Genotypic and Phenotypic Heterogeneity among Mycobacterium tuberculosis Isolates from Pulmonary Tuberculosis Patients

Abstract

Although the heterogeneity of Mycobacterium tuberculosis populations and the existence of mixed infections are now generally accepted, systematic studies on their relative importance are rare. In the present study, 10 individual colonies of each M. tuberculosis isolate (primary isolate) from 97 tuberculosis patients in a primarily human immunodeficiency virus-negative population were screened for heterogeneity and detectable mixed infections by spoligotyping, IS6110-based restriction fragment length polymorphism analysis, and mycobacterial interspersed repetitive unit-variable number of tandem repeat typing. The MICs of antituberculosis drugs for colonies with divergent fingerprints were determined. Infections with different bacterial subpopulations were detected in the samples from eight patients (8.2%), and the frequency of detectable mixed infections in the study population was estimated to be 2.1%. Genotypic variations were found to be independent of the drug susceptibilities, and the various molecular markers evolved independently in most cases. The predominant strains and the primary isolates always had concordant drug susceptibility and MIC testing results. These findings have implications on the interpretation of molecular epidemiology results for patient follow-up and in transmission studies.

Tuberculosis (TB) patients have often been assumed to be infected with a single Mycobacterium tuberculosis strain, and infection with one strain is thought to confer immunity to additional M. tuberculosis infections. Therefore, a recurrence of disease after treatment is often considered to be caused by the same strain that caused the original infection. However, cases of reinfection by a second M. tuberculosis strain and occasional infection with more than one strain have been documented (2, 5, 12, 22, 31), and the occurrence of mixed infections has now become generally accepted. Infection with multiple strains of M. tuberculosis can seriously confuse interpretation of drug susceptibility testing (DST) results and the detection of epidemiological links (2, 44). However, the frequency of mixed infections and the appreciation of their impact on the phenotypic and the genotypic properties of the primary isolates largely remain unknown, essentially because of the lack of systematic studies on the heterogeneity within M. tuberculosis populations from an individual patient.

Almost 30 years ago, Mankiewicz and Liivak (19) used phage typing to analyze the heterogeneity among individual colonies obtained from cultures of specimens isolated from 233 Eskimo patients, leading to the conclusion that 14.1% of the patients tested were simultaneously infected with more than one M. tuberculosis strain. The development of DNA fingerprinting tools has, meanwhile, considerably improved the capacity to distinguish M. tuberculosis strains. Among these tools, IS6110-based restriction fragment length polymorphism (IS6110-RFLP) analysis is considered the standard technique for comparison of M. tuberculosis isolates (32, 33). Other techniques, such as spoligotyping (17) and, more recently, typing based on variable numbers of tandem repeats (VNTRs) of mycobacterial interspersed repetitive units (MIRUs), have been shown to be useful alternatives for analysis of the genetic relatedness of M. tuberculosis isolates (11, 20, 30), especially in cases in which IS6110-RFLP analysis is not applicable or the results of IS6110-RFLP analysis are difficult to interpret.

Some of these tools have been applied to the study of mixed infections. However, most of these studies involved small numbers of isolates and/or used indirect approaches, such as analyses of serial isolates collected during one or more TB episodes (2, 5, 12, 22, 31). Mixed infections are defined as simultaneous infection by two or more M. tuberculosis strains, as evidenced by very distinct DNA fingerprints. In contrast, bacterial subpopulations are defined as strains with slight variations in fingerprints due to evolutionary events within an isolate from a patient originally infected by a single strain (2, 4, 22, 23).

Using these definitions, we have investigated mixed infections and the heterogeneity of bacterial subpopulations in 97 pulmonary TB (PTB) patients in a primarily human immunodeficiency virus-negative population from a high-incidence setting. This investigation was done by systematically analyzing 10 colonies isolated from each single M. tuberculosis isolate by using three different molecular biology-based tools combined with antibiotic resistance profiling.

MATERIALS AND METHODS

Study population.

The study was performed over a period of 18 months from September 2001 to February 2003. One hundred thirty-two consecutive smear-positive PTB patients enrolled from three hospitals in the greater Mymensingh District of Bangladesh were included in the study. All except one of the patients had previously received treatment with a combination of isoniazid, rifampin, pyrazinamide, ethambutol, or thiacetazone, according to the regimens of the International Union against Tuberculosis and Lung Disease, category 1 (2 months of treatment with ethambutol, isoniazid, rifampin, and pyrazinamide and 6 months of treatment with isoniazid and thiacetazone) (16), and the World Health Organization (WHO), category 2 (2 months of treatment with ethambutol, isoniazid, rifampin, pirazinamide, and streptomycin or 1 month of treatment with ethambutol, isoniazid, rifampin, and pyrazinamide; and 5 months of treatment with isoniazid, ethambutol, and rifampin) or category 3 (2 months of treatment with ethambutol, isoniazid, and pyrazinamide and 10 months of treatment with isoniazid and thiacetazone) (41).

