Summary
Background
Multidrug-resistant tuberculosis (MDR-TB) is a significant threat to tuberculosis elimination worldwide. Understanding the transmission pattern is crucial for its control. We used a genomic epidemiological approach to assess the recent transmission of MDR-TB and potential risk factors for transmission.
Methods
In a population-based retrospective study, we performed variable-number-of-tandem-repeat (VNTR) genotyping, followed by whole-genome sequencing (WGS) of isolates from all MDR-TB patients in Shanghai, China, 2009-2012. We measured strain diversity within and between genomically clustered patients. Genomic and epidemiologic data were combined to construct transmission networks.
Findings
367 (5%) of 7982 patients with tuberculosis had MDR tuberculosis and 324 (88%) of these had isolates available for genomic analysis. 103 (32%) of the 324 MDR strains were in 38 genomic clusters that differed by 12 or fewer single nucleotide polymorphisms (SNPs), indicating recent transmission of MDR strains. Patients who had delayed diagnosis or were older than 45 years had high risk of recent transmission. 235 (73%) patients with MDR tuberculosis probably had transmission of MDR strains. Transmission network analysis showed that 33 (87%) of the 38 clusters accumulated additional drug-resistance mutations through emergence or fixation of mutations during transmission. 68 (66%) of 103 clustered MDR strains had compensatory mutations of rifampicin resistance.
Interpretation
Recent transmission of MDR strains, with increasing drug-resistance, helps drive the MDR-TB epidemic in Shanghai, China. WGS provides a measure of the heterogeneity of drug-resistant mutations within and between hosts and enhances our ability to determine the transmission patterns of MDR-TB.
Funding
National Science and Technology Major Project, National Natural Science Foundation of China, and US National Insitutes of Health.
Keywords: Tuberculosis, multidrug-resistance, whole-genome sequencing, recent transmission, China
Introduction
The worldwide emergence of multidrug-resistance tuberculosis (MDR-TB) threatens the global eradication of tuberculosis. An estimated 480 000 cases of MDR tuberculosis occurred worldwide in 2015, but only one in ten cases were diagnosed, treated, and cured.1 Despite the acquisition of MDR-TB during treatment, person-to-person transmission of MDR strains also occurs and can be fueled by delayed diagnosis, prolonged treatment, and unfavorable treatment outcomes of MDR-TB patients. Therefore, understanding the cause and transmission patterns of MDR-TB will be crucial to inform effective public health actions to reduce MDR-TB.2
Molecular epidemiological methods that combine epidemiological investigations and genotyping of Mycobacterium tuberculosis strains provide the means to assess the recent transmission (within 2–3 years) of M tuberculosis and risk factors for transmission.3 However, traditional genotyping methods have limited discriminatory power and are limited in assessing homoplasy, which reduces their accuracy in identifying recent transmissions of M tuberculosis.4 High-throughput whole-genome sequencing provides increased resolution and accuracy over older methods, and is a powerful tool to study the transmission of M tuberculosis.4–9 Furthermore, phylogenetic networks based on whole-genome sequencing can be used to identify putative source cases, super-spreaders, and transmission directions in the absence of, or complementary to, extensive epidemiological data.8,9 However, few studies have applied whole-genome sequencing to address recent transmission of MDR tuberculosis at the population level.5,31
China has a high prevalence of drug-resistant tuberculosis and the second largest number of MDR cases worldwide.1 However, the high risk of MDR-TB among patients with previous treatment history has resulted in a common brief that most MDR-TB cases were due to the acquisition of resistance, and led to the allocation of resources to improve treatment. Although previous studies suggested that many MDR-TB cases resulted from the transmission of MDR strains,10-13 few studies present data that provide direct evidence of MDR-TB transmission at the population level. Thus, we hypothesized that the transmission of MDR strains plays a major role in the high prevalence of MDR-TB in China. To test this hypothesis, we conducted a population-based retrospective study in Shanghai, China, using WGS and epidemiological investigations. We quantified the magnitude of MDR-TB arising from the transmission of MDR strains, tracked the transmission patterns of MDR-TB, and identified the risk factors of transmitted MDR-TB in a major city in China.
Methods
Study population
Shanghai is the most populous city in China with an estimated population of 24 million, and has a relatively well-functioning tuberculosis control program. All suspected cases of pulmonary tuberculosis are referred to local designated hospitals, where the diagnosis is made by sputum smear and culture. Through a routine surveillance system, all of the tuberculosis cases were reported to Shanghai Municipal Centre for Disease Control and Prevention (Shanghai CDC). This study included all patients ≥15 years old with culture-confirmed MDR pulmonary tuberculosis reported by all local designated hospitals in Shanghai during January 1, 2009 through December 31, 2012. The institutional review boards of Shanghai CDC and the Institute of Biomedical Sciences, Fudan University, approved the study. Written informed consent was obtained before each of the epidemiologic investigations.
