Journal of Global Antimicrobial Resistance 16 (2019) 170–177

Contents lists available at ScienceDirect

Journal of Global Antimicrobial Resistance

journal homepage: www.elsevier.com/locate/jgar

Use of whole-genome sequencing to predict Mycobacterium

tuberculosis drug resistance in Indonesia
Lidya Chaidira,1,* , Carolien Ruesenb,1, Bas E. Dutilhc , Ahmad R. Ganiema,d,
Anggriani Andryanie, Lika Apriania , Martijn A. Huynenc, Rovina Ruslamia,d , Philip C. Hillf ,
Reinout van Crevelb , Bachti Alisjahbanaa,d
a
Infectious Disease Research Center, Faculty of Medicine, Universitas Padjadjaran, Eijkman 38, Bandung 40161, Indonesia
b
Department of Internal Medicine, Radboud University Medical Center, Geert Grooteplein-Zuid 10, Nijmegen 6525 GA, The Netherlands
c
Centre for Molecular and Biomolecular Informatics, Radboud University Medical Center, Geert Grooteplein 8, Nijmegen 6500 HB, The Netherlands
d
Faculty of Medicine, Universitas Padjadjaran/Hasan Sadikin Hospital, Pasir Kaliki 38, Bandung 40161, Indonesia
e
West Java Provincial Referral Laboratory, Sederhana 3-5, Bandung 40161, Indonesia
f
Centre for International Health, Dunedin School of Medicine, University of Otago, P.O. Box 56, Dunedin 9054, New Zealand

A R T I C L E I N F O A B S T R A C T

Article history: Objectives: Whole-genome sequencing (WGS) is rarely used for drug resistance testing of Mycobacterium
Received 20 October 2017 tuberculosis in high-endemic settings. Here we present the first study from Indonesia, which has the third
Received in revised form 6 June 2018 highest tuberculosis (TB) burden worldwide, with <50% of drug-resistant cases currently detected.
Accepted 23 August 2018
Methods: WGS was applied for strains from 322 human immunodeficiency virus (HIV)-negative adult TB
Available online 29 August 2018
patients. Phenotypic drug susceptibility testing (DST) was performed for a proportion of the patients.
Results: Using WGS, mutations associated with drug resistance to any TB drug were identified in 51
Keywords:
(15.8%) of the 322 patients, including 42 patients (13.0%) with no prior TB treatment (primary resistance).
Mycobacterium tuberculosis
Drug resistance
Eight isolates (2.5%) were multidrug-resistant (MDR) and one was extensively drug-resistant (XDR). Most
Whole-genome sequencing mutations were found in katG (n = 18), pncA (n = 18), rpoB (n = 10), fabG1 (n = 9) and embB (n = 9).
Resistance mutations Agreement of WGS-based resistance and phenotypic DST to first-line drugs was high for isoniazid and
rifampicin but was lower for ethambutol and streptomycin. Drug resistance was more common in Indo-
Oceanic lineage strains (37.5%) compared with Euro-American (18.2%) and East-Asian lineage strains
(10.3%) (P = 0.044), but combinations of multiple mutations were most common among East-Asian
lineage strains (P = 0.054).
Conclusions: These data support the potential use of WGS for more rapid and comprehensive prediction of
drug-resistant TB in Indonesia. Future studies should address potential barriers to implementing WGS,
the distribution of specific resistance mutations, and the association of particular mutations with
endemic M. tuberculosis lineages in Indonesia.
© 2018 International Society for Chemotherapy of Infection and Cancer. Published by Elsevier Ltd. All
rights reserved.

1. Introduction about 4-6 weeks after the isolation of Mycobacterium tuberculosis.

Furthermore, inconsistencies in DST results are widely reported,
Drug-resistant tuberculosis (TB) threatens the global control of especially for ethambutol (EMB) and second-line drugs [3,4],
TB in many parts of the world. There were an estimated 580 000 whilst phenotypic DST for pyrazinamide (PZA) requires different
cases of multidrug-resistant TB (MDR-TB) in 2015, with <50% being protocols [5]. Molecular-based DST testing using the Xpert MTB/
detected and even fewer receiving appropriate treatment [1,2]. RIF assay can rapidly detect most rifampicin (RIF) resistance [6],
Culture-based drug susceptibility testing (DST), the gold standard but this assay has incomplete sensitivity for RIF, does not examine
to diagnose drug-resistant TB, is technically difficult and takes resistance for other TB drugs and may fail to detect heteroresist-
ance [7,8].
Whole-genome sequencing (WGS) has been shown to be a
potential tool for reliable prediction of the drug susceptibility
* Corresponding author. Present address: Infectious Disease Research Center,
Faculty of Medicine, Universitas Padjadjaran, Eijkman 38, Bandung, Indonesia.
phenotype of M. tuberculosis isolates within a clinically relevant
E-mail address: [email protected] (L. Chaidir). timeframe [9]. However, so far WGS is mostly applied in well-
1
These two authors contributed equally to this article. resourced, low TB burden settings. Because of rapid advances in

https://doi.org/10.1016/j.jgar.2018.08.018
2213-7165/© 2018 International Society for Chemotherapy of Infection and Cancer. Published by Elsevier Ltd. All rights reserved.
 L. Chaidir et al. / Journal of Global Antimicrobial Resistance 16 (2019) 170–177 171

