Clonal Expansion of Multidrug-Resistant and Extensively Drug-Resistant Tuberculosis, Japan

Strain clustering suggests community transmission plays a critical role in incidence.

T he epidemic of drug-resistant tuberculosis (TB) has raised public health concern about the global control of TB. The World Health Organization estimated that 0.5 million cases of multidrug-resistant TB (MDR TB) (i.e., Mycobacterium tuberculosis resistant to >2 of the most potent TB drugs, rifampin and isoniazid) occurred in 2007 (1). Some countries have extraordinarily high rates of this disease, but the problem is universal, and the extent varies from 1 country to another.
Another recent alarming issue is the emergence of extensively drug-resistant TB (XDR TB) (i.e., M. tuberculosis with MDR plus resistance to any fl uoroquinolone and >1 injectable drug, thus posing even greater management challenges than MDR TB alone). The treatment outcome of XDR TB is worse than that of simple MDR TB, and the death rate is particularly high among HIV-infected patients (2). Also, because XDR TB is much more expensive to manage in terms of prolonged medication and prolonged period of infectivity to other persons (3), it has the potential to exhaust human and fi nancial resources of the public health system for TB control. Although this new life-threatening disease had been reported from 49 countries as of June 2008 (4), its transmissibility among immunocompetent persons is not well known (5).
In Japan, TB remains a major infectious disease; in 2008, a total of 19.4 cases/100,000 population were reported (6), and Japan is generally classifi ed as a country with intermediate TB incidence. According to the most recent nationwide drug-resistance survey, the prevalence of MDR TB and XDR TB were 1.9% and 0.5%, respectively (7). Approximately one third of MDR and XDR (MDR/ XDR) M. tuberculosis strains were isolated from new TB patients, implying ongoing transmission of MDR/XDR TB in Japan.
Our purpose was to evaluate the transmission dynamics of MDR/XDR TB by using strains from the most recent (2002) nationwide drug-resistance survey in Japan, an industrialized country with low HIV incidence and intermediate TB incidence. We did so by analyzing the MDR/XDR strains by molecular genotyping methods, i.e., insertion sequence 6110 restriction fragment length polymorphism (IS6110-RFLP), spacer-oligonucleotide genotyping (spoligotyping), and variable number tandem repeats (VNTR).

Materials and Methods
We used data and culture isolates obtained in the 2002 nationwide drug-resistance survey, as previously reported (7). Briefl y, during June-November 2002, a total of 3,122 Clonal Expansion of Multidrug-Resistant and Extensively Drug-Resistant Tuberculosis, Japan clinical strains were collected from different patients who had started treatment in 99 hospitals throughout Japan. The number of patients enrolled represented 36.0% of all new reported TB cases during the study period. The sampling of the hospital was not randomized but was based on voluntary participation. The survey identifi ed 60 MDR/ XDR M. tuberculosis strains, 55 of which were analyzed in this study; the other 5 strains were unavailable for use in this study.

Patient Information
We used patient information collected from a standard data collection form in the drug-resistance survey in 2002 (7). The information included demographic data (age, sex, nationality, hospital, geographic area of the hospital), clinical data (history of TB treatment, site of TB disease, chest radiograph fi ndings, underlying disease), and bacteriologic data (results of sputum smear test for acid-fast bacilli). Patients were classifi ed as new if they had never been treated for TB for >4 weeks and as previously treated if they had ever been treated for TB for >4 weeks. The survey protocol conformed to the national guidelines for epidemiologic research (8).

Defi nition of XDR TB
We defi ned XDR strains according to the World Health Organization defi nition of XDR (1) and drug availability for TB treatment in Japan. XDR strains were defi ned as M. tuberculosis strains with resistance to at least isoniazid, rifampin, levofl oxacin, and kanamycin.

