Introduction

Tuberculosis (TB) is the second leading cause of mortality by a single infectious agent, with 1.3 million deaths globally in 2022, almost twice as many deaths as HIV/AIDS1. Complex types and heteroresistance of drug-resistant TB increased difficulty in diagnosis and treatment, further complicating the progress towards ending TB, and personalized diagnosis and treatment of TB has become the key to controlling the disease2,3.

Sputum positive PTB have high sputum transmissibility3. Sputum conversion of PTB after 2 months of ATT is a reliable measure of the bactericidal activity of drugs, and has been identified as an important predictor of cure rate4. Delayed sputum conversion patients (DSCPs) were characterized by the persistence of sputum-positive PTB status upon completion of the intensive treatment phase, contributing to higher treatment cost and additional burden to healthcare services5,6,7,8. TB as a “social disease”, the solution to this dilemma depends on social, economic and environmental interventions, which also influence the occurrence of sputum conversion and sputum conversion rate (SCR)9,10. Besides, previous studies have proposed the main risk factors including baseline bacterial load, neutrophil-to-lymphocyte ratio, monocyte-to-lymphocyte ratio, but these conclusions may not be consistent due to differences in sample sizes and geographic regions10,11,12,13,14.

However, it remains uncertain whether the drug-resistance and genomic diversity of initial isolates impact sputum conversion. For DSCPs, the dynamics changes of molecular and phenotypic drug-resistance is also a matter of concern. Additionally, the genome of pathogenic bacteria's role in disease progression and treatment is increasingly distinct3,15,16,17,18, while the relationship between delayed sputum conversion and increased bacterial genomic heterogeneity intra-host requires further investigation. This study provided new perspective on the factors influencing the delayed sputum conversion of PTB and observed the dynamics of drug resistance and MTB genome changes of DSCPs during ATT.

Methods

Study design and ethics statement

A retrospective cohort study was carried out on data of 1782 PTB managed in Beijing, China from January 2021 to December 2022 to analyze the sputum conversions and epidemiological influences.

We selected PTB who were older than 18 years of age without any other immune diseases, and had preserved serial clinical isolates to constitute the delayed sputum conversion group (Case group). Due to the stringent conditions for strain preservation, only 24 DSCPs were ultimately included in the study. We conducted a case-matched study containing a Control group (24 TSCPs and their 24 clinical isolates) matched to a Case group (24 DSCPs and a total of 51 clinical isolates before and after ATT) based on age, gender, and comorbidities of cohort entry to explore the factors for delayed sputum conversion based on the genotype and phonotype drug-resistance of initial clinical isolates from PTB as well as the genetic diversity between TSCPs and DSCPs (Figs. 1, S1).

Figure 1
figure 1

Flowchart of Study Population. Abbreviations: PTB: pulmonary tuberculosis; ATT, anti-tuberculosis treatment; NTM, non-tuberculous mycobacteria; MTB, Mycobacterium tuberculosis.

Full size image

The 24 pre-treatment isolates of 24 DSCPs in Case group constitute Case_Baseline group (Case_B) and 27 post-treatment isolates constitute Case_Treated group (Case_T) including 18 strains isolated at the 2nd month of ATT and 9 strains isolated after 5 months of ATT. The Case_6 and Case_23 patients had two and three clinical isolates from different periods after ATT, respectively. This comparison was used to explore and elucidate the dynamics of DSCPs' drug resistance profiles, genomes, and dominant flora during ATT.

This study was approved by Beijing Center for Disease Prevention and Control (No.2023.16).

Bacteria subculture, species identification and DST

All clinical isolates were stored in 7H9 medium supplemented with 25% glycerin at − 80 °C refrigerator. Resuscitated clinical isolates were inoculated in LJ solid medium for subsequent DNA extraction and DST. DNA extraction of isolates was followed by strain identification using Mycobacterium nucleic acid detection kit (PCR-fluorescent probe method, CapitalBio technology, China). Microplate DST was used to get the MIC (Minimum Inhibitory Concentration) of MTB in this study, customized microtiter plates containing 13 anti-TB drugs (RIF, INH, PZA, EMB, LFX, MFX, BDQ, LZD, CFZ, DLM, DCS, KAN, CPM) sensitization tests (BASO, Zhuhai, China) was shown in the Fig. S2, experimental operation was carried out in full accordance with the instructions.

