Mixed-Strain Mycobacterium tuberculosis Infections among Patients Dying in a Hospital in KwaZulu-Natal, South Africa

Abstract

We performed spoligotyping and 24-locus mycobacterial interspersed repetitive-unit-variable-number tandem-repeat (MIRU-VNTR) typing on M. tuberculosis culture-positive biopsy specimens collected from adults dying in a hospital in KwaZulu-Natal. Of 56 culture-positive samples genotyped, we detected mixed strains in five (9%) and clonal heterogeneity in an additional four (7%).

The application of molecular approaches for detecting variation among Mycobacterium tuberculosis isolates has generated new appreciation for the diversity present within this relatively genetically conserved bacterial species (4, 17). Genotyping methods, such as insertion sequence typing (IS6110), spacer oligonucleotide typing (spoligotyping), and mycobacterial interspersed repetitive-unit-variable-number tandem-repeat (MIRU-VNTR) typing, have been used to identify transmission chains (1, 10, 18, 26), to classify strains into families and lineages (2, 5, 9, 18, 27, 29), to identify episodes of exogenous reinfection (3, 20, 25, 31), and, most recently, to detect the presence of within-host genetic heterogeneity (11, 19, 24, 33).

Genetic heterogeneity of M. tuberculosis within a host may arise by one of two mechanisms: (i) within-host diversification following a single infection event or (ii) reinfection resulting in a mixed infection with more than one strain (12, 16). The clinical consequences of within-host diversity are most obvious when manifesting as subpopulations of bacteria with resistance to tuberculosis (TB) antibiotics, either reflecting acquired drug resistance (mechanism 1) or transmitted drug resistance (mechanism 2). Furthermore, because individuals can be simultaneously infected by strains with different phenotypic characteristics (e.g., growth rates, drug resistance), the within-host competition between strains may influence the clinical outcomes for coinfected patients (30) and may affect the population dynamics of the pathogen in the community (6, 8).

While within-host M. tuberculosis genetic diversity has been documented in many settings, systematic efforts to measure the prevalence of these complex infections have rarely been attempted. In this study, we report the results of a genotyping analysis on isolates collected from young adults dying in a hospital in KwaZulu-Natal, South Africa.

We conducted limited autopsies on adults aged 20 to 45 years who died after admission to Edendale Hospital in KwaZulu-Natal, South Africa, between October 2008 and August 2009. The incidence of tuberculosis in KwaZulu-Natal is 1,094 cases/100,000 persons per year, and the HIV prevalence among women in antenatal clinical settings is 39% (14, 21). Our primary aim was to investigate the burden of tuberculosis as a contributing cause of death in this highly vulnerable community. Of the 240 decedents enrolled in the study, 94% were HIV seropositive. Fifty-eight percent of those on treatment for tuberculosis at the time of death were still infected with viable M. tuberculosis, while 42% of those not receiving treatment for TB also had positive M. tuberculosis cultures at the time of death. These results suggest that delayed and missing diagnoses of tuberculosis and the emergence of drug resistance contribute to the large burden of tuberculosis-related mortality in this setting. A more detailed account of the design and findings of this study is available in a related report (7).

Here we present a genotyping analysis of M. tuberculosis-positive specimens collected from a subset of M. tuberculosis culture-positive decedents included in the larger postmortem study; Table 1 displays the characteristics of decedents with samples included in this analysis. Briefly, for each decedent, we conducted limited autopsies to retrieve respiratory tract secretions as well as needle core biopsy specimens from lung, liver, and spleen. These specimens were pooled and grown in liquid culture medium (BACTEC MGIT 960; Becton-Dickinson, NJ). Positive cultures were identified as M. tuberculosis by the niacin-nitrate test. Mycobacterial DNA was then extracted from the liquid medium and shipped to a commercial laboratory (Genoscreen, Campus de l'Institut Pasteur de Lille, France) for spoligotyping and 24-locus MIRU-VNTR analysis (23). Of the 58 samples sent to the laboratory, 2 were of insufficient quality to perform genetic analysis.

TABLE 1.