Procedures
All clinical isolates were collected at diagnosis prior to treatment initiation and were sent to the Tuberculosis Reference Laboratory in Shanghai CDC for drug susceptibility testing (DST). We used the proportion method on Löwenstein-Jensen (L-J) medium for four first-line drugs (isoniazid, rifampin, ethambutol, and streptomycin). Pre-extensively drug resistant (pre-XDR) and XDR Mtb genotypes were defined based on the presence of genomic mutations that are associated with drug resistance, as reported in the scientific literatures (Supplementary Table 1).14
Genomic DNA was extracted from cultures of single sputum specimens (one per patient) following a previously published method.15 Beijing strains were identified by RD105 deletion-targeted multiplex PCR-based assay and were confirmed by use of sequencing to identify specific single nucleotide polymorphisms (SNPs).16 We first performed variable-number-of-tandem-repeat (VNTR) genotyping for all of the strains using the 9+3 loci set as previously described,17 and then integrated WGS to determine the genomic relatedness among the MDR strains that shared identical VNTR genotype patterns. Strains with cross-contamination were excluded as previously described.15
The genomic DNAs were sequenced using Illumina Hiseq 2000 with an expected coverage of 100×. Paired-end reads were mapped to the reference genome H37Rv (GenBank AL123456) with Bowtie2. The SAMtools/BCFtools suite was used for calling fixed SNPs (frequency ≥ 95%).18 SNPs were called at loci where the alternative basecalls were supported by at least five reads without strand bias (both strand were mapped by the reads). The fixed SNPs, excluding SNPs in the PE/PPE, PE-PGRS and drug-resistance associated genes, were combined into a concatenated alignment (Appendix data 1) that was used to construct a maximum-likelihood (M-L) phylogeny by MEGA (version 5·0, Tempe, USA). The sequences were also used to generate Median-joining (M-J) networks for each cluster with NETWORK. VarScan2 software was used to calling unfixed SNPs (frequency between 5% and 95%) in drug-resistant (DR) genes, which represent the emergence of DR mutations within individual patients.19 The sequencing data were deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (Accession number SRP058221, Appendix data 2).
Clinical, demographic, and epidemiological records for each MDR-TB patient were obtained during routine surveillance. A diagnostic delay was defined as the time that elapsed between the onset of symptoms and the date of diagnosis. For each MDR-TB patient that was in a genotypic cluster, we used a standardized questionnaire to obtain information about their social characteristics, their history of exposures to contacts with active TB disease, and the locations they frequented where transmission could have occurred.
Patients with epidemiological links were defined as: confirmed, if patients knew each other or lived at the same address or complex; probable, if patients did not know each other but shared locations where transmission likely occurred, including in a neighborhood complex or street in the same district; and without links if patients did not know each other and lacked a common neighborhood, location or setting. Putative transmission networks were constructed based on the structure of the genomic phylogeny and the epidemiologic links.
We used the chi-square test of proportions and the Wilcoxon non-parametric rank sum test to compare covariates between groups. We used univariate and multivariable logistic regression analysis to calculate the odds ratios (ORs) and 95% confidence intervals (CI) for the risk factors that were associated with genomic clusters. A forward, step-wise approach was used to add covariates to the multivariable model. Statistical analyses were performed using Stata (version 13·1/SE, Stata Corporation, College Station, Texas, USA).
Role of the funding source
The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
Results
During the four-year study period, 4·6% (367/7982) of all of the culture-positive tuberculosis cases in the study population had MDR-TB and 93·5% (343/367) had an isolate with genomic DNA available for analysis (Supplementary Figure 1). Of 343 cases with genotyping profiles, 19 were excluded for suspicion of laboratory contamination. The remaining 324 patients were used for the main analyses. Among these patients, 59·0% (191/324) were treatment-naïve cases (Figure 1), 73·4% (238/324) were male and the median age was 39 years old (range from 16-88 years).
Figure 1.
Classification of multidrug-resistant tuberculosis based on treatment history and genomic analysis. PTB, pulmonary tuberculosis; DST, drug susceptibility testing; MDR-TB, multidrug-resistant tuberculosis.
By VNTR genotyping analysis, 38·6% (125/324) of the MDR-TB patients had isolates with genotypic patterns that were identical to at least one other patient (Supplementary Figure 1). Useable DNA was obtained and whole-genome sequencing was performed for 122 of the 125 VNTR clustered strains. After excluding SNPs in highly repetitive regions and the mutations associated with drug-resistance, the smallest genomic differences between any two strains that shared identical genotypes ranged from 0 to 109 SNPs (Figure 2). A maximum-likelihood phylogeny tree was then constructed based on these SNPs (Figure 3). Beijing family strains accounted for 95·1% (116/122) of the sequenced strains and the predominant sub-lineage was of “modern” Beijing strains (67·2%, 78/116).
Figure 2.
Distribution of the number of single nucleotide polymorphisms (SNPs) in isolates from the closest multidrug-resistant tuberculosis cases within a cluster. Epidemiological links were indicated (Black).
Figure 3.
Maximum-likelihood whole-genome SNP-based phylogeny of multidrug-resistant isolates and mutations in RNA polymerase subunits rpoA, rpoB, and rpoC. Branches with colours indicate the sublineages: non-Beijing strains (yellow), “Ancient” Beijing strain (ice blue), “Modern” Beijing strain (sky blue);16 Square beside the nodes show the presence of pre-extensively drug resistant (pre-XDR, purple) and extensively drug resistant (XDR, orange) genotypes among genomically clustered MDR-TB strains. The amino-acid change caused by mutation of rpoABC is annotated along the periphery of the phylogeny.
