Abstract
Whole genome sequencing (WGS) has been proposed as a tool for the diagnosis of drug resistance in tuberculosis (TB); however, there have been few studies on its effectiveness in countries with significantly high drug resistance rates. This study therefore aimed to evaluate the effectiveness of WGS to identify mutations related to drug resistance in TB isolates from an endemic region of Mexico. The results showed that, of 35 multidrug-resistant isolates analyzed, the values of congruence found between the phenotypic drug susceptibility testing and polymorphisms were 94% for isoniazid, 97% for rifampicin, 90% for ethambutol, and 82% for pyrazinamide. It was also possible to identify eight isolates as potential pre-extensive drug resistant (XDR) and one as XDR. Twenty nine isolates were classified within L4 and two transmission clusters were identified.
The results show the potential utility of WGS for predicting resistance against first- and second-line drugs, as well as providing a phylogenetic characterization of TB drug-resistant isolates circulating in Mexico.
Introduction
According to the 2018 report of the World Health Organization, there were 10 million new cases and 1.3 million deaths associated with tuberculosis (TB), making it the infectious disease of greatest impact on human health in 2017. 1 Close to 25% of TB cases recorded annually worldwide show resistance to one of the four first-line antibiotics used against this disease: rifampicin (R), isoniazid (H), ethambutol (E), and pyrazinamide (Z). Close to 5% of the cases evolved to mixed resistance against H and R, giving rise to multidrug-resistant tuberculosis (MDR-TB), which can evolve into forms with resistance to fluoroquinolone and the second-line injectable drugs amikacin, kanamycin, and capreomycin, causing extensive drug-resistant tuberculosis (XDR-TB).
Whole genome sequencing (WGS) in TB has been shown to provide accurate information regarding changes to the prediction of drug susceptibility over a short time period, with clinical relevance,2–9 and is also of value for examining the dynamic of transmission and phylogeny.10–13 However, there is limited evidence regarding the utility of WGS in countries with a high burden of DR-, MDR-, and XDR-TB.14–17
TB is endemic in Mexico, with an incidence of 22 cases per 100,000 inhabitants, 18 of which 2.5% show resistance to any drug. Moreover, the numbers of MDR- and XDR-TB are increasing. 19 For these reasons, Mexico is ranked among the top five contributors of TB on the American continent. 20 However, there is no information about the usefulness of WGS in terms of characterizing drug-resistant mutations in TB. Considering this, the aim of this study was to evaluate the utility of WGS for identification of mutations related to resistance and phylogenetic groups in MDR-TB isolates circulating in Mexico.
Materials and Methods
Sample collection, isolation of Mycobacterium tuberculosis isolates, and drug susceptibility test
Clinical samples with a diagnosis of MDR-TB were randomly recovered from the bank of isolates of the Public Health Laboratory of Veracruz, considering the period January 2014 to December 2016. Strains were earlier isolated in Lowenstein–Jensen media, and drug sensitivity against the first-line drugs R, H, E, and Z were tested by fluorometric assay according to the supplier's instructions (MGIT 960, Becton-Dickinson). Demographic information of the patients was obtained from clinical laboratory records.
DNA extraction, WGS sequencing, and analysis
Extraction and purification of genomic DNA was performed according to van Soolingen et al. 21 Quantification was conducted with a NanoDrop (Thermo Scientific) and adjusted to a concentration of 0.2 ng/μL. Libraries were prepared according to the Nextera XT protocol (Illumina, CA), using 1 ng of DNA quantified by Qubit fluorometer (Invitrogen, CA). Quality was determined using TapeStation (Agilent Genomics) and the pool was normalized and loaded in 300-cycle mid-output cartridge at 1.8 pM. Sequencing was performed with NexSeq 500 (Illumina) using a 2 × 150 paired-end format.