The region has an estimated population of 14 million and a TB incidence rate of 106 per 100,000 population (26). Demographic data for each patient, including sex, age and previous history of TB, were collected upon hospitalization by using standard questionnaires for all patients aged 15 years and older. This study was approved by the Institutional Review Board of the Damien Foundation, Brussels, Belgium, and Dhaka, Bangladesh.

Samples and cultures.

One sputum sample was collected from each patient at the time of hospitalization and before the commencement of treatment. The samples were mixed with 1% cetylpyridinium chloride and shipped in batches to the Prince Leopold Institute of Tropical Medicine for analysis. Upon receipt at the Prince Leopold Institute of Tropical Medicine, each sample was cultured on Löwenstein-Jensen (L-J) medium and L-J medium without glycerol but supplemented with 0.4% sodium pyruvate to enhance the isolation of M. bovis after decontamination using the Petroff procedure. The cultures were incubated at 37°C and were read weekly for growth for a maximal duration of 16 weeks. Primary cultures were identified by classical methods.

In addition, the primary isolate of each patient was subcultured on Dubos agar plates to obtain isolated colonies. After incubation, 10 colonies per primary culture were picked and inoculated on L-J slants. The resulting subcultures, hereafter referred to as clones, were screened for genetic heterogeneity and drug susceptibility.

DST.

DST was performed with isoniazid (0.2 μg/ml), rifampin (40 μg/ml), streptomycin (4 μg/ml), and ethambutol (2 μg/ml) for all M. tuberculosis primary isolates by the proportion method on L-J slants, as described elsewhere (3). Briefly, a 1-mg/ml bacterial suspension was prepared in 5-ml screw-cap Bijoux bottles containing glass beads (diameter, 3.0 mm) in sterile distilled water. The suspension was homogenized on a vortex mixer for 1 min and left to stand for at least 10 min to reduce aerosol production in subsequent manipulations. Standard dilutions of 10⁻² and 10⁻⁴ mg/ml were prepared from the 1-mg/ml bacterial suspension. A 0.2-ml aliquot of the 10⁻²-mg/ml suspension was used to inoculate all L-J tubes containing the various drug concentrations, and two drug-free control L-J tubes were inoculated with the 10⁻²- and 10⁻⁴-mg/ml suspensions, respectively. The tubes were incubated overnight at 37°C in a slanted position with loosened caps to allow evaporation of the inoculum. After overnight incubation, the screw caps were tightened and the tubes were further incubated at the same temperature in an upright position. The initial reading of the tubes was performed on day 28 of incubation, while the final reading was done after 40 days of incubation. Resistance was defined as growth in the drug-containing tubes greater than 1% of the growth in drug-free control medium.

MICs.

The MICs of isoniazid (0.05, 0.2, 0.8, and 3.2 μg/ml), rifampin (10, 20, 40, and 80 μg/ml), and streptomycin (1, 2, 4, 8, and 16 μg/ml) for each clone were determined on L-J medium basically by previously described methods (42). Bacterial suspensions, inoculation, incubation conditions and lengths, and reading of the L-J tubes were prepared or performed as described above (see “DST” above), with the following exceptions: three dilutions (10⁻², 10⁻³, and 10⁻⁴) of the 1-mg/ml suspension were prepared for each clone, and the 1- and 10⁻³-mg/ml suspensions were inoculated into L-J tubes without drug and were used as extra controls, in addition to the L-J tubes inoculated with the 10⁻²- and 10⁻⁴-mg/ml suspensions. A strain was considered resistant when the MICs of isoniazid, rifampin, and streptomycin on L-J slants were >0.2, 40, and 4 μg/ml, respectively, which were the cutoff values used for standard DST.

DNA extraction.

DNA was extracted either by boiling of the mycobacterial cells for 5 min (only for spoligotyping or MIRU-VNTR typing) or by the internationally standardized method (34). Briefly, mycobacterial cells were heat inactivated for 30 min at 80°C, followed by lysis with lysozyme and sodium dodecyl sulfate-proteinase K, extraction with cetyltrimethyl ammonium bromide-sodium chloride, and chloroform-isoamyl alcohol purification. After the DNA was air dried, it was reconstituted in TE buffer (10 mM Tris HCl, 1 mM EDTA [pH 8.0]) and stored at −20°C until further use.

Spoligotyping.