WGS and analysis

Extraction and purification of genomic DNA were carried out following Bacterial DNA Extraction Kit (Gene-Optimal, 60,300 K-50 T) protocols. All WGS procedures were performed by Shanghai Gene-Optimal Science & Technology Co. Ltd. (Shanghai, China). Libraries were constructed on the Illumina platform using an FS DNA Lib Prep Kit V6 (RK20259), and we used an Illumina NovaSeq 6000 sequencer, PE 150 software for base-calling, Fastp (0.20.0) for quality control and removal of low-quality data in Raw Data, Cutadapt (v1.15) for trim adapter sequences at the tail of sequencing reads. Clean data was aligned to reference genome (H37Rv, NC000962.3) using BWA (v0.7.17) to evaluate the sequencing depth and coverage. kraken2 for species annotation and abundance detection, compared with the species database of NCBI (https://www.ncbi.nlm.nih.gov/)19,20. Variant sites were identified through alignment with the reference genome H37Rv using Freebayes 1.3.2 and SnpEff 4.3t. Snippy, 4.4.3 for mutation matrix; snp-dists 0.7.0 for matrix of SNP distances, and clustered strains were identified (≤ 12 SNPs)21,22. Drug-Phenotypes were predicted based on mutation and resistance relationships in the Resistance Locus Database (https://github.com/jodyphelan/tbdb). The phylogenetic tree was constructed with SAM-TB23, visualized and modified using iTOL (https://itol.embl.de/).

Statistical analysis

The basic information of patients was downloaded from the Beijing Municipal Tuberculosis Report system. Data were analyzed using SPSS 20.0 software. For continuous variables that were normally distributed, the mean ± standard deviation (SD) was used. Counting data was described in terms of rate or composition ratio (%). χ2 test was used to analyze the single factor correlation between groups. p < 0.05 was considered statistically significant.

Results

Statistics on sputum conversion at the end of ATT of 1782 PTB

1782 cases initially diagnosed with positive sputum and available for subsequent pathogen detection reports were recruited from the registered TB in Beijing. Sputum conversion at the end of a two-month intensive period of antituberculosis treatment is an important indicator for efficacy evaluation. PTB are required to undergo sputum test at the end of the 2nd, 5th and 6th month of ATT, but not all patients comply with this requirement. Based on the sputum reports, we can categorize the patients into the following groups. we can categorize patients into several groups: timely or delayed sputum conversion, reversion to positive after initial conversion, and persistent non-conversion. 1706 patients (95.7%) successfully achieved sputum conversion after complete ATT. 932 patients underwent sputum test at the end of the 2nd month of ATT, 89.6% of them achieved sputum conversion, but 9 patients subsequently relapsed with positive sputum. The cumulative population of PTB who achieved sputum conversion to negative at the end of the 5 and 6 months of ATT and maintained are 1545(86.7%), 1706(95.7%), respectively. Additionally, 76 (4.3%) patients fail to sputum conversion to negative at the end of ATT, including 10 patients who achieved sputum conversion during treatment but eventually relapsed and 65 patients (3.6%) who never reported negative sputum results (Fig. 2).

Figure 2
figure 2

Sputum test results at different time points during the ATT of 1782 PTB. Patients with positive sputum tests for the last time marked with red frames. Abbreviation: ATT: anti-tuberculosis treatment; M, month; CTN: conversion to negative.

Full size image

Basic description of the study population

A total of 932 patients underwent sputum testing at the end of the 2nd month of anti-tuberculosis treatment. As shown in Table 1, the population age structure, gender and household registration address were no statistically significant difference between sputum conversion patients at the end of 2nd month of ATT and patients remain sputum test positive, but those patients who achieve sputum conversion at the 2nd month of ATT have better prognosis and shorter treatment time notably.

Table 1 Socio-demographic and clinical characteristics of PTB with sputum-positive at first diagnosis.
Full size table

As for the case-matched design cohort, 24 patients were recruited in our Case group, only 4 pairs of study subjects were female, the median age of the whole group was 45 years (range: 24–80 years). Among the 24 pairs cases enrolled, diabetes mellitus (n = 10, 41.7%) was the most frequent comorbidity in patients with PTB. In addition, we examined the type of household registration as an epidemiological risk factor that could potentially impact delayed conversion. However, we didn’t find significant difference in our small sample. The number of cured patients was notably higher in Control group compared to the Case group (P < 0.05, Table 2). The phylogenetic tree of the enrolled isolates is in Fig. S3, which includes strain lineage and drug resistance information.

Table 2 Socio-demographic and clinical characteristics of the Case and Control groups.
Full size table

Delayed sputum conversion patients present more extensive and complex drug resistance

We conducted 13 anti-TB-drugs DST on all 75 clinical isolates. Compared to the Control group, the Case_B group emerged higher frequency of drug-resistance (Fig. 3A) and high-level of resistance24, as shown in Table S1, specifically, INH (2 in Control, 5 in Case), LFX (0 in Control, 1 in Case), and KAN (1 in Control, 0 in Case). There were significant differences in the frequency of resistance to both first-line and total tested 13-drugs between the two groups (P = 0.048, 0.045, Table 3). The overall resistance to multiple drugs is related to achieving sputum conversion at the end of 2 months of ATT, rather than a single drug. When we observed the overall changes in drug resistance before and after ATT from a longitudinal perspective of DSCPs, we found that the resistance rates to drugs RIF, EMB and DCS also increased significantly. Besides, comparing WGS resistance mutation detection and MIC, the missed detection rate of WGS was relatively low (7/107, 6.5%, Table S1), indicating WGS is profound for application in drug-resistance detection due to its shorter turnaround time, high detection rate, and ability to provide additional strain information.

