Discrepancies in Drug Susceptibility Test for Tuberculosis Patients Resulted from the Mixed Infection and the Testing System

To find the potential reasons for the discrepancies in the drug susceptibility test (DST) of M. tuberculosis isolates, twenty paired isolates with disputed drug susceptibilities to isoniazid (INH) were selected according to the MGIT960 testing and Löwenstein-Jensen (L-J) proportion methods. Their MICs were confirmed again by broth microdilution method and by L-J proportion method. The spoligotyping results showed that, of all the 20 paired strains, 11 paired isolates belonged to the Beijing genotype and 6 paired isolates belonged to SIT1634, and that each of the remaining 3 paired isolates had two genotypes, namely, SIT1 and SIT1634. Those 3 paired isolates with different intrapair spoligotypes were further confirmed as mixed infection by the results that those three pairs of isolates with different 12 locus MIRU intrapair types and one pair carried different base pair at codon 315 (AGC versus AAC). Totally mutations in the katG gene were identified in 13 paired isolates. No mutations were found in the regulatory sequences and open reading frames (ORF) of the inhA and ahpC genes in any of the tested isolates. Those results showed that the different test systems and the mixed infection with particular genotypes of M. tuberculosis strains contributed to the drug susceptibility discrepancies.


Introduction
Performance of drug susceptibility testing (DST) to measure drug resistance is important not only before treatment, but also in the course of therapy to identify acquired resistance, especially in the areas with a high incidence of MDR-TB [1]. Conventional DST methods rely on egg-based (Löwenstein-Jensen; L-J) or agar-based (Middlebrook) media, but these are laborious and time-consuming procedures requiring 3 to 8 weeks to obtain results [2]. A number of new methods for DST, including the mycobacterial growth indicator tube (MGIT) [3], test [4], and Alamar blue [5] methods, have been introduced over the last decade to detect mycobacteria rapidly and to improve their growth rates [6,7].
The BACTEC MGIT960 method has been assessed in many countries and its degree of agreement with conventional DST methods in M. tuberculosis has been assessed [8][9][10]. Meta-analysis of published results revealed high accuracy and high predictive value associated with the use of BACTEC MGIT960 [11]. However, there are still discrepancies in the DST results obtained for different anti-TB drugs between BACTEC MGIT960 and other DST methods. The discrepancies in INH susceptibility between the MGIT960 and L-J proportion methods, for example, varied from 0% to 1% [9]; however, few investigations have been reported that addressed the possible mechanisms underlying the discrepancies between the MGIT960 system and L-J proportion methods. Discrepancies can arise from many reasons, for example, different DST systems used, mixed infection with different M. tuberculosis strains, and last but not least, contamination. In this study, 20 paired isolates with disputed drug susceptibilities to INH were selected according to the MGIT960 testing and L-J proportion methods. The name of the "paired isolates" referred to the two isolates obtained separately from the cultures after the DST by MGIT960 and L-J proportion methods from the same sputum of the patient. The reasons for the DST discrepancies were analyzed by the spoligotyping and VNTR genotyping methods and drug resistance-related mutations tested in INH resistance-related genes.

Strains and Antibiotics.
A total of 20 paired M. tuberculosis isolates with DST discrepancies were collected in Tianjin Haihe Hospital in the year of 2006 from total 1412 isolates ( Figure 1). "paired isolates" were from the the culture of the MGIT960 and L-J proportion method, respectively, which was mentioned above. Meanwhile, 96 randomly selected isolates, whose MGIT960 and agar proportion DST results were in agreement, were also collected from the same hospital. The 20 paired M. tuberculosis isolates were determined to be sensitive to INH using the conventional L-J proportion method (1 g/mL) [12] but resistant to INH using the BACTEC MGIT960 method (0.1 g/mL, Becton Dickinson Microbiology Systems, MD, USA) [9]. M. tuberculosis H37Rv (ATCC 27294) obtained from the Chinese National Reference Laboratory was used as a control.