Spoligotyping was performed with a commercial kit (Isogen Bioscience BV, Maarssen, The Netherlands) by the method described previously (17). Either 4 μl of the denatured bacterial suspension or 2 μl of purified genomic DNA from each sample was used for amplification of the direct-repeat (DR) region with oligonucleotides DRa (5′ biotinylated) and DRb. The labeled amplicons were used as probes for hybridization with a set of 43 known oligonucleotide spacer sequences of M. tuberculosis H37Rv and M. bovis BCG P3 covalently bound to a nylon membrane (Amersham Biosciences Little Chalfont, United Kingdom). Bound PCR fragments were detected with a streptavidin-horseradish peroxidase-enhanced conjugate and an enhanced chemiluminescence (ECL) system, followed by exposure to ECL hyperfilms (Amersham Pharmacia-Biotech, Roosendael, The Netherlands). The spoligotyping patterns were compared by computer-assisted analysis (BioNumerics, version 3.0; Applied Maths, Sint-Martens-Latem, Belgium).

IS6110-RFLP typing.

IS6110-RFLP fingerprinting was performed by the international standard typing method for M. tuberculosis (32). Briefly, genomic DNA was digested with restriction endonuclease PvuII (MBI Fermentas, St. Leon-Rot, Germany), subjected to electrophoresis on a 0.8% agarose gel, and vacuum blotted onto a nylon membrane (Amersham Pharmacia-Biotech). Hybridization was performed overnight at 42°C with a 245-bp 3′ IS6110 PCR-generated fragment with primer pair INS1 and INS2. The probe was labeled with the ECL direct nucleic acid labeling kit according to the instructions of the manufacturer (Amersham Pharmacia-Biotech). Hybridization patterns were visualized with an ECL kit and hyperfilms (Amersham Pharmacia-Biotech). The IS6110-RFLP fingerprint patterns were compared by using BioNumerics (version 3.0) computer software (Applied Maths).

MIRU-VNTR typing.

VNTR typing relies on PCR amplification of different VNTR regions with primers specific for the flanking regions of the VNTRs and on the determination of the sizes of the amplicons, which reflect the numbers of amplified VNTR copies. The 13 loci used in this study contained VNTRs of MIRUs (11, 28, 29) or VNTRs of other interspersed genetic elements (11, 18, 25). This set included the following MIRU-VNTR loci, with the alternative designations given in parentheses: 4 (ETR-D), 10, 16, 26, 31 (ETR-E), and 40. In addition, seven VNTR loci with designations according to previous reports (11, 18, 25, 40) were included: 424, 577 (ETR-C), 1895 (QUB-1895), 1982 (QUB-18), 2401, 3690, and 4156 (QUB-4156). These loci are collectively designated the MIRU-VNTR loci in this report. The 13 loci were selected from a wider set of MIRU-VNTR loci on the basis of their distinctive variabilities in epidemiologically unrelated isolates from different M. tuberculosis genotype families and from various geographic origins and on the basis of their stabilities in epidemiologically linked isolates from different sources (27, 29; P. Supply et al., unpublished data). Four multiplex PCRs and one simplex PCR amplification of the 13 loci (see Fig. 2) were used, as described by Supply et al. (29). Briefly, the MIRU-VNTRs were amplified by 40 PCR cycles by use of the HotStartTaq DNA polymerase kit (Qiagen, Hilden, Germany). The fluorescent primers for MIRU-VNTR loci 4, 10, 16, 26, 31, and 40 were the same as those described previously (29), except that labeling with HEX (5′-hexachloro-fluorescein phosphoramidite) was replaced by labeling with VIC, as the latter displays less spectral overlap with the other dyes used. The fluorescent primers for the seven other loci are described in Table 1. The amplicons were subjected to electrophoresis on a 96-well ABI 377 sequencer. Sizing of the PCR fragments and assignment of the various VNTR alleles were done with the GeneScan and customized Genotyper software packages (PE Applied Biosystems), as described elsewhere (29). The genotypes are expressed as a numerical code representing the numbers of tandem repeats in each of the 13 genomic loci.

FIG. 2.

DNA fingerprinting results for representative M. tuberculosis clones and primary isolates with variant fingerprint types. The patterns presented were generated by computer analysis with Bionumerics software (version 3.0). The position and proportion of each spoligotype and IS6110-RFLP type are shown. Missing spacers and extra, deleted, or transposed bands are indicated by arrows. Only one MIRU-VNTR profile is shown for patients infected with heterogeneous populations, for whom the results for the primary isolate are identical to those for the isolate with either the predominant and/or the variant fingerprint type. Polymorphic MIRU-VNTR loci are shown in boxes (the individual MIRU-VNTRs are described elsewhere [11, 18, 25, 29, 30]). ^a, MIRU-VNTR loci are listed according to their position (in kilobase pairs) on the H37Rv genome; ^b, also called ETR D (11).

