In China, all suspected pulmonary TB patients are referred to local TB-designated hospitals for diagnosis, treatment, and registration management. Through a routine surveillance system, the Shanghai Municipal Centre for Disease Control and Prevention (CDC) collected and preserved all culture-positive M. tuberculosis isolates from local TB-designated hospitals. However, the isolates of patients who were first diagnosed in the Shanghai Pulmonary Hospital were not sent to the Shanghai CDC. Our study obtained all clinical isolates of TB patients in the Chongming district, which were collected from the Shanghai CDC.

This study included patients aged 15 years or older with culture-confirmed pulmonary TB reported in Shanghai, between January 1, 2009 and December 31, 2018. Demographic information of TB patients was collected from the Tuberculosis Information Management System, which has updated small modules according to the work needs during the 10 years. Despite this, the main data collected in our study were not affected, such as age, sex, residential address, diagnostic time, occupation, and census register (residents refer to local patients with census registration in the Chongming district rather than migrants from other provinces in China).

Phenotypic drug susceptibility testing (DST) against first-line anti-tuberculosis drugs was performed using the proportion method on Lowenstein–Jensen medium at the following concentrations: rifampicin 40.0 μg/ml, isoniazid 0.2 μg/ml, streptomycin 4.0 μg/ml, and ethambutol 2.0 μg/ml, respectively (World Health Organization [WHO], 2018).

Extraction and purification of genomic DNA were carried out following QIAamp DNA Mini Kit (Qiagen) protocols. Libraries were constructed by Nextera XT library prep (Illumina), and paired-end 150-bp DNA sequencing was performed on a Hiseq 2500 platform (Illumina) with an expected coverage of 100×. The bioinformatics analysis was performed following a previously validated pipeline (Yang et al., 2018). M. tuberculosis H37Rv strain (GenBank accession number NC_000962.3) was used as the reference template. Sequencing reads were mapped to the reference genome using Bowtie2 (v2.3.1). SAMtools (v1.6) was used for SNP calling with mapping quality greater than 30. Fixed mutations (frequency ≥ 75%) were identified using VarScan (v2.3.6) with at least 10 reads supporting and the strand bias filter option on. We excluded all SNPs annotated in regions, such as PPE/PE-PGRS family genes, phage sequences, insertion, or mobile genetic elements (Casali et al., 2016).

To assess the genetic drug resistance, the data of SNPs were compared with a list of the mutations related to anti-TB drug resistance (Coll et al., 2015). Excluding mutations in drug-resistance-associated genes, all SNP locations were combined into a non-redundant consensus list and were recalled with the mpileup2cns function of VarScan to build a phylogenetic tree. Nucleotide positions with missing calls present on more than 5% of the isolates were removed (Liu et al., 2018; Liu et al., 2020). A phylogenetic tree was constructed with MEGA X (version 10.1) using the maximum-likelihood method and a general time-reversible model of nucleotide substitution, with a gamma distribution (GTRGAMMA) and 500 bootstraps. Visualization of the bacteriological information was performed at Interactive Tree of Life.¹ We adapted a recently described hierarchical nomenclature to define nodes and subclades within the tree in the definition of sub-lineages (Coll et al., 2014).

Clustered cases were invited to participate in a contact investigation. After acquiring the informed consent, we have conducted a questionnaire in person or telephone interviews. To understand the social relationship between clustered cases, we collected data on close contacts, social contacts, and places they frequented before their diagnosis of TB, including the information of medical records, workplace, and residential address.

We compared the characteristics between “clustered” and “non-clustered” using the chi-square test. Multivariate logistic regression was conducted to identify the risk factors of TB transmission, such as demographic, bacteriological, phenotypic resistance, and clinical characteristics. Factors with a P-value less than 0.05 in the final model were independently associated with genomic clusters. The adjusted odds ratios (AORs) and 95% confidence intervals (95% CIs) were calculated. Our statistical analysis was conducted with SAS (version 9.4). The clustering rate was calculated as the percentage of clustered patients among the number of total patients (number of clustered patients/number of total patients).

This study protocol was approved by the Ethical Review Committee at the Shanghai CDC (No. 2020-14). All participants provided written informed consent after completing the description of the study.

All biological experimentations in our study were performed after acquiring approval from the Biosafety Commission, Shanghai CDC. Our study abides by National Rules on Biosafety Management of Pathogenic Microorganism Laboratory (State Council of the People’s Republic of China [SCPRC], 2018) to conduct all specific procedures, such as microscopy, culture, DST, and molecular testing. A set of essential biosafety measures was enacted to minimize risks in our TB laboratory, including risk assessment, plans for emergency preparedness and response, codes of practice, personal protective equipment and clothing, as well as disposal procedures for contaminated materials.

