Molecular Detection of Mixed Infections of Mycobacterium tuberculosis Strains in Sputum Samples from Patients in Karonga District, Malawi

Abstract

The occurrence of mixed infections of Mycobacterium tuberculosis is no longer disputed. However, their frequency, and the impact they may have on our understanding of tuberculosis (TB) pathogenesis and epidemiology, remains undetermined. Most previous studies of frequency applied genotyping techniques to cultured M. tuberculosis isolates and found mixed infections to be rare. PCR-based techniques may be more sensitive for detecting multiple M. tuberculosis strains and can be applied to sputum. To date, one study in South Africa has used a PCR approach and suggested that mixed infection could be common. We investigated mixed infections in northern Malawi using two lineage-specific PCR assays targeting the Latin American-Mediterranean (LAM) and Beijing lineages. Compared with spoligotyping, the specificity and sensitivity of both assays was 100%. From 160 culture-positive sputa, mixed LAM and non-LAM strains were detected in 4 sputa belonging to 2 (2.8%) patients. Both patients were HIV positive, with no history of TB. Cultured isolates from both patients showed only LAM by PCR and spoligotyping. In a set of 377 cultured isolates, 4 were mixed LAM and non-LAM. Only one showed evidence of more than one M. tuberculosis strain using IS6110-based restriction fragment length polymorphism (IS6110-RFLP) and spoligotyping analyses. Corresponding sputa for the 4 isolates were unavailable. Mixed Beijing and non-Beijing strains were not detected in this study. Mixed infections appear to be rare in our setting and are unlikely to affect findings based on DNA fingerprinting data. Molecular methods, which avoid the selective nature of culture and target distinct strains, are well suited to detection of mixed infections.

It used to be thought that tuberculosis in the human host was caused by infection with a single Mycobacterium tuberculosis strain (25). This belief was first challenged when phage typing identified the presence of more than one M. tuberculosis strain in isolates obtained from a single patient (2, 17). With the development of DNA fingerprinting (28) and other PCR-based genotyping methods (15, 27) to differentiate M. tuberculosis strains, the presence of multiple strains within a single patient has been demonstrated during routine genotyping studies (5, 19, 20, 24, 33). The results from these and further studies designed to specifically examine mixed infections (1, 10, 20, 22, 23, 29, 32) have shown that they occur in different geographical settings among both HIV-negative and HIV-positive tuberculosis (TB) patients.

While the occurrence of mixed infections of M. tuberculosis is no longer questioned, the rate at which they occur is unknown. Yet this is of fundamental importance, as much of our understanding of tuberculosis pathogenesis and epidemiology has assumed that mixed infection is rare. The frequency is likely to differ according to the level of M. tuberculosis transmission in different settings. If mixed infections are common, it would necessitate reinterpretation of studies on recent infection, the importance of reinfection as a mechanism of disease, the origin of recurrent episodes, and therefore, evaluations of the effectiveness of treatment. It also raises questions about strain-specific immunity.

A recent study in South Africa using M. tuberculosis lineage-specific PCR primers in sputum specimens suggests that mixed infections may be much more common than previously supposed (32). In Cape Town, 19% of individuals were infected with both non-Beijing and Beijing strains. More than half of the patients with Beijing strains were apparently coinfected with other (non-Beijing) strains, suggesting that the total proportion of mixed infections could be very high if other circulating strains were also identified. This has yet to be repeated elsewhere.

Spoligotype families from all major M. tuberculosis lineages have been found in northern Malawi, although the Latin American-Mediterranean (LAM) lineage has been the most common local genotype for the last 20 years and accounts for around 50% of M. tuberculosis strains in this population (12). Strains belonging to the Beijing lineage appear to have stabilized at 4% over the last 10 years (12). Against this background, we present data obtained from Malawi using 2 PCR assays to detect mixed infections involving strains from either the LAM or Beijing genotype.

MATERIALS AND METHODS

Setting.

The Karonga Prevention Study (KPS) in northern Malawi is a population-based epidemiological study of mycobacterial and other diseases. Karonga District is a rural area with a population of about 250,000. The annual risk of M. tuberculosis infection in this district is around 1% (6). Project staff based at peripheral clinics and at the local hospital screen individuals with chronic cough and other symptoms suggestive of TB. TB patients are diagnosed by following standard diagnostic algorithms for identifying pulmonary TB using smear microscopy and culture of biological samples. Treatment of TB follows Malawi government guidelines. Patient details, including age, sex, and history of TB, are recorded at diagnosis. Since 1988, diagnosed patients have been HIV tested, after counseling and if consent is given. Approximately, 60% of culture-confirmed TB patients are HIV positive, and currently, around half are already on antiretroviral therapy at diagnosis.