TABLE 1.

Conditions for multiplex PCRs of seven VNTR loci

Multiplex PCR	Locusa	VNTR length (bp)	MgCl2 concn (mM)	PCR primer pair sequences (5′ to 3′)b	Reference or source
Mix F	VNTR 0424	51	1.5	CTTGGCCGGCATCAAGCGCATTATT	40
				GGCAGCAGAGAGCCCGGGATTCTTC (FAM)
	VNTR 0577	58		CGAGAGTGGCAGTGGCGGTTATCT (VIC)	40
				AATGACTTGAACGCGCAAATTGTGA
	VNTR 1895	57		GTGAGCAGGCCCAGCAGACT (NED)	40
				CCACGAAATGTTCAAACACCTCAAT
Mix G	VNTR 2401	58	3.0	CTTGAAGCCCCGGTCTCATCTGT (FAM)	40
				ACTTGAACCCCCACGCCCATTAGTA
	VNTR 3690	58		CGGTGGAGGCGATGAACGTCTTC (VIC)	40
				TAGAGCGGCACGGGGGAAAGCTTAG
	VNTR 4156	59		TGACCACGGATTGCTCTAGT	40
				GCCGGCGTCCATGTT (NED)
Individual	VNTR 1982	78	1.5	CCGGAATCTGCAATGGCGGCAAATTAAAAG	This study
				TGATCTGACTCTGCCGCCGCTGCAAATA (FAM)

^aLocus designations are according to the position (in kilobase pairs) on the M. tuberculosis H37Rv chromosome. VNTRs 0577, 2401, 4156, and 1895 correspond to the designations ETRC, ETRB, QUB 4156, and 1895, respectively (11, 25).

^bLabeling is indicated in parentheses: FAM, 6-carboxy fluorescein.

Exclusion of laboratory cross-contamination.

In order to exclude the possibility of cross-contamination during sample processing as a potential cause for the heterogeneity observed, primary isolates with which sputum processing, DST, or plating was concurrently performed with one of the samples with mixed M. tuberculosis isolates were typed by MIRU-VNTR. None of these data showed evidence of cross-contamination.

RESULTS

Mycobacterial cultures and patients characteristics.

Of the 132 specimens from patients tested, 102 were positive for M. tuberculosis by culture, no isolates grew from 20 samples, 6 specimens were contaminated, and nontuberculous mycobacteria were isolated from 4 other specimens. Individual colonies obtained from 5 of the 102 patients positive for M. tuberculosis by culture did not yield sufficient material to obtain DNA fingerprinting results. Thus, the M. tuberculosis isolates from 97 patients were analyzed in this study. The characteristics of the patients included in this study are presented in Table 2 and are arranged according to the DNA fingerprinting results. Sixty-six patients were males and 31 were females; and they included 1 new patient, 41 patients who were defaulters, 44 patients with relapses, and 11 patients with treatment failures. In this study, a new patient was defined as a TB patient who either had no prior anti-TB treatment or was treated with any anti-TB drug for less than 1 month, whereas defaulters were defined as patients who interrupted their treatment for more than 2 months after having received anti-TB treatment for at least 1 month and who then returned with bacteriologically confirmed TB. Patients with relapses were considered those individuals who became smear positive again after having been treated for TB and declared cured after the completion of treatment, while patients with treatment failures were considered patients who began treatment for smear-positive PTB but who remained smear positive or became smear positive again 5 months or later during the course of treatment (43).

TABLE 2.

Patient characteristics

Patienta	Age (yr)	Sexb	Treatment historyc	Time lapse (mo) between onset of previous and present treatments
1	61	M	F1	8
2	47	M	D1	6
3	30	F	F1	6
4	47	M	R1	12
5	60	M	R1	72
6	60	M	D1	9
7	46	M	R1	12
8	43	M	R1	72
9	41	M	R2	30
10	22	F	R1	36
11-97	16-80	M (58), F (29)	D0 (13), D1 (22), D2 (4), F1 (7), F2 (2), N (1), R0 (2), R1 (32), R2 (3), R3 (1)

^aIsolates from patients 1 to 10 had variant fingerprints, while those from patients 11 to 97 had invariant fingerprints.

^bM; male, F; female. Values in parentheses are number of patients.

^cF1, failure after WHO category 1 treatment; F2, failure after WHO category 2 treatment; D0, default after non-WHO TB treatment; D1, default after WHO category 1 treatment; D2, default after WHO category 2 treatment; R0, relapse after non-WHO TB treatment; R1, relapse after WHO category 1 treatment; R2, relapse after WHO category 2 treatment; R3, relapse after WHO category 3 treatment; N, new case. Values in parentheses are number of patients.

