Selection of Mutations To Detect Multidrug-Resistant Mycobacterium tuberculosis Strains in Shanghai, China

Abstract

Novel tools are urgently needed for the rapid, reliable detection of multidrug-resistant (MDR) and extensively drug-resistant (XDR) strains of Mycobacterium tuberculosis. To develop such tools, we need information about the frequency and distribution of the mycobacterial mutations and genotypes that are associated with phenotypic drug resistance. In a population-based study, we sequenced specific genes of M. tuberculosis that were associated with resistance to rifampin and isoniazid in 242 phenotypically MDR isolates and 50 phenotypically pan-susceptible isolates from tuberculosis (TB) cases in Shanghai, China. We estimated the sensitivity and specificity of the mutations, using the results of conventional, culture-based phenotypic drug susceptibility testing as the standard. We detected mutations within the 81-bp core region of rpoB in 96.3% of phenotypically MDR isolates. Mutations in two structural genes (katG and inhA) and two regulatory regions (the promoter of mabA-inhA and the intergenic region of oxyR-ahpC) were found in 89.3% of the MDR isolates. In total, 88.0% (213/242 strains) of the phenotypic MDR strains were confirmed by mutations in the sequenced regions. Mutations in embB306 were also considered a marker for MDR and significantly increased the sensitivity of the approach. Based on our findings, an approach that prospectively screens for mutations in 11 sites of the M. tuberculosis genome (rpoB531, rpoB526, rpoB516, rpoB533, and rpoB513, katG315, inhA-15, ahpC-10, ahpC-6, and ahpC-12, and embB306) could detect 86.8% of MDR strains in Shanghai. This study lays the foundation for the development of a rapid, reliable molecular genetic test to detect MDR strains of M. tuberculosis in China.

Multidrug-resistant (MDR) tuberculosis (TB), defined as resistance to at least rifampin (RIF) and isoniazid (INH), and extensively drug-resistant (XDR) TB, defined as additional resistance to any fluoroquinolone and one injectable second-line drug, are among the most serious health threats of the 21st century. The epidemic of MDR TB is especially severe in China, a nation with the world's second largest number of TB cases and the largest number of MDR TB cases (39). A recent study reported that 9.3% of all TB cases in China are MDR, almost twice the worldwide MDR prevalence (4.8%) (14). While a lot of attention has been focused on acquired drug resistance among TB patients who receive an inadequate treatment regimen or who cannot adhere to their treatment regimen, several studies also showed that a large number of MDR TB cases are likely caused by transmission of MDR strains of Mycobacterium tuberculosis (2, 19). Therefore, there is an urgent need for new tools and approaches that will provide a rapid, reliable, and cost-effective diagnosis of MDR TB, particularly in resource-limited settings. This will help to prevent transmission of MDR strains and to optimize treatment regimens for MDR cases.

Drug susceptibility testing by the conventional solid medium culture method is highly sensitive and specific but extremely slow, due to the slow growth of M. tuberculosis. Liquid culture methods can reduce the turnaround time but require specialized instrumentation and reagents and are not feasible in most resource-limited settings. New molecular diagnostic methods represent a potentially rapid and sensitive alternative to conventional diagnostics. The molecular basis for phenotypic rifampin resistance is linked to mutations in the 81-bp core region of rpoB. Phenotypic isoniazid resistance has been associated with mutations in katG, particularly at codon 315, as well as with mutations in inhA, the promoter of mabA-inhA, and the intergenic region of oxyR-ahpC (3, 33, 34, 37, 43). Recently, a database of tuberculosis drug resistance mutations (TBDReaMDB) was established (27), and several genotypic diagnostic methods based on specific drug resistance-conferring mutations were developed. Two line probe assays, the INNO-LiPARif.TB assay (Innogenetics, Belgium) and the GenoType MTBDR Plus assay (Hain Lifescience, Nehren, Germany), have been approved by the World Health Organization (WHO) as tools for the rapid diagnosis of MDR TB (11, 20, 22). These tools are rapid and reproducible, but performance varies by geographic location, depending on the prevalent strains of M. tuberculosis and the type and frequency of drug resistance-conferring mutations in the population being tested (20, 22). Therefore, a thorough understanding of the diversity of the mycobacterial genetic mutations will form the foundation for new diagnostic methods.

