Analysis of Drug-Resistance Characteristics and Genetic Diversity of Multidrug-Resistant Tuberculosis Based on Whole-Genome Sequencing on the Hainan Island, China

Introduction

Tuberculosis (TB), caused by Mycobacterium tuberculosis (MTB), continues to pose a significant global public health challenge.^1–3 In 2021, there were 10.6 million new TB cases worldwide, with China accounting for 780,000 cases, making it the third highest burden country for TB globally.⁴ The World Health Organization (WHO) estimates that in 2021, there were 450,000 cases of rifampicin-resistant tuberculosis (RR-TB) globally, of which over 70% were multidrug-resistant tuberculosis (MDR-TB), characterized by resistance to both isoniazid and rifampicin simultaneously.⁴ Unfortunately, only one-third of these patients receive appropriate diagnosis and treatment. Given the increased difficulty, cost, and poorer prognosis associated with treating MDR-TB, ongoing vigilance is essential.^5,6 Furthermore, as early diagnosis and appropriate treatment can prevent the majority of TB-related deaths, the rapid and accurate identification of MDR-TB patients is crucial for guiding early clinical interventions and improving cure rates.

Since the successful completion of the WGS of MTB H37Rv and its publication in 1998,⁷ WGS has found extensive clinical applications, encompassing drug susceptibility prediction, epidemiological analysis, and transmission dynamics research.⁸ WGS enables the detection of transmission events missed by epidemiological investigations, differentiation between recurrent TB and reinfection, and more reliable classification of MTB lineages and sub-lineages.⁹ Moreover, it offers a comprehensive assessment of pathogen drug resistance profiles.^10–12

TB incidence in China displays regional disparities, with substantial variations in the degree and pattern of drug resistance across different areas.¹³ Hainan Island, comprising four provinces and fifteen counties, is home to an estimated population of 10.2 million.¹⁴ Positioned within the tropical and subtropical regions, Hainan Island experiences a tropical monsoon climate and maintains a low altitude. Separated from mainland China by the Qiongzhou Strait, the island has limited population flow compared to other provinces. In 2010, the prevalence of smear-positive TB on the island was 167 cases per 100,000 people, approximately three times the national average of 59 cases per 100,000 people.^15,16 Given the unique geographical environment of Hainan Island, it is reasonable to assume that Multidrug-resistant Mycobacterium tuberculosis (MDR-MTB) exhibits distinctive genetic characteristics in this region. However, due to the island’s lower socioeconomic development level and limited medical resources, research in this field has been scarce. Consequently, to comprehend the molecular resistance and genetic characteristics of MDR-MTB on Hainan Island, we employed WGS technology to analyze the strains isolated from patients diagnosed with MDR-TB admitted to the Second Affiliated Hospital of Hainan Medical University between July 2019 and June 2021. This study aimed to assess the traits of drug resistance-related mutations and genetic diversity on the island while providing a scientific foundation for MDR-TB prevention and control.

Materials and Methods

Study Subjects

According to the diagnostic criteria established by the National Health Commission for suspected TB patients, respiratory tract specimens such as sputum and alveolar lavage fluid were collected from individuals receiving treatment at the Second Affiliated Hospital of Hainan Medical University between July 2019 and June 2021. This hospital, designated as the Provincial Clinical Center for Tuberculosis and recognized as one of the designated TB hospitals in Hainan, is a 2100-bed general hospital that includes 92-bed open wards specifically dedicated to specialized TB treatment. The patients included in this study were permanent residents of Hainan Island. Clinical and epidemiological data pertaining to these patients were retrospectively retrieved from the hospital’s electronic medical record management system for further analysis.

The H37Rv (ATCC27294) standard sensitive strain, provided by the Tuberculosis Reference Laboratory of China CDC, was used as the reference strain. This study was approved by the Ethics Committee of the University of the Second Affiliated Hospital of Hainan Medical University, China (No. LW2023022). This study complies with the Declaration of Helsinki. It was considered exempt from informed consent because it was secondary research, for which consent is not required.

The inclusion criteria were as follows: 1) Patients who tested positive for MTB in at least two cultures, with the identification of MDR-MTB strains. Only the first strain cultured at the beginning of the study period was recorded; 2) MDR-TB patients who had been residents of Hainan Island for a minimum of 10 years.