Spoligotyping.

Initially, each primary isolate from a single sputum specimen was analyzed by spoligotyping. As depicted in Fig. 1, the majority of the isolates tested could be grouped into one of the major known genotype families (10). The East African Indian type comprised nearly half (49 of 97) of the samples tested, the Central Asian genotype was found in 10 (10.3%) of the 97 patients, the Beijing genotype was observed in 7 (7.2%) of the 97 patients, and the remaining 31 isolates did not belong to any of the families mentioned above.

FIG. 1.

Spoligotype dendrogram generated for representative clones of 97 clinical M. tuberculosis isolates from the Mymmensingh District of Bangladesh. The patterns obtained after computer analysis with Bionumerics software (version 3.0) are presented. The scale at the upper left indicates the band-based Dice similarity index, which was used with the unweighted pair group method with arithmetic averages algorithm. Spoligotype families, defined as described by Filliol et al. (10), are indicated on the right. Three known major spoligotype clusters were detected. EAI, East African Indian type; CAS, Central Asian type.

To screen for heterogeneity within the M. tuberculosis population from a single sputum specimen, 10 clones were obtained from each primary isolate after colony isolation on Dubos agar plates. All 10 clones from each of the 97 patients were tested by spoligotyping. All 10 clones of the strains from 89 (91.8%) of the 97 patients tested, including the 7 Beijing genotype strains, showed identical spoligotypes. Variations among the 10 clones were observed for the strains from the remaining eight patients (Fig. 2). A deletion of a single spacer was observed in some clones from five (5.2%) patients (patients 1 to 5), while a difference of more than five spacers was observed in clones from the other three (3.1%) patients. In the clones from the latter three patients, a deletion of spacers 8 to 20 in 2 of 10 clones from patient 6 and of spacers 1 to 21 in 1 clone from patient 7 were observed. Remarkably, four different spoligotypes were observed in clones from patient 8. With the exception of patients 1 and 8, the spoligotypes of the primary isolate were always identical to the predominant spoligotype among the respective clones. The primary isolate of patient 8 appeared to be a combination of the two most prevalent spoligotypes, resulting in weak reactions for two spacers (spacers 37 and 38), while the spoligotype of the primary isolate from patient 1 was identical to the variant spoligotype.

IS6110-RFLP typing.

The clones showing variations in their spoligotypes were further subjected to IS6110-RFLP analysis. Either no difference or minor differences in IS6110-RFLP types were observed among the clones from six of the eight patients whose isolates showed variant spoligotypes (Fig. 2). No differences were observed in the clones from patients 4 and 5. A deletion of two IS6110 bands was associated with the absence of a single spacer in the spoligotypes of the clones from patient 3, while a shift of one IS6110 band was associated with the absence of a single spacer in the spoligotypes of the clones from patient 2. A deletion of one IS6110 band was associated with a deletion of spacers 1 to 21 in the clones from patient 7. The clones from patient 6 revealed three slightly different IS6110-RFLP patterns, in contrast to the two spoligotypes observed: some of the variant clones (clones 5 and 7) showed a band shift, while others (clones 1, 3, and 6) showed an additional band compared to the bands for the predominant RFLP type.

For patients 1 and 8, however, two and four completely different IS6110-RFLP patterns, respectively, were observed among the clones, in agreement with the spoligotype diversities. Unfortunately, only five clones representing all four spoligotypes could be tested by IS6110-RFLP analysis for patient 8.

Five of the seven sets of clones belonging to the Beijing genotype family (35) observed in this study gave identical patterns, while a gain of one hybridization band was observed in one clone from patients 9 and 10 (Fig. 2). Finally, all 15 sets of control clones with identical spoligotypes yielded identical IS6110-RFLP profiles as well (data not shown). These controls were representative of the various spoligotypes observed in the study population (Fig. 1).

MIRU-VNTR typing.

In order to further investigate the heterogeneity within the M. tuberculosis population, the clones showing variations in their spoligotypes and/or IS6110-RFLP types were also subjected to MIRU-VNTR typing. Identical MIRU-VNTR types were observed among all clones from 8 of the 10 patients with minor variations in spoligotypes and/or IS6110-RFLP types (patients 2 to 7, 9, and 10) (Fig. 2). However, the clones from patients 1 and 8 that showed very distinct RFLP profiles also showed two and four different MIRU-VNTR types, respectively. All the clones from five patients with identical spoligotypes and IS6110-RFLP types used as controls were also identical by MIRU-VNTR typing (data not shown).

Combined DNA fingerprinting results.