Despite the large number of MDR TB cases in China, relatively few studies have determined the prevalence of different drug resistance-conferring mutations among MDR clinical isolates. In this study, we investigated the type and frequency of drug resistance-conferring mutations that occurred among M. tuberculosis clinical isolates that were phenotypically MDR. Our goal was to identify and select a limited, parsimonious number of mutation sites that can be used to prospectively and rapidly screen isolates to detect MDR TB in Shanghai.

MATERIALS AND METHODS

Study population and isolates.

We performed a retrospective cohort study using the existing data and specimens at the Shanghai Municipal Center for Disease Control and Prevention (Shanghai CDC) for all MDR TB patients who were diagnosed in Shanghai, China, from March 2004 through November 2008 (44). Of TB patients who had more than one isolate, only one isolate was included in this study. As control strains, we randomly selected 50 clinical isolates from different TB cases reported in Shanghai during the same period that were phenotypically pan-susceptible to all four first-line drugs used in China (isoniazid, rifampin, ethambutol, and streptomycin).

DST.

Phenotypic drug susceptibility testing (DST) was routinely performed in the Tuberculosis Reference Laboratory (TRL) at the Shanghai CDC. TRL participated in the World Health Organization/International Union against Tuberculosis and Lung Disease Global Project on Anti-Tuberculosis Drug Resistance Surveillance (38).

The DST was performed by the egg-based Löwenstein-Jensen (LJ) proportion method, with the following critical drug concentrations: isoniazid, 0.2 μg/ml; rifampin, 40.0 μg/ml; streptomycin, 4.0 μg/ml; and ethambutol, 2.0 μg/ml (8). Species identification of M. tuberculosis was performed by conventional biochemical methods and PCR (18, 24), and strains that were not M. tuberculosis were excluded.

DNA extraction and sequencing.

DNAs of M. tuberculosis strains were extracted as previously described (28). The following fragments were amplified and sequenced: 411 bp of the rpoB gene (including the 81-bp core region), 322 bp of the katG gene (including codon 315), 320 bp of the inhA promoter region, 457 bp of the inhA gene, 359 bp of the intergenic region of oxyR-ahpC, and 400 bp of the embB gene (Sunny Co., Shanghai, China). Primer sequences and PCR conditions were based on previously reported studies (1, 4, 23, 30, 34). Primers were synthesized by Invitrogen Bio Co. (China). The PCR mixtures were prepared using 2× Taq MasterMix (Tiangen Co., China). Sequence data were assembled and analyzed by CLUSTAL W software (European Bioinformatics Institute). For MDR strains in which drug resistance-conferring mutations for isoniazid, rifampin, or both were not detected by sequencing, the phenotypic drug susceptibility tests were repeated.

Genotyping.

Variable-number-tandem-repeat (VNTR) genotyping was performed on all isolates in the study, following the protocol described by Zhang et al. (40). VNTR-7 testing was performed, followed by VNTR-16 testing for isolates that had identical VNTR-7 patterns. To minimize bias due to oversampling, only one strain was randomly selected from each clonal population, defined as two or more strains with the same VNTR genotype and the same drug resistance-conferring mutations. We used the deletion-targeted multiplex PCR (DTM-PCR) method to identify the Beijing genotype strains (6).

Statistical analysis.

We determined the number and frequency, expressed as a percentage (%) of the total number of isolates screened, of each drug resistance-conferring mutation among the clinical isolates in our sample. We estimated the individual and cumulative sensitivities and specificities of each drug resistance-conferring mutation and their respective 95% binomial confidence intervals (95% CI). We also identified a small set of mutations to be used as a diagnostic algorithm for the detection of MDR TB.

RESULTS

MDR TB.

From March 2004 through November 2008, 322 MDR TB patients were diagnosed in Shanghai, and 260 (80.7%) had a stored clinical isolate that was successfully recovered. Of these, 11 isolates were determined to not be MDR by repeat DST; they were excluded, along with 7 clustered MDR isolates with the same genotype. The remaining 242 (93.1%; 242/260 isolates) MDR isolates were analyzed. Most of the MDR isolates (213/242 isolates [88.0%]) were Beijing genotype strains. Of the 242 MDR isolates, 77.3% (187/242 isolates) also had phenotypic resistance to streptomycin, ethambutol, or both. Among the 90 MDR isolates that were also resistant to ethambutol, 80.0% (72/90 isolates) were resistant to all four first-line drugs.