Since 1986, initial cultures of suspected Mycobacterium spp. have been sent to the Mycobacterium Reference Laboratory, United Kingdom, for full identification and drug susceptibility testing. At least one culture is sent at the time of diagnosis and any positive cultures obtained after 5 months of treatment (14). Confirmed M. tuberculosis isolates are then collected and processed for DNA fingerprinting (using IS6110-based restriction fragment length polymorphism [IS6110-RFLP] analysis) (28) at the London School of Hygiene & Tropical Medicine (LSHTM), United Kingdom, where they are also stored.

Specimens.

Three sets of specimens were used. First, strains from a variety of sources were used to assess the specificity of the primers. This first set consisted of DNA from the following: 48 M. tuberculosis isolates from Malawi with a known spoligotype collected from 2002 to 2003 by the KPS (13); 1 isolate each of M. bovis BCG Pasteur, M. tuberculosis H37Rv, and M. tuberculosis MT14323 from laboratory stocks; 7 M. tuberculosis Beijing family isolates (National Institute for Public Health and the Environment, Netherlands); and 17 nontuberculous mycobacterial (NTM) isolates (M. avium, M. chelonae, M. gordonae, M. heidelbergense, M. immunogen, M. intracellulare, M. liflandi, M. mageritense, M. marinum, M. neoaurum, M. nonchromogenicum, M. paraffinicum, M. peregrinum, M. smegmatis, M. szulgai, M. ulcerans, M. xenopi) supplied by the Institute of Tropical Medicine, Belgium.

The second set consisted of sputum samples from Karonga, Malawi. Three sputum samples were obtained from individuals with suspected TB, and a fourth one was obtained from a patient admitted for treatment. Two sputa were collected at 2 and 5 months and at the end of treatment. All sputum samples received at the KPS laboratory were decontaminated and inoculated onto Löwenstein-Jensen (LJ) slopes. What remained of the decontaminated sputum sample was stored at −20°C.

During the period from 1 November 2007 to 31 December 2008, the neutralized deposits from 469 whole-sputum samples that had been found culture positive for suspected Mycobacterium spp. were routinely collected and frozen. A total of 160 of these deposits were shipped to the LSHTM at ambient temperature. Sputa for shipment were selected in blocks of consecutive samples, ranging in size from 5 to 30, according to how many samples were found in each storage unit in the KPS laboratory and covered the entire period from November 2007 to December 2008. These were stored at −20°C in the category 3 laboratory at LSHTM until required. Of the 160 sputa, 140 were smear positive. The cultures from all 160 sputa had previously been sent to the reference laboratory in London, United Kingdom, for identification and drug sensitivity testing.

The third set of specimens used in this study consisted of 377 M. tuberculosis cultured isolates obtained from patients in Karonga diagnosed with TB between 2006 and 2009. IS6110-RFLP and spoligotyping data were available for this set of specimens. Nineteen isolates had a corresponding sputum specimen included in this study.

DNA isolation.

For the 48 M. tuberculosis cultures from the KPS, DNA was extracted using the standard IS6110-RFLP method as previously described (28). For all other cultures, colonies grown on LJ slopes were scraped off and resuspended in 400 μl of Tris-EDTA buffer. Suspensions were boiled for 20 min and centrifuged at 13,000 rpm for 2 min, and the resulting supernatants were used for PCR analysis.

The decontaminated sputum samples were first centrifuged at 3,000 × g for 30 min, and all but 1 ml of supernatant was removed. Samples were then transferred to screw-cap microcentrifuge tubes and heat killed by immersion in a boiling water bath for 20 min. Following centrifugation at 12,000 rpm for 10 min and removal of supernatant, DNA was extracted using a QIAamp DNA minikit (Qiagen, Crawley, United Kingdom) by following the manufacturer's instructions. Sputum samples were extracted in batches of 17. As a positive extraction control, a few colonies of M. tuberculosis H37Rv were taken from an LJ slope and added to molecular biology-grade water (Flowgen Bioscience Ltd., Nottingham, United Kingdom). In addition, 2 negative extraction controls (molecular biology-grade water) were included with every extraction. Extraction controls were processed in exactly the same way as the sputum samples.

