Patient clinical records
A 40-yr-old HIV-negative woman – who was born in the northwestern region of Argentina and worked as a cleaner in a private health center in Buenos Aires − was diagnosed with pulmonary TB in 2008, based on clinical and radiological data (multiple cavitary lesions) together with the presence of acid-fast bacilli (AFB) in sputum smear examination at Hospital-1. The patient started standard anti-TB treatment consisting of 2 months RIF, INH, pyrazinamide (PZA), and ethambutol (EMB), followed by 4 months RIF and INH (Fig. 1A). In 2009, she returned to Hospital-1 with AFB-positive sputum and lung cavitation. The treatment started again with RIF, INH, PZA, and EMB plus levofloxacin (LVX) for 10 months. In 2010, the patient became AFB smear-negative and remained so for about a year. In August 2011, she relapsed with AFB smear-positive TB. At this point, the first isolate was obtained (TB1) at Muñiz Hospital, showing M. tuberculosis resistant to INH, RIF and STR, susceptible to kanamycin (KAN), amikacin (AMK), ofloxacin (OFX), and EMB. LVX minimal inhibitory concentration (MIC) was < 0.5 mg/L. A new drug combination scheme was started, consisting of terizidone, para-aminosalicylic acid (PAS), ethionamide (ETH), OFX, and KAN for 9 months. Three months later, the patient remained AFB smear-positive and culture-positive (TB2), with a DST profile identical to the first one. In March 2012 (TB3, TB4), the patient was still bacteriologically positive, and the isolate showed additional resistance to OFX and LVX (MIC 12 µg/ml). The isolate remained susceptible to PZA, EMB, KAN, AMK, capreomycin (CP), ETH, cycloserine (Cs), PAS, and moxifloxacin (MFX). Based on these results, a new therapy was established with PZA, EMB, KAN, ETH, Cs, PAS, and linezolid (LNZ). In July 2012, the patient persisted AFB smear-positive and was hospitalized in Muñiz Hospital. In December (TB5), the isolate was resistant to INH, RIF, STR, OFX, and LVX and persisted susceptible to KAN, AMK, CP, PZA, PAS, Cs, ETH, LNZ, and MFX.
In September 2013 (TB6), the patient was hospitalized again and was prescribed a second-line treatment scheme with MXF, PAS, Cs, KAN, PZA, and ETH. The ETH was taken irregularly due to intolerance until it was suspended. After several months, the patient was discharged from hospital. As an outpatient, she did not adhere properly to therapy.
In 2015 (TB7), the patient returned symptomatic and was readmitted to Muñiz Hospital where she was treated empirically with MXF, PAS, Cs, KAN, PZA, ETH, LZN, and meropenem-clavulanate while waiting for results of a new DST. Once again, the ETH had to be suspended due to intolerance. The DST results showed resistance to INH, RIF, STR, OFX, LVX, PZA and MXF, this latter at 0.5 µg/ml. In view of these results, the drug scheme was adjusted to high-dose MXF, Cs, PAS, KAN, meropenem-clavulanate, and LNZ. This treatment was discontinued due to a supply shortage in the hospital pharmacy. In May 2016 (TB8), the therapy was resumed, but meropenem-clavulanate was replaced by imipenem-clavulanate and AMK was administered instead of KAN. This drug scheme was accomplished under supervision for 7 months. After 152 days of AMK and 209 days of imipenem-clavulanate with favourable evolution, the patient was discharged and injectable drugs were discontinued. She continued treatment as an outpatient under regular controls. By the end of 2016, the patient remained asymptomatic and negative to AFB smear and culture. (Fig. 1A).
Clinical isolates
We analyzed 8 isolates obtained from the same patient between August 2011 and January 2016 at Hospital Muñiz. Identification and DST to first- and second-line anti-TB drugs was performed at Muñiz Hospital and confirmed at the TB National Reference Laboratory (NRL) at the National Institute of Infectious Diseases Dr. Carlos Malbrán (ANLIS). The first 4 primary sputum cultures (TB1 to TB4) were collected at Hospital-1, the rest (TB5 to TB8) were collected at Hospital Muñiz; both hospitals are located in Buenos Aires. A previous TB episode occurred in 2008, and by 2010 the patient had become smear-negative and was discharged. Unfortunately, no isolate was available from that first episode, and we cannot determine if the following TB episode in August 2011 was caused by the same strain.