Mutation patterns.

Specific resistance genes or regulatory regions were sequenced, and the sequencing results were compared with the DST results. Table 1 shows detailed information about the mutations that were detected in the 411-bp fragment of rpoB (including the 81-bp core region). Altogether, 96.3% (233/242 isolates) of the MDR isolates harbored mutations in 12 codons of the 81-bp core region of rpoB. The most frequently mutated rpoB codons were codons 531, 526, 516, and 533, with mutation frequencies of 61.2%, 19.4%, 7.4%, and 5.0%, respectively. No mutations in rpoB were detected in the 50 drug-sensitive isolates.

TABLE 1.

Distribution of mutations in the rpoB gene among 242 MDR isolates from tuberculosis cases in Shanghai, China

Locus	Nucleotide change	Amino acid change	No. (%) of isolates (n = 242)
rpoB531	TCG→TTG	Ser→Leu	143 (59.1)
	TCG→TGG	Ser→Trp	2 (0.8)
	TCG→TTT	Ser→Phe	1 (0.4)
rpoB531 and rpoB526	TCG→TGG and CAC→AAC	Ser→Trp and His→Asn	1 (0.4)
	TCG→GGG and CAC→CGC	Ser→Gly and His→Arg	1 (0.4)
rpoB526	CAC→GAC	His→Asp	13 (5.4)
	CAC→TAC	His→Tyr	15 (6.2)
	CAC→CGC	His→Arg	6 (2.5)
	CAC→AAC	His→Asn	3 (1.2)
	CAC→CCC	His→Pro	1 (0.4)
	CAC→CTC	His→Leu	1 (0.4)
rpoB526 and rpoB511	CAC→AAC and CTG→CCG	His→Asn and Leu→Pro	2 (0.8)
rpoB526 and rpoB533	CAC→AAC and CTG→CCG	His→Asn and Leu→Pro	2 (0.8)
rpoB526 and rpoB534	CAC→AAC and GGG→CGG	His→Asn and Gly→Arg	1 (0.4)
rpoB526 and rpoB529	CAC→CGC and CGA→CTG	His→Arg and Arg→Leu	1 (0.4)
rpoB516	GAC→GTC	Asp→Val	12 (5.0)
	GAC→CCC	Asp→Pro	1 (0.4)
rpoB516 and rpoB511	GAC→GGC and CTG→CCG	Asp→Gly and Leu→Pro	1 (0.4)
	GAC→GTC and CTG→CCG	Asp→Gly and Leu→Pro	2 (0.8)
rpoB516 and rpoB518	GAC→GGC and CTG→CCG	Asp→Gly and Leu→Pro	1 (0.4)
rpoB516 and rpoB533	GAC→GCC and CTG→CCG	Asp→Ala and Leu→Pro	1 (0.4)
rpoB533	CTG→CCG	Leu→Pro	7 (2.9)
rpoB533 and rpoB515	CTG→CCG and ATG→GTG	Leu→Pro and Met→Val	1 (0.4)
rpoB533 and rpoB518	CTG→CCG and ATG→GTG	Leu→Pro and Met→Val	1 (0.4)
rpoB513	CAA→CCA	Gln→Pro	4 (1.7)
	CAA→AAA	Gln→Lys	1 (0.4)
rpoB513 and rpoB511	CAA→CCA and CTG→CCG	Gln→Pro and Leu→Pro	2 (0.8)
rpoB515 and rpoB511	ATG→CTG and CTG→CCG	Met→Leu and Leu→Pro	1 (0.4)
rpoB522	TCG→TGG	Ser→Trp	1 (0.4)
	TCG→TTG	Ser→Leu	3 (1.2)
rpoB519	AAC→AAG	Asn→Lys	1 (0.4)
Wild type			9 (3.7)