LAM and Beijing PCR assays.

The primers used in this study targeted the Latin American-Mediterranean (LAM) and Beijing lineages. The specificity of each PCR is due to the unique presence of an IS6110 insertion sequence: an IS6110 insertion in Rv2920 which is unique to the Beijing lineage (31), and the presence of an IS6110 insertion, position 932204 according to the H37Rv whole-genome sequence, which is unique to members of the LAM family (21). The primer sequences are given in Table 1. Three primers were added to each PCR. The interpretation of the PCR result is dependent upon the number and size of the amplified products. The presence of both possible PCR products indicates a mixed DNA population. For example, the amplification of a 203-bp product and a 141-bp product in the Beijing PCR would detect the presence of both Beijing and non-Beijing DNA in one sample.

TABLE 1.

Primers used for each PCR and interpretation of amplified products^a

PCR	Primer	Primer sequence	Expected PCR product size (bp)	Interpretation
Beijing	F29F	AGA TCG GCT TTC TTC CCG C	203	Beijing
	F29R	CGT CTT TCT TTC CGT TGT GGA	141	Non-Beijing
	XhoI	TTC AAC CAT CGC CGC CTC TAC
LAM	LAMF	TAG CCC ACC ACC ACA GCT TC	205	LAM
	LAMR	ACC ACC CTG CCT AAC CAA TTC	141	Non-LAM
	XhoI	TTC AAC CAT CGC CGC CTC TAC

^aThree primers are used in both the Beijing and LAM PCR assays.

Amplification reactions were set up using a Qiagen HotStarTaq Plus kit (Qiagen, Crawley, United Kingdom). Both the LAM and Beijing reactions used 25-μl volumes containing 5 μl of DNA, 25 pmol of each primer, 0.2 mM each deoxynucleoside triphosphate (dNTP), 3.5 mM MgCl₂, 1× PCR buffer, 1× Q buffer, and 0.5 U Taq polymerase.

The LSHTM laboratory operates a strict three-room, one-way procedure for performing PCR, and this was adhered to throughout the study. PCR positive controls were included in each run. Six negative controls (molecular biology-grade water added instead of DNA) were also included in every PCR setup. Three negative controls were set up in the PCR clean room, and three were set up in the room in which the DNA was added.

Cycling parameters used for both the LAM and Beijing PCR were as follows: 95°C for 5 min to activate the Taq polymerase, followed by 40 cycles at 94°C for 1 min, 62°C for 1 min, and 72°C for 1 min, and a final extension at 72°C for 10 min.

Agarose gel electrophoresis of PCR products.

Ten microliters of each PCR product was separated in 1.5% agarose (Electran; VWR, Lutterworth, England) and visualized by staining with ethidium bromide. Digital images of the gels were captured using the MultiDoc-It digital imaging system (UVP Inc., Cambridge, England). Amplicon size was estimated with Gene Tools version 3.07 software (Syngene, Cambridge, England) using 100-bp ladder markers (Sigma, Poole, United Kingdom).

Limit of detection.

The limit of detection for each PCR was determined using single and mixed DNA templates. For the single DNA template assays, 10-fold serial dilutions of DNA from a representative Beijing and LAM strain were tested. M. tuberculosis H37Rv was used to represent the nonlineage DNA. The concentration of each DNA was estimated using NanoDrop ND-100 (Nanodrop Technologies, Wilmington, DE), and each DNA was diluted to give approximately an 10⁵ to 10⁰ genome copy number.

For the mixed DNA template assays, Beijing, LAM, and M. tuberculosis H37Rv DNA were each diluted to give 10⁵ to 10⁰ genome copy numbers. These preparations were then mixed in a checkerboard format such that the Beijing and LAM DNA dilutions were each mixed with dilutions of M. tuberculosis H37Rv DNA.

All PCR assays were performed in triplicate. A visible PCR product in at least two out of three replicates was considered to be reliable detection.

Detection of mixed M. tuberculosis strains.

DNA extracted from the 160 decontaminated sputum samples was tested undiluted (neat) and in duplicate against each PCR. Samples which failed to amplify in either or both PCR assays were retested neat and also diluted 1/10 and 1/100. All samples were processed blind; only the sample numbers were provided. Patient, smear, and culture data were not matched to sample numbers until all PCR results had been finalized.

