunized guinea pigs, with 2.1 ± 0.3 log (mean ± standard error [SE]) fewer CFU of M. tuberculosis in the lung and 2.6 ± 0.1 log fewer CFU of M. tuberculosis in the spleen (P < 0.0001 by both ANOVA and K-W analysis). More importantly, in the three experiments, guinea pigs immunized with rBCG(mbtB)30 had on average 0.3 log fewer CFU of M. tuberculosis in the lung (P < 0.04 by ANOVA and K-W analysis) and 0.6 log fewer CFU of M. tuberculosis in the spleen ( P ≤ 0.001 by ANOVA and K-W analysis) than did guinea pigs immunized with BCG (Table 2). Greater protective efficacy was achieved using 10⁶ CFU than 10³ CFU of rBCG(mbtB)30. This is in contrast to the case for BCG, where increasing the dose to 10⁶ CFU from 10³ CFU did not increase protective efficacy (Fig. 7B and D) (nonsignificant differences in the lung and spleen for both experiments), as previously reported (24).

TABLE 2.

Statistical analysis of the protective efficacy of rBCG(mbtB)30 versus BCG

Expt	Strain	Log protection vs sham infectiona		Log protection vs BCGb
		Lung	Spleen	Lung (± SE)	Spleen (± SE)
1	BCG	1.20	1.37
	rBCG(mbtB)30	1.61	2.64	0.41 ± 0.31	1.27 ± 0.39*
2	BCG	1.46	1.93
	rBCG(mbtB)30	2.07	2.47	0.61 ± 0.18**	0.54 ± 0.22***
3	BCG	2.51	2.79
	rBCG(mbtB)30	2.49	2.73	−0.02 ± 0.09	−0.06 ± 0.09
Mean ± SE	BCG	1.72 ± 0.40	2.03 ± 0.41
	rBCG(mbtB)30	2.06 ± 0.25	2.61 ± 0.08	0.33 ± 0.19****	0.58 ± 0.38*****

^aMean log CFU (sham) − mean log CFU [BCG or rBCG(mbtB)30].

^bMean log CFU (BCG) − mean log CFU [rBCG(mbtB)30]. *, P = 0.001 by ANOVA and P = 0.009 by K-W analysis; **, P = 0.001 by ANOVA and P = 0.008 by K-W analysis; ***, P = 0.02 by ANOVA and P = 0.01 by K-W analysis; ****, P = 0.03 by ANOVA and P < 0.04 by K-W analysis, calculated for the three combined experiments normalized to the sham-immunized group (see Materials and Methods); *****, P = 0.0005 by ANOVA and P = 0.001 by K-W analysis, calculated for the three combined experiments normalized to the sham-immunized group (see Materials and Methods).

rBCG(panCD)30, like rBCG(mbtB)30, was more efficacious when it was administered at a dose of 10⁶ CFU than 10³ CFU (Fig. 7C and D). rBCG(panCD)30 administered at a dose of 10⁶ CFU provided protection similar to that of BCG in one experiment and somewhat less than that of BCG in a second experiment; however, a dose of 10⁸ CFU provided protection similar to that of BCG in the second experiment. The increased protection observed in the lung and spleen with increasing doses of rBCG(panCD)30 (10⁸ > 10⁶ > 10³ CFU) was statistically significant (P < 0.0001); importantly, the 10⁸-CFU dose of rBCG(panCD)30 was well tolerated, comparable to 10³ CFU of BCG. In an attempt to increase the in vivo multiplication of rBCG(panCD)30 and, by extension, its protective efficacy, we fed some immunized animals a pantothenate-supplemented diet. The supplemented diet had no apparent effect on the protective efficacy of rBCG(panCD)30 in the first experiment but resulted in a modest improvement in protective efficacy that was statistically significant (P = 0.008) in the second experiment. A second auxotroph, rBCG(trpD)30, which is similar to rBCG(panCD)30 but requires l-tryptophan instead of pantothenate (see Fig. S1C in the supplemental material), was tested along with rBCG(mbtB)30 and rBCG(panCD)30 in one experiment (Fig. 7D). At a dose of 10⁶ CFU, rBCG(mbtB)30 was clearly superior to both of the auxotrophs [rBCG(panCD)30 and rBCG(trpD)30]; the latter mutants provided similar levels of protection.

Attenuation in immunocompromised SCID mice.