Molecular Genotyping
Three molecular genotyping methods based on IS6110-RFLP, spoligotyping, and VNTR were performed on the 55 MDR/XDR strains. IS6110-RFLP typing was performed according to the standardized protocol (9). The RFLP band patterns were compared by using the BioNumerics software package version 5.1 (Applied Maths, Kortrijk, Belgium). Strains with an identical band pattern were classifi ed as an RFLP cluster. Spoligotyping was also performed according to the standard protocol (10). Classifi cation of the spoligotype family was performed according to the international database, SpolDB4 (11). The VNTR analysis was conducted as described elsewhere (12) by using the standard 15 mycobacterial interspersed repetitive unit-VNTR proposed by Supply et al. (13), i.e., VNTRs 0424, 0577, 0580, 0802, 0960, 1644, 1955, 2163b, 2165, 2401, 2996, 3192, 3690, 4052, and 4156 plus 4 other loci, VNTRs 2074, 2372, 3155, and 3336. The latter 4 loci were added to increase discrimination for the Beijing family strains because of their high prevalence in Japan (12).

Statistical Analysis
Statistical analysis was performed with Epi Info software 3.51 (Centers for Disease Control and Prevention, Atlanta, GA, USA), by using χ 2 test or Fisher exact test for the comparison of proportions. A p value <0.05 was considered signifi cant.

Epidemiologic Characteristics of Patients
A total of 55 MDR/XDR cases were analyzed. The characteristics of patients with MDR/XDR TB are summarized and compared with those of patients with pansusceptible strains (n = 2,782) in Table 1. Patients with MDR/ XDR TB were signifi cantly more likely to be younger (odds ratio [OR] 5.69 for age 21-40 years; 4.11 for age 41-60 years) and foreigners (OR 6.41) and to have been previously treated (OR 10.55). All MDR/XDR TB patients had pulmonary disease, and these patients were signifi cantly more likely to have cavitary lesions (OR 3.24) and to have positive sputum smear test results (OR 2.20).
Of the 55 MDR/XDR TB cases, 17 (31%) were XDR TB. We found no signifi cant differences between patients with XDR TB and patients with MDR but not XDR (non-XDR MDR) TB. XDR TB patients tended to be female, although the difference was not statistically signifi cant (p = 0.06, Fisher exact test).
The proportion of the Beijing family, which is frequently reported to be associated with drug resistance (14), did not signifi cantly differ between the non-XDR MDR M. tuberculosis strains and XDR M. tuberculosis strains (68% vs. 47%; p = 0.14), and distribution of the other spoligotype families did not differ signifi cantly (data not shown).

Cluster Analysis by IS6110-RFLP
All 55 MDR/XDR strains were genotyped by IS6110-RFLP, and 21 (38%) were classifi ed into 9 clusters, each with identical RFLP patterns (clusters 1-9) (online Appendix Figure, www.cdc.gov/EID/content/16/6/948-appF. htm). All members of each cluster belonged to the same spoligotype family. The remaining 34 strains exhibited unique RFLP patterns. Cluster size ranged from 2 to 4, and 7 clusters had 2 members, 1 with 3 members and 1 with 4 members. IS6110 copy numbers ranged from 1 to 18. Four (7%) strains had <5 copies of IS6110, and these strains were discriminated as unique strains by the subsequent VNTR analysis. Of the 21 RFLP-clustered strains, 13 (62%) were classifi ed as XDR M. tuberculosis, and 8 (38%) were isolated from new TB patients. Of the 17 XDR M. tuberculosis strains, 4 were resistant to all drugs tested, and 2 belonged to cluster 8. Among the 9 RFLPclusters, 7 were from Japanese patients exclusively; the other 2 clusters were from both Japanese and foreign-born patients.

Cluster Analysis by VNTR
The results of the 19-locus VNTR (15 mycobacterial interspersed repetitive unit-VNTR + additional 4 loci) analysis showed that 7 of the 9 RFLP-clusters were identical according to the 19-locus VNTR (online Appendix Figure). A difference in only 1 locus was observed in the remaining 2 RFLP-clusters: 1 at VNTR1955 in cluster 6, and 1 at VNTR2163b in cluster 9 (online Appendix Figure).