The exclusion criteria: Patients with incomplete clinical data and those who were non-permanent residents of Hainan Island were excluded from the analysis.

Treatment history: 1) Previously treated patients was defined as irregular anti-tuberculosis treatment for ≥1 month, failure of initial treatment and recurrence of pulmonary tuberculosis. 2) New patients were defined as those receiving anti-tuberculosis treatment for the first time. The treatment history was based on the clinical diagnosis.

Experimental Methods

Isolation, Culture, and Screening of MDR-MTB

During the study period, clinical respiratory tract specimens were utilized to isolate mycobacteria, and preliminary screening for MDR-MTB strains was conducted using the P-Nitrobenzoic acid (PNB)/2-Thiophenecarboxylic acid hydrazide (TCH) identification medium and proportional drug-susceptibility test.

Isolation and Culture of MTB

Respiratory tract specimens, including 2–3 mL of sputum, alveolar lavage fluid, and other relevant samples from suspected pulmonary tuberculosis patients, were collected. These specimens were mixed with an equal volume of 4% NaOH solution through shaking to obtain pretreatment specimens. The pretreated specimens were then left at room temperature for 15 minutes. Next, 150 µL of the pretreated specimen was inoculated into two acidic Roche media (BASO Company, Zhuhai, China) slants. The covers of the media were slightly loosened, and they were incubated at 36±1°C for 24 hours with the slants facing upward. Subsequently, the media covers were tightened and placed vertically. The growth of colonies was observed on the third and seventh day of culture, primarily to assess contamination or the growth of fast-growing mycobacteria. Weekly observations were continued until the eighth week to record culture and contamination.

MDR-MTB Screening and Drug-Susceptibility Test (DST) of Anti-Tuberculosis Drugs by Roche Proportion Method

Fresh colonies on L-J medium were scraped using a disposable sterile inoculation loop and transferred to a grinding bottle, which was then tightly closed. The colonies were mixed using a vortex shaker for 0.5 to 1 minute and left to stand for a period of time. The supernatant was then drawn and added to a culture tube containing sterilized normal saline until the turbidity matched that of a MacFarland standard turbidity tube (MacFarland No.1). This process resulted in a bacterial suspension with a concentration of 1 mg/mL, which was further serially diluted to concentrations of 10⁻² and 10⁻⁴ mg/mL with sterilized saline.

Using a disposable sterile inoculation loop, 0.01 mL of the 10⁻² mg/mL and 10⁻⁴ mg/mL bacterial suspensions was uniformly inoculated into blank control medium, drug-containing media with rifampicin (RIF) at 40.0 μg/mL, isoniazid (INH) at 0.2 μg/mL, ethambutol (EMB) at 2.0 μg/mL, streptomycin (SM) at 4.0 μg/mL, ofloxacin (OFX) at 2.0 μg/mL, kanamycin (KM) at 30.0 μg/mL, capreomycin (CPM) at 40.0 μg/mL (BASO Company, Zhuhai, China), and ethionamide (ETO) at 40.0 μg/mL (Yinke Company, Zhuhai, China). Additionally, the PNB/TCH identification medium (BASO Company, Zhuhai, China) was included. The culture media were incubated at 36±1°C, and the results were observed on the third day and subsequently once a week until the fourth week. Percentage of resistance =(number of colonies grown on drug containing medium)/(number of colonies grown on control medium)×100%. If the percentage of patients with drug resistance ≥1% reported drug resistance (R), and susceptibility (S) was reported when the percentage of drug resistance was less than 1%. If the bacterial suspension at the concentration of 10–4 mg/mL grew less than 20 colonies on the control medium, the culture should be subcultured from the control medium and the experiment should be repeated.

The standard sensitive strain H37Rv (ATCC27294) should be used as the sensitive control for each batch of drug-containing medium, and all drug-containing medium should not grow and be judged as “sensitive”. Roche proportional method drug susceptibility test participated in and passed the proficiency assessment of anti-tuberculosis drug susceptibility test organized by the Tuberculosis Reference Laboratory of China CDC every year.