Determination of the MIC of INH by Middlebrook 7H9
Broth Microdilution and L-J Agar Dilution. Resazurin was used as an indicator to test the MIC of INH in the Middlebrook 7H9 broth microdilution method [13]. Briefly, a 100 L volume of Middlebrook 7H9 broth containing 0.05% Tween 80 and 10% OADC (Sigma, USA) was dispensed into the wells of a 96-well cell culture plate (Corning Coast). INH concentrations, in Middlebrook 7H9 medium, were as follows: 0.1, 0.2, 0.4, 0.8, 1.0, 1.2, 1.6, and 1.8 mg/L. Recovered isolates were collected from L-J slants and homogenized. Turbidity was adjusted to the number 1 McFarland standard (approximately 1 × 10 7 CFU/mL) and the suspension is diluted 1 : 10 and 100 L of the dilution is added in each well that contains 100 L of the appropriate INH dilution. The final inoculum concentration was 5 × 10 4 CFU/mL. The plates were sealed and incubated at 37 ∘ C for one week. Twentyfive microliter of 0.02% resazurin (Sigma Chem. Co., USA) solution was then added to each well and the plates were incubated for an additional 2 days. A change in color from blue to pink indicated the growth of bacteria and the MIC was read as the minimum INH concentration that prevented the color change in the presence of resazurin.
Determination of the MIC of INH using the L-J proportion method followed the protocol of the Chinese Anti-Tuberculosis Association [12]. INH concentrations used in the L-J medium were 2.0, 1.8, 1.6, 1.2, 1.0, 0.8, 0.6, 0.4, and 0.2 mg/L. About 10 5 CFU were inoculated on the INHcontaining medium slants and results were recorded after 5-6 weeks.

Genomic DNA Isolation, Polymerase Chain Reaction (PCR), and Sequence Analysis.
Colonies were first removed from the recovering slants by scraping, resuspended in 500 L of TE (10 mM Tris, 1 mM EDTA (pH 8.0)), and killed by heating at 80 ∘ C for 30 min. The DNA extraction method, primers (from CyberSyn Co. Beijing, China), and PCR conditions were as described previously [14]. The primers were designed to amplify the katG gene, including the region around codon 315, the inhA regulatory region, the inhA ORF, and oxyR-ahpC regions (Table 1) [15,16]. Both strands were sequenced for confirmation. Mutations were identified by BLAST comparisons with M. tuberculosis H37Rv as the reference (GenBank number NC 000962.3).

Molecular Typing by Spoligotyping and the 12-Locus MIRU
Method. Spoligotyping was performed with a commercial kit (Isogen Bioscience BV, Maarssen, The Netherlands) according to the manufacturer's instructions. Amplification of the direct variant regions for spoligotyping was performed essentially as described previously [17]. Interpretation of spoligotype patterns and assignment of octal codes were based on SITVIT2 database (Pasteur Institute of Guadeloupe, Parris, France), which is an updated version of the  [18]. The numbers of tandem repeats (TRs) at each locus in the isolates were determined on the basis of the number of whole repeats in a PCR product of the size estimated from the gel [19]. Polymerase chain reaction assays for the 12 chosen loci were repeated and compared within and between gels to ensure consistent estimation of size and TR copy number [20].

Genotyping Analysis.
Genotyping analysis can determine not only whether an infection results from transmission of the given tuberculosis isolate, but also whether the infection involves more than one strain of M. tuberculosis. Results from our genotyping analysis showed that 10 paired isolates belong to the Spoligotype International Type SIT1 (Beijing genotype, 000000000003771) and 6 paired isolates belong to the Spoligotype International Type SIT1634 (MANU2, 777777777723771) (Table 2), a spoligotype that was not found in the 96 randomly selected clinical isolates (Table 3). Three paired isolates were mixtures of the SIT1 and SIT1634 spoligotypes, and one pair was a mixture of SIT1 and the SIT269 (Beijing genotype, 000000000000771) spoligotypes. Compared with our set of 96 randomly selected isolates from Tianjin, only the Beijing and MANU genotypes were present and the percentage of the MANU genotype was extremely high (20 paired isolates: 15/40, 37.5%; 96 random clinical isolates: 3/96, 3.125%).
Results obtained by using the 12-locus MIRU method [19] showed that 20 pairs of isolates had 14 MIRU patterns. Both the spoligotyping and the MIRU patterns were different in the isolates named as 6, 12, and 18 pairs, individually. The isolates named as 7 pairs had different spoligotypes, but the same MIRU type ( Table 2).

MICs of the Tested Strains.
To identify the differences between the liquid Middlebrook 7H9 and L-J proportion methods in DST, we tested the MICs of each of the 16 paired INH-resistant isolates and 4 pairs of isolates which consisted of different genotypes using both Middlebrook 7H9 broth microdilution and L-J proportion methods. The MICs of all the 24 tested isolates were determined to be greater than 0.1 g/mL (0.1 to 0.6 g/mL) using the Middlebrook 7H9 broth microdilution method and greater than 0.3 g/mL (0.4 to 1.8 g/mL) using the L-J proportion method ( Table 4). The MICs of 5 pairs of the tested isolates using the L-J proportion method were higher than 1 g/mL, the cutoff concentration for determining drug susceptibility in the L-J agar proportion method in this study (Table 4).