To detect isoniazid resistance, we sequenced fragments of katG, inhA, the promoter region of mabA-inhA, and the intergenic region of oxyR-ahpC. A total of 89.3% (216/242 strains) of the MDR strains had mutations in one or more of these four regions (Table 2). The most frequent mutation site was katG315; among the 179 katG mutants, 98.3% (176/179 strains) had mutations in katG315. In the inhA gene, only two strains were detected with mutations at codon 25, and both of these strains had other INH resistance mutations. In contrast, 9.9% (24/242 isolates) of the MDR isolates harbored mutations in the promoter region of inhA. Eleven of these 24 isolates had no additional mutation, 7 had an additional mutation in katG315, and 6 had additional mutations in the intergenic region of oxyR-ahpC. Among 242 MDR isolates, 13.6% (33/242 isolates) of the isolates had nucleotide substitutions in the intergenic region of oxyR-ahpC. Seven of these 33 isolates had an additional katG mutation, and 6 had a nucleotide substitution in the promoter region of inhA. The remaining 60.6% (20/33 isolates) of the isolates lacked an additional mutation. All 50 pan-susceptible isolates had the wild-type sequence in rpoB, katG315, the promoter region of inhA, and the intergenic region of oxyR-ahpC.

TABLE 2.

Distribution of mutations in katG, inhA, and two regulatory regions (oxyR-ahpC intergenic region and promoter of mabA-inhA) among 242 MDR isolates from tuberculosis cases in Shanghai, China

Locus	Nucleotide change	Amino acid change	No. (%) of isolates (n = 242)
katG315	AGC→ACC	Ser→Thr	150 (62.0)
	AGC→ACG	Ser→Thr	2 (0.8)
	AGC→AAC	Ser→Asn	8 (3.3)
katG295	CAG→CCG	Gln→Pro	1 (0.4)
katG299 and ahpC-15	GGC→AGC and C→T	Gly→Ser	1 (0.4)
katG322	ACG→GCG	Thr→Ala	1 (0.4)
katG315 and inhA-15	AGC→ACC and C→T	Ser→Thr	6 (2.5)
katG 315 and inhA-8	AGC→ACC and T→C	Ser→Thr	1 (0.4)
katG315 and ahpC-15	AGC→ACC and C→T	Ser→Thr	1 (0.4)
katG315 and ahpC-32	AGC→ACC and G→A	Ser→Thr	1 (0.4)
katG315 and ahpC-52	AGC→ACC and C→T	Ser→Thr	1 (0.4)
katG315 and ahpC-10	AGC→ACC and G→A	Ser→Thr	1 (0.4)
katG315 and ahpC-6	AGC→ACC and G→A	Ser→Thr	1 (0.4)
katG315 and inhA3	AGC→ACC and GGA→GGC	Ser→Thr and Gly→Gly	2 (0.8)
katG315, inhA3, and ahpC-39	AGC→AAC, GGA→GGC, and C→T	Ser→Asn and Gly→Gly	2 (0.8)
inhA3 and ahpC-10	GGA→GGC and C→T	Gly→Gly	2 (0.8)
inhA25, inhA-8, and ahpC-6	ATC→ACC, T→C and, G→A	Ile→Thr	2 (0.8)
inhA-15	C→T		11 (4.6)
inhA-15 and ahpC-10	C→T and C→T		3 (1.2)
inhA-8 and ahpC-6	T→C and G→A		1 (0.4)
ahpC-39	C→T		2 (0.8)
ahpC-32	G→A		1 (0.4)
ahpC-12	C→T		6 (2.5)
ahpC-10	C→T		3 (1.2)
ahpC-9	G→A		3 (1.2)
ahpC-6	G→A		3 (1.2)
Wild type			26 (10.7)

We previously showed that embB306 is a highly specific (92.6%) genetic marker associated with MDR TB (30). In the present study, 38.0% (92/242 isolates) of MDR isolates had a mutation at embB306. The most common mutations were Met306Val (ATG→GTG) (n = 56) and Met306Ile (ATG→ATA) (n = 25). All of the strains with a mutation in embB306 also had mutations in rpoB, and 92.4% (85/92 strains) of them had a mutation in katG, inhA, or ahpC. Seven strains were not associated with any other isoniazid resistance-conferring mutation. No mutations in embB306 were detected in the 50 pan-susceptible isolates.

Detection of MDR TB.