Geographic Distribution of Hospitals with MDR/XDR TB Patients among Each Cluster
To estimate the possible clonal expansion of MDR/ XDR TB in communities, we compared the geographic areas of patients' hospitals among each cluster ( Table 3). The 55 MDR/XDR TB patients were treated by 23 hospitals, which were located in 16 of the 47 prefectures in Japan. Of the 9 clusters, 7 consisted of patients whose hospitals were located in the same or neighboring prefectures, and the remaining 2 clusters consisted of patients whose hos- pitals were located in the same region (i.e., a geographic and administrative area with several prefectures). Clusters with geographic links suggest the clonal expansion of MDR/XDR TB occurred in local areas. Although patients in clusters 4 and 7 were from the same hospital, further information about their contact was unavailable because an epidemiologic survey was not performed.

Characteristics of Clustered Cases
We analyzed the associations of genetic clustering by IS6110-RFLP with patients' characteristics and bacteriologic factors (Table 4). XDR TB occurred more frequently among the clustered cases than among the unique cases (OR 7.73), but differences between the 2 groups were not signifi cant for any of the remaining factors i.e., age, sex, nationality, treatment history, site of TB disease, chest radiographic fi ndings, sputum-smear test results, underlying disease, or member of the Beijing family genotype.

Discussion
Our study described the transmission dynamics of MDR/XDR TB in Japan on a national scale. Analysis of the 55 MDR/XDR TB cases showed that MDR/XDR TB was more frequent among younger patients, previously treated patients, and foreign-born patients than among patients with drug-susceptible TB (Table 1). In addition, the XDR TB cases, which represented 31% of the total MDR TB cases, were more likely to be associated with clustering than were the non-XDR MDR TB cases ( Table 4), suggesting that ongoing transmission plays a critical role in new cases of XDR TB. We also found that the Beijing family genotype predominated in MDR/XDR M. tuberculosis (Table 2) and in drug-susceptible M. tuberculosis in this setting (12).
Although IS6110-RFLP analysis has been the standard for strain typing in studies of M. tuberculosis transmission, false clusters comprising strains without any epidemiologic link, and thus not refl ecting clonal transmission, have been reported (15)(16)(17). This use of IS6110-RFLP analysis is especially problematic in area in which strains with stable RFLP patterns are endemic (18,19). In this context, because of its post hoc nature, a limitation of our study is its lack of information about epidemiologic links among clustered patients.
We therefore evaluated the genetic clonality of RFLPclustered strains more rigorously by using the 19-locus VNTR. Our previous study demonstrated that most of the RFLP-clustered strains without epidemiologic links were discriminated by the 19-locus VNTR (12). In this study, the results correlated strongly between RFLP and VNTR in terms of cluster formation; 7 of the 9 RFLP-clusters had completely identical 19-locus VNTR profi les (online Appendix Figure). Each of the remaining 2 clusters contained a single-locus variant, i.e., the difference was at only 1 locus of 1 strain in each cluster. This minor difference in VNTR profi le is highly likely to have occurred as a random variation in strains from the same source, as argued by several investigators (13,20). In addition to analyzing by RFLP and VNTR, we confi rmed that all of the 9 clusters shared identical mutations on drug resistance genes for rifampin and isoniazid (i.e., rpoB and katG, respectively, data not shown).
Geographic distribution of the hospitals also supports the clonal expansion of MDR/XDR TB ( Table 3). Isolation of most of the clustered strains from hospitals in the same or neighboring prefectures may indicate that transmission is occurring in the communities. Furthermore, we assume that some of the clustering in our study did not result from direct transmission among the members but rather resulted from exposure of the members to a common infection source or infection with different sources sharing a near ancestor. The discordance of drug resistance other than isoniazid and rifampin resistance among clusters implies that the stepwise acquisition of drug resistance had occurred chronologically during successive transmissions (online Appendix Figure). Thus, all these fi ndings support the assumption of ongoing community transmissions of MDR/XDR TB.