Epidemiological investigations were conducted with the 122 MDR-TB cases, except 10 could not be completed due to death or lost follow-up. Confirmed or probable epidemiological links were identified in 57·1% (64/112) MDR patients (Table 1). None of the MDR-TB patients with epidemiological links had strains that were separated by more than 12 SNPs (Figure 2). Therefore, we defined a genomic cluster in this study as a group of strains that differed by no more than 12 SNPs, similar to the threshold used to define or rule out a transmission event in other studies.6,8,20 Thus, 31·8% (103/324) strains were in 38 different genomic clusters with a size ranged from two (26 clusters) to eight patients (one cluster), indicating the recent transmission of MDR strains. MDR-TB among treatment-naïve patients indicated the transmission of MDR strains. Thus, if we combine the MDR-TB cases among treatment-naïve cases and in genomically clustered cases, up to 72·5% (235/324, 95% CI 67·3%-77·3%) of MDR-TB patients were likely caused by transmission of MDR strains (Figure 1).
Table 1.
Characteristics of multidrug-resistant tuberculosis genomic clusters based on whole-genome sequencing analysis (n=103)
| Number of cases in cluster | Median age (years [IQR]) | Number of patients living in the same district | Number of patients with positive sputum smear result | Number of new cases | Treatment outcome (number of patients) | Known risk factors (number of patients) | Epidemiologically linked | Nature of epidemiological link [number of patients, n=64] | |
|---|---|---|---|---|---|---|---|---|---|
| Cluster 33 | 8 | 49 (29-52) | 3, 4* | 6 | 1 | Cured (5),died (1),default (1),unknown (1) | Retreatment (7); game room (3) | Yes | Social (same street [2], Neighborhood street [2] and community game room [3]) |
| Cluster 35 | 7 | 59 (47-61) | 6 | 6 | 5 | Cured or completed (3), died (2), failed(1), unknown (1) | Hospitalization patient (5) | Yes | Nosocomial (Health center [2]), Social (Resident community [5]) |
| Cluster 09† | 7 | 53 (51-58) | 7 | 7 | 5 | Cured (1), died (3), default (1), unknown (2) | Drug misuse (2); game room (5) | Yes | Household [2], social (game room [4]), drug user [1]). All of them lived in the same residential complex. |
| Cluster 21 | 5 | 63 (48-65) | 4 | 4 | 1 | Cured (2), default (1), unknown (1), died (1) | ND | Probable | Social (Neighborhood street [5]) |
| Cluster 08 | 3 | 30 (27-53) | 2 | 3 | 2 | Cured (2), failed (1) | ND | Yes | Social (Community, Same Street [2]) |
| Cluster 22 | 3 | 37 (25-53) | 3 | 3 | 3 | Cured or completed (2), failed (1) | HIV infection (1) | Yes | Social (Community, Same Street [2]) |
| Cluster 23 | 3 | 64 (54-66) | 0 | 3 | 3 | Cured (1), failed (2) | ND | Unknown | Unknown |
| Cluster 25.2 | 3 | 37 (27-38) | 3 | 3 | 2 | Cured (3) | ND | Yes | Social (Same Street [3]) |
| Cluster 31 | 3 | 30 (27-37) | 3 | 3 | 3 | Cured (1), default (1), unknown (1) | ND | Probable | Social (Community, former residence place [2]) |
| Cluster 32 | 3 | 28 (25-50) | 2 | 3 | 2 | Cured (2), failed (1) | ND | Yes | Social (Community, known to each other [2]) |
| Cluster 37 | 3 | ND | 2 | ND | 2 | Cured (1), default (1), unknown (1) | ND | Yes | Social (Community, residential place [2]) |
| Cluster 38 | 3 | 61 (50-78) | 3 | 3 | 3 | Cured (3) | Old age (3) | Probable | Social (Community, food market [3]) |
| Other clusters‡ | 52 | 43 (27-61) | ND | 43 | 28 | Cured or completed (32), died (3), default (6), unknown (11) | ND | ND | Household [2], social (residential complex [8], community [4], neighborhood Street [8]) |
Four cases were from one district, and three cases were from another neighbour district.
Includes two variable-number tandem repeat clusters.
Patients in clusters with two cases of multidrug-resistant tuberculosis. Cured was defined as a negative sputum smear result at the end of treatment and on one other occasion among patients who had previously had a positive sputum smear result.
ND=not determined.
The genomic clusters may represent transmission of a MDR strain, or initial transmission of a non-MDR strain that developed resistance afterwards. To validate the transmission of a MDR strain, we constructed the M-J network of each genomic cluster and mapped drug-resistance mutations onto the M-J network (Figure 4 and Supplementary Figure 2). 89·5% (34/38) of the genomic clusters had mutations of isoniazid and rifampin that were consistent among all of the strains within a cluster, confirming the recent transmission of MDR strains rather than acquired resistance. Many other drug-resistance mutations were also consistent within all of the strains in these 34 clusters, including 16 clusters that contained strains with shared rifampin-resistance compensatory mutations (Supplementary Figure 2). Taken together, 95 of the 103 genomically clustered MDR-TB cases resulted from recent transmission of MDR strains.
Figure 4.