WGS technology and its decreasing cost and turnaround time, WGS measured using a NanoDropTM 2000 spectrophotometer (Thermo
is now becoming accessible in limited-resourced, high TB burden Fisher Scientific, Waltham, MA) and the intactness of DNA was
countries [10]. In the current study, WGS was applied to predict checked by agarose gel electrophoresis.
drug resistance in a selection of M. tuberculosis isolates from M. tuberculosis DNA was sequenced on an Illumina HiSeq 2000
Indonesia, which has the third-highest TB burden in the world. In instrument (Illumina Inc., San Diego, CA) using 2  100-bp paired-end
Indonesia, diagnosis of drug-resistant TB is challenging as M. reads. Sequencing was performed at BGI (Hong Kong). After
tuberculosis culture is not routinely performed and DST is only sequencing, the raw FASTQ sequence reads were filtered, including
available in certain reference laboratories. To date, the contribution removal of adapter sequences, contamination, and low-quality reads
and concordance of phenotypic and genotypic DST in Indonesia is that had more than 10% N base calls or had a quality score 4 in more
unknown. This study aimed to describe resistance-conferring than 40% of the bases. Five TB meningitis strains and four pulmonary
mutations to first- and second-line TB drugs in Indonesia using TB strains were contaminated, based on a low GC content and were
WGS and to examine their concordance with phenotypic DST and excluded from further analyses. Sequencing coverage was deter-
their distribution among different M. tuberculosis lineages. mined using the FastQC v.0.10.1 (http://www.bioinformatics.babra-
ham.ac.uk/projects/fastqc/) quality control tool. The proportion of
2. Materials and methods bases sequenced with a sequencing error rate of 1% per base ranged
from 93% to 97% per genome. The average depth of coverage for the
2.1. Selection of M. tuberculosis isolates and drug susceptibility remaining 322 sequenced strains was 121.1, and the average
testing percentage of bases covered by at least one read was 98.9%. The
sequence reads were aligned to reference strain M. tuberculosis H37Rv
WGS was performed on a random sample of archived M. (GenBank accession no. NC_000962.3) and variants were called
tuberculosis isolates from 322 human immunodeficiency virus using Breseq software v.0.27.1 [13]. Mutations with low-quality
(HIV)-negative adult patients (216 with pulmonary TB and 106 evidence (i.e. possible mixed read alignment) were not included.
with meningeal TB) with complete medical information. Pulmo-
nary TB patients had been diagnosed in Hasan Sadikin Hospital 2.3. In silico determination of drug resistance
(Bandung, Indonesia) between 2012–2013 and in the TB-HIV
Research Centre of Universitas Padjadjaran (Bandung, Indonesia) Raw FASTQ sequencing files were uploaded to TB Profiler, an
between 2013–2015. TB meningitis patients had been diagnosed at online tool to determine drug resistance in silico [10]. It uses raw
Hasan Sadikin Hospital between 2006–2013 [11]. Specimens from sequence data as input and compares identified single nucleotide
each patient were processed accordingly and were inoculated on polymorphisms (SNPs) and indels to a curated list of 1325 drug
solid Ogawa medium or in MODS (Microscopic Observation of Drug resistance mutations and displays related output. The precision of
Susceptibility) liquid medium [12]. Positive cultures from each TB Profiler’s curated mutation catalogue for predicting resistance
method were subcultured and aliquots were archived at 80  C had been assessed using six geographically distinct data sets from
prior to DNA extraction. China, Pakistan, Malawi, Portugal, Russia and Canada [10], and its
Xpert MTB/RIF was available in the study setting only after 2012 accuracy compared with other in silico drug resistance prediction
and was accessible only for patients with suspected MDR-TB tools has been proven recently [14].
according to Indonesian national guidelines. Phenotypic DST was
not performed routinely for all patients but only when requested 2.4. Phylogeny construction
by treating physicians. DST was performed in the provincial
referral laboratory using the proportion method on Löwenstein– A phylogenetic tree was constructed to determine the evolution-
Jensen (LJ) medium at concentrations of 40.0 mg/mL for RIF, 0.2 mg/ ary relationship of the isolates. All 29 199 variable positions
mL for isoniazid (INH), 2.0 mg/mL for EMB and 4.0 mg/mL for identified by breseq across the 322 M. tuberculosis sequences were
streptomycin (STR). Briefly, a 1.0 McFarland standard isolate extracted and concatenated into a single alignment. Solely for the
suspension was serially diluted 10-fold (from 10 1 to 10 5) in purpose of creating the phylogenetic tree, SNPs occurring in PE/PPE
sterile distilled water. Dilutions 10 3 and 10 5 were inoculated, genes and genes related to mobile elements (listed in Supplemen-
respectively, onto LJ slants with and without drugs, and were tary Table S1) were excluded to avoid any concern about
incubated at 37  C. The results were read at 28 days and up to 42 inaccuracies in the read alignment in these parts of the genome.
days, depending on the control growth. An isolate was considered In addition, SNPs in an additional 40 genes previously associated
resistant to a given drug when growth of 1% above the control with drug resistance [10] were also removed to exclude the
was observed in drug-containing medium. DST for second-line possibility that homoplasy of drug resistance mutations would
drugs was not available during patient inclusion for this study. significantly affect the phylogeny [15]. After applying these filters to
Phenotypic DST was repeated for several isolates that had tested the initial set of 29 199 SNPs, the 28 544 remaining SNPs were used
INH-susceptible but that harboured mutations conferring resis- to construct the phylogenetic tree using PhyML v.3.0 [16] using the
tance to INH in the katG315 gene. In these cases, M. tuberculosis was HKY85 model with four categories for the gamma model of rate
subcultured from frozen isolates onto Ogawa slopes prior to DST. distributions and 100 bootstraps.
The study protocols for the inclusion of patients and for bioanalysis To examine possible associations between M. tuberculosis
were approved by the Ethical Committee of the Faculty of Medicine lineage and drug susceptibility, the lineage was determined for
of Universitas Padjadjaran. each of the 322 strains using a 62-SNP barcode [17]. The resulting
classification in the main M. tuberculosis lineages also served as a
2.2. Whole-genome sequencing and analysis quality check for the generated maximum-likelihood phylogenetic
tree, as it enabled us to validate that isolates belonging to the same
Frozen isolates were subcultured on Ogawa slopes and lineage clustered together in the tree.
mycobacterial DNA was extracted for sequencing using Ultra-
Clean1 Microbial DNA Isolation Kit (MO BIO Laboratories, 2.5. Statistical analysis
Carlsbad, CA) following the manufacturer’s protocol, or using
the cetyltrimethylammonium bromide (CTAB) method of DNA The χ2 test was used to statistically test the association between
purification. The concentration and purity of extracted DNA were M. tuberculosis lineage and drug susceptibility. Cohen’s k was used
172 L. Chaidir et al. / Journal of Global Antimicrobial Resistance 16 (2019) 170–177