WGS of MTB

A single ring of bacteria was scraped using a disposable sterile inoculation ring and placed in 500 µL TE buffer for high-temperature inactivation at 80°C for 30 minutes. The inactivated strains were promptly removed and frozen at −80°C. These samples were then transported on dry ice to Tb Healthcare Biotechnology (Guangdong) Co., Ltd. (China) for high-throughput sequencing and bioinformatics analysis, which allowed the acquisition of variation information such as single nucleotide polymorphisms (SNPs) and insertions/deletions (InDels) in the target genome compared to the reference genome. The sequenced MDR-MTB strains were analyzed to determine genetic distance, lineage, and drug resistance.

Extraction and Sequencing of Genomic DNA

For the extraction of genomic DNA, 400 µL of inactivated MTB bacterial suspension was dispersed by sonication and then subjected to the cetyltrimethylammonium bromide-lysozyme (CTAB) method.¹⁷ The purity, concentration, and integrity of DNA were determined using an Agilent Fragment Analyzer 5400 (Agilent, USA).

Library Preparation

(1) Ultrasonic Fragmentation: Genomic DNA was extracted, quantified, and subjected to quality checks. Subsequently, it was fragmented into nucleic acid fragments ranging from 200 to 300 base pairs using a Covaris ultrasonic fragmentation machine.

(2) Library Construction: The interrupted DNA fragments were processed for library construction. The 5’ protruding end or the 3’ protruding end of the DNA fragment was blunted, and T4 DNA ligase was employed to ligate the connection points at both ends of the small DNA fragment. Fragment selection and purification were achieved using magnetic bead purification. PCR amplification was performed for library enrichment. To eliminate interference such as primer dimers in the reaction system, a magnetic bead purification method was utilized for further purification, resulting in a DNA library suitable for sequencing.

(3) Library Quantification: The concentration and length distribution of the libraries were determined using the Qubit Fluorometer 3.0 fluorometer and Agilent 2100, respectively.

Machine Sequencing

The qualified libraries underwent testing, and different libraries were pooled together based on the required effective concentration and target data amount. Subsequently, paired-end sequencing (PE150) was carried out on an Illumina NovaSeq 6000 machine (Illumina, USA).

Bioinformatics Analysis

Using FastQC software (version 0.11.9), reads containing adaptor sequences or exhibiting low quality were eliminated to obtain clean reads for subsequent analysis. The resulting filtered clean reads were aligned to the MTB reference genome H37Rv (NC_000962.3) using BWA (0.7.17).¹⁸ Freebayes (1.3.2) and SnpEff (4.3t) were used for mutation site identification and mutation annotation respectively. An in-house script was employed to obtain the SNP sequence and calculate the pairwise distance between each two samples. Pairwise distance was calculated using only fixed SNPs (frequency≥75%) that were not in drug-resistance associated genes or repetitive regions of the genome (eg PPE/PE-PGRS family genes, phage sequence, insertion or mobile genetic elements). Samples with MTB genomes differing by no more than 12 SNPs in cluster analysis were defined as belonging to the same genomic cluster.^2,19,20

TB Profiler software (version 2.8.12, https://github.com/jodyphelan/TB Profiler)^21,22 was employed to employed the MTB lineages and genes associated with resistance against eight anti-TB drugs (RIF, INH, SM, EMB, OFX, CPM, KM, ETO). After loading the SNP sequence in MEGA, the function “Construct/Test Neighbor-Joining Tree”under the “PHYLOGENY”tool was select to construct our phylogenetic tree.²³ The resulting phylogenetic trees were visualized and modified using the iTOL program (version 6.6) (https://itol.embl.de/).²⁴

To verify the location, the residential addresses of patients within each cluster were geocoded using Baidu Map. The spatial distance between two patients represented the distance between their respective residential locations on the map.

Statistical Analysis

The statistical analysis was conducted using SPSS 26.0 software to examine the data of 202 patients diagnosed with MDR-TB. As the age distribution and clustered patient mean paired distance data did not follow a normal distribution, they were presented as medians and quartiles. A comparison of the clustering rates between lineage 2 (East Asian) and lineage 4 (Euro-American) was performed using the Chi-square test, the differences were considered statistically significant at a P-value of < 0.05.