Microbiological and molecular studies
All isolates were grown on Löwenstein-Jensen slants and identified as M. tuberculosis by biochemical and molecular tests. DST was performed by the reference standard proportion method in Löwenstein-Jensen medium and/or BACTEC MGIT 960 (Becton Dickinson, MD) under international standards [13]. A multiplex allele-specific PCR (MAS-PCR) for detection of mutations conferring resistance to INH and RIF (codons katG315, inhA-15, rpoB450, 445, and 425) was performed on all isolates according to a modified protocol described elsewhere [14].
Genotyping was performed by spoligotyping and MIRU-VNTR according to standard procedures [15, 16], followed by comparison with SITVITWEB [17] and MIRU-VNTRplus database [18].
Genome sequencing
To perform WGS, we re-cultured the isolates on Löwenstein-Jensen slants. The DNA was extracted according to a standard protocol for mycobacteria [19]. Genomic libraries were constructed using Nextera® XT DNA Sample Preparation Kit (Illumina) following the manufacturer's instructions. Individual libraries were indexed with Nextera® XT Index Kit. Paired-end reads from all the isolates were obtained using the Illumina MiSeq platform at ANLIS. All reads were deposited in SRA NCBI, Accession number PRJNA646920.
Variant calling
Raw reads were checked for quality using FastQC version v0.11.5 [20] and processed with the PRINSEQ lite version 0.20.4 [20, 21] ensuring quality assessment. The reads were aligned to the H37Rv reference genome (NCBI access number NC_000962.3) using BWA [22]. Variant calling was made with GATK [23]. After filtering problematic regions, we applied 3 different approaches to the analysis: first, we called variants at a low frequency (0.25) to explore short-term mutations, then we called variants at a higher frequency (0.75) to infer phylogenetic trees. Finally, we called even more frequent variants (0.9) to identify fixed mutations within bacillary populations. The resulting variants were annotated with SnpEff [24] and compared with our database of drug resistance-associated variants which combines data of TB Profiler [25], KvarQ [26], CARD [27], and bibliographic data. The analysis of the variant-containing genes was performed with Target Pathogen [28].
Phylogenetic analyses
The phylogenetic analysis was performed by comparing the samples TB1 to TB8 with representative samples of the main M. tuberculosis lineages and sublineages [29] downloaded from NCBI. The reads were processed using the variant calling pipeline mentioned above. The resulting variants from all samples were combined in a single VCF using GATK. To confirm that all the patient’s isolates composed a single monophyletic clade, we constructed a maximum-likelihood genome phylogenetic tree using RaXML [30] dataset used includes 1 or more representative genomes for each M. tuberculosis sublineage.
Nineteen mutations emerged among the 8 time-serial isolates over 57-month of intermittent therapy. Nine out of these 19 variants reached fixation in the bacilli population. The other 10 were categorized as short-term mutations, i.e. single nucleotide polymorphism (SNPs) and Insertion–deletion mutations (InDels) that arise and fade-out within months or years (Supplementary Table 1). The most recent common ancestor was MDR and evolved to a pre-XDR status within 7 months (Fig. 1B).
The clonal nature of the isolates was supported by identical MIRUs-VNTR and spoligotyping patterns (Fig. 2A). Spoligotyping results showed that all the samples belonged to SIT450, an ambiguous clade between X1 and T5. These sublineages belong to the 4 Euro-American lineage, which is the predominant in Latin America [31].
To place the patient’s isolates in a global phylogenetic context, and confirm they form a single monophyletic clade, we constructed a maximum-likelihood core genome phylogenetic tree. According to a recent SNP-based phylogenetic classification [32], the 8 isolates composed a monophyletic cluster assigned to the sub-lineage 4.1.1, which corresponds to the X type (Fig. 2B).
Drug-resistance conferring mutations and putative compensatory mutations
In August 2011, the TB NRL received the first isolate (TB1), and 3 months later the second one (TB2). Both isolates were found to be already resistant to INH and RIF – that is, they qualified as MDR – with additional resistance to STR. These isolates harbored katG S315T, rpoB D435V, and rrs C517T, mutations related to resistance to these three drugs, respectively. These 3 mutations remained fixed in all 8 isolates until the end of the treatment. The genotypic prediction of resistance from WGS data was in line with the results of the phenotypic and molecular techniques applied at the TB NRL (Table 1 and Supplementary Table 2).
Table 1
Drug resistance conferring-mutations and their correlation with phenotypic drug resistance.