to determine the level of agreement between WGS and phenotypic 432 and 441 were found together in one isolate. One MDR-TB
DST for first-line drugs. In addition, enrichment values were isolate with rpoB Ser450Leu also harboured the well-known
calculated for drug resistance per lineage based on the ratio of compensatory mutation at rpoC Leu527Val.
lineage-specific observed and expected occurrence of drug For EMB, nine isolates (2.8%) showed mutations in the embB
resistance. The ratios were visualised in a heat map as a measure gene, with the mutation Met 306Val being most common (n = 4),
of association between M. tuberculosis lineage and drug suscepti- and rare mutations at codon 406 and codon 497 in three MDR
bility. isolates. PZA resistance-conferring mutations were identified in 18
isolates (5.6%). All mutations occurred as a single mutation in five
3. Results codons in the pncA gene. For fluoroquinolones (FQs), mutations in
gyrA codon 90 and 94 were identified in 4 isolates (1.2%), three with
3.1. Patient characteristics FQ monoresistance and one XDR-TB isolate. With regard to
resistance to STR and injectable agents, mutations were found
Patients were mostly young (median age 33 years, interquartile in the rpsL and rrs loci in nine isolates (2.8%). Mutations at rrs codon
range 23.5–45.0 years), 51.9% were male and 16.1% reported a 492 were only found in STR-monoresistant strains. The XDR-TB
history of TB treatment. Using WGS, mutations associated with isolate harboured a mutation at the eis promoter linked to
drug resistance to any TB drug were identified in 51 (15.8%) of the kanamycin resistance.
322 patients; 42 (13.0%) were identified to have primary
resistance, MDR-TB was present in 8 patients (2.5%), including 3.3. Agreement of phenotypic and genotypic drug susceptibility testing
primary MDR-TB in 5 patients (1.6%) and primary extensively drug- to first-line drugs
resistant TB (XDR-TB) in 1 patient (0.3%) (Table 1; Supplementary
Table S2). Overall, there was considerable agreement between genotypic
and phenotypic DST. Phenotypic DST for the first-line drugs RIF,
3.2. Mutations associated with drug resistance EMB and STR was available for 103 isolates with WGS data and for
INH it was available for 102 isolates, with resistance found to INH
Genetic variants in multiple genes associated with drug (n = 17), RIF (n = 7), EMB (n = 6) and STR (n = 7). Concordance
resistance in M. tuberculosis were identified by WGS (Table 2). A between WGS and phenotypic DST was high for RIF (k = 0.865;
total of 29 isolates (9.0%) had mutations in genes associated with P < 0.001) and INH (k = 0.814; P < 0.001) but was low for EMB
resistance to INH, including katG (n = 18), kasA (n = 3) and the (k = 0.712; P < 0.001) and STR (k = 0.136; P = 0.148). Agreement of
promoter region of fabG1 C-15T (n = 9). Nine isolates with genotypic and phenotypic DST for each drug is shown in Fig. 1.
mutations at the fabG1 promoter site were predicted to have co- The common mutations katG Ser315Thr and fabG1 C-15T were
resistance to INH and ethionamide. The most common mutation in highly predictive of phenotypic INH resistance in this setting
katG was Ser315Thr (n = 14), but the uncommon mutation katG (89.5% of the isolates with either one of these mutations were
Ser315Met was also found in two isolates, and katG Trp191Arg and phenotypically resistant). Similarly, mutations in the RIF resis-
katG Ala106Val were found in one isolate each. The katG Ser315Thr tance-determining region (RRDR) of rpoB were highly predictive of
mutation was found in seven of eight MDR-TB isolates (Table 2). resistance to RIF (100% of the isolates with a mutation were
RIF resistance-conferring mutations were identified in ten phenotypically resistant). Nine isolates showed drug resistance-
isolates (3.1%), all in the rpoB gene. Six common mutations associated mutations but were susceptible by phenotypic DST. In
distributed over codons 435, 445 and 450 were found, mostly as those isolates, mutations were found for STR at rrs C492T (n = 3)
single mutations at one of these positions. Rare mutations at codon and A514C (n = 1) and rpsL Lys43Arg (n = 1); for INH at katG
Ser315Thr (n = 1), katG Trp191Arg (n = 1), fabG1 C-15T (n = 1) and
kasA Gly312Ser (n = 3); and for EMB at embB Met306Val (n = 1) and
Table 1
Presence of drug resistance mutations in 322 clinical Mycobacterium tuberculosis Met306Ile (n = 1). Conversely, nine other isolates that were drug-
isolates detected by whole-genome sequencing. resistant according to phenotypic DST (n = 7 for STR, n = 2 for RIF
and n = 1 for EMB) showed no known drug resistance mutations
Resistance pattern No. (%) of strains
using WGS. However, one isolate with phenotypic RIF resistance
Susceptible to all drugs 271 (84.2)
but without well-known drug resistance mutations harboured
Resistant to any drug 51 (15.8)
Resistant to first-line drugs
other mutations in rpoB (rpoB Cys681Gly and rpoB Pro1014Ser).
Any first-line drug 48 (14.9)
Isoniazid 29 (9.0) 3.4. Phylogenetic distribution of drug resistance
Rifampicin 10 (3.1)
Ethambutol 8 (2.5)
Drug resistance rates and patterns differed significantly between
Pyrazinamide 18 (5.6)
Streptomycin 9 (2.8) different M. tuberculosis lineages (Fig. 2; Supplementary Table S3).
Resistant to second-line drugs Among 116 isolates belonging to the East-Asian lineage, 12 (10.3%)
Ethionamide 9 (2.8) were genotypically drug-resistant compared with 36 (18.2%) of 198
Fluoroquinolones 4 (1.2) Euro-American strains and 3 (37.5%) of 8 Indo-Oceanic strains
Amikacin 1 (0.3)
Kanamycin 1 (0.3)
(χ2 = 6.258; P = 0.044). Although fewer strains belonging to the East-
Monoresistance Asian lineage had drug resistance mutations, they more often had
Isoniazid 10 (3.1) multiple drug resistance mutations (χ2 = 5.844; P = 0.054). From the
Rifampicin 1 (0.3) phylogeny, it was observed that most of the pncA mutations in
Pyrazinamide 15 (4.7)
isolates harbouring PZA resistance clustered together. The nine
Streptomycin 3 (0.9)
Fluoroquinolones 3 (0.9) isolates with a pncA Thr87Met mutation were adjacent in the tree,
Resistance to multiple drugs and the same goes for the six isolates carrying the pncA His82Arg
Multidrug-resistant (MDR) 8 (2.5) mutation, suggesting transmitted resistance. However, the isolates
Extensively drug-resistant (XDR) 1 (0.3) differed by more than 12 SNPs, the commonly used threshold for
Polyresistanta 10 (3.1)
(recent) transmission [18]. The other three mutations occurred only
a
Defined as resistance to multiple drugs but not MDR/XDR. once and in genetically distant strains.
 L. Chaidir et al. / Journal of Global Antimicrobial Resistance 16 (2019) 170–177 173