Transmission networks of MDR-TB based on genetic distances and drug-resistance mutations. (A) The Median-joining (M-J) networks for four main MDR-TB clusters. For each network, the arrow indicates the root and the circles represent M. tuberculosis (Mtb) isolates (numbered according to patient identification). Mtb isolates were separated by lines with length (as suggested by the dots) representing genetic distance. Isolates with identical genomes were grouped in the same circle. The color of each circle represents multidrug-resistant (blue), pre-XDR (yellow), and XDR (pink) tuberculosis patients. The colored shapes on the branches or within the circles represent resistance mutations. (B) Matrix of resistance mutations for the clusters in panel A. Fixed mutations were highlighted with colors, heterogeneous mutations (simultaneous existence of mutant and wild-type alleles) have a white background and wild-type alleles are indicated as dots in a light-blue background. The heterogeneous mutations were mapped into corresponding circles in the M-J networks to indicate emerging resistance. The fixed mutations were mapped onto the branches by assuming parsimonious acquisition with no reversion. Different shapes of the same color were used to indicate different mutations for the same resistance phenotype within each network. Abbreviations: TB, tuberculosis; MDR, multi drug-resistant; XDR, extensively drug-resistant; INH, isoniazid; RIF, rifampin; SM, streptomycin; EMB, ethambutol; PZA, pyrazinamide; FLQ, fluoroquinolones; AMI, amikacin; RIFc=RIF compensatory.
On the basis of epidemiological investigations, 64 (69%) of the 93 genomically clustered cases of MDR tuberculosis had identifiable epidemiological links, whereas 29 (31%) cases were not epidemiologically linked. The most common links were living in the same residential complex or community, or on the same neighborhood street, or using shared public facilities such as food markets (70·3%, 45/64). However, household links were identified in only four patients. Table 2 shows the results of the univariate analysis of the risk factors associated with genomic clustering of MDR-TB. In the multivariable model having a diagnostic delay of at least two months (adjusted OR, [aOR] 2·02, 95% CI 1·08-3·81), and being 45-64 years old (aOR 1·90, 95% CI 1·04-3.46) or ≥ 65 years old (aOR 2·94, 95% CI 1·28-6.79) were independently associated with genomically clusters of MDR-TB.
Table 2.
Univariable analysis of risk factors associated with multidrug-resistance in genomic clusters (n=299)*
| Characteristics | Genomic clusters by whole-genome
sequencing |
Odds ratio (95% CI)† | P value | ||||
|---|---|---|---|---|---|---|---|
| Non-cluster (n=203) | Cluster (n=96) | ||||||
| Sex | |||||||
| Male | 65 | (32.0) | 21 | (21.9) | 1.00 | ||
| Female | 138 | (68.0) | 75 | (78.1) | 1.68 | (0.95-2.97) | 0.07 |
| Age, years | |||||||
| 15-34 | 95 | (46.8) | 28 | (29.2) | 1.00 | ||
| 35-44 | 32 | (15.8) | 15 | (15.6) | 1.59 | (0.75-3.37) | 0.18 |
| 45-64 | 60 | (29.5) | 38 | (39.6) | 2.15 | (1.18-3.90) | 0.009 |
| ≥ 65 | 16 | (7.9) | 15 | (15.6) | 3.18 | (1.36-7.41) | 0.004 |
| Previous treatment | |||||||
| No | 106 | (52.2) | 55 | (57.3) | 1.00 | ||
| Yes | 97 | (47.8) | 41 | (42.7) | 0.81 | (0.49-1.33) | 0.41 |
| Diagnostic delay | |||||||
| < Two months | 175 | (86.2) | 70 | (72.9) | 1.00 | ||
| ≥ Two months | 28 | (13.8) | 26 | (27.1) | 2.20 | (1.19-4.07) | 0.005 |
| Sputum smear positive | |||||||
| No | 37 | (18.2) | 16 | (16.7) | 1.00 | ||
| Yes | 164 | (80.8) | 80 | (83.3) | 1.13 | (0.59-2.15) | 0.58 |
| Unknown | 2 | (1.0) | 0 | (0.0) | - | - | - |
| Treatment outcomes | |||||||
| Cured/Treatment completed | 119 | (58.6) | 58 | (60.4) | 1.00 | ||
| Treatment Failure | 11 | (5.4) | 6 | (6.2) | 1.11 | (0.39-3.18) | 0.83 |
| Default/Moved/Lost | 34 | (16.8) | 11 | (11.5) | 0.66 | (0.31-1.41) | 0.28 |
| Died | 21 | (10.3) | 10 | (10.4) | 0.98 | (0.43-2.21) | 0.96 |
| Still on treatment/Unknown | 18 | (8.9) | 11 | (11.5) | 1.25 | (0.55-2.83) | 0.58 |
| Beijing strains | |||||||
| No | 27 | (13.3) | 6 | (6.3) | 1.00 | ||
| Yes | 176 | (86.7) | 90 | (93.7) | 2.30 | (0.91-5.81) | 0.06 |
| Cavitary disease‡ (n=296) | |||||||
| No | 97/201 | (48.3) | 53/95 | (55.8) | 1.00 | ||
| Yes | 104/201 | (51.7) | 42/95 | (44.2) | 0.74 | (0.45-1.21) | 0.22 |
Data are n (%) unless otherwise specifi ed. OR=odds ratio.