Sequence data associated with this study were deposited in the Sequence Read Archive (SRA) of NCBI under project accession PRJNA760838.

As shown in Figure 1, 1,708 TB cases were reported in the Chongming district between 2009 and 2018, including 685 bacteriologically confirmed TB cases. There were 525 culture-positive isolates and 160 smear-positive but culture-negative isolates. Of those, 370 patients (70.48%, 370/525) with culture-positive M. tuberculosis isolates were enrolled in this study. The rest of patients (155 culture-positive isolates) who were first diagnosed in the Shanghai Pulmonary Hospital were not included. In total, 16 isolates were excluded due to DNA extraction failure, contamination, or insufficient quality of WGS data. Seven patients had a relapse of pulmonary TB and one patient relapsed twice, so we excluded the nine isolates of relapsed patients. For the final analysis, 345 M. tuberculosis isolates were included.

The median age of the TB patients was 60 years (range 38–74). A total of 277 patients (80.29%, 277/345) were residents with census register in the Chongming district. About 86.09% of the patients (297/345) were new TB cases. The percentage of patients with resistance to isoniazid (INH) was 12.46% (43/345). The median time of health system delay was 3 days (range 2–9 days).

Most of the M. tuberculosis isolates belonged to the lineage 2 (n = 298, 86.38%) followed by lineage 4 (n = 46, 13.33%), and lineage 3 (n = 1, 0.29%). Sub-lineage 2.2.1 (n = 289, 83.76%) was the main sub-lineage. The phylogenetic tree revealed that three lineages of M. tuberculosis isolate circulated in the Chongming district (Figure 2). The tree also displays the drug resistance profile for 12 anti-TB drugs based on the presence of validated resistance-conferring mutations. The mutations in drug resistance genes mainly occurred in rpsL, katG, and rpoB.

Among 345 TB cases, 94 were grouped into 42 genomic clusters. The clustering rate was 27.25%, which indicated TB transmission in this district. The genomic clusters consisted of two to five strains, and the majority of the clusters had two strains, accounting for 74.47% (Table 1) of the clustered cases (70/94).

Beijing genotype strains were more frequent in clustered cases than in non-clustered cases (P < 0.05). Resistance to isoniazid was associated with genomic clusters (P < 0.05). Additionally, positive sputum smears and cavitation on chest radiography did not differ between clustered cases and non-clustered cases (P > 0.05) (Table 2).

In the multivariable logistic analysis, factors independently associated with genomic cluster included age, census register, health system delay, genotype, and resistance to isoniazid. As shown in Table 3, patients under 65 years old were more likely to be clustered than patients aged 65 years or older (AOR = 3.11, 95% CI: 1.76–5.49). Residents (AOR = 2.43, 95% CI: 1.18–4.99) and Beijing genotype strains (AOR = 3.35, 95% CI: 1.32–8.53) had a higher risk of clustering. Interestingly, patients with resistance to isoniazid were more likely to be clustered than others (AOR = 2.36, 95% CI: 1.15–4.88). Compared to a health system delay < 14 days, a health system delay ≥ 14 days increases by 2.57 (95% CI: 1.34–4.95) times the risk to form genomic clusters.

Thirty-three clustered cases were investigated, with a response rate of 35.11% (33/94), and those cases covered 52.38% of the genomic clusters (22/42). We found four clusters with confirmed epidemiological links, in which one cluster included two prisoners from the same prison and three clusters included six family members. Based on the residential address, we found that several patients in the same clusters lived in the same towns. Thus, probable epidemiological links could be established for 28 patients from 12 clusters. No epidemiological links were found in eight genomic clusters.

Many clustered patients reported a history of exposure to TB with family members (C11, C12, and C25) or classmates (C02, C15, C16, and C26). There were five patients of four clusters (C02, C15, C24, and C29) that worked in other places before they were diagnosed. The WGS data indicated that they were in the same clusters with other residents. Several clustered cases were from the same sex, shared the same occupation, and had similar age. The young people often went to internet cafes and school, while the older group often went to markets, chess, and card rooms. It means that casual contact transmission could occur in those same groups due to their lifestyles.

