Determinants of Multidrug-Resistant Tuberculosis Clusters, California, USA, 2004–2007

TOC summary: Type of isoniazid resistance–conferring mutation may be a determinant of genotypic clustering.

I n 2007, >500,000 cases of multidrug-resistant (MDR) tuberculosis (TB), defi ned as resistance to at least isoniazid and rifampin, occurred worldwide (1). Although demographic and clinical risk factors for transmission and pathogenesis of both drug-susceptible and drug-resistant Mycobacterium tuberculosis have been well described (2, 3), little is known about the microbial factors that infl uence the generation of secondary MDR TB cases (4,5).
Community-and population-based molecular epidemiologic studies of isoniazid-monoresistant M. tuberculosis (6)(7)(8) have shown that specifi c resistance-conferring mutations are associated with variable degrees of genotypic clustering, a measure of strain pathogenicity that incorporates host factors, transmissibility of the organism, and capacity of the organism to cause active disease. For example, isoniazid-monoresistant strains with a serine-to-threonine substitution at position 315 (S315T) are more often associated with secondary cases than are strains without the S315T mutation (6,7), likely because of reduced or absent catalase-peroxidase production (9). However, the effects of specifi c isoniazid resistance-conferring mutations on genotypic clustering in multidrug resistance are less well characterized. The studies reported to date have been limited by inadequate genotypic discrimination (10,11) and/or nonrepresentative sampling of cases (10,(12)(13)(14).
California reports the highest annual number of TB cases (15), more than one fourth of all MDR TB cases (16), and the highest immigration rates in the United States (17). We conducted a population-based cohort study of all incident MDR TB cases in California during a 4-year period (January 2004-December 2007) to 1) describe demographic and clinical characteristics of clustering and 2) estimate the effect of specifi c isoniazid resistance-conferring mutations and strain lineage on genotypic clustering of MDR M. tuberculosis.

Methods
We analyzed culture-positive cases of MDR TB reported to the California TB registry from January

Characterization of Mutations Associated with Isoniazid and Rifampin Resistance
For each isolate, genomic DNA was extracted from solid media (Lowenstein-Jensen slants, Middlebrook 7H10 or 7H11 agar), liquid media (BACTEC 12 B or MGIT [Becton Dickinson]), or smear-positive sputum sediments. A real-time PCR assay with 6 molecular beacon probes was performed by using an iQ5 iCycler instrument (Bio-Rad, Hercules, CA, USA) to screen for mutations associated with isoniazid and rifampin resistance (19). Two molecular beacons that targeted katG (codon 311-317) and the inhA promoter were used to detect isoniazid resistance-conferring mutations, and 4 molecular beacons that targeted the core region of rpoB were used to detect rifampin resistance-conferring mutations. Isolates with mutations in katG detected by the wild-type probe were further tested with another molecular beacon that specifi cally targeted katG S315T (AGC-ACC). When molecular beacon analysis did not show katG S315T or -c15t inhA promoter mutations, the entire furA-katG locus (H37Rv: 2153626-2156657, 3,031 bp) was sequenced as described (6). Sequence data were generated by using ABI BigDye v3.1 dye terminator sequencing chemistry and the ABI PRISM 3730xl capillary DNA analyzer (Applied Biosystems, Foster City, CA, USA) at the Genomic Core Facility, University of California, San Francisco (www.genomics.ucsf.edu/Sequencing/ index.aspx), and were analyzed with ClustalW (www.ebi. ac.uk/Tools/clustalw/index.html).

Genotyping and Lineage Determination
Spacer oligonucleotide typing (spoligotyping) and mycobacterial interspersed repetitive unit (MIRU) typing were performed in accordance with the Centers for Disease Control and Prevention Universal Genotyping Program procedures (20). Spoligotyping was performed by using Luminex-based methods to detect 43 known spacer sequences in the direct repeat locus (21). MIRU typing was performed by using the protocol described by Cowan et al. (22). A capillary sequencer, CEQ 8000 (Beckman, Fullerton, CA, USA), was used to analyze the number of repeated sequences at each of the 12 loci. IS6110-based restriction fragment length polymorphism (RFLP) genotyping was performed following standardized methods (23). RFLP patterns were compared by using Bioimage Whole Band Analyzer software version 4.2.1 (Bioimage Corp., Ann Arbor, MI, USA) (24). RFLP patterns with <20 identical bands, or >20 bands and differing by no more than a single band, were considered matched. IS6110 RFLP band assignment was edited by 2 independent readers, and the cluster assignment was confi rmed visually.
The phylogeographic lineage of strains was determined from spoligotyping results. Spoligotype families H, LAM, and T, X, S were considered to be of Euro-American lineage; Beijing of East-Asian lineage; EAI of Indo-Oceanic lineage; and CAS of East African-Indian lineage (5).