Mixed infections were previously considered simultaneous infections with two or more isolates showing very distinct DNA fingerprints, i.e., fingerprints that differed by more than three IS6110-RFLP bands. In contrast, slight variations in IS6110-RFLP patterns were considered to represent subpopulations with the same clonal origin (2, 4, 7, 22, 23). In accordance with these definitions, we considered the eight patients (patients 2 to 7, 9, and 10) whose clones had limited variations in IS6110-RFLP and/or spoligotyping patterns but identical MIRU-VNTR profiles to harbor bacterial subpopulations that resulted from a single original clone. The remaining two patients (patients 1 and 8), whose clones showed more than three IS6110 bands differences and differences at two or more MIRU-VNTR loci, were considered to be infected with different M. tuberculosis strains.

Drug susceptibility.

Seventy-three (75.3%) of the 97 M. tuberculosis primary isolates tested were pansusceptible to all drugs tested (Table 3). The remaining 23 (23.7%) isolates were resistant to at least one drug. Drug resistance was observed more frequently among Beijing genotype isolates (5 of 7) than non-Beijing genotype isolates (19 of 89) (P < 0.01; Fisher exact test). Multidrug resistance was found in 5 of 97 (5.2%) isolates but was not present significantly more often among the Beijing genotype isolates (2 of 7) than the other isolates (P = 0.07; Fisher exact test).

TABLE 3.

MICs for M. tuberculosis isolates from the 10 patients with variant fingerprint types among their clones

Patientb	Strain identity	DST result or MIC (μg/ml)a
		INH	RIF	STR
1	Primary isolate	Sensitive	Sensitive	Sensitive
	Strain 1	<0.05	<10	<1
	Strain 2	<0.2	<10	<1
2	Primary isolate	Sensitive	Sensitive	Sensitive
	Predominant spoligotype	<0.2	<10	<1
	Variant spoligotype	<0.05	<10	<1
3	Primary isolate	Sensitive	Sensitive	Sensitive
	Predominant spoligotype	<0.2	<10	<2
	Variant spoligotype	<0.05	<10	<4
4	Primary isolate	Sensitive	Sensitive	Sensitive
	Predominant spoligotype	<0.05	<10	<1
	Variant spoligotype	<3.2	<10	<1
5	Primary isolate	Sensitive	Sensitive	Sensitive
	Predominant spoligotype	<0.2	<20	<1
	Variant spoligotype	<0.05	<10	<1
6	Primary isolate	Sensitive	Sensitive	Sensitive
	Predominant spoligotype	<0.05	<10	<1
	Variant spoligotype	<0.05	<10	<1
7	Primary isolate	Sensitive	Sensitive	Sensitive
	Predominant spoligotype	<0.05	<10	<1
	Variant spoligotype	<0.2	<10	<1
8	Primary isolate	Resistant	Sensitive	Resistant
	Strain 1	<0.2	<10	>16
	Strain 2	<0.05	<10	<1
	Strain 3	<0.8	<20	>16
	Strain 4	<0.2	<10	<1
9	Primary isolate	Resistant	Resistant	Resistant
	Predominant RFLP type	>3.2	>80	>16
	Variant RFLP type	>3.2	>80	>16
10	Primary isolate	Sensitive	Sensitive	Sensitive
	Predominant RFLP type	<0.2	<10	<1
	Variant RFLP type	<0.2	<10	<1

^aOnly normal DST results (on L-J slants) obtained with critical concentrations, 0.2 μg/ml (and 1.0 μg/ml) for isoniazid, 40 μg/ml for rifampin, and 4 μg/ml for streptomycin, are reported for the primary isolates. MICs were determined on L-J slants. INH, isoniazid, RIF, rifampin, STR, streptomycin.

^bPatient designations are the same as those in Table 2.

MICs.

The MICs of isoniazid, rifampin, and streptomycin for all clones from all sets of primary isolates with divergent fingerprint types were measured. As shown in Table 3, for all patients infected with heterogeneous bacterial populations, the MIC results for the predominant clones were consistent with the DST results for the parent strain. Furthermore, with the exception of the clones from patient 4, no variation in MICs for the clones was observed. For patient 4, the isoniazid MIC for one clone with a variant spoligotype also varied (3.2 μg/ml), whereas the isoniazid MICs for the clones with the predominant spoligotype (the remaining nine clones) as well as for the primary isolate were <0.2 μg/ml. The IS6110-RFLP and MIRU-VNTR profiles of this variant clone were identical to those of the predominant clones.

The MICs for the clones from the two patients with mixed infections showed no variation in one patient but wide variations in the second patient. Both the primary isolate and the 10 clones from patient 1 (who was infected with two different strains) were susceptible. In contrast, the MICs for the different clones from patient 8 (who was infected with four different strains) varied from susceptible to streptomycin resistant and isoniazid and streptomycin resistant. Isoniazid and streptomycin resistance was also detected in the primary isolate.