To determine the safety of rBCG(mbtB)30 and rBCG(panCD)30, we challenged SCID mice with 10⁶ or 10⁸ CFU of BCG, rBCG(mbtB)30, or rBCG(panCD)30 and monitored their survival (Fig. 8A and B). In a parallel experiment, we challenged the mice with 10⁶ CFU and monitored spleen weight and bacterial burdens in three organs, the lung, spleen, and liver (Fig. 8C, D, E, and F). Both rBCG(mbtB)30 and rBCG(panCD)30 were highly attenuated compared with BCG. rBCG(mbtB)30 was preloaded with a high concentration of mycobactin prior to challenge, and under these conditions, it was capable of substantial growth in SCID mouse organs (Fig. 8C, D, and E), as in broth culture (Fig. 3B and C) and human macrophages (Fig. 3E). However, there was a significant lag in the growth of rBCG(mbtB)30 compared with that of BCG, and growth was slower after the lag phase. Consequently, rBCG(mbtB)30-immunized animals survived much longer than BCG-immunized animals (mean survival time, 266 days versus 97 days at 10⁶ CFU and 223 days versus 32 days at 10⁸ CFU; P < 0.0001 for both doses), with 25 to 30% surviving the 280-day experiment, whereas all BCG-immunized animals died by 156 days.

FIG. 8.

Attenuation of rBCG(mbtB)30 and rBCG(panCD)30 in SCID mice. (A and B) SCID mice in groups of 20 were injected with 10⁶ (A) or 10⁸ (B) CFU of BCG, rBCG(mbtB)30, or rBCG(panCD)30 or sham treated with PBS via the tail vein, and survival was monitored over a 40-week period. (C to F) SCID mice in groups of 72 were injected with 10⁶ CFU of BCG, rBCG(mbtB)30, or rBCG(panCD)30 via the tail vein, and the bacterial burden was assessed at various time points over a 40-week period. At each time point, six mice per group were euthanized, their spleens were weighed (F), and CFU in their (C) lungs, (D) spleens, and (E) livers were assayed. Data are mean log CFU ± SE. The bacterial burdens of animals infected with BCG that succumbed to disease were very high in all three organs (in lungs, 8.38 ± 0.04 log CFU; in spleens, 8.51 ± 0.06 log CFU; and in livers, 8.18 ± 0.06 log CFU [mean ± SE]) (n = 37 [20 mice from the survival study and 17 early deaths from the bacterial burden study]). The data for rBCG(panCD)30 at the 112-day time point (undetectable CFU in the spleen and liver) are presumably flawed due to premature scoring of the plates for CFU. At later time points, a longer incubation period was used prior to scoring of the plates and small colonies of rBCG(panCD)30 were detected. The slow starvation of rBCG(panCD)30 in vivo may have resulted in less vigorous growth than that of BCG and rBCG(mbtB)30 when cells were plated.

As in broth culture, rBCG(panCD)30 was much more replication limited in SCID mice than rBCG(mbtB)30 was, and CFU slowly declined in mouse organs. Consistent with the low CFU levels, most of the mice administered rBCG(panCD)30 survived even the extremely high dose of 10⁸ CFU, similar to sham-infected animals (no statistically significant difference in survival), demonstrating that rBCG(panCD)30 was even more attenuated than rBCG(mbtB)30.

DISCUSSION

This study demonstrates the first recombinant BCG strain, rBCG(mbtB)30, that is both significantly more attenuated than BCG in immunocompromised SCID mice and more efficacious than BCG in the stringent outbred guinea pig model of pulmonary tuberculosis. rBCG(mbtB)30 expresses r30 at a similar level to that in rBCG30, the highly potent recombinant BCG vaccine we previously developed (22, 24-26), and was made safer than BCG for immunocompromised persons by specifically engineering the strain to undergo limited replication in vivo. This was achieved by deletion of mbtB, resulting in a strain that cannot synthesize the mycobactin and exochelin molecules that are needed by mycobacteria for the acquisition of iron under low-iron conditions, such as occurs at sites of mycobacterial replication in the host (10, 42). By preloading rBCG(mbtB)30 with its required nutrient, ferric mycobactin, significant yet strictly limited growth was obtained in the absence of mycobactin. This critical attribute of the vaccine allows it to induce an especially potent immune response while greatly increasing the safety of the vaccine for immunocompromised persons.