Characteristics of individuals with genotyped samples

Characteristic	Result
Total no. of samples genotyped	56
No. (%) of male subjects	36 (64)
Median age in yr (range) at time of death	34 (21-45)
No. (%) of HIV-infected subjects	54 (96)
No. (%) on ARVsa	8 (15)
No. (%) with MDRTB	12 (21)
No. (%) on TB treatment at time of death	34 (60)
No. of days (range) of final hospitalization	2 (0-40)

^aNumber and percentage of HIV-infected subjects on antiretroviral drugs (ARVs).

Consistent with existing definitions (16, 24, 28), we categorized infections as mixed-strain infections when we identified more than one allele at more than one MIRU-VNTR locus and as clonal population infections when there was more than one allele at a single locus. By this classification approach, we found that five decedents (9%) had infection with more than one strain and four (7%) had evidence of clonal heterogeneity (Table 2). Previous examination of worldwide samples of M. tuberculosis isolates found that certain MIRU-VNTR loci were more variable than others (28). In our sample, heterogeneity in the four polyclonal populations we detected was due to variation at four different loci: 4052 (Qub26), 2996 (MIRU 26), 960 (MIRU 10), and 802 (MIRU 40). These loci were (respectively) the first, fourth, sixth, and seventh (out of 24) most variable loci in the international sample. Each individual with a complex infection was HIV coinfected.

TABLE 2.

Samples categorized as complex infections

Sample population type and study ID no.	Allele(s) at MIRU-VNTR locus:
	580	2996	802	960	1644	3192	424	577	2165	2401	3690	4156	2163b	1955	4052	154	2531	4348	2059	2687	3007	2347	2461	3171
Mixed
BN093	2	4	3 + 4b	—a	—	2	2	4	3	2	3	—	2	2	3	2	5 + 6b	2 + 3b	2	1	2	5	2	3
BN099	2	6	1	—	—	2 + 3b	2 + 3b	3 + 4b	3	4	2	—	4	2	—	2	3 + 5b	2	1	1	2 + 3b	2 + 4b	2 + 3b	2
BN112	2	4 + 5b	1 + 4 + 5b	3 + 4b	1	3	2	3 + 4b	2 + 3b	2 + 4b	3	—	2	2 + 4b	4	2	5	2	2	1	3	4	1 + 2b	3
BN123	2	7	3	2	3	5	4	3 + 4b	3	4	4	2	2	5	6	2	5	4	2	1	3	4	2	2 + 3b
BN154	3	4	4	3	3	3	3	4	1 + 4b	2	1 + 3b	2	3	1	—	3	5	2	2	1	3	4	2	2
Clonal
BN090	2	5	3	1 + 4b	—	3	4	4	2	2	1	2	2	4	—	2	6	2	2	1	3	4	2	1
BN160	2	5	3	5	3	3	2	3	1	3	3	3	4	3	6 + 7b	2	3	2	1	1	3	2	2	3
BN177	4	5 + 6b	3	4	3	3	4	4	2	2	1	2	3	4	—	2	6	2	2	1	3	4	2	1
BN220	2	5	3 + 4b	4	3	3	3	4	2	1	2	3	2	3	—	1	6	2	2	1	3	4	2	3

^a—, nonamplified markers after two rounds of independent and monoplex PCR.

^bDouble alleles that were confirmed by two independent rounds of PCR.

After grouping mixed-strain infections and clonal heterogeneity into a single category of complex infection, we investigated if any measured clinical and demographic variables were associated with this outcome. Univariate logistic regression did not identify any host or bacterial variables that were statistically significantly associated with this outcome, and neither did male gender (odds ratio of complex infection [OR], 2.41; 95% confidence interval [CI], 0.45 to 12.84; P, 0.30), increases in age (OR, 0.98 per year; 95% CI, 0.87 to 1.10; P, 0.75), antiretroviral use (OR, 1.95; 95% CI, 0.33 to 11.69; P, 0.46), or multidrug-resistant (MDR) TB (OR, 0.47; 95% CI, 0.05 to 4.33; P, 0.51). While it did not reach the level of statistical significance, there is some suggestion that being on treatment for TB at the time of death reduced the probability of complex infection (OR, 0.27; 95% CI, 0.06 to 1.20; P, 0.08). This may reflect the selective pressure of standard drug treatments, which we expect should rapidly eliminate drug-sensitive bacteria; this hypothesis is supported by the strong positive association we observed between being on treatment at the time of death and the detection of MDRTB among these individuals (OR, 9.0; 95% CI, 1.06 to 76.48; P, 0.04). This finding is important because it indicates that in areas where complex infections occur, individuals who appear to have acquired drug resistance while on therapy (i.e., individuals whose drug sensitivity tests indicated sensitivity prior to treatment and resistance after treatment has begun) may instead have experienced transmitted drug resistance which was subsequently “unmasked” by treatment (30). Such unmasking may, in turn, contribute to ongoing transmission of drug-resistant tuberculosis during first-line treatment.