DISCUSSION

Although the heterogeneity of M. tuberculosis isolates and the existence of mixed infections are now generally accepted, systematic studies of their relative importance are rare. Of the few studies reported in the literature, most were based exclusively on IS6110-RFLP analysis. In the present study, we analyzed 10 individual colonies from each of the primary M. tuberculosis isolates from 97 hospitalized PTB patients for heterogeneity and detectable mixed infections using a stepwise combination of different standard techniques to distinguish clinical isolates, including spoligotyping, IS6110-RFLP analysis, and MIRU-VNTR typing (11, 29, 32).

In the greater Mymensingh District of Bangladesh, a low proportion of strains have identical spoligotypes but variable IS6110-RFLP patterns (e.g., the Beijing genotype) (unpublished observations), as was confirmed with the specific population evaluated in the present study. Therefore, spoligotyping was used as the initial screening method (14, 15) to determine the minimal proportion of single isolates with heterogeneous populations or mixed infections. IS6110-RFLP analysis proved to be valuable for further analysis of the genotypes of the sets of clones selected after the initial screening, as all but one of the isolates typed had high IS6110 copy numbers. This finding is in contrast to most reports on Asian isolates harboring single or no IS6110 elements (9, 21). MIRU-VNTR typing was used together with IS6110-RFLP analysis, as it has been shown to be highly discriminatory for M. tuberculosis isolates with both high and low (less than five) IS6110 copy numbers (20, 29). On the other hand, spoligotyping is less discriminative than IS6110-RFLP analysis, and therefore, the frequencies of heterogeneous populations and mixed infections might have been underestimated by this strategy. However, since all 15 sets of control clones with identical spoligotypes drawn from the different genogroups showed identical IS6110-RFLP profiles, the potential underestimation due to our typing strategy was probably negligible.

In previous studies, mixed infections were considered simultaneous infections with two or more isolates showing very distinct DNA fingerprints, i.e., fingerprints that differ by more than three IS6110-RFLP bands. In contrast, slight variations in IS6110-RFLP patterns were considered to represent subpopulations with the same clonal origin (2, 4, 7, 22, 23). On the basis of these definitions, we detected infections with bacterial subpopulations in samples from eight patients (8.2%) and estimated the frequency of detectable mixed infections in the study population to be 2.1% (2 of 97 patients). The possibility that laboratory cross-contamination caused the presence of mixed strains among the isolates from these two patients was ruled out, as no other isolate processed on the same day as these study isolates shared MIRU-VNTR genotypes with the different strains detected among these isolates. One of the two patients was found to be infected with four strains simultaneously. Such a multiclonal infection has not been described before and is especially remarkable given the low seroprevalence of human immunodeficiency virus in the study population. Unfortunately, complete clinical information about this patient was unavailable.

In accordance with the definitions presented above, the clones with variant spoligotypes and differences of more than three IS6110-RFLP bands were also found to be different at at least two MIRU-VNTR loci. Consistent with previous findings based on IS6110-RFLP analysis (28, 38) and MIRU-VNTR typing (27), these clones were therefore considered different strains, and the corresponding infections were designated true cases of mixed infections. In contrast, we considered that infection with bacterial subpopulations was indicated when two or more clones showed a change in one spoligotype spacer or a block of spoligotype spacers or an addition, deletion, or shift of one or two IS6110-RFLP bands but were identical by MIRU-VNTR analysis. However, in these situations, some cases of infection as a result of the independent acquisition of very closely related strains cannot be totally ruled out. A comprehensive analysis of the strain distribution over time in the region studied would be required to further address this question.

All specimens with minor variations in IS6110-RFLP profiles among their clones showed identical MIRU-VNTR patterns for all 10 clones. The spoligotypes of the isolates from patients 6 and 7, for which a single-band difference in the IS6110-RFLP profiles was observed, differed by the deletion of a block of spacers (deletion of spacers 1 to 21 and spacers 8 to 20, respectively, in some clones). These deletions of contiguous spacers in association with a single IS6110 band deletion could be due to homologous recombination between IS6110 repeats, mediated by the insertion of IS6110 into the DR region (8, 36). Such simultaneous changes in IS6110-RFLP types and spoligotypes can thus reflect a single mutational event that generates a clonal variant and, hence, do not necessarily indicate the presence of a mixture of unrelated strains. On the other hand, the example of patient 1 shows that clones that differ by a single spoligotype spacer can represent two different M. tuberculosis strains, as defined by different IS6110-RFLP and MIRU-VNTR types.