Data Availability Statement

The raw WGS data have been submitted to the NCBI SRA database. The entry number of biological project was PRJNA1000054. The SRA entry numbers of 202 MDR-MTB strains were from SAMN36757384 to SAMN36757585. The average sequencing depth of genome was 94.6–362.7 (mean: 193.5), the 1-fold coverage of genome was 98.7–100 (mean: 99.3), and the 10-fold coverage of genome was 98.6–99.9 (mean: 99.2). The total number of SNPS ranged from 695 to 1818 (mean: 1175.7). All 202 MDR-MTB strains were interpretable by WGS.

Results

Patient Demographics

Between July 2019 and June 2021, a comprehensive investigation was conducted, yielding a total of 1522 mycobacterial strains isolated from respiratory specimens obtained from 3519 individuals suspected of having TB. Among these isolates, 1318 strains were identified as MTB, while 204 strains belonged to non-tuberculous mycobacteria (NTM). Subsequently, a meticulous screening process utilizing the proportion method for drug susceptibility testing (DST) was employed to identify 253 MDR-MTB strains. However, a number of strains were excluded due to various reasons: 12 strains were associated with non-permanent residents of Hainan Island, three strains lacked complete clinical data, 29 strains failed to undergo successful transfer and resuscitation, four strains were contaminated, and three strains did not meet the predetermined quality control parameters for WGS data. Consequently, out of the initial 253 MDR-MTB strains, a total of 202 strains (79.8%; 95% CI, 74.4–84.6%) were included in the subsequent WGS analysis. The study’s overall workflow is depicted in Figure 1.

Figure 1 Flow chart summarizing the screening of strains for WGS analysis.

The cohort of MDR-MTB strains analyzed via WGS consisted predominantly of male individuals within the middle-aged and elderly populations. Among the included strains, 175 were derived from males with a median age of 52 (interquartile range: 42, 61) years, while the remaining 27 strains were from females with a median age of 46 (interquartile range: 25, 54) years. Type 2 diabetes mellitus was the most prevalent comorbidity, affecting 60 out of the 202 patients (29.7%; 95% CI, 23.5–36.5%). In terms of treatment history, 55 cases (27.2%; 95% CI, 21.2–33.9%) were classified as new patients, while 147 cases (72.8%; 95% CI, 66.1–78.8%) had received prior treatment. Radiographic assessments revealed the presence of lung cavities in 158 patients (78.2%; 95% CI, 71.9–83.7%) (see Table 1 for details).

Table 1 Demographic Characteristics of 202 MDR-TB Patients

Identification of the Lineage and Analysis of Genetic Distance

A total of 202 MDR-MTB strains analyzed and identified through phylogenetic analysis of transmission clusters, three major lineages were observed. The most prevalent lineage was lineage 2 (East Asian), comprising 169 strains (83.7%; 169/202; 95% CI, 77.8–88.5%). 55 strains (27.2%, 55/202; 95% CI 21.2–33.9%) were lineage 2.1 (non-Beijing). And 114 strains (56.4%, 114/202; 95% CI 49.3–63.4%) were lineage 2.2 (Beijing), of which 94 strains (46.5%, 94/202; 95% CI 39.5–53.7%) were lineage 2.2.1 (modern Beijing), 8 strains (4.0%, 8/202; 95% CI 1.7–7.7%) were lineage 2.2.1.1 (modern Beijing), 12 strains (5.9%, 12/202; 95% CI 3.1–10.1%) were lineage 2.2.2 (ancestral Beijing). Lineage 4 (Euro-American) accounted for 27 strains (13.4%; 27/202; 95% CI, 9.0–18.8%), included 6 strains (3.0%, 6/202; 95% CI 1.0–6.4%) were lineage 4.2.2, 13 strains (6.4%, 13/202; 95% CI 3.5–10.8%) were lineage 4.4.2, 8 strains (4.0%, 8/202; 95% CI 1.7–7.7%) were lineage 4.5, while lineage1 (Indo-Oceanic) was represented by 6 strains (3.0%; 6/202; 95% CI, 1.1–6.4%) (Table 2).