To find a parsimonious set of mutations that could be used to rapidly and reliably diagnose MDR strains of M. tuberculosis in Shanghai, we estimated the individual and cumulative sensitivities of all drug-resistant mutation sites that were associated with each drug resistance phenotype (Table 3). Taken together, mutations in 12 codons of the 81-bp core region of rpoB and 14 mutation sites in four isoniazid resistance-conferring regions identified 96.3% of the strains that had phenotypic rifampin resistance and 89.7% of the strains that had phenotypic isoniazid resistance, respectively (Table 3). In this study, mutations in embB306 also identified 38.0% (92/242 isolates) of the phenotypically MDR isolates. In order to limit the number of mutations for an MDR TB screening test, we selected 11 mutations. In this basic algorithm, a mutation at rpoB531, rpoB526, rpoB516, rpoB533, or rpoB513 is associated with phenotypic rifampin resistance. A mutation at katG315, inhA-15, ahpC-10, ahpC-6, or ahpC-12 is associated with isoniazid resistance. Finally, a mutation at embB306 alone is a diagnostic marker for MDR TB (Table 4). This algorithm could be used to prospectively detect 93.8% of RIF resistance and 89.3% of INH resistance. By screening for rifampin and isoniazid resistance-conferring mutations and mutations in embB306, 86.8% (210/242 isolates; 95% CI, 81.8% to 90.8%) of the phenotypic MDR isolates in Shanghai, China, could be diagnosed (Table 5).

TABLE 3.

Individual and cumulative sensitivities of drug-resistant mutation sites in Mycobacterium tuberculosis compared with phenotypic resistance to isoniazid and rifampin in 242 phenotypically MDR isolates from Shanghai, China

Mutation site	Individual sensitivity		95% CI	% Sensitivity (cumulative)	95% CI	% Specificity (individual)
	No. of isolates detected/total no. of isolates	% Sensitivity
Mutations associated with phenotypic isoniazid resistance
katG315	176/242	72.7	66.7-78.2	72.7	66.7-78.2	100
inhA-15	20/242	8.3	5.1-12.4	78.5	72.8-83.5	100
ahpC-10	9/242	3.7	1.7-6.9	80.6	75.0-85.4	100
ahpC-6	7/242	2.9	1.2-5.9	83.1	77.7-87.6	100
ahpC-12	6/242	2.5	0.9-5.3	85.5	80.5-89.7	100
ahpC-39	4/242	1.7	0.5-4.2	86.4	81.4-90.4	100
inhA-8	4/242	1.7	0.5-4.3	86.4	81.4-90.4	100
ahpC-9	3/242	1.2	0.2-3.6	87.6	82.8-91.5	100
inhA25	2/242	0.8	0.1-3.0	87.6	82.8-91.5	100
ahpC-32	2/242	0.8	0.1-3.0	88.0	83.2-91.8	100
ahpC-15	2/242	0.8	0.1-3.0	88.4	83.7-92.2	100
katG295	1/242	0.4	0.0-2.3	88.8	84.2-92.5	100
katG299	1/242	0.4	0.0-2.3	88.8	84.2-92.5	100
katG322	1/242	0.4	0.0-2.3	89.3	84.7-92.9	100
Mutations associated with phenotypic rifampin resistance
rpoB531	148/242	61.2	54.7-67.3	61.2	54.7-67.3	100
rpoB526	47/242	19.4	14.6-25.0	79.8	74.1-84.6	100
rpoB516	18/242	7.4	4.5-11.5	87.2	82.3-91.1	100
rpoB533	12/242	5.0	2.5-8.5	90.9	86.6-94.2	100
rpoB513	7/242	2.9	1.1-5.9	93.8	90.0-96.5	100
rpoB511	8/242	3.3	1.4-6.4	94.2	90.5-96.8	100
rpoB522	4/242	1.7	0.4-4.2	95.9	92.5-98.0	100
rpoB518	2/242	0.8	0.1-3.0	95.9	92.5-98.0	100
rpoB515	2/242	0.8	0.1-3.0	95.9	92.5-98.0	100
rpoB534	1/242	0.4	0.0-2.3	95.9	92.5-98.0	100
rpoB529	1/242	0.4	0.0-2.3	95.9	92.5-98.0	100
rpoB519	1/242	0.4	0.0-2.3	96.3	93.1-98.3	100
Additional mutation associated with MDR
embB306	92/242	38.0	31.9-44.5	38.0	31.9-44.5	100

TABLE 4.