Data were available for 299 patients.
Comparison between genomically clustered and non-clustered multidrug-resistant tuberculosis cases.
Cavitary disease was based on the presence or absence of a cavity on the chest radiograph at the time of diagnosis.
Our previous study indicated there is considerable heterogeneity (measured as unfixed SNPs) in drug-resistance genes within patients during treatment.19 In the present study, we identified unfixed SNPs in DR genes in 37 (35·9%) of the 103 genomically clustered MDR strains, suggesting the emergence of DR mutations within the host. By mapping these mutations onto the putative transmission networks, eight clusters showed between-patient fixation of these newly emerged DR mutations, evidence of the direction of the transmission chain (Figure 4 and Supplementary Figure 2, Appendix data 3). More importantly, 86·8% (33/38) of the MDR-TB clusters accumulated additional drug-resistance mutations along the transmission chain. 43 (41·7%) of the 103 genomically clustered MDR strains were pre-XDR and 11 (10·7%) were XDR (Figure 3 and Supplementary Figure 2).
One cluster, C-09, represents both the emergence and fixation of DR mutations during transmission (Figure 5). There were two putative index cases, a husband (P1614) and wife (P0659) who were first diagnosed with MDR-TB in March 2006, prior to the present study. Both of the putative index cases were non-compliant with their treatment and likely transmitted Mtb to all of the other individuals in this cluster after the putative index cases moved to this residential complex in 2008. However, WGS-based analysis showed that there was a sub-index case diagnosed in 2010 (P2010), based on the within-host emergence of a pyrazinamide related resistant mutation (pncA p·Val155Gly) and one ethambutol related resistant mutation (embB p·Met306Val) in this patient; the two mutations were fixed during the subsequent infection of three hosts. The MDR strain within patient P2010 likely evolved into at least two clonal subpopulations and one of the clonal subpopulations was transmitted to the subsequent hosts, with increased drug-resistance. These transmission events likely occurred in the game room of the residential complex.
Figure 5.
Results of the detailed epidemiologic investigation and genomic analysis of cluster C-09. (A) Putative transmission network based on epidemiological links. (B) Time of onset of symptoms, diagnosis, and treatment. Based on Walker, T.M., et al.8 (C) Median-joining (M-J) network and drug-resistance mutations of clusters. Each circle represents clustered patients who were infected with strains separated by no SNPs; M. tuberculosis (Mtb) isolates were separated by lines with length (as suggested by the dots) representing genetic distance. The color of each circle represents multidrug-resistant (blue), pre-XDR (yellow) and XDR (pink) tuberculosis patients. The heterogeneous mutations were mapped into corresponding circles in the M-J networks to indicate emerging resistance. The fixed mutations were mapped onto the branches by assuming parsimonious acquisition with no reversion. Different shapes of the same color were used to indicate different mutations for the same resistance phenotype within each network.
Fitness reduction due to drug-resistant mutations can be restored through compensatory evolution, which may be associated with the spread of MDR strains.21 Mutations in rpoA, rpoC and regions outside the rifampin-resistance determining region (RRDR) were associate with compensation for rifampin resistance.22 We first investigated the occurrence of compensatory mutations in rpoA and rpoC genes in strains carrying rifampin resistance mutations. 37 different nonsynonymous SNPs in rpoA (four SNPs) and rpoC (33 SNPs) were identified (Supplementary Table 2 and Supplementary Figure 3). Most of the nonsynonymous SNPs (73·0%, 27/37) were in strains containing the rpoB p·Ser450Leu substitution, and they occurred more frequently in MDR strains with the rpoB p·Ser450Leu substitution mutation than in MDR strains without this substitution (64/73 versus 15/49, p<0·001, Supplementary Table 2). We identified an additional 21 nonsynonymous SNPs in the regions that outside the RRDR of rpoB gene. The 21 nonsynonymous SNPs were significantly more likely to occur in strains without alternative compensatory mutations in rpoA and rpoC (20/58 versus 4/64, p<0·001, Figure 3 and Supplementary Figure 3). Overall, 66·0% (68/103, 95% CI, 56·0%-75·0%) of the genomically clustered MDR strains harbored compensatory mutations in rpoABC genes and the proportion reached 86·3% (63/73, 95% CI, 76·2%-93·2%) in strains with the rpoB p·Ser450Leu resistance mutation (Figure 3 and Appendix data 4).
Discussion
Our strategy combined traditional genotyping, whole genome sequencing, and epidemiological investigation. Transmission of MDR strains accounted for most (73%) cases of MDR tuberculosis overall. Residential communities or complexes, and related public facilities were the most common transmission settings. Whole genome sequencing analyses were highly informative to determine the transmission patterns of MDR strains and the ongoing acquisition of additional antibiotic resistance during transmission.