Table 2
Distribution of drug resistance-associated mutations in 51 Mycobacterium tuberculosis isolates with any drug resistance identified by whole-genome sequencing.

Drug Gene Amino acid change No. of isolates No. of MDR-TB isolates
Isoniazid katG Ser315Thr 14 7a
Ser315Met 2 0
Trp191Arg 1 0
fabG1 C-15T promoter 8 1
kasA Gly312Ser 3 0
Double loci fabG1 + katG C-15T promoter + Ala106Val 1 1
Rifampicin rpoB His445Tyr 1 1
His445Asp 1 1
Asp435Val 1 1
Asp435Tyr 2 2
Ser450Leu 2 2a
His445Cys 1 1
Gln432Lys + Ser441Leu 1 0
Double loci rpoB + rpoC Ser450Leu + Leu527Val 1 1
Ethambutol embB Met306Ile 2 1a
Met306Val 4 1
Gly406Asp 1 1
Gln497Lys 1 1
Met306Val + Gly406Asp 1 1
Streptomycin rrs C492T 3 0
A514C 1 1
rpsL Lys43Arg 5 3a
Pyrazinamide pncA His82Arg 6 0
Thr87Met 9 1
Ser66Pro 1 1
Pro62Leu 1 0
Ala171Val 1 1a
Ethionamide fabG1 C-15T promoter 9 2
Fluoroquinolones gyrA Asp94Gly 1 0
Ala90Val 1 0
Asp94Asn 1 0
Asp94Ala 1 1a
Amikacin rrs A514C 1 1
Kanamycin eis G-14A promoter 1 1a

MDR-TB, multidrug-resistant.
a
Mutation occurred in the extensively drug-resistant (XDR) isolate.