Table 2 Lineages of 202 MDR-MTB Strains

Based on the pairwise genetic distance analysis of MTB strains, 15 clusters consisting of 42 MDR-MTB strains were identified, resulting in a clustering rate of 20.8% (42/202; 95% CI, 15.4–27.0%). Each cluster comprised 2 to 5 MDR-MTB strains. Among the 42 MDR-MTB strains, 22 strains were isolated from new TB patients (55) and the clustering rate was 40.0% (22/55; 95% CI, 27.0–54.1%) and 20 strains were from previously treated TB patients (147) was 13.6% (20/147; 95% CI, 15.4–27.0%).

The clustering rate of lineage 2 (East Asian) was determined to be 19.5% (33/169; 95% CI, 13.8–26.3%), whereas lineage 4 (Euro-American) exhibited a clustering rate of 33.3% (9/27; 95% CI, 16.5–54.0%), with no significant difference observed between the two lineages (χ2 = 2.636, P = 0.104) (refer to Table 3 for detailed results). Notably, four patients which from two pairs of clustered patients differed by 0–2 SNPS (Figure 2, G15: M180 vs M212, G2:M135 vs M144), indicating a recent direct transmission events.² Eighteen patients showed a genetic distance of 3–5 SNPs, suggesting direct transmission. Additionally, 20 patients exhibited a genetic distance of 6–12 SNPs, indicating earlier transmission events (Figure 2).

Table 3 Comparison of Clustering Rates Between lineage 2 and lineage 4

Figure 2 Phylogenetic tree of 202 MDR-MTB strains. Notes: The inner circles marked in yellow are 15 clusters (G1-G15) with an SNP distance of no more than 12. The drugs labeled from the inside to the outside of the outer circle resistance spectrum are: RIF, INH, EMB, SM, OFX, KM, CPM, ETO.

Upon comparing the residential addresses of patients within each cluster, it was observed that the pairwise geographic distances between patients ranged from 0.5 km to 220.5 km, with an average distance of 71.9 km (31.9 km, 138.9 km).

Considering a threshold of living within two kilometers to define the same com munity, three pairs of patients within the clusters were identified as having community transmission (G4: M77 vs M108; G7: M120 vs M148; G13: M89 vs M219). Among these, only one pair of clustered patients demonstrated direct transmission, as indicated by a difference of 3 SNPs among the strains (G4: M77 vs M108) (Figure 3).

Figure 3 Spatial distribution of clustered strains.

Phenotypic Drug Resistance Profile of MDR-MTB Strains

Out of the 202 MDR-MTB strains analyzed in this study, a total of 99 strains (49.0%, 99/202; 95% CI, 41.9–56.1%) were classified as pre-XDR strains. Pre-XDR refers to tuberculosis caused by MTB strains that meet the criteria for multidrug resistance or rifampin resistance, while also displaying resistance to any fluoroquinolone (Figure 4).

Figure 4 Phenotypic drug resistance rate of 202 MDR-MTB strains. Abbreviations: RIF, rifampicin; INH, isoniazid; EMB, ethambutol; SM, streptomycin; OFX, ofloxacin; KM, kanamycin; CPM, capreomycin; ETO, ethionamide.

Drug-Resistance Mutation Spectrum of MDR-MTB Strains

The drug resistance profile of the 202 MDR-MTB strains analyzed in this study was determined for four first-line anti-TB drugs (RIF, INH, EMB, and SM) and four second-line anti-TB drugs (OFX, KM, CPM, and ETO). Table 4 provides a summary of the findings.

Table 4 Drug Resistance Gene Mutation Spectrum of 202 MDR-MTB Strains

Regarding the first-line anti-TB drugs, RIF-resistance genes were identified in 200 strains (99.0%, 200/202), with 98.5% (197/200) displaying mutations in the rifampicin-resistance determining region (RRDR) of the rpoB gene. The most prevalent mutation type was rpoB S450L, observed in 56.5% (113/200) of the RIF-resistant strains. Among these 200 strains, 42 (21.0%) exhibited combined mutations at two or more sites. Furthermore, two strains (1.0%) presented insertions or deletions in the rpoB gene (rpoB 1295_1303_delAATTCATGG and rpoB 1299_1304_delCATGGA).