Sequence Analysis of the Putative INH-Target Genes.
Mutations in the katG gene were identified in 13 paired isolates, of which each of 12 paired isolates carried the same mutations and one pair which showed a DST discrepancy by MGIT960 and L-J proportion methods carried different base pair at codon 315 (AGC versus AAC). The AGC315AAC mutation was found in 4 paired isolates, while 9 paired isolates carried the mutation AGC315ACC. The AGC315AAC and AGC 315ACC mutations were not associated with specificity to the Beijing or MANU2 genotypes among the tested isolates. Seven paired isolates did not contain mutations in the katG gene and no mutations were found in the regulatory sequences and open reading frames (ORF) of the inhA and ahpC genes in any of the tested isolates (Table 4).   Sparks, MD), and BACTEC MGIT 960 (BBL Becton Dickinson Microbiology Systems, Cockeysville, MD) systems [5,[21][22][23]. The DST results would be influenced by many steps of the protocol, including the culture and the DST methods. In this study, we analyzed the discrepancy of the drug susceptibility test by the MGIT and L-J methods for the isolates collected from the culture by MGIT and L-J, respectively. Except for the median time to report the DST results the M. tuberculosis complex culture positivity rates were also greatly different in MGIT and L-J [24], which indicated the possible culture preference to somewhat. And the detection time, accuracy, and performance capacity are also variable by different DST methods. Studies reported that the reasons for the different performance capacity among these methods mainly resulted from the different DST systems [23,25]. The most obvious difference is the drug concentrations used for the DST. In MGIT system, the sensitive strains were susceptible to the INH less than the 0.1 g/mL, while the concentration of the INH was 1 g/mL in L-J system in this study [12,26,27]. Of all the 20 paired cases 15 cases had MIC in borderlines between the MGIT and the DST methods, which was a usual reason for the discordant.

Discussion
Many reports showed that there was a good concordance between DST on L-J and MGIT for INH in DST [25][26][27]. In this study, we still found that 20 paired isolates with the same genotypes individually showed the discrepancy in the drug susceptibilities to INH according to the MGIT960 testing and L-J proportion methods. Lawson et al. demonstrated that there was a substantial degree of agreement between the two methods, with similar INH and rifampicin DST patterns, but more frequent detection of streptomycin resistance and less frequent detection of ethambutol with L-J than MGIT-960. However, the differences were not statistically significant [25]. A multiple center evaluation showed that the discrepancies in INH susceptibility between the MGIT960 and L-J proportion methods varied from 0% to 1% [9].
Mixed infection with the different genotypes of M. tuberculosis in the same patient also affected the DST results even by the same testing systems [28,29]. In this study heterogeneous genotypes were found in the isolates from each of the 4 patients. Three patients were infected by the different  (Table 4). Some mycobacterial characteristics might be associated with particular genotypes. A well-known but controversial example is that the Beijing family strains of M. tuberculosis are often associated with relapse [30], drug resistance [31], and an increased ability to cause disease, to be transmitted within certain geographic settings [32,33]. The isolates with particular genotypes, such as Spoligotype International Type SIT1634 (Manu2) in this study, showed higher rate of resistance in MGIT960 system than in L-J system. In this study, we found that the percentage of "MANU" genotype strains was markedly increased in paired isolates whose DST results showed discrepancies (37.5%) compared to the randomly selected clinical isolates (3.125%). An unusually high proportion of strains belonging to the "Manu" clade (27.15%) were also reported by Helal et al. [18]. Interestingly, Manu2 strains (SIT1634) have rarely been reported in Tianjin or even in China as a whole [34,35] or in the SPOLDB4 database (excluding this study, = 3, 1, from India and 2 from the USA).
In this study, all the 40 isolates were determined as resistant by MGIT and sensitive by L-J, of which twentyseven isolates were found with mutations in katG315 and 13 isolates were found with no mutations in katG315 (Table 4). Those results of the mutations found in the INH-targeted genes supported that the DST result by the MGIT was more accurate than that by the L-J, and we also found that the MICs of some isolates by L-J agar method were very higher than those in the first execution in clinic, which indicated, to some extent, that the operation needs to be improved in proportion method on L-J agar.

Conclusion
Our study confirmed that the discrepancies of the DST in M. tuberculosis clinical isolates did exist for INH. One of the reasons for the discrepancy is the different test systems between the BACTEC MGIT960 system and the traditional L-J proportion method. Mixed infection by the strains with MANU2 and Beijing genotype patterns could also contributed to drug discrepancies.