4. Discussion strains belonging to the Indo-Oceanic lineage compared with the

Euro-American and East-Asian lineages. On the other hand, strains
This is the first study to report the use of WGS on clinical M. belonging to the East-Asian lineage more often harboured multiple
tuberculosis isolates from Indonesia, showing its potential for mutations. One possible explanation is that the East-Asian lineage,
clinical management and TB control in Indonesia. Indonesia has a which includes Beijing strains, appears to have a higher mutation
huge gap between MDR-TB incidence and MDR-TB treatment. rate that could lead to accelerated acquisition of drug resistance
Together with China, India, Nigeria and the Russian Federation, it mutations. It may be one of the reasons why this lineage has been
accounted for >60% of this gap on a global scale in 2015 [2]. Of the repeatedly associated with drug resistance [22,23]. However, this
10 000 estimated MDR- or RIF-resistant notified pulmonary TB would also lead to more resistance against a single drug.
cases in Indonesia in 2015, only an estimated 15% were started on Published studies from Indonesia have clearly shown dispar-
treatment [2]. A large part of this gap can be explained by poor ities in the relative distribution of M. tuberculosis genotypes across
detection of drug resistance. Phenotypic DST is poorly accessible in the archipelago. In this article, we reported resistance mutations in
remote parts of the country. WGS could offer a rapid and a urban setting in Java where modern lineage 2 (East-Asian) and
comprehensive diagnostic solution, especially with the introduc- lineage 4 (Euro-American) are highly prevalent. These findings
tion of portable platforms and the decreasing price and turnaround should be confirmed in other regions, especially eastern Indonesia,
time of WGS [18], leading to quicker and more appropriate where lineage 1 (Indo-Oceanic) occurs at a substantially higher
treatment. frequency [24–26]. Current catalogues of drug resistance muta-
In a random collection of patient isolates, it was found that tions rely predominantly on data from modern lineages 2 and 4
15.8% carried mutations associated with drug resistance to any TB [10,27]. A recent study from India, where lineage 1 and 3
drug, often in patients without a history of TB treatment. This is predominate, reported putative novel resistance-conferring muta-
much lower than reported in an earlier study in Indonesia where tions in lineage 1 that had not previously been implicated in
resistance to first-line drugs was reported in 38.2% of 262 culture- resistance [27], suggesting that the mutations underlying geno-
positive samples, although these samples were collected over the typic drug resistance differ by lineage.
whole country and resistance was determined by phenotypic DST WGS offers the possibility to learn more about the influence of
[19]. In the current study, resistance mutations to all first-line the genetic background of strains on all aspects of drug resistance
drugs, FQs and second-line injectable drugs were found. MDR-TB evolution in M. tuberculosis. WGS could provide insight into
was present in eight isolates (2.5%), which is lower than expected secondary compensatory mutations, not conferring resistance but
based on recent survey data [2], and XDR-TB was present in one reducing the fitness cost of the resistance mutation by interacting
isolate (0.3%). epistatically with it. The rate of acquiring new drug resistance
In contrast to the findings from previous studies [19–21], in the mutations and the fitness costs of these mutations may vary as a
current study drug resistance was observed to occur more often in function of the strain genetic background [27,28].
174 L. Chaidir et al. / Journal of Global Antimicrobial Resistance 16 (2019) 170–177

Fig. 1. Comparison of phenotypic drug susceptibility testing (DST) and drug resistance testing by whole-genome sequencing (WGS) for Mycobacterium tuberculosis isolates to
(A) any first-line drug, (B) isoniazid (INH), (C) rifampicin (RIF), (D) ethambutol (EMB) and (E) streptomycin (STR). Light grey, drug resistance determined by WGS; dark grey,
drug resistance determined by phenotypic DST. Data are presented as number of strains. n = 103 for RIF, EMB and STR and n = 102 for INH.

Most mutations associated with INH resistance were found in mutations and INH-, RIF- or multidrug-resistant TB [32]. This
katG (62.1%) and the promoter region of fabG1 (31.0%). Testing only finding suggests that embB306 mutations may have a selective
for katG and fabG1 promoter mutations would detect INH advantage upon treatment with multiple drugs [33]. Regarding
resistance in 89.6% of the INH-resistant isolates. All mutations STR, isolates carried mutations in rpsL (n = 5) and rrs (n = 4). A
associated with RIF resistance were found only in the RRDR of rpoB. common variant in rpsL (Lys43Arg) has been associated with high-
This therefore confirms the usefulness of these three specific level STR resistance [32] and this variant was indeed the most
regions for prediction of INH and RIF resistance [29]. Conversely, common in this study. Phenotypic STR resistance was only
kasA Gly312Ser and katG Trp191Arg mutations were found in confirmed in two of seven genotypically STR-resistant isolates
phenotypically INH-susceptible isolates. KasA (β-ketoacyl ACP with DST results available. Three isolates had a single mutation in
synthase), coded by the kasA gene, is an enzyme involved in rrs codon 492; their relevance should be carefully interpreted since
mycolic acid synthesis. Mutations in the kasA gene have been this mutation has been reported as a marker for the LAM3 genetic
associated with low-level INH resistance [30,31] but the role of lineage of M. tuberculosis rather than for STR resistance [34].
these mutations in INH resistance remains unclear. Therefore, the In line with previous studies, concordance between WGS and
significance of this gene in conferring drug resistance in Indonesia phenotypic DST was good for INH and RIF [29] but had low
should be carefully assessed. Two isolates were phenotypically RIF- agreement for EMB and STR [35,36]. There are several possible
resistant but harboured mutations in rpoB that have not been explanations for this finding. First, discordance was mainly found
associated with drug resistance before (Cys681Gly and Pro1014S- with uncommon genotypic mutations, which may be associated
er), therefore their possible role in RIF resistance should be with low-level resistance that can be missed by conventional
confirmed in other studies. phenotypic DST. Second, the difference between the epidemiologi-
Regarding EMB, three loci of the embB gene (loci 306, 406 and cal breakpoint and the minimum inhibitory concentration (MIC)
497) were identified, with mutations at locus 306 being the most for EMB and STR is relatively small [35,36], which complicates
frequent. Mutations in embB were present only in combination phenotypic DST. Third, detection of phenotypic STR resistance in
with other drug resistance mutations, in line with several studies the absence of known mutations for STR resistance suggests the
that have demonstrated a strong association between embB306 existence of other resistance mechanisms such as efflux pumps
 L. Chaidir et al. / Journal of Global Antimicrobial Resistance 16 (2019) 170–177 175

Fig. 2. Phylogenetic tree of 322 Mycobacterium tuberculosis isolates indicating drug resistance profiles and main lineages based on single nucleotide polymorphism (SNP)
barcoding. INH, isoniazid; RIF rifampicin; EMB, ethambutol; PZA, pyrazinamide; STR, streptomycin; ETH, ethionamide; FQS, fluoroquinolones; AMK, amikacin; KAN,
kanamycin.