This genomic epidemiological study systematically delineates the relative contribution of TB transmission and identifies principal risk factors that may be responsible for TB transmission in Shanghai, from 2009 to 2018. Our study concludes that health system delay and resistance to isoniazid are principal contributing factors for the dissemination of the disease. A total of 27.25% (94/345) clustered cases were identified, implying that local transmission is a threat to end TB. WGS was performed for all available M. tuberculosis isolates in a 10-year period in the Chongming district, Shanghai, which could accurately reflect the level of TB transmission.

Our study shows the capability of WGS to identify exact clusters. The epidemiological investigation has further validated that confirmed or probable epidemiological links exist in some genomic clusters. We have found 16 clusters with confirmed/probable epidemiological links by contact investigation. Based on the medical records, WGS data, and epidemiological information, the transmission relationships of clustered patients were explored, and the characteristics of TB transmission were summarized. Household transmission is an important characteristic in this area. Several patients in the same cluster share family relationships, and many clustered patients have a history of family contact with TB.

Our study suggested that patients under 65 years old or residents had a higher risk for transmission. To halt the spread of TB, active case finding should be focused on these high-risk groups. A study in the United Kingdom suggests that most patients who migrate from high-incidence countries do not lead to local transmission (Walker et al., 2014). However, our results are distinct from a study in the Songjiang district, Shanghai, which inferred that TB transmission from residents to migrants occurs as commonly as transmission from migrants to residents (Yang et al., 2018). One possible explanation is that older residents in the Chongming district have lower awareness of TB and have more frequent social contact with others, which increases the chance to be infected and transmit to others. Also, most migrants in the Songjiang district could be younger groups with stronger bodies.

This study noted that Beijing genotype strains had a higher risk of clustering. This finding is consistent with previous studies (Yang et al., 2015; Jiang et al., 2019), indicating that the Beijing genotype may have increased transmissibility. However, it contrasts with a study (Lalor et al., 2017) conducted in England, which noted that the Beijing genotype strains are not associated with transmissibility among household contacts.

We detected a significant association between resistance to isoniazid and genomic clusters. A previous study (Dheda et al., 2017) has showed that close contacts of drug-resistant TB cases are more likely to be infected with a more infectious M. tuberculosis strain. A recent study (Theron et al., 2020) reported similar results, showing that drug-resistant TB patients have higher numbers of culturable M. tuberculosis in an aerosol. Our results also showed that 20.21% (19/94) of clustered isolates were resistant to isoniazid. A systematic review reported that treatment of isoniazid-resistant TB with the standardized first-line regimen led to treatment failure in 11% of the cases, compared with 2% among drug-susceptible TB patients (Gegia et al., 2017).

In our study, patients with a health system delay ≥ 14 days have a higher risk to transmit the disease. Chongming district is at the mouth of the Yangtze River, with relatively inconvenient traffic and a low economic level. This delay may be due to the unavailability of rapid and accurate diagnostic technologies required to confirm TB. A study (Segagni Lusignani et al., 2013) indicated that living in rural areas is associated with prolonged health system delay. It is worth noting that the diagnostic delay could prolong the infectious period. A meta-analysis (Bao et al., 2019) revealed that the case finding delay is a major contributing factor to the spread of TB in outbreaks. Also, the diagnostic delay might result in worse treatment outcome and increased risk of transmission (Cheng et al., 2013; Shiferaw and Zegeye, 2019). However, the main diagnostic tests for TB are rapid molecular assays, smear microscopy, and culture-based methods. Culture-based methods are the reference standard, which require a higher level of TB laboratory and can take up to 8 weeks to get results (World Health Organization [WHO], 2020). To achieve early diagnosis and promote early treatment, health system delay should be reduced by introducing rapid and accurate diagnostic tools such as sequencing technologies.

Our study has several limitations and strengths. First, the level of TB transmission may be underestimated. About 30% (155/525) of culture-positive isolates were not collected in our study. Also, it is unable to analyze the TB transmission of patients without clinical strains. Second, we obtained limited information about clustered cases due to the retrospective study design. Therefore, epidemiological links between clustered cases were difficult to interpret clearly. However, this was the first population-based study using WGS to explore the transmission of TB in the Chongming district, Shanghai. Also, the long-span study timeframe facilitates the capture of all TB cases and provides insights into TB transmission dynamics. We have traced the infection source of the population studied at the molecular level, and the transmission chains of TB patients have been detected.

Health system delay is a crucial factor for the transmission of TB in rural areas. The public health strategies should be redirected to reduce health system delay by introducing rapid and accurate diagnostic tools. Additionally, it could be feasible to take targeted interventions on patients with resistance to isoniazid.