Defi nitions
Cases were defi ned as clustered if >2 isolates from cases reported during the study period shared the same MIRU and spoligotype, matched IS6110 RFLP, and had specifi c drug resistance-conferring mutations for isoniazid and rifampin. Clustering was assumed to represent both transmission of M. tuberculosis and progression to active disease, leading to secondary case generation within the period of the study. Cases not in a cluster were considered to be the result of reactivation of latent infection. Patients with the earliest case report date within a cluster were regarded as index cases.

Statistical Analysis
The proportion of clustered MDR TB cases was analyzed as the number of clustered cases divided by the total number of culture-positive cases that occurred during the study period. Because of limited sample size, the independent effects of katG S315T and phylogeographic lineage on clustering were estimated by using exact logistic regression methods. Refugee status and sputum smear positivity were included in the model as relevant host risk factors. Refugee resettlement during the study period could bias our results in that overrepresentation of ethnic groups or geographic locations with a high prevalence of particular strain-specifi c factors (phylogeographic lineage or drugresistance mutations) could confound the association under study. To examine this infl uence and potential clustering of TB cases within households or communities related to refugee resettlement, sensitivity analysis was conducted by reestimation of study results after excluding 1) patients immigrating from refugee settings within the past 3 years and 2) the single largest patient cluster, which accounted for 40% of all clustered cases. In a separate analysis, standard logistic regression was used to estimate the effect of katG S315T and phylogeographic lineage on sputum smear positivity, again controlling for refugee status.
Odds ratios (ORs) and 95% confi dence intervals (CIs) were calculated to measure associations of interest. Categorical data (e.g., sex, foreign vs. US birth, homelessness) were compared by using Fisher exact tests. The Wilcoxon rank-sum test was used to determine differences in the distribution of continuous variables (e.g., age, time from entry into the United States to TB diagnosis). Interaction between the independent variables was assessed separately for each factor by stratifi cation and statistical testing (Breslow-Day with the Tarone correction [25] and Zelen test [26]), with p<0.2 assumed to indicate the presence of interaction. All p values were 2-sided with α = 0.05 as the signifi cance level. Analyses were performed by using Stata 10 (StataCorp., College Station, TX, USA) and StatExact 8 (CYTEL Software Corp., Cambridge, MA, USA).

Results
During the study period, 11,395 cases of TB were reported in California, of which 9,037 (79%) had positive cultures. Of these, 8,899 (98%) had isoniazid and rifampin susceptibility results available. Of the 141 (2%) incident MDR TB cases, 123 (87%) had isolates available for MIRU, spoligotyping, and IS6110 RFLP analysis. Isolates unavailable for genotyping (n = 18) were more often from Los Angeles County; other demographic and clinical characteristics were similar to those of analyzed cases.
Twenty-fi ve MDR M. tuberculosis isolates were aggregated in 8 clusters (1 cluster of 10 cases, 1 cluster of 3 cases, and 6 clusters of 2 cases) for an overall cluster proportion of 20% (25/123). Excluding the 8 index cases, 14% (17/123) of all cases were considered to have resulted from recent transmission and rapid progression to disease (secondary cases). Within clusters, a median of 3 months elapsed between successive secondary cases (range 0-20 months). Of the 123 total cases, 113 (92%) occurred among foreign-born patients, with more than half (56%) occurring among immigrants from Mexico, Philippines, the People's Republic of China, or Vietnam ( Younger persons were more likely than older persons to harbor strains involved in MDR TB clusters (Table 1). HIV infection was unusual in this patient population; only 3 (4%) of 75 patients with known HIV status were HIV-infected. Twenty-eight percent (35/123) of patients reported a history of active TB; this proportion did not vary between clustered and nonclustered cases (p = 0.75). Eight patients had documented previous treatment in California. Time to culture negativity (2.4 vs. 2.8 months, p = 0.95), treatment failure, or death did not differ between clustered and nonclustered cases (p = 0.57) among 105 (85%) of 123 cases for which data were available.
Clustering was independently associated with katG S315T (OR 11.2, 95% CI 2.2-∞; p = 0.004) and refugee status (OR 6.0, 95% CI 1.2-36.2; p = 0.03) in exact multivariate analyses in which strain lineage and sputum smear positivity were controlled for ( Table 5). The estimated association between katG S315T and case clustering did not appreciably change in sensitivity analyses that excluded either all recently arrived refugees or the single largest patient cluster.