[36]. Finally, we cannot exclude errors in phenotypic DST despite resistance is common in MDR-TB strains [40] although significant
good quality control in the laboratory. In this regard, the fact that rates of PZA monoresistance have been reported in some settings
one isolate was phenotypically drug-susceptible with a high [40–42]. A study in China showed that PZA monoresistance
confidence katG Ser315Thr mutation for INH in WGS is a matter of contributes to delay in the resolution of lung cavitation without
concern since this mutation is associated with high-level resis- affecting the sputum conversion and lesion elimination rates [41].
tance [29,37]. Given this high specificity, mutations at katG315 The non-essential nature of the pncA gene, such that it can
might even be used to assess laboratory quality for DST. accumulate various mutations without affecting the viability of the
Second-line drug resistance was observed in a number of organism, might explain the spread of PZA-monoresistant strains.
isolates. At present, there is no standard phenotypic DST for PZA Two pncA mutations exclusively occurred in strains adjacent in the
and the second-line drugs in Indonesia, and evidence is limited on phylogenetic tree, suggesting possible transmission. However, the
the performance of DST for these drugs [38], indicating once more genetic difference between these strains was too large to conclude
the potential added value of WGS in this setting. Also, reported that this was indeed the case.
prior treatment for TB poorly corresponded with genotypic drug This study has several limitations. First, WGS was performed on
resistance, and primary drug resistance was common in this study a random sample of archived M. tuberculosis isolates so we cannot
population of Indonesian patients. Of concern, the only XDR-TB conclude that the proportions of drug resistance shown in this
strain was found in a patient who had reported no prior TB study are representative. This was a convenience sample collected
treatment. This suggests that selective use of genotypic or from patients with pulmonary and meningeal TB. We estimate that
phenotypic DST, targeting only those with previous treatment or several thousands of patients are treated for TB each year in
other factors presumed to be associated with drug resistance, will Bandung. However, culture is not routinely performed and isolates
result in many undetected resistant cases and ongoing transmis- are not archived. Therefore, our sampling fraction is likely to be
sion of drug-resistant strains [39]. This is further supported by the <10%. Second, phenotypic DST was available for most genotypically
observation that 10 of 51 isolates with mutations associated with resistant strains but only for a fraction of isolates without
resistance to any drug were not phenotypically tested for drug resistance mutations. As a consequence, specificity estimation
susceptibility because they did not have these risk factors. WGS for genotypic resistance was not possible. Third, sequencing was
could help in early identification of resistance in these patients, performed retrospectively on archived isolates, therefore it was
limiting the spread of drug-resistant TB. not possible to evaluate time to diagnosis of drug resistance using
Molecular drug resistance for PZA mostly involves mutations in WGS. Nevertheless, we for the first time highlight the potential
the pncA gene. There is no predominant drug resistance mutation benefit of using WGS to generate an in silico drug susceptibility
in pncA but a range of diverse mutations across the gene, each profile in Indonesia and show that mutations associated with drug
associated with a different MIC [5]. Five different mutations in resistance are highly predictive for phenotypic resistance to RIF
pncA were found in the isolates in this study, almost all in PZA- and INH in the region. Larger studies are needed to confirm the
monoresistant strains. Most previous studies reported that PZA clinical relevance of several uncommon mutations found in this
176 L. Chaidir et al. / Journal of Global Antimicrobial Resistance 16 (2019) 170–177