The diversity of strain lineages present in our sample is qualitatively similar to the diversity of other strains that have been genotyped in Africa and submitted to international collections (5) (Table 3). The most common lineages and sublineages in our sample included LAM (16 isolates, 33% of all classified strains), T (11 isolates, 23%), Beijing (8 isolates, 17%), and S (8 isolates, 17%). Six of the eight strains in the S lineage had spoligotypes matching the F28 family identified by Warren et al. in Cape Town, South Africa (32). In Table 3, while we have provided the shared international type (SIT) grouping assigned to those samples in which we detected mixed infections, we note that these assignments may be misleading. In a mixed infection, the spoligotype pattern may in some cases reflect the cumulative presence of spacers in all strains present or in other cases may accurately report the spoligotype from the dominant strain. Accordingly, in areas where mixed infections are common, the usefulness of spoligotyping may be compromised.

TABLE 3.

Spoligotypes of samples and association with complex infections

SITa	Lineage/sublineage	Frequency	No. of samples showing complex infection
1	Beijing	8	0
33	LAM3	8c	1 (clonal heterogeneity)
34	Sb	5c	1 (mixed infection)
37	T3	1	0
39	T4-CEU1	2	0
51	T1	1	0
52	T2	1c	1 (mixed infection)
54	MANU2	1c	1 (mixed infection)
60	LAM4	5	1 (clonal heterogeneity)
62	H1	1	1 (clonal heterogeneity)
71	Sb	1	0
73	T2-T3	2	0
88	S	1	0
92	X3	1	0
207	H3	1	0
336	X1	1	0
719	T1	4	1 (clonal heterogeneity)
815	LAM11_ZWE	3	0
1196	Undesignated	1c	1 (mixed infection)
1915	S	1	0
No SIT type		7c	1 (mixed infection)

^aSIT, shared international type (according to SpolDB4 [5]).

^bThese spoligotypes are the same as those for the F28 family (32).

^cFrequency reported includes one sample in which mixed strains were detected by MIRU-VNTR; thus the SIT assignments for these particular samples may not be accurate.

We note several limitations to this study. First, nearly half of the M. tuberculosis culture-positive samples (47%) from the postmortem exams were not stored at the laboratory and thus were not available for genotyping. Samples collected from males (OR, 2.43; 95% CI, 1.13 to 5.25; P value = 0.02) and samples that were MDR (OR, 3.74; 95% CI, 1.13 to 12.39; P value = 0.03) were more likely to be saved and thus are overrepresented in this analysis. Although this limits the generalizability of our results, the finding that a substantial fraction of these individuals had active infection with more than one strain is noteworthy. While MIRU-VNTR appears to be a specific test for mixed infections, its sensitivity is likely to be quite limited, since detection of a mixed infection requires that clinical samples collected from the patient actually contain some of each strain present, that the culture procedures are permissive enough to allow growth of minority strains (15), and that the DNA isolation procedures from these cultures capture material from all strains present. While our strategy for pooling all autopsy material from a single individual before culturing may have increased the sensitivity for detecting mixed infections (especially if different strains were present in different organs [13]), we were not able to identify the anatomic distribution of these strains or strain variants.

In conclusion, we found mixed-strain infections and clonal heterogeneity in a substantial fraction of young adults dying in a hospital in KwaZulu-Natal, South Africa. Because of the limited sensitivity of our testing procedures (22), the frequency of complex infections is probably higher than we found in this setting. The high frequency of HIV coinfection in our study population and the absence of timely TB treatment may contribute to the risk of both mixed M. tuberculosis infection and clonal heterogeneity in these patients; however, the effect of HIV on the risk of complex infection cannot be teased out directly from our data, since nearly everyone in our sample had HIV infection. Future studies that compare the risk of mixed infection among those with and without severe immunosuppression within the same community will help clarify the impact of HIV on the within-host diversity of TB infection.