About one third of the MDR-TB cases in our study population were attributable to recent transmission, a higher percentage than that reported in low-burden countries such as the United States or the United Kingdom.10,12 In a multi-setting study in China, 43% of the MDR-TB cases were in genotypic clusters and MDR was an independent risk factor for the recent transmission of Mtb strains.15 Combining the number of treatment naïve MDR-TB cases in our study population, 72% of the MDR-TB cases were due to the transmission of MDR-TB strains, rather than acquisition of resistance. However, we may still underestimate the real burden of MDR-TB transmission. Patients with genomically unique strains could also have disease due to reinfection with MDR strains. For example, our previous study demonstrated that 84% of the drug-resistant tuberculosis cases with a previous treatment history were attributed to exogenous reinfection with a new drug-resistant strain.23
The transmission of MDR-TB in other areas in China could be much more serious. The national survey of drug resistant tuberculosis in China in 2008 suggested that more than 80% of MDR-TB were new cases resulted from the transmission of MDR strains.24 The rural areas in China have limited or no resources to perform culture and DST, WHO estimated that less than 10% of MDR-TB in China were diagnosed or treated.1 Since most MDR-TB cases are undetected or inappropriately treated, the transmission from their strains will result in a much more serious MDR-TB epidemic.
WGS currently offers several advantages over traditional genotyping methods to differentiate clinical Mtb isolates, and allows for effective studies of the transmission dynamics of Mtb strains.6,8,20,25-27 In this study, WGS provides more definitive evidence of the emergence and fixation of drug-resistance mutations along the transmission chains. Inappropriate therapies could accelerate the selection of new mutations and hence increased drug resistance that can be transmitted to others. Unfortunately, inappropriate treatment of MDR-TB has been a serious problem in China, particularly in hospitals.24,28 Therefore, treating patients with MDR tuberculosis on the basis of results of drug susceptibility testing should become a mandatory policy.
Notably, most of the clustered MDR strains in our study harbored rpoABC compensatory mutations that potentially reduce the fitness cost associated with drug resistant mutations.5,6,22,29 The percentage of compensatory mutations detected in the present study (66%) was much higher than the percentage reported globally (20%) and in high burden regions (31%),22 but is comparable to the percentage observed in Samara, Russia (87%), where MDR-TB is already endemic.5 One study in South Africa observed that rpoC compensatory mutations were associated with recent transmission of drug-resistant tuberculosis.30 Larger, prospective studies are needed to determine the transmissibility of MDR strains with compensatory mutations.
Our study has several limitations that could lead to underestimation of the true magnitude of the transmission of MDR strains. First, due to the retrospective study design and the study time frame, it is possible that we did not capture all of the culture positive MDR-TB case and strains in the population. Second, strains could be misclassified as unique if they were in fact clustered with strains outside the study period and geographic setting. Finally, this is a retrospective study and some patients died or were lost to follow up before the epidemiologic investigation was completed. We may have missed the identification of some addtional transmission settings and epidemiological links.
In summary, we used a combined genomic epidemiology approach to assess the transmission patterns of MDR-TB in the population level. Transmission of MDR strains plays a significant role in the burden of MDR-TB in China. Furthermore, most strains develop increased drug resistance during the expansion of MDR-TB clusters. Interventions, such as early detection of MDR-TB cases, infection control and evidence-based treatment of MDR-TB, are urgently needed to stem the epidemic of MDR-TB in China.
Research in context
Evidence before the study
We searched the PubMed with the search terms “tuberculosis”, “multidrug resistance”, “whole genome sequencing”, and “transmission” for molecular epidemiology study of Mycobacterium tuberculosis that used whole genome sequencing (WGS) published in English before December 2015. We identified 18 studies after 2011 that used WGS analysis to investigate the transmission of M. tuberculosis. In 2011, Gardy and colleagues applied WGS-based analysis to a large MIRU-VNTR cluster of M. tuberculosis and showed that WGS had higher resolution compared to traditional genotyping.4 Walker and co-workers evaluated the potential of WGS to delineate outbreaks of M. tuberculosis and to measure the recent transmission of M. tuberculosis in the UK.8,9 In a large-scale population-based study using WGS, Guerra-Assunção and colleagues reported that most cases of tuberculosis in a high-incidence setting in Malawi were caused by just one lineage of M. tuberculosis.25 An additional study investigated the long-term evolution and endemic spread of MDR-TB in a Russia population.5 We also searched the China Knowledge Resource Integrated Database for relevant papers published in Chinese but did not identify any references.
Added value of this study
Although several studies evaluated the application of WGS to track the transmission of M. tuberculosis, few studies applied WGS to address the transmission of MDR-TB at the population level. To our knowledge, we describe the first population-based study combing genomics with detailed epidemiological data to identify the transmission pathways of MDR-TB in aregion over time in China, the country with the highest number of MDR-TB cases in the world. We aimed to measure and quantify the burden of recent transmission and to construct the transmission networks of MDR strains. We further applied WGS to integrate the analysis of resistance-associated mutations and the within- and between-host heterogeneity of drug-resistant mutations and related compensatory mutations, to delineate the transmission patterns of MDR strains and to identify specific person-to-person transmission events. We provide direct evidence of the emergence of the diversity of M. tuberculosis sub-populations within hosts and the fixation of certain sub-populations of M. tuberculosis between hosts along a chain of transmission. In the past, inappropriate treatment was considered to be the most common way of developing MDR-TB. However, our data show that most of the MDR-TB patients in our study population were due to infection with transmitted MDR strains, rather than acquired drug resistance. We also determined that in our study population, the majority of the transmission events occurred in settings such as residential communities or complexes and related public facilities. Our results could be inferred to other areas of China and other countries with a high burden of MDR-TB.