strain collection. Given the fact that phenotypic DST is complex [8] Zetola NM, Shin SS, Tumedi KA, Moeti K, Ncube R, Nicol M, et al. Mixed
and slow and is poorly accessible in large parts of Indonesia, WGS Mycobacterium tuberculosis complex infections and false-negative results for
rifampin resistance by GeneXpert MTB/RIF are associated with poor clinical
could offer a rapid and comprehensive diagnostic solution. This outcomes. J Clin Microbiol 2014;52:2422–9.
technology is now more accessible with the introduction of [9] Witney AA, Cosgrove CA, Arnold A, Hinds J, Stoker NG, Butcher PD. Clinical use
portable platforms and an automatic bioinformatic pipeline as well of whole genome sequencing for Mycobacterium tuberculosis. BMC Med
2016;14:46.
as the decreasing price and turnaround time of WGS [43]. It could [10] Coll F, McNerney R, Preston MD, Guerra-Assuncao JA, Warry A, Hill-Cawthorne
lead to quicker and more appropriate treatment in low-income G, et al. Rapid determination of anti-tuberculosis drug resistance from whole-
settings where many still rely on empirical treatment regimens. genome sequences. Genome Med 2015;7:51.
[11] van Laarhoven A, Dian S, Ruesen C, Hayati E, Damen MSMA, Annisa J, et al.
This pilot study offers a good starting point to further evaluate the Clinical parameters, routine inflammatory markers and LTA4H genotype as
impact of WGS in the diagnosis, treatment, surveillance and predictors for mortality among 608 tuberculous meningitis patients in
control of drug-resistant TB in Indonesia. Indonesia. J Infect Dis 2017;215:1029–39.
[12] Chaidir L, Annisa J, Dian S, Moore DAJ, Muhsinin S, Parwati I, et al. MODS
culture for primary diagnosis of tuberculous meningitis and HIV-associated
Funding pulmonary tuberculosis in Indonesia. Int J Trop Dis Health 2013;3:346–54.
[13] Deatherage DE, Barrick JE. Identification of mutations in laboratory-evolved
This study is a result of collaborative projects between Indonesia microbes from next-generation sequencing data using breseq. Methods Mol
Biol 2014;1151:165–88.
and the Netherlands. Inclusion of patients and collection of M. [14] Farhat MR, Shapiro BJ, Sheppard SK, Colijn C, Murray M. A phylogeny-based
tuberculosis isolates in the TB meningitis cohort were financially sampling strategy and power calculator informs genome-wide associations
supported by the Royal Academy of Arts and Sciences (KNAW), study design for microbial pathogens. Genome Med 2014;6:101.
[15] Guindon S, Dufayard JF, Lefort V, Anisimova M, Hordijk W, Gascuel O. New
Foundation for Advancement of Tropical Research (NWO-WOTRO), algorithms and methods to estimate maximum-likelihood phylogenies:
and IMPACT, a 5-year HIV programme supported by the European assessing the performance of PhyML 3.0. Syst Biol 2010;59:307–21.
Commission. Inclusion of patients and collection of M. tuberculosis [16] Coll F, Preston M, Guerra-Assuncao JA, Hill-Cawthorn G, Harris D, Perdigao J,
et al. PolyTB: a genomic variation map for Mycobacterium tuberculosis.
isolates in the pulmonary TB cohort were supported by TANDEM (TB Tuberculosis (Edinb) 2014;94:346–54.
and Diabetes Mellitus) from the European Community’s Seventh [17] Walker TM, Ip CLC, Harrell RH, Evans JT, Kapatai G, Dedicoat MJ, et al. Whole-
Framework Programme [FP7/2007-2013] and Universitas Padjad- genome sequencing to delineate Mycobacterium tuberculosis outbreaks: a
retrospective observational study. Lancet Infect Dis 2013;13:137–46.
jaran, Indonesia. Whole-genome sequencing was financially
[18] Witney AA, Gould KA, Arnold A, Coleman D, Delgado R, Dhillon J, et al. Clinical
supported by a VIDI grant from the Netherlands Organisation for application of whole-genome sequencing to inform treatment for multidrug-
Scientific Research [NWO 017.106.310], The Netherlands. LC has a resistant tuberculosis cases. J Clin Microbiol 2015;53:1473–83.
[19] Lisdawati V, Puspandari N, Rif’ati L, Soekarno T, M Melatiwati, K Syamsidar, et al.
fellowship from the Indonesian Ministry of Research, Technology
Molecular epidemiology study of Mycobacterium tuberculosis and its suscepti-
and Higher Education (Indonesia) and Radboud University Medical bility to anti-tuberculosis drugs in Indonesia. BMC Infect Dis 2015;15:366.
Center (The Netherlands), and CR has a fellowship from Radboud [20] Black PA, de Vos M, Louw GE, van der Merwe RG, Dippenaar A, Streicher EM, et al.
University Medical Center. BED was supported by the Netherlands Whole genome sequencing reveals genomic heterogeneity and antibiotic
purification in Mycobacterium tuberculosis isolates. BMC Genomics 2015;16:857.
Organisation for Scientific Research [NWO VIDI grant 864.14.004]. [21] Aung HL, Tun T, Moradigaravand D, Koser CU, Nyunt WW, Aung ST, et al.
The funders had no role in study design, data collection and analysis, Whole-genome sequencing of multidrug-resistant Mycobacterium tuberculosis
decision to publish, or preparation of the manuscript. isolates from Myanmar. J Glob Antimicrob Resist 2016;6:113–7.
[22] Fenner L, Egger M, Bodmer T, Altpeter E, Zwahlen M, Jaton K, et al. Effect of
mutation and genetic background on drug resistance in Mycobacterium
Competing interests tuberculosis. Antimicrob Agents Chemother 2012;56:3047–53.
[23] Trauner A, Borrell S, Reither K, Gagneux S. Evolution of drug resistance in
tuberculosis: recent progress and implications for diagnosis and therapy.
None declared. Drugs 2014;74:1063–72.
[24] Chaidir L, Sengstake S, de Beer J, Krismawati H, Lestari FD, Ayawaila S, et al.
Mycobacterium tuberculosis genotypic drug resistance patterns and clustering
Ethical approval in Jayapura, Papua, Indonesia. Int J Tuberc Lung Dis 2015;19:428–33.
[25] Chaidir L, Sengstake S, de Beer J, Oktavian A, Krismawati H, Muhapril E, et al.
Health Research Ethics Committee, Faculty of Medicine, Predominance of modern Mycobacterium tuberculosis strains and active
transmission of Beijing sublineage in Jayapura, Indonesia Papua. Infect Genet
Universitas Padjadjaran (Bandung, Indonesia).
Evol 2016;39:187–93.
[26] Sasmono RT, Massi MN, Setianingsih TY, Wahyuni S, Anita, Halik H, et al.
Appendix A. Supplementary data Heterogeneity of Mycobacterium tuberculosis strains in Makassar, Indonesia.
Int J Tuberc Lung Dis 2012;16:1441–8.
[27] Gygli SM, Borrell S, Trauner A, Gagneux S. Antimicrobial resistance in
Supplementary data associated with this article can be found, in Mycobacterium tuberculosis: mechanistic and evolutionary perspectives. FEMS
the online version, at https://doi.org/10.1016/j.jgar.2018.08.018. Microbiol Rev 2017;41:354–73.
[28] Manson AL, Abeel T, Galagan JE, Sundaramurthi JC, Salazar A, Gehrmann T, et al.
Mycobacterium tuberculosis whole genome sequences from Southern India
References suggest novel resistance mechanisms and the need for region-specific
diagnostics. Clin Infect Dis 2017;64:1494–501.
[1] Falzon D, Jaramillo E, Wares F, Zignol M, Floyd K, Raviglione MC. Universal [29] Rodwell TC, Valafar F, Douglas J, Qian L, Garfein RS, Chawla A, et al. Predicting
access to care for multidrug-resistant tuberculosis: an analysis of surveillance extensively drug-resistant Mycobacterium tuberculosis phenotypes with
data. Lancet Infect Dis 2013;13:690–7. genetic mutations. J Clin Microbiol 2014;52:781–9.
[2] World Health Organization. Global tuberculosis report. 2016 http://apps.who. [30] Unissa AN, Subbian S, Hanna LE, Selvakumar N. Overview on mechanisms of
int/medicinedocs/en/d/Js23098en/. [Accessed 27 December 2018]. isoniazid action and resistance in Mycobacterium tuberculosis. Infect Genet
[3] Lee RS, Behr MA. The implications of whole-genome sequencing in the control Evol 2016;45:474–92.
of tuberculosis. Ther Adv Infect Dis 2016;3:47–62. [31] Lee ASG, Lim IHK, Tang LLH, Telenti A, Wong SY. Contribution of kasA analysis
[4] Salamon H, Yamaguchi KD, Cirillo DM, Miotto P, Schito M, Posey J, et al. to detection of isoniazid-resistant Mycobacterium tuberculosis in Singapore.
Integration of published information into a resistance-associated mutation Antimicrob Agents Chemother 1999;43:2087–9.
database for Mycobacterium tuberculosis. J Infect Dis 2015;211(Suppl 2):S50–7. [32] Jagielski T, Ignatowska H, Bakula Z, Dziewit L, Napiorkowska A, Augustyno-
[5] Ramirez-Busby SM, Valafar F. Systematic review of mutations in pyrazina- wicz-Kopec E, et al. Screening for streptomycin resistance-conferring
midase associated with pyrazinamide resistance in Mycobacterium tuberculo- mutations in Mycobacterium tuberculosis clinical isolates from Poland. PLoS
sis clinical isolates. Antimicrob Agents Chemother 2015;59:5267–77. One 2014;9:e100078.
[6] World Health Organization (WHO). Xpert MTB/RIF assay for the diagnosis of [33] Bakula Z, Napiorkowska A, Bielecki J, Augustynowicz-Kopec E, Zwolska Z,
pulmonary and extrapulmonary TB in adults and children. Policy update. Jagielski T. Mutations in the embB gene and their association with ethambutol
Geneva, Switzerland: WHO; 2013. resistance in multidrug-resistant Mycobacterium tuberculosis clinical isolates
[7] Steingart KR, Schiller I, Horne DJ, Pai M, Boehme CC, Dendukuri N. Xpert1 from Poland. Biomed Res Int 2013;2013:167954.
MTB/RIF assay for pulmonary tuberculosis and rifampicin resistance in adults. [34] Villellas C, Aristimuno L, Vitoria MA, Prat C, Blanco S, Garcia de Viedma D, et al.
Cochrane Database Syst Rev 2014(1)CD009593. Analysis of mutations in streptomycin-resistant strains reveals a simple and
 L. Chaidir et al. / Journal of Global Antimicrobial Resistance 16 (2019) 170–177 177