INH-resistance genes were detected in 191 strains (94.5%, 191/202), with the most frequent mutation being katG S315T, observed in 74.9% (143/191) of the INH-resistant strains. Nineteen strains (9.9%, 19/191) exhibited combined mutations at two or more sites, primarily involving katG and fabG1 promoter mutations. Additionally, 13 strains (6.8%, 13/191) presented katG frameshift mutations, 1 strain (0.5%, 1/191) had a katG gene deletion, and 2 strains (1.0%, 2/191) displayed katG 1_1251_del and katG 1080_2156111_del mutations.

EMB-resistance-related genes were detected in 125 strains (61.9%, 125/202), primarily associated with mutations in the embB gene and embA promoter region. Among these, embB M306I (30.4%, 38/125) and M306V (24.8%, 31/125) were the most common mutations observed. Moreover, 20 strains (16.0%, 20/125) exhibited combined mutations at two or more sites.

Mutations associated with SM resistance were identified in 100 strains (49.5%, 100/202), predominantly affecting the rpsL and rrs genes. The primary mutation type observed was rpsL K43R, accounting for 58.0% (58/100) of the SM-resistant strains. Two SM-resistant strains (2.0%) exhibited combined mutations, while six strains (6.0%) displayed insertions or deletions in the gid gene.

Among the second-line drugs, 97 strains (48.0%, 97/202) harbored resistance genes related to fluoroquinolones (FQs), such as OFX. The mutation rate in the quinolone-resistance-determining region (QRDR) for FQs was as high as 99.0% (96/97). The primary mutation types identified in the gyrA gene were D94G, observed in 44.3% (43/97) of the strains, and A90V, observed in 33.3% (29/97) of the strains. Furthermore, 13 strains (13.4%, 13/97) exhibited combined mutations at two sites. For aminoglycosides, mutations in the rrs promoter region were prevalent, with rrs A1401G accounting for 93.3% (14/15) and 77.8% (14/18) of KM and CPM-resistant strains, respectively. ETO-resistant mutations primarily affected the ethA frameshift (47.4%, 18/38) and fabG1 C-15T (39.5%, 15/38) genes. Combined mutations were observed in five ETO-resistant strains, and one strain exhibited an insertion or deletion in the ethA gene.

Consistency of Phenotypic and Genotypic Drug Resistance in MDR-MTB Strains

Drug sensitivity was determined through WGS by assessing the presence or absence of mutations in resistance-related genes. A strain was classified as genotype resistant if a drug resistance gene mutation was detected, while it was considered genotype sensitive if no such mutation was found. The accuracy of WGS predictions for drug resistance was evaluated using phenotypic drug susceptibility results as the reference standard, as depicted in Table 5. When compared to phenotypic results, WGS exhibited a sensitivity of 90% or higher for detecting resistance to each of the six drugs (RIF, INH, EMB, OFX, KM, CPM). Except for EMB, the specificity of WGS predictions for the remaining drugs exceeded 88%. Overall, agreement between the two methods in determining strain sensitivity or resistance was observed in more than 85% of cases for all drugs, excluding EMB.

Table 5 Comparison of Genotypic and Phenotypic Resistance Results of WGS

Discussion

The identification and treatment of close contacts with TB-infected patients play a crucial role in controlling the spread of the disease.²⁵ Obtaining epidemiological data for TB in China can be challenging due to its high burden. However, WGS can aid in identifying transmission events in patients even without epidemiological data.^26,27 In this study, WGS was employed to detect drug resistance patterns, assess genetic diversity, and analyze the drug resistance of potential MDR-MTB strains, thereby shedding light on the molecular epidemiology of MDR-MTB on Hainan Island, China. Consistent with previous reports from various regions in China,^6,27–29 lineage 2 (East Asian) was found to be the predominant genotype among MDR-MTB strains on Hainan Island, followed by lineage 4 (Euro-American). Interestingly, a small proportion (3.0%) of MDR-MTB strains belonged to lineage 1 (Indo-Oceanic), resembling the genetic profile observed in Indonesia, a tropical island in the South China Sea.³⁰ A case of MDR-TB with lineage 1.1.1.1 was also identified, which was only considered to be an endemic lineage in Vietnam.³¹ Although the lineage 2.2 (Beijing) was the dominant lineage 2 (East Asian) sublineage (56.4%), we were more surprised to find that lineage 2.1 (non-Beijing) accounted for 27.2% of the MDR-MTB strains in Hainan Island, which was unheard of before. Only 9.6% reported in Guangxi³² and 5.35% reported in Thailand³³ were found in other areas of China and neighboring countries of Hainan Island, which indicated that Hainan Island had unique genetic diversity and was obviously different from other areas of China and neighboring countries.