In addition, DNA from 377 clinical M. tuberculosis isolates obtained from the KPS was also assayed with both the LAM and Beijing PCR. These isolates had previously been analyzed by the IS6110-RFLP method (28). Spoligotyping (15) was also performed on these isolates with the aid of a commercially available kit (Isogen Biosciences BV, Maarssen, Netherlands). IS6110-RFLP and spoligotype patterns were analyzed using BioNumerics version 4.0 (Applied Maths, Belgium) and checked by eye. Spoligotype patterns were converted into an octal code (8) and compared to the SPOLDB4 international database (http://www.pasteur-guadeloupe.fr/tb/bd_myco.html) to determine the corresponding shared type (ST) number and lineage/sublineage. Patterns that were not found in the SPOLDB4 database were classified as unique. Where possible, unique patterns were reassigned to a lineage based on SPOLDB4 classification (4) and the model-based program Spotclust (30).

Detection of mixed strains by MIRU-VNTR analysis.

Thirty DNA samples (28 sputum DNAs and 2 culture DNAs) were sent to a commercial company (Genoscreen, Lille, France) for 24-locus mycobacterial interspersed repetitive-unit-variable-number tandem-repeat (MIRU-VNTR) analysis (26). The 30 samples included all samples obtained from the two patients (A and B) who had been identified with possible mixed infections. The remaining 21 samples were randomly selected to represent all PCR result variations identified (LAM positive, non-Beijing positive; non-LAM positive, non-Beijing positive; and Beijing positive, non-LAM positive) from both HIV-positive and HIV-negative patients.

Ethical approval.

The studies were approved by the Health Sciences Research Committee in Malawi and by the LSHTM ethics committee. All specimens were collected as part of routine tuberculosis clinical activities. HIV testing included pre- and posttest counseling and informed consent. Written consent was sought for all studies.

RESULTS

Validation of LAM and Beijing PCR assays.

Each PCR primer set was initially tested against the first set of specimens. Compared to the known spoligotype patterns, specificity was 100% for both the Beijing and LAM PCR. No amplification was noted for any of the 17 NTM isolates in either PCR.

When DNA templates were diluted and tested alone, the limit of detection for the Beijing PCR was approximately 39 Beijing genome copies and 45 non-Beijing genome copies. The LAM PCR could detect approximately 19 LAM copies and 45 non-LAM copies. From the checkerboard mixed template assays, the Beijing PCR was able to detect a minimum of approximately 1 Beijing copy in the presence of 40 non-Beijing copies and 1 non-Beijing copy in the presence of 26 Beijing copies. For the LAM PCR, 1 LAM copy was detectable in the presence of 40 non-LAM copies and 1 non-LAM copy in the presence of 26 LAM copies.

Detection of mixed infections in sputum.

Of the 160 samples received from the KPS, the total number amplified by the LAM and the Beijing PCRs was 125 and 131, respectively. DNA from 44 (35.2%) of 125 samples was positive for the presence of LAM DNA, 77 (61.6%) were positive for non-LAM DNA, and 4 (3.2%) were positive for mixed LAM and non-LAM DNA. The number of samples that were Beijing or non-Beijing DNA positive was 5 (3.8%) or 126 (96.2%), respectively. None that contained mixed Beijing and non-Beijing DNA were detected, and none which contained DNA from both LAM and Beijing strains were detected.

The 160 sputum samples belonged to 72 patients. A total of 25 patients had just one sputum sample included in this study, 18 patients had two, 20 patients had three, 7 patients had four, 1 patient had five, and 1 patient had six. The ages of the patients ranged from 4 to 84 years, with a median age of 36 years. Thirty-eight patients (52.8%) were female. HIV results were available for 59 patients, of which 37 (62.7%) were HIV positive. Ten patients (13.9%) had received treatment for a previous TB episode.

The four samples in which both LAM and non-LAM DNA were detected were obtained from two different patients. Patient and sample details are given in Table 2. Both patients were HIV positive, and neither of them had a history of previous TB.

TABLE 2.