Figure 3
figure 3

Drug resistance of Control and Case_Baseline Case_Treated groups. (A) Drug susceptibility test results of clinical isolates from the control group and the experimental baseline group; (B) Drug susceptibility test results of clinical isolates before and after 2 months of ATT (n = 18); (C) Drug susceptibility test results of clinical isolates before and after ≥ 5 months of ATT(n = 9). Anti-tuberculosis drugs with pink backgrounds are first-line drugs. Gray: drug-sensitive; Orange: MIC-only reported resistance; Green: MIC and WGS both reported resistance; Blue: only WGS reported resistance. Strains labeled with red are those reported as NTM by WGS.

Full size image
Table 3 Drug susceptibility of Control、Case_Baseline、Case_Treated.
Full size table

Changes of drug resistance among delayed sputum conversion patients after ATT

We further analyzed the overall changes in drug resistance before and after ATT from a longitudinal perspective of DSCPs, the frequencies of strains acquired phenotypic drug resistance within first-line drugs and total drugs were significantly increased (P < 0.001, Table 3, Fig. 3B, C). Notably, there was a significant increase in the frequency of resistance to RIF and DCS (PRIF = 0.042; PDCS = 0.024). Additionally, the incidence of high-level INH resistance and KAN resistance also increased (INH: 5 in Case_B, 11 in Case_T; KAN: 0 in Case_B, 1 in Case_T). The number of multidrug-resistant bacteria in the Case group increased after treatment.

The statistics of drug-resistance changes in DSCPs before and after ATT are shown in Table 4. 44.4% of patients acquired drug-resistance within two months of treatment, larger compared to after five months of treatment, revealing the initial two months of ATT play a crucial role in ensuring proper drug administration to prevent the emergence of new drug-resistance. Among all patients, 55.6% of isolates showed no change in drug-resistance, while 33.3% developed phenotypic drug-resistance cluster in first-line drugs RIF and INH, with frequencies more than 25%. These results underscore the importance of regular DST for TB throughout the treatment process, facilitating medication adjustments during the course of treatment to prevent the development of worse drug-resistance. Interestingly, re-sensitization occurred in several strains: CPM (Case_7), INH (Case_6), DLM (Case_7, 10). The mechanisms of acquired resistance and re-sensitization require further study.

Table 4 Changes in drug resistance after ATT of Delayed Sputum Conversion Patients.
Full size table

Genomic changes in series clinical isolates from delayed sputum conversion patients

We further analyzed the differences in SNPs of serial isolates from delayed-sputum-conversion PTB to confirm whether intra-host evolution or multiple/mixed infections occurred during treatment. The SNP differential matrix plot (Fig. 4A) revealed that 16/24 patients (71%) remain in a cluster after ATT (SNP < 12), while case 11 exhibited no change in bacterial species but occurred a high SNPs count of 557, indicating possible multiple infections. Surprisingly, 7 patients (29%) dominant organism changed to nontuberculous mycobacteria (NTM) on the second sputum test (Fig. S4, Table S2).

Figure 4
figure 4

Differential SNPs and Their Frequencies within Groups among Control, Case_Baseline, and Case_Treated Strains. (A) The digital matrix chart of the differential single-nucleotide polymorphisms (SNPs) of 24 pairs. NTM marked in red. (B) The frequency of the differential SNPs of MTB within the Control, Case_Baseline, and Case_Treated groups; (C) Statistics of frequency and gene distribution of differential SNPs in groups. Below is the genomic map of H37Rv for reference.

Full size image

Excluding 7 NTM strains and nonsense mutation, we further analyzed the differential SNPs characteristics and their frequencies within three cohorts: Control, Case_B and Case_T (n = 20) groups. The SNPs blank regions exhibited distinct distribution patterns across the three groups. Specifically, 11% of the SNPs were uniquely present in either the Control or Case_B (existing in only one group), while the Case_B and Case_T exhibited more similar in SNPs distribution, with 94% showing no differences (Fig. 4B).

We aimed to identify high-frequency SNPs specific to each group. As shown in Fig. 4C, 55–68% of the differential SNPs between groups have a low frequency within-group (0/24, or 1/24). The differential SNPs with high frequency in the Control and Case_B are relatively high, indicating that there may be SNPs-set distinguish two groups; conversely, the genetic changes of Case_T are more complex and diverse. Obtaining a specific and efficient SNPs-set that can distinguish different groups require further investigation with larger sample size and comprehensive research.