Implications of all the available evidence
Whole genome sequencing provides greater precision than traditional genotyping method to infer the recent transmission of MDR strains of M. tuberculosis. WGS is also useful to detect and use the heterogeneity of strain diversity within- and between- hosts to infer the transmission trajectory of MDR strains. Our findings also suggest that strategies and interventions to halt the ongoing transmission of MDR strains should be a priority for TB control program in China and other settings with a high burden of MDR-TB.
Supplementary Material
Acknowledgements
We thank the tuberculosis public health teams in Shanghai Municipal Center for Disease Control and Prevention for their support. We thank the reviewers and Dr. Ted Cohen from Yale University for their valuable comments on this manuscript.
Funding
This study was supported by the National Science and Technology Major Project [2013ZX10004903-006]; National Natural Science Foundation of China [81402727 to C.Y.] and Ministry of Science and Technology of China [2014DFA30340]; the International Postdoctoral Fellowship Program [20150058 to C.Y.] funded by China Postdoctoral Science Foundation, and the GloCal Health Fellowship Program sponcored by the National Insititutes of Health (NIH) Fogarty International Center/ University of California Global Health Institute [R25 TW009343 to C.Y.]; NIH grants [D43TW007887 to T.L., Q.G. and K.D., and DP2OD006452 to K.D.]; and Shanghai Municipal Commission of Health and Family Planning grants [GWTD2015S02 to X.S.].
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Contributors
C.Y., T.L., X.S., J.M., K.D. and Q.G. designed the study. C.Y., X.S., Z.W., S.L. and Z.Y. conducted the epidemiological investigations; J.W., P.X. and J.T. performed the laboratory work and generated the laboratory data. C.Y., T.L., M.G. and Q.L. performed the data analyses. C.Y., T.L., X.S., J.M., K.D. and Q.G. drafted and revised the manuscript. All coauthors reviewed and approved the final manuscript prior to submission.
Conflict of interest
We declare that we have no conflict of interest.
References
- 1.WHO [Oct 13, 2016];Global tuberculosis report 2016. http://www.who.int/tb/publications/global_report/en/
- 2.Nathanson E, Nunn P, Uplekar M, et al. MDR tuberculosis--critical steps for prevention and control. N Engl J Med. 2010;363(11):1050–8. doi: 10.1056/NEJMra0908076. [DOI] [PubMed] [Google Scholar]
- 3.Barnes PF, Cave MD. Molecular epidemiology of tuberculosis. N Engl J Med. 2003;349(12):1149–56. doi: 10.1056/NEJMra021964. [DOI] [PubMed] [Google Scholar]
- 4.Gardy JL, Johnston JC, Ho Sui SJ, et al. Whole-genome sequencing and social-network analysis of a tuberculosis outbreak. N Engl J Med. 2011;364(8):730–9. doi: 10.1056/NEJMoa1003176. [DOI] [PubMed] [Google Scholar]
- 5.Casali N, Nikolayevskyy V, Balabanova Y, et al. Evolution and transmission of drug-resistant tuberculosis in a Russian population. Nat Genet. 2014;46(3):279–86. doi: 10.1038/ng.2878. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Cohen KA, Abeel T, Manson McGuire A, et al. Evolution of Extensively Drug-Resistant Tuberculosis over Four Decades: Whole Genome Sequencing and Dating Analysis of Mycobacterium tuberculosis Isolates from KwaZulu-Natal. PLoS Med. 2015;12(9):e1001880. doi: 10.1371/journal.pmed.1001880. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Luo T, Yang C, Peng Y, et al. Whole-genome sequencing to detect recent transmission of Mycobacterium tuberculosis in settings with a high burden of tuberculosis. Tuberculosis (Edinb) 2014;94(4):434–40. doi: 10.1016/j.tube.2014.04.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Walker TM, Ip CL, Harrell RH, et al. Whole-genome sequencing to delineate Mycobacterium tuberculosis outbreaks: a retrospective observational study. Lancet Infect Dis. 2013;13(2):137–46. doi: 10.1016/S1473-3099(12)70277-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Walker TM, Lalor MK, Broda A, et al. Assessment of Mycobacterium tuberculosis transmission in Oxfordshire, UK, 2007-12, with whole pathogen genome sequences: an observational study. Lancet Respir Med. 2014;2(4):285–92. doi: 10.1016/S2213-2600(14)70027-X. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Anderson LF, Tamne S, Brown T, et al. Transmission of multidrug-resistant tuberculosis in the UK: a cross-sectional molecular and epidemiological study of clustering and contact tracing. Lancet Infect Dis. 2014;14(5):406–15. doi: 10.1016/S1473-3099(14)70022-2. [DOI] [PubMed] [Google Scholar]