Patient and sample characteristics for the two patients with mixed infection detected in sputa^d

Patient/sample no.a	Collection day	Smearb	Culture day	Colony count	LAM PCR result	Spoligotyping result by SpolDB4	MIRU-VNTR typing result (no. of double loci/no. of loci amplified)
A/1s	1	2+	6	Confluent	LAM	NA	0/6
A/1c	1	NA	6	Confluent	LAM	LAM11_ZWE (SIT 59)	2/24
A/2s	1	1+	6	Innumerable	LAM	NA	1/11
A/3s	9	0	22	2 colonies	LAM+non-LAM	NA	1/4
B/1s	1	2+	62c	Many colonies	LAM+non-LAM	NA	6/22
B/1c	1	NA	62	Many colonies	LAM	LAM11_ZWE (SIT 1468)	0/24
B/2s	2	2+	7	Innumerable	LAM+non-LAM	NA	5/11
B/3s	2	3+	12	Innumerable	LAM+non-LAM	NA	5/12
B/4s	7	3+	13	3 colonies	non-LAM	NA	0/3

^as, sputum; c, cultured isolate. Patient A was a 33-year-old male, while patient B was a 26-year-old female. Both patients were HIV positive, with no history of TB.

^b0, no bacilli in at least 100 fields; 1+, 1 to 10 bacilli in 100 fields; 2+, 1 to 10 bacilli in 10 fields; 3+, 1 to 10 bacilli in 1 field.

^cFirst cultured on day 7, culture was overgrown by other organisms. Recultured on day 62.

^dNA, not applicable.

Patient A, a 33-year-old male, had a total of three sputum samples included in this study. The two samples that were collected on the day of diagnosis (day 1) were positive for LAM only. The third sample, collected 8 days later when treatment was started, was positive for both LAM and non-LAM DNA. No bacilli were seen in the smear, and only two colonies grew in culture from this third specimen. A culture from this patient, from day 1, was shown to be LAM by PCR and spoligotyping. When studied by MIRU-VNTR analysis, the three sputum specimens from patient A amplified only 4 to 11 loci and showed only one double allele. The culture from patient A had two double alleles from 24 loci, one of which was at the same locus as that identified in one of the patient's sputum samples.

Patient B was a 26-year-old female. Four sputa from this patient were included in this study. LAM and non-LAM DNA were detected in the three samples collected on day 1 and day 2. Only non-LAM DNA was detected in the sputum sample collected on day 7, which was 3 days after treatment had started. Culture from the sputum sample collected on day 1, and recultured 2 months later following initial contamination, showed only LAM by PCR and spoligotyping. By MIRU-VNTR analysis, the three sputum specimens that were mixed for PCR had double alleles at 5 or 6 loci from the 11 to 22 loci that amplified. Three loci with double alleles were common to 3 sputa, and the other three were common to 2 sputa. The nonmixed sputum sample obtained from this patient amplified only at three loci, and the culture from this patient showed no double alleles from 24 loci.

Detection of LAM and Beijing strains among cultured isolates.

A total of 377 subcultured M. tuberculosis isolates obtained from 347 patients were assayed with both the LAM and Beijing PCRs. These isolates were also spoligotyped and had previously been characterized by IS6110-RFLP analysis. Table 3 shows the PCR results compared to the spoligotype results of each isolate. Only the 17 (4.5%) spoligotype-defined Beijing isolates amplified a 203-bp product. The remaining 360 isolates were all identified as non-Beijing in the PCR. No samples with mixed Beijing or non-Beijing DNA were detected.

TABLE 3.

Spoligotyping versus Beijing and LAM PCR results for 377 clinical M. tuberculosis isolates used in this study