Molecular evolution studies have suggested that lineage 2 (East Asian) has a higher propensity for propagation compared to lineage 4 (Euro-American).^34,35 However, despite no significant association between MDR-MTB genotypes and clustering rate (χ2=2.636, P=0.104) in this study, lineage 4 (Euro-American) exhibited a higher clustering rate (33.3%) than lineage 2 (East Asian) (19.5%).

Among the 202 MDR-MTB strains analyzed in this study, only 42 (20.8%) exhibited clustering, including 22 strains from new TB patients and 20 strains from previously treated TB patients. Furthermore, 33 MDR-MTB strains without unique clusters were identified among new patients. According to this rule,²⁷ it can be inferred that 75 (37.1%) MDR-TB patients on Hainan Island potentially transmitted MDR-MTB strains during the study period, consistent with findings from other studies where more than one-third of MDR-TB cases were suspected to be transmission events.²⁰ Several factors, including sample size, study duration, and different genomic cluster thresholds, may contribute to variations in these results.³⁶

In addition, this study investigated the correlation among 42 patients with MDR-TB in 15 clusters by integrating conventional epidemiological data with WGS information. We found that the geographical distance of clustered patients in Hainan Island was uneven, with an average pairing distance of 71.9 km. Among these clusters, only patients within two specific clusters were identified as having a community transmission connection, as they were known to each other (G4: M77 vs M108; G13: M89 vs M219), these two pairs of clustered patients also had clear epidemiological investigation evidence. No transmission associations were observed in the remaining clusters, suggesting that these cases may have resulted from incidental transmission events. This indicates that with the construction of international tourism island and free trade port in Hainan Island in the past decade, economic development and increased personnel flow have also brought pressure to local tuberculosis prevention and control.

Numerous studies have demonstrated that over 96% of rifampicin (RIF)-resistant strains exhibit mutations in the rifampicin-resistance determining region (RRDR) of the rpoB gene.³⁷ Consistent with these findings, our study revealed RRDR mutations in the rpoB gene in 97.5% of the MDR-MTB strains on Hainan Island, China, with only three MDR-MTB strains lacking RRDR mutations. The predominant mutation observed was rpoB S450L (56.5%). Resistance to isoniazid (INH) can be attributed to various gene mutations, and the specific mutations determine the extent of resistance. Notably, the katG gene mutation indicates high-level INH resistance, while mutations in the fabG1 promoter region are associated with low-level resistance to INH. Our study identified the katG S315T mutation (74.9%) as the most prevalent in MDR-MTB strains, and previous research has established its association with multidrug resistance.³⁸ The primary mutation types of RIF and INH in MDR-MTB strains on Hainan Island were similar to those reported in other regions of China, although the mutation frequencies differed. For example, rpoB S450L (58.2%) and katG S315T (73.8%) in Beijing,³⁹ rpoB S450L (55.1%) and katG S315T (75.9%) in Chongqing,²⁷ rpoB S450L (54.8%) and katG S315T (86.0%) in Shanghai,²⁸ rpoB S450L (45.2%) and katG S315T (56.3%) in Hunan,⁶ rpoB_S450L (57.7%) and katG S315T (53.7%) in Hebei,⁶ rpoB S450L (44.2%) and katG S315T (55.8%) in Jiangxi.⁴⁰