Algorithm for detection of monoresistance and MDR TB^a

Genotype	Mutation profile			Prediction
	rpoB531, rpoB526, rpoB516, rpoB533, or rpoB513	katG315, inhA-15, or ahpC-12	embB306
1	Y	Y	Y	MDR
2	Y	Y	N	MDR
3	Y	N	Y	MDR
4	N	Y	Y	MDR
5	N	N	Y	MDR
6	Y	N	N	Mono-RIFr
7	N	Y	N	Mono-INHr
8	N	N	N	Susceptible

^aAny mutation in RIF-related sites is indicative of RIF resistance. Any mutation in INH-related sites is indicative of INH resistance. A mutation at embB306 alone is indicative of MDR TB. Y, at least one mutation; N, no mutation; r, resistant.

TABLE 5.

Minimal set of mutations that can be used to detect MDR strains of Mycobacterium tuberculosis^a

Mutation site	% Sensitivity (% CI) for resistance detection
	RIF	INH	MDR
rpoB531, rpoB526, rpoB516, rpoB533, or rpoB513	93.8 (90.0-96.5)		82.6 (77.3-87.2)
katG315, inhA-15, or ahpC-10, ahpC-6, or ahpC-12		85.5 (80.5-89.7)
embB306	93.8 (90.0-96.5)	89.3 (84.7-92.9)	86.8 (81.8-90.8)

^aA screening test that uses mutations in rpoB531, rpoB526, rpoB516, rpoB533, and rpoB513, katG315, inhA-15, ahpC-10, ahpC-6, and ahpC-12, and embB306 can detect 86.8%(81.8% to 90.8%) of MDR strains(tested using 242 isolates).

DISCUSSION

New tools for the rapid, reliable, and cost-effective diagnosis of MDR TB are urgently needed to control MDR epidemics globally, particularly in countries such as China, where most of the world's MDR TB patients reside. Molecular diagnostics could potentially fill this need but require data about the type and frequency of specific drug resistance-conferring mutations. Our study adds to this growing body of knowledge and allows the selection of a parsimonious set of mutations to screen in a highly sensitive and specific diagnostic test for MDR TB.

Mutations in the 81-bp core region of rpoB detected 96.3% of the phenotypic rifampin resistance among MDR isolates. Although the frequencies of particular mutations may differ, our results are consistent with the majority of surveys in both China and other regions of the world. Rifampin-monoresistant strains are rare, and according to our recent retrospective study of all pulmonary TB cases in Shanghai, 70.6% of rifampin-resistant strains were also resistant to isoniazid (29). Therefore, rifampin resistance and mutations in rpoB are sensitive indicators of MDR in Shanghai.

Isoniazid resistance-conferring mutations were most frequently detected at katG315. Mutations in this site are thought to introduce only a slight fitness cost to the bacterium (10, 25). According to Gagneux et al., this mutation may be more prevalent in mycobacterial strains of the Euro-American lineage, while the East Asian lineage may be associated with katG mutations other than those at katG315 (10). However, 98.1% (159/162 strains) of katG mutations in the Beijing genotype strains in our sample were in codon 315. A study of MDR isolates in central China also found a high percentage (100%) of katG mutations at codon 315 (42). Discrepancies could be explained by geographically based prevalence and mutation frequency differences associated with different sublineages of M. tuberculosis within the East Asian lineage (15, 21, 35). Another explanation is that there may be a higher selective pressure for katG315 mutations in MDR isolates. Supporting this hypothesis is a study of INH monoresistance in five Chinese provinces which found only 68.0% of katG mutations at codon 315 (41, 42).

We also sequenced katG and the promoters of inhA and ahpC in 95 isoniazid-monoresistant strains and the 81-bp core region of rpoB of 43 rifampin-monoresistant strains. A total of 81.1% of the isoniazid-monoresistant strains contained mutations in katG, inhA, or ahpC (see Table S1 in the supplemental material). As expected, only 52.6% of strains contained katG315 mutations, which is significantly lower than the frequency in MDR strains (72.7%; P < 0.01). A total of 88.4% of isolates contained mutations in the 81-bp core region of rpoB (see Table S2 in the supplemental material). However, the mutation frequency at rpoB531 (39.5%) was much lower than that found in MDR strains (61.2%; P < 0.01). One possible explanation for the difference in the frequencies of rpoB531 mutation is a higher selective pressure for rpoB531 in MDR isolates (9).