reliable genetic marker for identification of the Mycobacterium tuberculosis Mycobacterium tuberculosis clinical isolates from northwestern Russia, 1996 to
Beijing genotype. J Clin Microbiol 2013;51:2124–30. 2001. Antimicrob Agents Chemother 2002;46:1417–24.
[35] Horne DJ, Pinto LM, Arentz M, Lin SY, Desmond E, Flores LL, et al. Diagnostic [40] Parsons LM, Salfinger M, Clobridge A, Dormandy J, Mirabello L, Polletta VL,
accuracy and reproducibility of WHO-endorsed phenotypic drug susceptibility et al. Phenotypic and molecular characterization of Mycobacterium tuberculosis
testing methods for first-line and second-line antituberculosis drugs. J Clin isolates resistant to both isoniazid and ethambutol. Antimicrob Agents
Microbiol 2013;51:393–401. Chemother 2005;49:2218–25.
[36] Martin LJ, Roper MH, Grandjean L, Gilman RH, Coronel J, Caviedes L, et al. [41] Kurbatova EV, Cavanaugh JS, Dalton T, Click ES, Cegielski JP. Epidemiology of
Rationing tests for drug-resistant tuberculosis—who are we prepared to miss? pyrazinamide-resistant tuberculosis in the United States, 1999–2009. Clin
BMC Med 2016;14:30. Infect Dis 2013;57:1081–93.
[37] Cui Z, Li Y, Cheng S, Yang H, Lu J, Hu Z, et al. Mutations in the embC–embA [42] Tan S, Rao Y, Guo J, Tan Y, Cai X, Kuang H, et al. The influence of pyrazinamide
intergenic region contribute to Mycobacterium tuberculosis resistance to monoresistance on treatment outcomes in tuberculosis patients from
ethambutol. Antimicrob Agents Chemother 2014;58:6837–43. southern China. J Tuberc Res 2016;4:9–17.
[38] Ssengooba W, Meehan CJ, Lukoye D, Kasule GW, Musisi K, Joloba ML, et al. [43] Cheng SJ, Thibert L, Sanchez T, Heifets L, Zhang Y. pncA mutations as a major
Whole genome sequencing to complement tuberculosis drug resistance mechanism of pyrazinamide resistance in Mycobacterium tuberculosis: spread
surveys in Uganda. Infect Genet Evol 2016;40:8–16. of a monoresistant strain in Quebec, Canada. Antimicrob Agents Chemother
[39] Mokrousov I, Narvskaya O, Otten T, Limeschenko E, Steklova L, Vyshnevskiy B. 2000;44:528–32.
High prevalence of KatG Ser315Thr substitution among isoniazid-resistant