Lineage	Spoligotyping results from culture			No. of indicated isolates detected by LAM and Beijing PCRs
	Total no. of isolates	Sublineage	No. of sublineage isolates	Beijing	Non-Beijing	LAM	Non-LAM	Mixeda
Beijing	17	None	17	17	0	0	17	0
CAS	59	CAS1-DELHI	12	0	12	0	12	0
		CAS1-KILI	44	0	44	0	44	0
		CAS, other	3	0	3	0	3	0
EAI	55	EAI1-SOM	15	0	15	0	15	0
		EAI5	28	0	28	0	28	0
		EAI5 or EAI3	2	0	2	0	2	0
		EAI, other	10	0	10	0	10	0
Haarlem	4	H1	2	0	2	0	2	0
		H2	1	0	1	0	1	0
		Haarlem, other	1	0	1	0	1	0
LAM	166	LAM1	10	0	10	10	0	0
		LAM2	1	0	1	1	0	0
		LAM3/S	2	0	2	1	1	0
		LAM4	3	0	3	3	0	0
		LAM9	4	0	4	4	0	0
		LAM11_ZWE	103	0	103	102	0	1
		LAM, other	43	0	43	38	1	0
MANU	6	MANU1	6	0	6	0	6	0
T	47	T1	21	0	21	0	20	1
		T2	5	0	5	0	5	0
		T2-T3	13	0	13	0	13	0
		T, other	8	0	8	0	8	0
U	1	None	1	0	1	0	1	0
X	6	X1	3	0	3	0	0	0
		X1-LAM9	1	0	1	1	0	0
		X3	2	0	2	0	2	0
Unique	16	None	16	0	16	7	7	2
Total (%)	377 (100)		377 (100)	17 (4.5)	360 (95.5)	167 (44.3)	199 (52.8)b	4 (1.1)

^aMixed, LAM and non-LAM isolates detected.

^bNot included in these data are 7 isolates for which nonspecific amplification by the LAM PCR was obtained. These include 4 isolates assigned to “LAM, other” and 3 assigned to the X1 sublineage.

A total of 167 (44.3%) isolates were identified as LAM positive in the LAM PCR. Based on spoligotyping results, 159 of these were assigned to the LAM lineage, 1 was assigned to the X1-LAM9 sublineage, and 7 were unique. Non-LAM DNA was detected in 199 (52.8%) isolates.

LAM DNA and non-LAM DNA were detected in a total of 4 isolates (1.1%), which belonged to four different patients. The spoligotype and IS6110-RFLP patterns for these were reexamined. Only one showed any evidence of a potential mixture of M. tuberculosis strains; low-intensity bands were seen in both the spoligotype and IS6110-RFLP patterns. In addition, the IS6110-RFLP type for this isolate was unique within our database.

Sputum DNA versus culture DNA results.

Nineteen M. tuberculosis culture isolates were matched to sputa that was tested in this study. Among these were isolates from patient A (n = 1) and patient B (n = 1), as mentioned previously.

For the remaining 17 isolates, the LAM and Beijing PCR results for the cultured isolates were concordant with the results obtained from applying both PCR assays to the corresponding sputa.

Detection of mixed infections by MIRU-VNTR genotyping.

Thirty DNA were sent to Genoscreen for independent analysis by 24-locus MIRU-VNTR genotyping. The presence of double alleles at more than one locus is taken to indicate the presence of two M. tuberculosis strains in a sample. A number of loci were not successfully amplified from the 28 sputum DNA samples, possibly due to low DNA concentration (<10 ng/μl). In addition, it was not possible to confirm double alleles (by two independent rounds of PCRs).

The results of MIRU-VNTR analysis of the samples obtained from patient A (n = 4) and patient B (n = 5) are described above. The remaining 21 DNA samples came from 19 different patients. Three samples showed double alleles at two loci. None showed double alleles at more than 2 loci; however, only 4 samples were amplified at all 24 loci.

DISCUSSION

The detection rate of mixed infections is influenced by both the sampling approach and the genotyping method used to differentiate between strains. We used a PCR-based approach and applied two PCR assays to sputum samples obtained from TB patients in Malawi. We investigated the following two different genotypes: Beijing and LAM. Mixed infection was identified in the sputa of two (2.8%) patients, much lower than the rate found in Cape Town, South Africa, and similar to the 2% rate found in Bangladesh from examination of single colonies (23). Differences between the study settings may partly account for this; the annual risk of infection with tuberculosis is around 3% in Cape Town (16) and 1% in Malawi (6).

Other important differences lie in the laboratory methods. In our study, DNA was extracted directly from processed sputum samples, while in the South African study, sputum samples were first cultured in Bactec medium for 7 days. This may have enhanced the sensitivity of the assay, although overgrowth of one strain by another is also possible. In addition, the Beijing/non-Beijing PCR assay used in South Africa was based on noncompetitive primers and used an extra 5 PCR cycles, and either or both of which may have contributed to assay sensitivity. As with any method currently used to investigate mixed infections, the presence of an underlying strain can be demonstrated only when sufficient numbers (or DNA copies) of that strain are present in the sample being analyzed. It is difficult to know what proportion of the total DNA extracted from each sputum sample in our study came from M. tuberculosis cells and to what extent multiple strains contributed to this proportion. We used two different PCR assays and, since LAM is present in about half of our population, this should have been sufficiently discriminatory to detect mixed infection if present. However, it is possible that mixed infections are occurring in our setting with multiple LAM (or multiple non-LAM) strains, which would not be detected given the dichotomous nature of the PCR assays we applied. The development of strain-specific, or sublineage-specific, PCR assays or probes would allow this issue to be addressed.