Regarding streptomycin (SM) resistance, our study revealed rpsL K43R (58.0%) as the primary mutation in MDR-MTB strains, consistent with findings in Shanghai, China,²⁸ whereas rpsL L43A (76.7%) was predominant in Hunan.⁶ The significance of embB gene mutations, particularly at codon 306, has been subject to debate due to their presence in both ethambutol (EMB)-resistant and EMB-sensitive strains. In our study, the most common embB gene mutations were M306I (30.4%) and M306V (24.8%), whereas M306V was the prevailing mutation type in Hunan and Shanghai strains.^6,28 The primary fluoroquinolone-resistance gene identified was gyrA mutation. In our study, the mutation rate in the quinolone-resistance-determining region (QRDR) reached 99.0%, with the main mutation types being D94G (44.3%) and A90V (33.3%) in the gyrA gene, although previous studies reported the highest frequency of A94G mutation.^6,39,40 Cross-resistance between ethionamide (ETO) and INH has been well-documented, and fabG1 mutation serves as a primary mechanism of ETO resistance.^41–43 Notably, ethA frameshift (47.4%) and fabG1 C-15T (39.5%) were identified as major drug-resistant genes, distinguishing them from other studies conducted in China.⁴⁴ The rrs A1401G mutation site plays a crucial role in the resistance to injectable aminoglycosides. In our study, we found that 14 strains with rrs A1401G mutation exhibited resistance to kanamycin (KM) and capreomycin (CPM). Overall, the primary mutation types associated with anti-tuberculosis drug resistance in different regions of China are generally consistent, albeit with potential variations in mutation frequencies that may be influenced by sample size, genotype differences, medication practices, and patient adherence.

In this study, we evaluated the performance of WGS in predicting drug resistance using phenotypic drug susceptibility as the reference standard. Our findings demonstrate that WGS exhibited high sensitivity and specificity in predicting resistance to rifampicin (RIF), isoniazid (INH), streptomycin (SM), ofloxacin (OFX), kanamycin (KM), and capreomycin (CPM), surpassing the 80.0% threshold. Notably, the concordance between WGS predictions and phenotypic susceptibility results for RIF, INH, OFX, KM, and CPM exceeded 90%, consistent with previously reported findings.^22,44–46 However, it is important to highlight that the specificity of WGS in predicting ethambutol (EMB) resistance was notably lower compared to other drugs. This discrepancy may arise from the presence of single mutations, such as embA or M306I/G406A in embB, which can be found in both EMB-resistant and EMB-sensitive strains. The reliance on phenotypic drug susceptibility testing may be less reliable for EMB when single gene locus mutations are involved.^44,47,48 Moreover, the relatively low sensitivity of WGS in predicting ethionamide (ETO) resistance could be attributed to the inherent limitations of phenotypic drug susceptibility testing and the complex nature of resistance mechanisms.^49–52

In this study, we employed WGS data from MDR-TB strains obtained from Hainan Island, the sole tropical island in China, to perform cluster analysis based on single nucleotide polymorphism (SNP) differences. This approach allowed for more accurate identification of clustered strains. Nevertheless, it is important to acknowledge certain limitations associated with this study. Firstly, the number of recent transmission events may have been underestimated since only patients with culture-positive MDR-TB over a two-year period were included, and transmissibility among culture-negative MDR-TB patients was not assessed. Secondly, the lack of recorded treatment outcomes for TB patients precluded the analysis of correlations between resistance characteristics and clinical outcomes. Thirdly, the limited resistance rates of certain drugs resulted in inadequate data, such as the low resistance rates of aminoglycosides, leading to a higher margin of error. Consequently, the high specificity observed for these drugs may not be entirely representative. It is important to acknowledge the study’s limitations, particularly regarding drugs with low resistance rates. Lastly, the small sample size prevented the examination of correlations between MTB strain sub-lineages and drug resistance characteristics.

Conclusion

This study focused on the examination of genetic diversity and drug-resistant mutation characteristics in prevalent MDR-MTB strains on Hainan Island, China, utilizing WGS. The findings revealed that lineage 2 (East Asian) and its sublineage lineage 2.2 (Beijing) played a dominant role in the MDR-TB prevalence on the Hainan Island. It was also found that the proportion of lineage 2.1 (non-Beijing) was significantly higher than that in other regions of China and surrounding countries of Hainan Island, which has unique genetic characteristics. Providing insights into the primary types of drug-resistant mutations observed in MDR-MTB strains for both first-line and second-line anti-tuberculosis drugs. Furthermore, the study demonstrated the favorable performance of WGS in predicting resistance to anti-TB drugs, with the exception of ethambutol. Given the significant impact of recent transmission on MDR-TB incidence, targeted interventions such as improved TB case management and active case detection are urgently required to prevent further transmission of MDR-TB on the Hainan Island. With its potential to predict resistance and identify strain clusters, WGS emerges as a promising tool for guiding treatment regimen design and informing tailored public interventions.