The fact that the efficacy of rBCG(mbtB)30 was demonstrated in the outbred guinea pig model is noteworthy because tuberculosis in the guinea pig so closely resembles human disease clinically, immunologically, and pathologically. In contrast to mice and rats, but like humans, guinea pigs are susceptible to low doses of M. tuberculosis, exhibit strong reactivity to tuberculin, develop a cutaneous DTH reaction characterized by a dense mononuclear cell infiltrate, and exhibit Langhans giant cells and caseous necrosis in tuberculous lesions (35). Moreover, in contrast to models utilizing an inbred mouse strain, the outbred guinea pig model utilizes animals with diverse major histocompatibility complex types and genetic backgrounds. Finally, the nonrodent guinea pig is evolutionarily closer to humans than are mice and rats. For these reasons, efficacy in the guinea pig model is now widely considered, including by a WHO consensus panel, to be a prerequisite for the testing of live mycobacterial vaccines against tuberculosis in humans, and toward this end, both the European Union and the United States have established guinea pig testing facilities (30, 51).

rBCG(mbtB)30 is dependent upon exogenous mycobactin for growth in broth and human macrophages. The extent of loading with ferric mycobactin determines the amount of growth achievable in the absence of mycobactin, which can be substantial. The mycobactin dependence of rBCG(mbtB)30 is very similar to that seen in Mycobacterium avium subsp. paratuberculosis, a naturally occurring mycobactin-dependent mycobacterium. M. avium subsp. paratuberculosis is capable of substantial growth in the absence of mycobactin, such that the strain no longer appears mycobactin dependent under some conditions, due to carryover of cell-associated mycobactin from the inoculating culture (32). Interestingly, the mycobactin dependence of rBCG(mbtB)30 differs from that of the previously described M. tuberculosis H37Rv mbtB mutant, which was reported to grow normally in iron-replete broth without added mycobactin (10). However, in unpublished studies, we have shown that an M. tuberculosis Erdman mbtB mutant is mycobactin dependent in iron-replete broth, just like rBCG(mbtB)30. The reason for this difference requires further study.

Our second vaccine strain, rBCG(panCD)30, is a pantothenate auxotroph due to deletion of panCD, similar to an M. tuberculosis panCD mutant (45). Like rBCG(mbtB)30, rBCG(panCD)30 expresses r30 at a level similar to that in rBCG30. Replication of rBCG(panCD)30 in the absence of pantothenate is severely restricted in both broth and human macrophages. Furthermore, growth in the absence of pantothenate could not be increased by preloading the strain with pantothenate, which evidently cannot be stored, in contrast to the case with rBCG(mbtB)30, where growth was enhanced by preloading cells with ferric mycobactin. This severely replication-limited phenotype of rBCG(panCD)30 was also observed in guinea pigs and SCID mice, demonstrating that rBCG(panCD)30 is highly attenuated. The restricted growth of rBCG(panCD)30 in SCID mice differed greatly from that of an M. tuberculosis panCD mutant, which grew slowly to high levels in the lung and killed the mice (45).

Both rBCG(mbtB)30 and rBCG(panCD)30, at doses of 10⁶ CFU or higher, induced a strong cutaneous DTH response to r30 in guinea pigs, similar to the response obtained with rBCG30. Guinea pigs vaccinated with 10³ CFU of rBCG(mbtB)30 and rBCG(panCD)30 had no DTH response to r30, and protective efficacy was compromised compared with that in guinea pigs vaccinated with 10⁶ CFU. Importantly, rBCG(mbtB)30 at a dose of 10⁶ CFU provided greater protective efficacy than BCG did. It remains to be determined if increasing the dosage of rBCG(mbtB)30 to 10⁷ or 10⁸ CFU would enhance its protective efficacy to the level seen with rBCG30. The more restricted replication of rBCG(panCD)30 resulted in a somewhat less efficacious vaccine, although it was comparable to BCG. We attempted to increase the replication of rBCG(panCD)30 in vivo by supplementing the diet of vaccinated guinea pigs with pantothenate. A modest improvement in protective efficacy was observed in one of two experiments. Despite the large excess of pantothenate provided in the diet, it may not have been accessible by rBCG(panCD)30 in sufficient quantities for robust growth. However, increasing the dose of rBCG(panCD)30 from 10⁶ to 10⁸ CFU increased protection. Thus, increasing the dose compensates at least in part for any deficiency of the vaccine in inducing cell-mediated and protective immunity as a result of its limited replication in vivo. Moreover, in contrast to BCG, which is poorly tolerated at intradermal doses exceeding 10⁶ CFU, the replication-limited rBCG(panCD)30 strain is as well tolerated at a dose of 10⁸ CFU as the standard 10³-CFU dose of BCG. The observation that high doses of rBCG(panCD)30 are well tolerated in the guinea pig, a species that is more susceptible than humans to M. bovis and M. tuberculosis infection, bodes well for the safety of high doses of this vaccine in humans. In their dose dependence, these vaccines differ from BCG and rBCG30, which induce cell-mediated and protective immunity independent of dose over 5 orders of magnitude (24). To summarize the impact of increasing the dose of the vaccine on tolerability and protective efficacy, (i) much higher doses of the replication-limited vaccine rBCG(panCD)30 than of the replication-proficient vaccine BCG are tolerated and (ii) increasing the dose of the replication-limited vaccines rBCG(mbtB)30 and rBCG(panCD)30 enhances their protective efficacy, compensating for their limited replication, whereas increasing the dose of the replication-proficient vaccines BCG and rBCG30 does not enhance their protective efficacy.