A high proportion (71%, 12/17) of the XDR strains were involved in clusters, a fi nding consistent with the results of a hospital-based study in Osaka, Japan (21). Of the 12 clustered cases, 4 were new cases and 8 were among previously treated patients. The clustered XDR TB cases among the new cases strongly indicated that these persons had been primarily infected with XDR M. tuberculosis strains. Also, TB may have developed in the clustered XDR TB patients with previous TB treatment as a result of exogenous reinfection with XDR TB, as described in a report of fatal TB in a patient infected with an MDR strain when his disease had been almost cured (22). One possible explanation for the high clustering rate of XDR TB is that new cases of XDR TB are more likely to be caused by transmission than by acquisition of multiple drug resistance as a result of treatment failure. Shah et al. reported that patients with XDR TB were more likely than those without XDR TB to be infectious in terms of duration and proportion of sputum smear positivity (3). Furthermore, at least 142 non-XDR MDR TB and 180 XDR TB patients were reported to be in Japan as of 2000 (23). All were culture positive, and a considerable number were treated as outpatients despite their infectiousness and drug resistance. Thus, we assume that some of those patients with chronic MDR/XDR TB may have played a role as a source of transmission, as described in this study.
Because these fi ndings and our cluster analysis suggest that the current TB control strategy is insuffi cient to prevent community transmission of MDR/XDR TB, more detailed investigations of MDR/XDR TB transmission based on contact tracing are urgently needed to improve the infection control strategy, including an isolation policy for highly infectious patients with refractory drug-resistant strains. At the same time, ethical issues, such as the human rights of these patients, should be carefully considered.
Only a few cases of XDR TB transmission have ever been reported. A large group of XDR TB cases, mainly among HIV-infected patients, was reported in South Af-rica (2). Another study in South Africa reported 12 cases of exogenous reinfection with XDR TB (24). In Iran, DNA fi ngerprinting analysis suggested 2 outbreaks of XDR TB involving 12 patients in 1 family and their close contacts (25). In Norway, a patient who was lost to follow-up has been transmitting an XDR M. tuberculosis strain for 12 years, and that transmission has ultimately resulted in 15 XDR TB cases (26). Consistent with these previous reports, our study has added novel evidence for clonal expansion of XDR M. tuberculosis strains even in an industrialized country with intermediate TB incidence and low incidence of HIV.
Another limitation of this study is a possible sampling bias, which could be caused by the following factors. First, the sampling of the survey participants was not randomized but was based on voluntary participation of the hospitals, which may be likely to include more serious TB cases (7). Second, the sampling fraction is low (38%) and the study period is short (6 months), either of which may produce a biased subset of the total cases. In addition, the low sampling fraction and the short study period may lead to reduced clustering of cases (27,28). A more complete understanding of transmission dynamics of MDR/XDR TB requires a real-time DNA fi ngerprinting method such as VNTR on a nationwide scale.
Four strains from the 55 MDR/XDR TB cases were identifi ed as totally drug-resistant strains (29), indicating that they were resistant to all 10 drugs tested (online Appendix Figure). Of these strains, 2 in 1 cluster (cluster 8) were from new cases, and both patients were women in their 20s who were full-time workers and had no underlying disease.  Both patients were registered in the same area (no further data available); TB as formidable as totally drug-resistant TB can affect otherwise healthy young adults even on a mass basis. The results of this study showed that transmission of MDR TB, especially XDR TB, is currently occurring in communities of Japan. Further studies to prospectively identify the transmission route through contact tracing and real-time DNA fi ngerprinting methods on a population basis are required. Although the MDR/XDR TB problem is not great in Japan, our fi ndings highlight the relevance of proper infection control, as well as effective treatment, to further contain highly developed drug-resistant TB.