It has been suggested that the analysis of one isolate per patient can lead to an underestimation of mixed infections (22). Of the 72 patients from whom we analyzed sputum specimens, only 25 (34.7%) provided a single specimen. We had more than one sputum specimen, and a corresponding culture, from the two patients for whom mixed infection was detected. For one of these patients, patient B, mixed infection was detected in more than one sputum sample. The single sputum sample in which only non-LAM DNA was present was an admission sample taken after 3 days of treatment. While the bacillary load in this sputum sample was high, only three colonies grew. It has been suggested that the relative survival of M. tuberculosis strains may be different once treatment has begun (20), and it is possible that, as a result of treatment, the LAM strain in this specimen had been reduced to a level that was undetectable by our assay. For patient A, from whom only 1 of 3 sputum specimens showed evidence of mixed infection, cross-contamination is a possibility.

Neither of the two M. tuberculosis isolates cultured from the sputa obtained from patient A and patient B showed any evidence of more than one strain type when assayed with the LAM and Beijing PCRs. However, the provisional MIRU-VNTR typing results did indicate the presence of double alleles in 2 out of 24 loci for the isolate from patient A. As all the clinical M. tuberculosis isolates used in this study had been subcultured, it is unlikely that they were representative of the clonal composition present in the sputum specimen from which they were isolated. Any culture technique introduces an element of strain selection, and it has recently been shown that culture affects the clonal complexity of M. tuberculosis (18), as in vitro culture conditions preferentially select some strains over others. Therefore, it is not surprising that the PCR assays of culture detected only one DNA type in the M. tuberculosis isolates from the two mixed infection patients. The double loci that were observed in one of these isolates may be due to the high sensitivity of the MIRU-VNTR technique; the detection limit of mixed templates for our PCR assays was estimated to be 1:26 to 1:40, compared with a reported 1:99 detection limit for MIRU-VNTR typing (11).

The increased sensitivity of MIRU-VNTR typing may also explain the detection of the double alleles seen in DNA from 3 of 21 “control” sputa that were also sent to Genoscreen. The 3 smear-positive sputa did not amplify as mixed in either the LAM or the Beijing PCR assay. Unfortunately, the double alleles detected were unable to be confirmed due to low DNA concentrations. However, it is possible that, had we the means to investigate all samples in this study by also using MIRU-VNTR typing, we may have identified more mixed infections. Of the 3 most commonly used techniques for genotyping M. tuberculosis isolates (IS6110-RFLP, spoligotyping, and MIRU-VNTR typing), it is MIRU-VNTR typing that has the greatest capacity to detect the presence of multiple strains, as has been previously demonstrated (11, 22-24).

The detection of LAM and non-LAM DNA in four of the cultured isolates included in this study was unexpected, as most of the isolates had been subjected to more than one round of subculture. For one of the four isolates, the presence of more than one strain could also be seen in the spoligotype and IS6110-RFLP patterns. The low-intensity bands were not noted when these patterns were first examined, and their detection was dependent upon the exposure time of the autoradiogram, as has been previously reported for IS6110-RFLP typing (9). However, as the isolates had been cultured within 3 different laboratories, laboratory error or contamination cannot be ruled out as an explanation for the detection of mixed LAM and non-LAM M. tuberculosis isolates.

DNA fingerprinting data have been collected for M. tuberculosis isolates in Karonga, Malawi, for 15 years. The ability to differentiate M. tuberculosis strains with this technique has supported a number of studies looking at the rate of relapse and reinfection in this setting (6, 7, 14). The possibility that mixed infections may play a role in reinfection (29, 32) has raised a number of questions concerning TB control and the interpretation of molecular epidemiologic studies (3). Using the current methodology, the detectable rate of mixed infections observed in our study is low. The impact of mixed infections on the interpretation of the molecular epidemiological data in this setting appears to be minimal.