The rBCG(mbtB)30 vaccine used in this study was constructed on a Tice background and compared with its parental strain for efficacy against M. tuberculosis aerosol challenge. In recent unpublished studies, we determined that the protective efficacy of BCG Tice against M. tuberculosis aerosol challenge in the guinea pig model is equal to or greater than that of five major BCG strains in widespread use worldwide. Hence, the greater efficacy of rBCG(mbtB)30 than of BCG is not limited to its parental strain.

Both rBCG(mbtB)30 and rBCG(panCD)30 were attenuated in SCID mice compared with BCG. Although rBCG(mbtB)30 was able to multiply in SCID mouse organs, there was a considerable lag in growth and the strain grew more slowly than BCG, resulting in much longer survival times than those with BCG. The growth of rBCG(mbtB)30 in SCID mice may be due at least in part to preloading of the vaccine with ferric mycobactin; however, it is also possible that growth was promoted by an iron acquisition pathway that does not require mycobactin and that was not apparent under in vitro growth conditions. Hemin uptake is one possibility for an alternative iron acquisition pathway. Recently, we demonstrated that an M. tuberculosis mbtB mutant can grow efficiently in broth containing 1 μM hemin as a source of iron (N. Chim, L. McMath, M. Tullius, J. The, M. Sawaya, J. Whitelegge, D. Eisenberg, M. Horwitz, and C. Goulding, presented at the 7th International Conference on Pathogenesis of Mycobacterial Infections, Stockholm, Sweden, 26 to 28 June 2008). While BCG mbtB grows minimally with 1 μM hemin, greater growth is possible with 10 to 100 μM hemin, although it is still substantially less than growth obtained with mycobactin. Perhaps the level of hemin in the SCID mouse is sufficient for suboptimal growth of rBCG(mbtB)30. Because rBCG(panCD)30 was unable to multiply in SCID mice, this vaccine is even more attenuated than rBCG(mbtB)30. However, rBCG(mbtB)30 was preloaded with a high concentration of ferric mycobactin prior to the SCID mouse challenge, as it was for the guinea pig protection studies, to maximize its growth in vivo. Greater safety may be possible for rBCG(mbtB)30 by preloading the strain with a lower concentration of mycobactin, resulting in less growth in vivo in an immunocompromised host. While this theoretically could negatively impact protective efficacy, in the one experiment in which we preloaded rBCG(mbtB)30 with low and high concentrations of mycobactin, protection levels were comparable. Of course, if growth of rBCG(mbtB)30 in the SCID mouse was due to an alternative iron acquisition pathway, reducing mycobactin loading would not further enhance safety in this model.