The WGS analysis, including mapping and variant calling (SNP and INDELS), was conducted following a previously described pipeline. 22 To detect drug resistance-associated mutations with different frequencies within the population, single polymorphisms with at least 10 reads in both strands and a quality score of 20 were classified into two categories. Variant-SNPs were considered as mutations that involved a frequency of 10% to 89%, whereas constant-SNPs were considered those with a frequency of >90%. An INDEL was considered, where the mutation was present with at least 10 × depth coverage. Annotation variant analysis was performed using Mycobacterium tuberculosis H37Rv annotation reference (AL123456.2).
To classify the isolates and determine whether they had a specific genotype and if they were related, in silico typing was performed. All of the strains were classified according the presence of the specific lineages or sublineages of the phylogenetic variants proposed by Coll et al. 23 and Stucki et al. 24 To detect highly likely genomic transmission clusters (TCs), analysis of pairwise genetic distances was conducted based on a concatenated SNP alignment obtained from every fixed SNP from each isolate. A ≤12 SNP threshold was applied to delineate the TCs, as proposed by Walker et al. 25
Ethical concerns
No physical interventions took place with the patients and all of the information was treated as confidential. The ethics committees of the Public Health Institute at University of Veracruz oversaw and approved the ethical issues involved in this study. No experiments were conducted on animals.
Results
Isolates recovered, patient data, and resistotyping
Of the 35 MDR-TB isolates recovered in the study period, 24 were male, mean age was 45 (±14), and 17 individuals had type 2 diabetes mellitus as a comorbidity. Twenty-six individuals were diagnosed for the first time, while the rest were relapses. In addition to the resistance found to H and R, resistance to E was observed in 11 strains (31%), with resistance to Z in 18 (51%), and to all first-line drugs in 6 (17%) (Table 1).
Concordance of Phenotypic Resistance Versus Genotypic Prediction by Whole Genome Sequencing of Multidrug-Resistant Tuberculosis Isolates from Veracruz
R, resistance.
Characterization of mutations and correlation between resistances identified by phenotypic and genotypic analyses
The most frequently observed polymorphisms in the sequenced isolates were: katG315S/T present in 22 (62%) isolates; rpoB450 S/L/W in 28 (80%); embA4243221 c/t in 8 (22%); and pnzA120 L/P in 12 (34%) (Tables 2 and 3).
Polymorphisms in Genes Related with Resistance to E and Z in Isolates from Mexico Identified by Whole Genome Sequencing Analysis
X indicates occurrence of the mutation.
Transmission cluster one.
Transmission cluster two.
Polymorphisms in Genes Related with Resistance to H and R in Isolates from Mexico Identified by Whole Genome Sequencing Analysis
X indicates occurrence of the mutation.
Transmission cluster one.
Transmission cluster two.
CAS, Central Asian; EAI, East African–Indian.
Table 1 shows the concordance between phenotypic resistance to each drug and prediction of resistance using the PhyResSE online system 9 and the pipeline described in the methods. In relation to H resistance, a 94% concordance was observed. Of the 35 isolates with phenotypic resistance included in the study, 2 (781_16, 588_14 and 906_16) showed no mutations related to resistance against this drug (Table 1). Isolate 882_16 presented the mutations katG621A/V and ahpC-OxyR 2726139 c/t, which are not reported as being related to H resistance (Table 2).
Regarding resistance to R, a 97% concordance was observed between the phenotypic and genotypic procedures. Only one of the isolates with phenotypic resistance (906_16) showed no mutation (Table 2).
Ten of the 11 isolates with phenotypic resistance to E were in concordance with the prediction of resistance by WGS analysis (90%) (Table 1). However, nine isolates, considered as sensitive, were identified as carriers of highly reliable mutations in embA4243221, embA4243225, embB306M/V/I, and embB1002 H/R (Table 3).