Although most katG mutations in the MDR strains are at codon 315, the frequency of mutations in katG315 (72.7%) in our MDR sample was lower than that found in the Russian Federation (93.6%) and Brazil (87.1%) (1, 32). The prevalence we detected and those reported for other populations in China, including Beijing (60.6%), Eastern China (64.4%), and Hong Kong (41.5%), indicate that it is not sufficient to screen for mutations in katG315 alone as a marker of phenotypic isoniazid resistance in China (5, 16, 41). Mutations in the promoter region of inhA were found in 9.9% (24/242 isolates) of the MDR isolates in our sample. Mutations in this region result in the overexpression of the InhA protein, which is sufficient to confer the isoniazid resistance phenotype (33). In our sample, 70.8% (17/24 isolates) of the isolates with a mutation in the promoter region of inhA did not have a mutation in katG. The combination of mutations in katG315 and inhA-15 can detect 78.5% of INH-resistant isolates in the MDR population.

Mutations in the intergenic region of oxyR-ahpC that result in the upregulation of ahpC have been associated with the absence of catalase-peroxidase activity and may be compensatory mutations for the survival of mycobacteria (31, 33). However, many studies have reported that mutations in the intergenic region of oxyR-ahpC are rare among isoniazid-resistant strains that also have mutations at katG315 (10, 17, 26). The katG315 mutation diminishes rather than inactivates the catalase-peroxidase activity, thereby allowing in vivo survival of mycobacteria in the absence of compensatory ahpC mutations (17). In our study, mutations in ahpC were detected among 13.6% (33/242 strains) of the MDR strains, 66.7% (22/33 strains) of which could not be detected by katG315 or inhA-15 screening. A majority of these strains had no other mutation in the sequenced fragment of katG. We also found that 7.4% of 94 isoniazid-monoresistant strains had a single mutation in ahpC. It is likely that there are additional katG mutations outside the 322-bp fragment we sequenced that encode the loss of catalase-peroxidase activity. Although there is no direct relationship between mutations in the intergenic region of oxyR-ahpC and phenotypic isoniazid resistance, from a practical, diagnostic point of view, we highly recommend including this mutation region for rapid diagnosis of MDR in Shanghai.

Using a limited number of mutation sites to detect MDR is the approach for the development of cost-effective commercial products such as the INNO-LiPARif.TB assay and the GenoType MTBDR Plus assay. Although both of these products have a high sensitivity and specificity to detect rifampin resistance, there are large geographic variations in the sensitivity of the products to detect isoniazid resistance and MDR. In our sample, the three most common mutations (rpoB531, rpoB526, and katG315) detected only 59.9% of the MDR strains. In contrast, these three codons (rpoB531, rpoB526, and katG315) can detect up to 90% of the MDR strains in North Africa (36). The GenoType MTBDR Plus assay, which screens for five sites in the rpoB core region, katG315, inhA-15, inhA-16, and inhA-8 (T/C and T/A), can detect 79.8% of the MDR isolates in Shanghai, based on our results. The sensitivity of the diagnostic assay is not as high in China as it is in many other regions of the world (20).

The inclusion of embB306 can also increase the sensitivity of MDR detection, with high specificity. Mutations in embB306 were originally thought to be associated with resistance to ethambutol. However, several studies reported embB306 mutations in ethambutol-susceptible isolates and found an association between embB306 mutations and multidrug resistance (12, 13, 30). In the present study, we detected a mutation in embB306 among 38.0% (92/242 strains) of the 242 MDR strains. Our previous study showed that mutations in embB306 can predict MDR in isolates, with a specificity as high as 92.6% (30). From the study of Hazbon et al., we estimated the specificity of embB306 for MDR to be approximately 90% worldwide (12). Therefore, we propose that embB306 be included in a screening assay to detect MDR in China and other regions of the world.

It is estimated that at least 70% of all MDR TB cases need to be diagnosed and that 80% need to be cured to effectively control MDR TB epidemics (7). In the present study, the 11 most common drug resistance-conferring mutation sites would detect 86.8% of the MDR TB cases, with 100% specificity. This specificity may be overestimated because non-MDR drug-resistant strains were not included in our control group. Therefore, the next step will be to confirm our findings in prospective studies by screening all isolates from TB suspects for mutations in the 11 drug resistance-conferring mutations we identified. Screening for additional mutation sites only marginally increased the sensitivity of the assay in Shanghai but may be worthwhile in future studies for different populations in diverse geographical areas. Given the severity of the MDR TB epidemic in China, our findings should lead to the development of rapid, reliable, point-of-care diagnostic methods which are easily accessible to TB programs and their patients in high-burden countries.