- 11.Dobler CC, Korver S, Batbayar O, et al. Multidrug-Resistant Tuberculosis in Patients for Whom First-Line Treatment Failed, Mongolia, 2010-2011. Emerg Infect Dis. 2015;21(8):1451–4. doi: 10.3201/eid2108.141860. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Moonan PK, Teeter LD, Salcedo K, et al. Transmission of multidrug-resistant tuberculosis in the USA: a cross-sectional study. Lancet Infect Dis. 2013;13(9):777–84. doi: 10.1016/S1473-3099(13)70128-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Kendall EA, Fofana MO, Dowdy DW. Burden of transmitted multidrug resistance in epidemics of tuberculosis: a transmission modelling analysis. Lancet Respir Med. 2015;3(12):963–72. doi: 10.1016/S2213-2600(15)00458-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Flandrois JP, Lina G, Dumitrescu O. MUBII-TB-DB: a database of mutations associated with antibiotic resistance in Mycobacterium tuberculosis. BMC Bioinformatics. 2014;15:107. doi: 10.1186/1471-2105-15-107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Yang C, Shen X, Peng Y, et al. Transmission of Mycobacterium tuberculosis in China: A Population-Based Molecular Epidemiologic Study. Clin Infect Dis. 2015;61(2):219–27. doi: 10.1093/cid/civ255. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Luo T, Comas I, Luo D, et al. Southern East Asian origin and coexpansion of Mycobacterium tuberculosis Beijing family with Han Chinese. Proc Natl Acad Sci U S A. 2015;112(26):8136–41. doi: 10.1073/pnas.1424063112. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Luo T, Yang C, Pang Y, Zhao Y, Mei J, Gao Q. Development of a hierarchical variable-number tandem repeat typing scheme for Mycobacterium tuberculosis in China. PLoS One. 2014;9(2):e89726. doi: 10.1371/journal.pone.0089726. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Li H, Handsaker B, Wysoker A, et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics. 2009;25(16):2078–9. doi: 10.1093/bioinformatics/btp352. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Sun G, Luo T, Yang C, et al. Dynamic population changes in Mycobacterium tuberculosis during acquisition and fixation of drug resistance in patients. J Infect Dis. 2012;206(11):1724–33. doi: 10.1093/infdis/jis601. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Bryant JM, Harris SR, Parkhill J, et al. Whole-genome sequencing to establish relapse or re-infection with Mycobacterium tuberculosis: a retrospective observational study. Lancet Respir Med. 2013;1(10):786–92. doi: 10.1016/S2213-2600(13)70231-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Cohen T, Murray M. Modeling epidemics of multidrug-resistant M. tuberculosis of heterogeneous fitness. Nat Med. 2004;10(10):1117–21. doi: 10.1038/nm1110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Comas I, Borrell S, Roetzer A, et al. Whole-genome sequencing of rifampicin-resistant Mycobacterium tuberculosis strains identifies compensatory mutations in RNA polymerase genes. Nat Genet. 2012;44(1):106–10. doi: 10.1038/ng.1038. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Li X, Zhang Y, Shen X, et al. Transmission of drug-resistant tuberculosis among treated patients in Shanghai, China. J Infect Dis. 2007;195(6):864–9. doi: 10.1086/511985. [DOI] [PubMed] [Google Scholar]
- 24.Zhao Y, Xu S, Wang L, et al. National survey of drug-resistant tuberculosis in China. N Engl J Med. 2012;366(23):2161–70. doi: 10.1056/NEJMoa1108789. [DOI] [PubMed] [Google Scholar]
- 25.Guerra-Assuncao JA, Crampin AC, Houben RM, et al. Large-scale whole genome sequencing of M. tuberculosis provides insights into transmission in a high prevalence area. Elife. 2015;4:e05166. doi: 10.7554/eLife.05166. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Kato-Maeda M, Ho C, Passarelli B, et al. Use of whole genome sequencing to determine the microevolution of Mycobacterium tuberculosis during an outbreak. PLoS One. 2013;8(3):e58235. doi: 10.1371/journal.pone.0058235. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Perez-Lago L, Comas I, Navarro Y, et al. Whole genome sequencing analysis of intrapatient microevolution in Mycobacterium tuberculosis: potential impact on the inference of tuberculosis transmission. J Infect Dis. 2014;209(1):98–108. doi: 10.1093/infdis/jit439. [DOI] [PubMed] [Google Scholar]
- 28.Xue He G, van den Hof S, van der Werf MJ, et al. Inappropriate tuberculosis treatment regimens in Chinese tuberculosis hospitals. Clin Infect Dis. 2011;52(7):e153–6. doi: 10.1093/cid/cir075. [DOI] [PubMed] [Google Scholar]
- 29.Gagneux S, Long CD, Small PM, Van T, Schoolnik GK, Bohannan BJ. The competitive cost of antibiotic resistance in Mycobacterium tuberculosis. Science. 2006;312(5782):1944–6. doi: 10.1126/science.1124410. [DOI] [PubMed] [Google Scholar]
- 30.de Vos M, Muller B, Borrell S, et al. Putative compensatory mutations in the rpoC gene of rifampin-resistant Mycobacterium tuberculosis are associated with ongoing transmission. Antimicrob Agents Chemother. 2013;57(2):827–32. doi: 10.1128/AAC.01541-12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Casali N, Broda A, Harris SR, Parkill J, Drobniewski F. Whole genome sequence analysis of a large isoniazid-resistant tuberculosis outbreak in London: a retrospective observational study. PLoS Med. 2016;13:e1002137–32. doi: 10.1371/journal.pmed.1002137. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.