Fourteen isolates with resistance to Z presented concordance with the WGS analysis (82%) (Table 1). However, seven phenotypically sensitive isolates showed highly reliable mutations that confer resistance to Z (Table 1). The isolate 971_16 presented the change pncA136D/Y, while isolate 1292_16 presented the change pncA128V/G, isolates 214_15, 626_14 and 743_14 had the mutation pncA120 L/P, strain 1213_14 had the change pncA155V/G, and isolate 1289_16 showed the polymorphism pncA132G/A (Table 3).
Mutations related to second-line drug resistance and identification of pre- and XDR-TB isolates
Eleven mutations in four genes associated with resistance to second-line drugs were identified in nine isolates (Table 4). The 1292_16 isolate showed a change in the gene rrs 1473246 a/g. Two isolates, 1288_16 and 1290_16, showed the change gyrA94 N/A, while isolate 1120_16 showed simultaneous changes in gyrB 500T/N and rrs1473246a/g. Isolate 061_14 showed two mutations at gyrB500 T/N and gyrA90 A/V. Isolate 1305_14 showed the change gyrA94 N/A. Isolate 1289_16 presented the change eis2715342g/a. The strain 480_14 had the mutation gyrA94 N/A and, finally, isolate 1008_16 presented the polymorphism gyrA90A/V. No changes were observed in genes related to resistance against linezolid (rrl and rplC), aminoglycosides (tlyA), and ethionamide (ethR, mshA, inhA, ndh, and ethA) (Table 4).
Polymorphisms in Genes Related to Second-Line Drugs in Isolates from Mexico Identified by Whole Genome Sequencing Analysis
X indicates occurrence of the mutation.
Transmission cluster one.
Transmission cluster two.
AM, amikacin; CM, capreomycin; KM, kanamycin.
The phenotypic profile of sensitivity to second-line drugs was not determined, so it was not possible to analyze the concordance between both methods; however, analysis of the mutations in genes related to first- and second-line drug resistance predicted eight isolates as pre-XDR-TB, whereas isolate 1120_16 was predicted as XDR-TB (Table 4).
Identification of lineages and TCs
Four lineages (L) were identified (Table 2): lineage L1 (East African–Indian [EAI]), with one isolate (3%), lineage L2 (Beijing) with three isolates (9%), and L3 (Central Asian [CAS]) with two isolates (6%). The remaining 29 strains (82%) were classified within L4, and included eight sublineages (Table 2), the most abundant of which was 4.1.1.3 (X), which was observed in 12 (34%) isolates. Finally, isolates included in the two TCs were observed: TC1 included two isolates with L3 (CAS), sharing specific polymorphisms related to MDR and pre-XDR-TB characters. Finally, in TC2, isolate 061_14 had the mutations that defined it as pre-XDR, whereas isolate 1120_16 was considered as potential XDR-TB (Table 4).
Discussion
Through WGS, it was possible to identify the polymorphisms that would confer resistance with a phenotypic concordance of 94% for H, 97% for R, and 90% for MDR-TB. These values show the potential utility of WGS as a tool for screening resistance to the two most important drugs used against TB. This could be extremely valuable in terms of the early diagnosis of MDR-TB isolates in Mexico, a country that is one of the most important contributors of these types of cases in Latin America. 20
It is important to mention that only isolate 906_16 showed no mutation type, despite the fact that it was MDR-TB. Coincidentally, this isolate showed one of the lowest levels of coverage and depth, in terms of the quality of readings obtained. As a consequence, these factors should be closely monitored in further assays to ensure proper identification of the polymorphisms involved.
In accordance with reports from larger collections of strains, the congruence of WGS with R resistance was high. 7 For H resistance, one isolate (882_16) had an unknown nonsynonymous mutation in katG together with a mutation in the ahpC promoter region. The combination of these changes is usually predictive of isoniazid resistance, despite the fact that the individual mutations have not been described as of high mutation confidence. Isolate 480_14 showed an unusual mutation at ahpC. Isolates 906_16 had the isoniazid phenotypic profile, but we were unable to identify any causative mutation. Altogether, the results suggest that prediction for H is not as effective as has been observed elsewhere, 7 and may reproduce problems similar to those described in India. 26 Additional studies are clearly necessary to identify the particularities of the mutations causing MDR-TB in this setting.
