A total of 34 MDR and XDR M. tuberculosis isolates were collected by the Office of Disease Prevention and Control Region 10 (DPC 10) from patients with MDR/XDR-TB in the upper north of Thailand from 2005 to 2012. The identification of MDR/XDR-TB isolates was performed conventionally using the indirect proportion method on agar based medium with concentrations of isoniazid (0.2 µg/ml), rifampicin (40 µg/ml), etambutol (2 µg/ml) and streptomycin (4 µg/ml) as shown in Figure 1 [16] . When they resisted isoniazid and rifampicin, then ofloxacin (2.0 µg/ml) and kanamycin (30 µg/ml) susceptibility was tested on middle brook medium using the proportion method. All 34 isolates were subcultured in 5 ml of 7H9 broth with PANTA supplementation and 3% Ogawa- based medium. In all, 34 isolates were re-grown and tested for their phenotypes using first and second-line drugs (isoniazid, rifampicin, ofloxacin and kanamycin) by the proportion method with LJ medium as shown in Figure 1 [17] .

Drug susceptibility testing (DST) was conducted by the indirect agar proportion method on two of LJ medium and bacilli suspension of 10⁻² and 10⁻⁴ mg/ml for inoculation. The LJ medium were inoculated with following drugs isoniazid (0.2 µg/ml), rifampicin (40.0 µg/ml), ofloxacin (2.0 µg/ml), and kanamycin (30

µg/ml), para-nitrobenzoic acid and no drug (drug-free).

Totally, 34 isolates were cultured on solid media (Löwenstein-Jensen and OGAWA). Chromosomal DNA was extracted by commercial kit using the MolecuTech REBA MTB-MDR 2011 method. The purified DNA pellets were stored at 4˚C until sending to Macrogen, Korea for nucleotide sequencing. After that, the two loci were amplified by PCR: gyrA (ofloxacin) and rrs (kanamycin). The primers for gyrA PCR amplification followed that of Leung Kl, et al. in 2010 that were designed using gyrA including the QRDRs of this gene: forward, 5'-AAGAGCGCCACCGACATC-3' and reverse, 5'-CAGCATCTCCATCGCCAA-3'. The following steps were performed: one denature cycle at 95˚C for 2 min, followed by 30 cycles of 1 min at 95˚C, 1 min at 65˚C and 1 min at 72˚C, followed by elongation at 72˚C for 10 min. PCR product comprised 320 base pairs in size(11). To detect kanamycin resistance, the primers were designed to identify mutations in codon 1401 of rrs gene: 16s-1F (5'-CGTGGCCGTTTGTTTTGTC-3') and 16s-1F (5'-TGGTGCTCCTTAGAAAGGAGG-3') following the method described by Leung Kl, et al. in 2010. PCR product comprised 665 base pair fragments, according to the following steps: beginning denaturation at 95˚C for 2 min, followed by 35 cycles of 94˚C for 1 min, 60˚C for 1min, 68˚C for 2 min, followed by elongation at 68˚C for 10 min [11] . The amplified products were purified with a multiscreen filter plate (Millipore Corp., Bedford, MA, USA) and DNA sequencing data were produced by the ABI 3730xl DNA analyzer at Macrogen.

The sequencing data were used to determine the confidence level with scanners and ABI sequence chromatograms. The analysis was used to determine the presence and/or absence of gene mutations using the H37Rv sequence to make comparison.

The resistance gene, genetic site mutation data and DST results were compiled using excel sheet within a Microsoft Office Excel 2010 database. Statistical analysis was performed using STATA version 11.0. Univariate comparison of drug resistance and resistance gene was performed using Fisher’s exact test.

We analyzed 34 isolates in total: 24 MDR-TB isolates (70.5%), 9 Pre XDR-TB isolates (26.5%) and 1 XDR-TB isolate (3.0%). Phenotypic DST and DNA sequencing were studied in all isolates for two drugs, ofloxacin and kanamycin, to predict pre-XDR and XDR-TB [1] [6] . DNA sequencings were performed for two resistance genes: gyrA, and rrs. The distribution of genetic drug resistance with the second line anti-tuberculosis drug resistance was shown in Table 1. The frequency of gyrA and rrs gene mutation was found among patients with MDR, Pre-XDR and XDR TB in Table 2.