Discussion
In this 4-year population-based molecular epidemiologic study, transmission followed by secondary case generation contributed to ≈14%, or 1 of every 7, MDR TB cases in California. Clustered cases occurred more often among younger persons and persons who had emigrated from Mexico and refugee settings in Southeast Asia. In addition, pathogen-specifi c factors were associated with clustering of MDR TB cases, independent of traditional clinical and demographic risk factors.
The proportion of MDR TB cases attributed to transmission in California was higher than that reported by other investigators in most low incidence settings (28)(29)(30)(31)(32). The largest clusters in our study resulted from MDR TB out-breaks in California after resettlement of Hmong refugees in 2005-2006 (33) and resettlement of Tibetan refugees in 2001-2006. The associations between pathogen-specifi c factors and case clustering could be due to regional differences in strain prevalence and preferential migration of persons with specifi c strains to California. We attempted to control for these factors by using highly stringent criteria to defi ne clustered cases and by adjusting for refugee status in our multivariate model. However, without detailed contact information and contact tracing, we cannot be certain of the extent to which transmission or progression to active disease occurred within or outside California. Given that most clustered cases occurred among persons residing in the United States for >3 years and that US-born persons were involved in 3 of 8 clusters, at least some proportion of MDR TB transmission is likely to have occurred in California. This observation suggests that although most MDR TB cases in the United States are related to the migration of persons already infected with drug-resistant M. tuberculosis, a small but notable proportion may be due to ongoing transmission. Heterogeneity in the reproductive success of drugresistant M. tuberculosis is now well established (34,35). In this study, the katG S315T mutation was the only isoni-  (5), and Ukraine (1). ¶Date of US entry was missing for 2 persons. #Nonclustered cases: cervical lymph node (5), bone (1), other (1); clustered cases: pleural (1). **Culture positive after >8 months of treatment; limited to pulmonary TB patients who were alive at diagnosis and had an initial positive sputum culture. † †Treatment outcome available for 105 (85%) cases.
azid resistance-conferring mutation found among clustered MDR TB cases. The high prevalence of katG S315T among MDR strains (12,14,(36)(37)(38) and the association of this mutation with increased secondary case generation among isoniazid-monoresistant strains (6,7) have been documented. We report that the katG S315T isoniazid resistance-conferring mutation retains an independent effect on clustering of MDR TB cases despite the presence of mutations that confer resistance to additional drugs. In particular, secondary case generation did not signifi cantly vary according to site of rpoB mutations that confer rifampin resistance, which supports the hypothesis that these are no-cost mutations or that compensatory mutations commonly exist (39).
The katG S315T mutation is thought to preserve fi tness through the relative preservation of catalase-peroxidase production (9), although whether this mutation is associated with different clinical phenotypes is unknown. In this study, we noted an inverse association between presence of the katG S315T mutation and sputum smear positivity. In addition, the katG S315T mutation was associated with case clustering, independent of sputum smear status. These fi ndings suggest that the katG S315T mutation may preserve the ability of M. tuberculosis to transmit and cause secondary cases through mechanisms unrelated to conventional indices of disease severity, such as the presence of abundant acid-fast bacilli in sputum.
Our fi ndings are clinically useful for at least 2 reasons. First, MDR TB in California is occurring predominantly among patients who were not born in the United States, with some cases from recent transmission and rapid progression to disease. Our study suggests that in California, younger persons and persons who have emigrated from Mexico and from refugee settings may be at higher risk for transmitting MDR M. tuberculosis. Likewise, our fi ndings reinforce the need for giving priority to screening and prevention activities in immigrant communities and US investment in international TB control (40). Second, if our results are verifi ed in other settings, TB-control programs should consider pathogen-specifi c factors such as isoniazid resistance-conferring mutations when planning the intensity of contact investigation and secondary case-fi nding activities.
This study has several limitations. First, our estimates of case clustering are imprecise because of the limited num-  ber of MDR TB cases observed during the study period. However, these estimates are the best currently available, given that the data make up the largest population-based TB registry in the United States. Second, although our definition of genotypic clustering was highly rigorous, the lack of detailed epidemiologic information precluded confi rmation of transmission within California. Third, because our study did not include pan-susceptible or isoniazid-monoresistant strains, we cannot comment directly on MDR M. tuberculosis pathogenicity relative to these groups. Lastly, our analyses implicitly assume independence of outcome events, and household or community-level factors potentially associated with clustering were not available. Future studies should be designed so that statistical methods can be used that are able to accommodate the possible effects of within-household clustering. We found a substantial proportion of MDR TB cases and case clustering in California among non-US-born persons, and the katG S315T mutation was independently associated with clustering. Validation of these fi ndings in larger cohorts and in different population settings may have crucial public health consequences.