Ten isolates phenotypically sensitive to E, seven to Z, and five isolates with simultaneous sensitivity to E and Z, showed a diverse set of mutations; some with high and others with low reliability, while some others were unreported. This diversity of changes in isolates from the region supports those found in previous studies.27–30 The presence of isolate 1008_16 bearing the deletion pncA472289101/-gttgc should be noted. This “local” change was described 5 years ago (2012) in isolates circulating exclusively in this region, 27 confirming that isolates bearing this deletion currently remain in active circulation in the population. The presence of this diversity and the occurrence of local polymorphisms identified by WGS in isolates sensitive to E and Z raises questions about the effectiveness of the phenotypic diagnostic test for resistance and highlights the need to develop further studies exploring these unreported changes, to confirm their role, and thus increase the effectiveness of WGS in predicting TB drug resistance in the region.
One of the major limitations of this study is the fact that was not possible to obtain the phenotypic profiles against second-line drugs in the isolates analyzed. Nevertheless, through WGS, eight isolates were predicted as pre-XDR and one as XDR-TB. WGS has been described as a suitable method by which to identify these isolate types with sensitivity and specificity values higher than 85%.4,6 The possibility of generating a faster screening diagnosis of aggravated forms of TB could have important implications in countries such as Mexico, where an increasing number of these isolate types are observed and where, under the national protocols, a considerable time is required to establish sensitivity against these second-line drugs.
Through WGS, it was possible to assign a specific lineage to all of the isolates analyzed and to identify L4 as the most frequently observed lineage. This is in contrast to previous reports, where the use of traditional techniques, such as MIRU-VNTR and spoligotyping, have described the presence of important numbers of orphan isolates, although they do agree with the presence of L4 (Euro-American) as the most important lineage.31–34 It was also possible to identify two TCs, of which TC2 included ten isolates with sublineage 4.1.1.3 (X3-X1). This sublineage, associated with a MDR-TB condition, has scarcely been referenced in Mexico,31–34 so a greater number of isolates analyzed by this procedure will be required to determine the true level of transmission. Considering the above, WGS could also represent a useful technique for the development of studies of phylogeny and genomic epidemiology in TB isolates from Mexico.
Another limitation of this study was related to the small number of samples analyzed. It will therefore be necessary to examine a greater number of isolates to obtain a better description of the lineages and TC circulating in the population, as well as to determine the frequencies of the polymorphisms that are circulating in the drug-resistant isolates from the area and to assess its value as a diagnostic probe.
In Mexico, the cost of determining sensitivity to first- and second-line drugs is close to 200 USD and, if Xpert MTB/RIF analysis is utilized, this increases to 350–400 USD per sample per patient. In contrast, the cost for sequencing a Mycobacterium genome is close to 250 USD and decreasing. This slight difference in terms of cost is a factor for consideration if WGS is to be used as a tool for characterization and diagnostic of DR-TB.
In conclusion, the information obtained in this study demonstrates the utility of WGS for identifying mutations that will help to predict resistance against first- and, potentially, second-line drugs in TB, and will contribute to increasing our understanding of both phylogeny and transmission. There is no doubt that this procedure can have a positive impact on the TB health program in Mexico.
Footnotes
Disclosure Statement
No competing financial interests exist.
Funding Information
E.F. was a fellow of CONACYT-Maestría en Ciencias de la Salud, Universidad Veracruzana No. 850447. D.V. was a fellow of CONACyT-Doctorado en Ciencias Biomédicas Universidad Veracruzana No. 661113. R.Z.-C. and I.C. were partially funded by I-COOP-2017- COOPB2032.