A number of auxotrophic strains of BCG and M. tuberculosis have been generated and tested for the ability to induce protective immunity in small-animal models of tuberculosis, primarily the mouse model (5, 14, 21, 29, 34, 39, 43-48). Protection was less than or comparable to that provided by BCG, although in some of these instances BCG did not protect very well or multiple doses of the auxotroph were required. Of the few vaccines tested in guinea pigs, leucine and purine auxotrophs of BCG and a purine auxotroph of M. tuberculosis were less protective than BCG, while an M. tuberculosis leucine and pantothenate double auxotroph provided protection comparable to that of BCG (5, 29, 46). In general, most of these experimental vaccines are safer than BCG in SCID mice due to the severely limited replication of auxotrophic mycobacteria in vivo. Their severely limited replication likely accounts for their reduced protective efficacy compared with BCG. In contrast to BCG mutants, attenuated M. tuberculosis vaccines, including auxotrophs, pose substantial safety concerns because of the possibility of reversion to virulence, necessitating more than one major attenuating mutation in an M. tuberculosis-based vaccine prior to consideration for use in humans (30). While a few M. tuberculosis single-gene mutants (drrC, phoP, and secA2 mutants) have demonstrated protection better than that of BCG under some conditions and the phoP mutant was demonstrated to be safer than BCG in SCID mice (the drrC and secA2 mutants, while attenuated compared with M. tuberculosis, appear to be more virulent than BCG, as the drrC mutant multiplies more than BCG in immunocompetent mice and the secA2 mutant kills immunocompetent mice), these strains will all require at least one additional major attenuating mutation for testing in humans; the addition of a major attenuating mutation to these vaccines is likely to severely limit their growth in vivo and consequently may have a negative impact on their protective efficacy (20, 37, 41).

Because rBCG(mbtB)30 harbors antibiotic resistance genes, it is possible that regulatory authorities in some countries would require removal of these markers for clinical approval. Our unpublished studies indicate that removal of such markers can be accomplished without impacting either the expression of r30 or the limited replication phenotype of the vaccine and therefore should have no impact on the safety or efficacy of the vaccine.

As an alternative to auxotrophs, Kaufmann and coworkers constructed a ΔureC hly⁺ rBCG strain that, evidently due to improved cross-priming, provides superior protective efficacy to that of BCG in the mouse, although thus far not in the guinea pig (13, 51). Although this vaccine was designed for greater protective efficacy, it also provides improved safety over that of BCG in SCID mice. However, rBCG(mbtB)30 has an even greater margin of safety. At a dose of 10⁸ CFU, all of the mice infected with ΔureC hly⁺ rBCG died in less than 90 days (13), whereas rBCG(mbtB)30-infected animals had a mean survival time of 223 days and 25% of the animals survived the duration of the experiment.

Live attenuated vaccines often induce potent immune responses but carry the risk of excessive replication and dissemination in the host, resulting in illness and even death. Auxotrophs may avoid the consequences of excessive in vivo growth but restrict growth so severely that efficacy is impaired. The novel approach described in this paper, in vitro preloading of the vector with an essential nutrient that is otherwise unattainable from the host, allows a sufficient amount of replication to induce a potent immune response, but not so much as to compromise safety. This approach should be generally applicable to generating safe but effective live attenuated vaccines against other intracellular pathogens. As in the case of recombinant BCG vaccines, many intracellular pathogens depend on siderophores to acquire iron from the host, e.g., Salmonella species, Francisella tularensis, Brucella abortus, Legionella pneumophila, Yersinia pestis, and Histoplasma capsulatum, and therefore are directly amenable to the approach used in this paper, i.e., engineering attenuated siderophore-deficient vaccines and preloading them with a high concentration of iron-siderophore complex. Alternatively, enhanced iron storage might be achieved in siderophore-deficient vaccines by use of a low concentration of iron-siderophore complex combined with overexpression of iron storage proteins (28).

In summary, we have constructed two recombinant BCG vaccines, modeled after the rBCG30 vaccine, that are more attenuated than BCG in SCID mice and induce potent cell-mediated and protective immunity in guinea pigs. rBCG(mbtB)30 provides protection better than that of BCG, while the more replication-restricted rBCG(panCD)30 vaccine provides protection comparable to that of BCG. The increased protective efficacy of rBCG(mbtB)30 compared with that of BCG or typical auxotrophs, such as rBCG(panCD)30 and rBCG(trpD)30, was achieved by combining the novel approach of preloading the vaccine vector with a required nutrient to enhance replication in vivo with the approach of overexpressing a known immunoprotective antigen. These vaccines potentially offer HIV-positive individuals a safe alternative to BCG and, in the case of rBCG(mbtB)30, a more effective alternative. For maximal benefit, we envision the vaccines being administered to HIV-positive infants at birth and to others early in the course of HIV infection, before the immune system has deteriorated. Especially in concert with antiretroviral therapy to help maintain the integrity of the immune system, vaccines such as those described herein have the potential to provide substantial protection against the devastating consequences of tuberculosis in this highly susceptible population. In view of its enhanced safety and efficacy compared with those of BCG, rBCG(mbtB)30 may potentially also be appropriate in areas of high HIV prevalence where the HIV status of newborns is unknown.