The presence of bacterial subpopulations within some patients is highly unlikely to have arisen as a result of infection with many clonally varied bacteria from a single source, given the usually low infectious doses of tubercle bacilli (1). Instead, the presence of such bacterial subpopulations might be due to gradual evolutionary changes in M. tuberculosis isolates that occur postinfection, following adaptation in the new host environment, as suggested by De Boer and coworkers (6). Gradual changes could have been favored in our case, as nearly all the patients were being retreated and had harbored the bacteria for at least 6 to 72 months. However, in contrast to the study of De Boer et al. (6), we did not find any association between the age of the patient and the occurrence of heterogeneous bacterial populations from the same ancestral clone, as there was no significant difference in age between patients infected with homogeneous bacterial populations and patients with heterogeneous bacterial populations (P = 0.909; Fisher exact test). Nevertheless, the presence of subpopulations of single isolates may explain the previously observed sporadic appearance of clonal variants among serial isolates from given patients or the emergence of clonal variants between serial isolates within a few days (6, 37). Because of this sporadic appearance of clonal variants, it is possible that consecutive sampling of isolates might have revealed an even higher frequency of heterogeneous subpopulations in our study patients.

Among the patients that contained heterogeneous M. tuberculosis populations in their sputum, variant MICs were observed for isolates from patient 4. Both the primary isolate and the clones with the predominant genotype were susceptible to all drugs tested, whereas the MIC for the single clone with a divergent genotype was above the cutoff for resistance. This indicates that a proportion of 10% resistant bacteria (1 of 10 clones) in some patients may actually constitute a proportion of the entire bacterial population too small to be detected by routine DST. For the remaining isolates with heterogeneous bacterial populations, the genotypic variations observed by spoligotyping and/or IS6110-RFLP typing did not reflect variant MICs, consistent with previous findings that DST profiles do not change with changes in fingerprints (13, 24).

For the two patients with mixed infections, identical MICs were observed for the clones from patient 1. The MICs for the clones from patient 8, whose primary isolate was found to be isoniazid and streptomycin resistant, varied among the clones typed. Patient 8 had a category 1 treatment relapse and had thus been treated with isoniazid and rifampin but not streptomycin. Although we did not test the strain from the initial episode of this patient, the DST results reported here suggest that patient 8 was initially infected with strain 3 with a low level of isoniazid and streptomycin resistance. Infection with the other strains probably occurred after treatment, as otherwise, the susceptible strains would have been expected to be eliminated by the category 1 treatment. The DST profile of the primary isolate represents a sum of all possible resistances of the individual strains, including isoniazid resistance, which was found in only a single clone (strain 3 [isoniazid MIC < 0.8 μg/ml]). Therefore, in this particular patient, the presence of a mixed infection with drug-resistant and -sensitive strains may not have been problematic and therefore did not require a change in the anti-TB treatment.

Since a primary culture step was needed to obtain maximal growth after sample transport and before plating of the primary culture for single colonies, some selection may have occurred and may have reduced the initial heterogeneity in the samples. The PCR strategy used to detect the presence of Beijing genotype and non-Beijing genotype strains simultaneously directly from sputum specimens from South African patients (39) was not directly applicable in the present study because of the low prevalence of Beijing genotype strains in our setting (see above). In the South African study (39), the proportion of mixed infections reported was 10-fold higher than that in this study. This difference may be explained in part by the severalfold higher incidence of TB in South Africa, thus increasing the probability of multiple infections, although the increased sensitivity of sputum analysis may have played a role.

Variations in fingerprint patterns between paired isolates have been used to calculate the rate of change of the IS6110-RFLP patterns in order to assess the adequacy of IS6110 fingerprinting to track the recent transmission of TB (6, 23, 37). These calculations have been made by implicitly assuming genotypic homogeneity within each isolate within the pair. In this study we have detected genetic differences in individual colonies from single isolates, which indicates that the assumed homogeneity of single isolates is an oversimplification. Therefore, we recommend screening of isolates used in studies for heterogeneity to calculate the half-lives of molecular markers.

All isolates harboring bacterial subpopulations defined by slightly different IS6110-RFLP patterns and/or spoligotypes had identical MIRU-VNTR patterns, whereas isolates from mixed infections defined by very distinct IS6110-RFLP patterns and/or spoligotypes had different MIRU-VNTR patterns. This observation suggests that MIRU-VNTR typing may thus represent a helpful method that can be used to assess the occurrence of infections with multiple strains in epidemiological studies.

In summary, the findings presented here have implications for the interpretation of molecular epidemiology results for patient follow-up and in transmission studies, although we are aware that the selection criteria of the population sample used here (retreatment cases) may have introduced a bias. Therefore, it would be interesting to extend this study to analysis of the general population prior to any treatment both in high-incidence areas and in low-incidence areas.