Antibiotic | Gene | Amino acid change | Phenotypicresistence | Genome coverage | Mutation frequency | Isolate |
INH | katG | S315T | yes | 95x | 1 | TB1-8 |
RIF | rpoB | D435V | yes | 102x | 1 | TB1-8 |
STR | Rrs | C517T | yes | 147x | 1 | TB1-8 |
OFX- LVX- MXF | gyrA | D94H | yes | 71x | 1 | TB3-5, TB7-8 |
OFX- LVX | gyrA | A90V | yes | 77x | 0.85 | TB6 |
PZA | pncA | T177fs | yes | 47x | 1 | TB7-8 |
ETH | ethA | T88fs | no | 53x | 0.6 | TB7-8 |
AMGs | 5´utr-whiB7 | 3568733A > C | no | 73x | 1 | TB7-8 |
CFZ- BDQ* | Rv0678 | D47fs | yes (CFZ) | 45x | 0.5 | TB7-8 |
INH: isoniazid, RIF: rifampicin, STR: streptomycin, OFX: ofloxacin, LVX: levoflaxin, MXF: moxifloxacin, PZA: pyrazinamide, ETH: ethionamide, AMGs: aminoglycosid, CFZ: clofazimine, BDQ: bedaquiline. |
By March 2012 (TB3), phenotypic resistance to OFX and LVX was detected, with borderline resistance to MXF, which changed the strain status to pre-XDR. This phenotypic resistance was supported by the mutation gyrA D94H. The strain remained phenotypically resistant to fluoroquinolones until the end, but the frequency of this mutation fluctuated between isolates. In month 25 of treatment (TB6), we registered a decrease in the frequency of clones with gyrA D94H mutation and the rise of a new population of clones carrying gyrA A90V mutation (Fig. 3). Taking advantage of the closeness of the 2 variant positions, we examined whether the same clonal variant harboured these 2 SNP simultaneously. We did not find evidence of a single clonal variant harbouring these 2 variants together in any of the 97 reads that shared both positions. Clones carrying gyrA A90V vanished in subsequent isolates along with the implementation of adequate therapy. Although we detected both mutations 15 months apart, these resistance conferring-mutations emerged independently from the parental MDR strain (Fig. 1B and Fig. 3).
When the therapy was resumed (52 months of treatment, TB7), we observed an increased genetic heterogeneity, involving both, short-term and fixed drug-resistance conferring mutations. Among the short-term SNPs, we found 2 mutations related to drug resistance, ethA (T88fs) and Rv0678 (N47fs). Mutations in the ethA gene have been reported to confer resistance to ETH [33]. More than half of the reads of TB7 and TB8 harboured an ethA frameshift mutation but these last 2 isolates remained notably susceptible to ETH. The patient’s drug scheme included ETH, but she took it irregularly due to intolerance until it was suspended. The Rv0678 gene – where we found a frameshift – is a transcriptional repressor of the genes encoding the MmpS5-MmpL5 efflux pump, which has been associated with bedaquiline (BDQ) and clofazimine (CFZ) resistance.
Among the drug resistance-conferring mutations that became fixed, one appeared towards the end of the treatment at the pncA gene (T177fs), confirming the phenotypic resistance to PZA. Another was gyrA D94H, which finally re-emerged and became fixed in the bacillary population (TB8) (Fig. 3). A mutation in the 5′ untranslated region of whiB7 gene (T-54G) was found in the last 2 isolates (TB7 and TB8). The whiB7 gene is a transcriptional activator of the eis gene. Mutations in its 5' untranslated region have been related to whiB7 overexpression and the subsequent increase in eis expression, ultimately conferring resistance to KAN [34].
Within-host diversity/microevolution
To explore the within-host diversity among the 8 time-serial isolates, we called variants at a frequency of 25%. Along with the aforementioned short-term and fixed mutations related to drug resistance phenotypes, we found other 12 genomic variants: 6 changes were miss-sense mutations, 3 were InDels and 3 were synonymous mutations. (Supplementary Table 1). The overall substitution rate (5.68 SNPs per genome per year) was higher than the estimate for M. tuberculosis (0.4–0.5 SNPs per genome per year) [35]. This is in line with previous reports suggesting that antibiotics can distort the mutation rate, as random SNPs emerging in the genetic background of resistant clones could potentially fix together with well-established resistance-conferring mutations during the course of the treatment [4].
Looking at the identity of relevant variants, we found miss-sense mutations and InDels related to intermediary metabolism and respiration (oplA V674fs, Rv2141c T178A), a stress protein (Rv1636 D15A), lipid metabolism (pks6 G807C), and information pathways (recC A773D) (Supplementary Table 1), reinforcing the idea of within-host coexistence of different bacillary sub-populations of a single strain.