Eight isolates (23.5%) resisted to ofloxacin. Twenty isolates (79.4%) were positive for any gyrA mutation (Table 2). Overall sensitivity and specificity of any gyrA mutation to detect ofloxacin resistance were 75% and 19.2%, respectively (Table 3). No significant correlation was found between mutation in gyrA gene and ofloxacin resistance. gyrA 162 resistance was more sensitive (50%) to ofloxacin resistance than gyrA 21 (0%), gyrA 70 (12.5%), gyrA 87 (37.5%), gyrA 102 (0%), and gyrA 187 (25%).

Six isolates (17.6%) resisted to ofloxacin. Fifteen isolates (44.1%) were positive for any rrs mutation in (Table 2). The overall sensitivity and specificity of any rrs mutation to detect kanamycin resistance were 33.3% and 53.6%, respectively (Table 4). No significant correlation was observed between the rrs gene and kanamycin drug resistance (Table 4).

The genotypic distribution of gyrA mutations in both ofloxacin resistant and ofloxacin susceptible groups were not different. We found that the gyrA primer can detect mutation in gyrA 21, gyrA 70, gyrA 87, gyrA 102, gyrA 162, and gyrA 187 at 0%, 12.5%, 37.5%, 0%, 50.0% and 25.0% sensitivity, respectively, and at 96.2%, 96.2%, 20.1%, 96.2%, 57.7% and 61.5% specificity, respectively. Mutations in gyrA have been found in ofloxacin resistance at codons 88, 90, 91, and 94 [18] [19] [20] . One study presented that ofloxacin susceptibility was associated with gyrA mutations at codon 88 to 94 in East Asian strains (55.6%) and Indo-Oceanic strains (44.4%), but these mutations could not be found in Euro-American strains [20] . One related study showed that gyrA mutations at positions 88 to 94 had a high level of ofloxacin resistance [20] , especially gyrA 94 that exhibited high level of resistance to moxifloxacin and gatifloxam [21] . Some studies found that mutations in gyrA at codons 21, 95, and 668 were not associated with ofloxacin resistance [14] . Only gyrA mutation at codon 95 could be detected by commercial molecular tests [20] . Further studies are required to determine real lineage-specific differences regarding the ability of ofloxacin resistance in pre-XDR and XDR-TB strains to develop commercial molecular testing.

The rrs mutation were found in rrs 512, rrs 241, rrs 223, rrs 414, and rrs 408 at 16.7%, 0%, 0%, 16.7% and 16.7% sensitivity, respectively, and 96.4%, 92.9%, 82.1%, 82.1% and 71.4% specificity, respectively, in this study. A systematic review showed that the combination of M401G and C517T mutation sites in rrs could predicted kanamycin resistance with high sensitivity and specificity [9] . One study in 2001 reported that streptomycin resistance was related to mutations in the 500 region of rrs [22] . Our study followed primer patterns from one related study in Hong Kong that could predict 96.8% specificity and 64.2% sensitivity for kanamycin from rrs on gene marker 1401 [11] . The most frequently reported kanamycin associated rrs mutations included A1401G, C1402T and G1484T [11] [23] [24] . The rrs A1401G mutation varied from 50% to 90% in these groups [10] [11] [23] [25] [26] [27] [28] . A study from Pakistan also showed 78% of kanamycin resistance correlated to SNPs in rrs at nt1401 [28] and concluded that the commercial diagnostic tool determining the aminoglycoside drug resistance target at rrs nt 1401 could detect 78% of kanamycin resistance in XDR-TB isolates [28] . One related study in Thailand showed that the designed rrs primer using 3 positions and 1680 PCR product size could predict 100% specificity and 72.4% sensitivity of kanamycin resistant M/XDR-TB [29] . However, we found that the ability of rrs 1401 mutation to detect kanamycin resistance was quite low with sensitivity of 33.3% and specificity of 53.6%. From the result, this rrs primer is not useful to detect kanamycin resistance in this area.