Discussion

The conversion of sputum in PTB during the intensive treatment period is an effective indicator for evaluating treatment outcomes, affected by various factors. This study shows that PTB who achieved sputum conversion at the end of 2nd month of ATT have better treatment outcome, which is identical to the previous reports5,6,7,25. Differing from existing research perspectives about factors affecting sputum conversion12,13,25,26, we provide an innovative and comprehensive information on bacteriology and genomics of clinical isolates from DSCPs.

Phenotypic DST of 13 anti-tuberculosis drugs demonstrates that DSCPs more extensive and complex drug resistance undergo dynamic changes during the ATT, which may increase the difficulty of treatment. Compared to general TB patients, people who developed RR/MDR-TB have poorer anti-tuberculosis treatment outcomes, higher treatment costs, and longer duration of the infectious period1. During our exploration of drug resistance changes in the treatment process, the difference between phenotypic drug resistance and molecular drug resistance may also be a factor affecting the judgment. For example, Case6_T was reported INH-resistant by MIC without INH-related drug-resistant mutation that stresses the importance of paying attention to specific MIC values such as “low concentration resistance”, “high concentration resistance”, “critical concentration resistance”24, et. It is necessary to expand regional studies on genotype and phenotype drug-resistant consistency of MTB, and update the catalogs of drug-resistant mutation based on WGS2,27,28.

There are two cases were DLM re-sensitization in our cohort. DLM is included in Group C, which is recommended for use in longer multidrug-resistant (MDR)-TB regimens29, more attention has been paid to the exploration of DLM resistance mutation sites and mechanism30,31. Besides, changes in the resistance phenotype of DSCPs may be caused by mixed/multiple infections, where the dominant strain in the patient changes along with the course of treatment, resulting in drug-resistant changes. Complex infections possibly lead to multidrug resistance32. Strict implementation of sputum bacteriological and drug-resistant monitoring during ATT are crucial for effectively valuing treatment progress. It aids in enhancing our understanding of the local prevalence, evolution, and transmission of pulmonary tuberculosis.

In addition, a certain percentage of PTB whose dominant strain turned to NTM in our study, drawing our attention to mix infection and personalized treatment. While minor NTM known can cause human disease, the incidence of NTM infections appears risen significantly in recent decades due to factors like population aging, immunosuppression, and the use of broad-spectrum antibiotics33. Whether our findings suggest that patients on ATT are more susceptible to other pathogens remains uncertain.

The whole-genome analysis of isolates reveals variations in SNPs site distribution regions and frequencies between Control and Case groups, showing differences in genetic backgrounds of the two groups, suggesting the possible presence of group-specific SNPs-set. The effect of pathogenic microbial (such as MTB) genome SNPs is profound in microbial genetics and infectious diseases34. The association of SNPs with drug-resistance implying that SNPs can affect an individual's susceptibility, treatment response and disease prognosis35. The SNPs-set approach deciphered the complexity of both genotype and phenotype as well as their complex relationships. For technical and sample size limitations, we did not explore the impact of small insertions and deletions or large chromosomal changes (such as repeat regions in the genome) and SNPs-set analysis following this study36.

MTB is characterized by low mutation rate, and NGS-related studies have also shown that the MTB genome is highly stable during transmission between individuals, with an evolutionary mutation rate of approximately 0.5 SNPs/genome/year37. However, MTB evolution within patients exhibited considerable genetic diversity38, more SNPs observed in series isolates MTB populations from individual patients, these SNPs occurred concurrently with drug-resistance, leading to mutations referred as "hitchhiker SNPs"37. With the treatment duration increases, the variation level decreases, suggesting that the selective pressure from drug exposure has led to a purifying effect21,37,38. Intra-patient microevolution of MDR-MTBC strains under longitudinal treatment is more complex than previously anticipated39. Rapid expansion and collapse of different clones led to a difference of 11 SNPs between two strains separated by only three months37. The number of SNPs between initial and treated isolates in this study was concentrated in the range of 0–6, with only Case_11 experiencing an explosive change with a high SNPs count of 557, which may involve the rapid expansion of different clones and possibly the occurrence of multiple infections.

Conclusion

This study illustrates that PTB achieved sputum conversion at the end of 2 months of ATT had better treatment outcome through the cohort of 1782 PTB. We also provide valuable insights into the understanding risk factors and dynamics in therapy of delayed sputum conversion PTB from the perspective of drug resistance, genomic diversity, intra-host evolution. Delayed sputum conversion patients present more extensive and complex drug resistance and show dynamic changes resistance (re-sensitization, acquired phenotypic drug resistance) during the treatment process, which may be caused by intra-host heterogeneity and faster microevolution.