High rate of drug resistance among tuberculous meningitis cases in Shaanxi province, China

The clinical and mycobacterial features of tuberculous meningitis (TBM) cases in China are not well described; especially in western provinces with poor tuberculosis control. We prospectively enrolled patients in whom TBM was considered in Shaanxi Province, northwestern China, over a 2-year period (September 2010 to December 2012). Cerebrospinal fluid specimens were cultured for Mycobacterium tuberculosis; with phenotypic and genotypic drug susceptibility testing (DST), as well as genotyping of all positive cultures. Among 350 patients included in the study, 27 (7.7%) had culture-confirmed TBM; 84 (24.0%) had probable and 239 (68.3%) had possible TBM. DST was performed on 25/27 (92.3%) culture positive specimens; 12/25 (48.0%) had “any resistance” detected and 3 (12.0%) were multi-drug resistant (MDR). Demographic and clinical features of drug resistant and drug susceptible TBM cases were similar. Beijing was the most common genotype (20/25; 80.0%) with 9/20 (45%) of the Beijing strains exhibiting drug resistance; including all 3 MDR strains. All (4/4) isoniazid resistant strains had mutations in the katG gene; 75% (3/4) of strains with phenotypic rifampicin resistance had mutations in the rpoB gene detected by Xpert MTB/RIF®. High rates of drug resistance were found among culture-confirmed TBM cases; most were Beijing strains.

Scientific RepoRts | 6:25251 | DOI: 10.1038/srep25251 Phenotypic DST assesses the ability of isolated bacilli to grow in the presence of a "critical concentration" of the test drug 7 . The critical concentration is defined as the lowest drug concentration that inhibits ≥ 95% of wild-type M. tuberculosis strains 7 , but testing is time consuming to perform. The detection of specific mutations associated with drug resistance (genotypic DST) may overcome some of the limitations, especially the long time delays, associated with phenotypic DST 8 . Drug resistance related mutations have been found in the katG, inhA, rpoB, embB, gyrA, gyrB, rrs-KAN, eis, rpsL and gidB genes 9 . The novel Xpert MTB/RIF ® assay has the ability to simultaneously detect M. tuberculosis and rpoB gene mutations in a critical 81 base pair region (amino acid 507-533), referred to as the rifampicin-resistance-determining region (RRDR) 10 . In general, the presence of rifampin resistance is highly suggestive of MDR tuberculosis, although rifampin mono-resistance is well documented and fairly common in some settings 11 .
Molecular epidemiological analysis helps to characterize the M. tuberculosis population structure and transmission dynamics in a particular setting. Mycobacterial interspersed repetitive unit (MIRU) typing, is currently the most reproducible and widely used method. MIRU typing assesses variable number tandem repeats found in 41 loci across the M. tuberculosis genome; its discriminatory power varies depending on the number and combination of loci used 12 . Spacer oligonucleotide typing (spoligotyping) analyzes genomic polymorphisms in the short direct repeat (DR) region of M. tuberculosis, consisting of identical 36-bp DRs interspersed with 35-41 base pair non-repetitive spacer sequences 13 . A combination of spoligo-and MIRU typing is often used for enhanced strain resolution.
The molecular epidemiology and drug resistance profile of M. tuberculosis strains isolated from TBM patients have not been documented in northwestern China. Neither has a comparison of the disease phenotype observed in drug resistant and drug susceptible cases, or a detailed description of the mutations associated with particular drug resistance patterns, been performed.

Methods
Study participants and setting. We performed a prospective descriptive study over a 2-year study period (September 2010 to December 2012) in Shaanxi province, China. Patients with meningitis (fever, headache and nuchal rigidity or abnormal CSF parameters) were recruited from all tertiary referral centres in Shaanxi province (12 hospitals in total), China; coordinated by the Fourth Military Medical University and Xijing hospital. Shaanxi province is the economic and political centre of the northwestern part of China with a population of 36.7 million people. The TB incidence is high (notification rate 109/100,000 population) 14,15 and HIV-1 prevalence is low (< 1% of TB patients) 16 . Two trained interviewers collected clinical data with the use of a standard questionnaire. Additional clinical information was gathered from reviewing relevant clinical notes. Patients were scored according to the revised British Medical Research Council TBM severity grade 17 and classified according to uniform TBM diagnostic criteria 18 , as 'definite' , 'probable' or 'possible' TBM. All patients were tested for HIV-specific antibodies (Roche Elecsys HIV Combi).
Specimen processing. All CSF specimens were sent to a central laboratory at Xijing hospital, where study tests were performed. Conventional and modified Ziehl-Neelsen stain 19 , standard CSF biochemical and microbiological tests, including Gram stain, India ink stain and ESAT-6 (early secreting antigen target 6KD) immunocytochemical stain, bacterial culture and liquid medium culture, were performed on all specimens. MGIT 960 culture (Becton, Dickinson and Co., Franklin Lakes, NJ) was performed using 0.5 ml of CSF. Culture positive isolates were sent to the National Tuberculosis Reference Laboratory (NTRL) at the Chinese Center for Disease Control and Prevention, Beijing, for M. tuberculosis confirmation by 16S rRNA gene sequencing 20 (Table S1), drug susceptibility testing (DST) and genotyping.
DNA was extracted from freshly cultured M. tuberculosis colonies on solid (L-J) medium and processed using standard reagents (Tiangen Biotech ltd., Beijing) and methodology for spoligo 13 and 25-loci MIRU-25 typing 12 . Given that the Hunter and Gaston Discriminatory Index (HGDI) of some loci included in standard 24-loci MIRU is low 30 , we chose 25 loci with maximal (HGDI = 99.7%) discriminatory power (Table S3).
Patient management. TBM patients were given 3 months of daily oral isoniazid (300-400 mg/d), rifampin (450-600 mg/d), pyrazinamide (25-40 mg/kg/d) and ethambutol (15-20 mg/kg/d) or intramuscular streptomycin (15 mg/kg/d) during the intensive phase of treatment. The same daily dose of isoniazid and rifampin were given for a further 6 months during the continuation phase of treatment. Additional treatment with levofloxacin 1000 mg/d or moxifloxacin 800 mg/d was included at the discretion of the treating physician, as were adjunctive treatment with other antibiotics (meropenem, cefoperazone and/or vancomycin) or supportive therapy (acetazolamide, corticosteroids, mannitol, or CSF shunting). Data analysis and ethics. Statistical analyses were performed with SPSS (V21.0, IBM) and GraphPad Prism v.5.0 (GraphPad Software Inc., La Jolla, CA). Continuous variables were described with medians and ranges, categorical variables with numbers and percentages. Univariate analysis was performed using the Mann-Whitney U test for nonparametric data and between group differences assessed using the Chi square test. Differences were considered significant with p-values < 0.05. MIRU-25 and spoligotyping clustering patterns were analyzed with BioNumerics software version 5.0 (Applied Maths, Sint-Martens-Latem, Belgium). The discriminatory power of various MIRU loci was calculated using the HGDI 31 . M. tuberculosis H37Rv (ATCC 25177, obtained from US-CDC) was used as the reference strain and for quality control purposes.
Ethics Statement. The study protocol was approved by the Ethics Committee of Xijing Hospital (Fourth Military Medical University) and written informed consent was obtained from all participants or their legal surrogates. The methods were carried out in accordance with the approved guidelines.

Results
In total 350 patients were included in the analysis (Fig. 1); 27 (7.7%) with definite TBM, 84 (24.0%) with probable and 239 (68.3%) with possible TBM. None of the patients were HIV-infected. The majority were male (205/350; 58.6%) with a median age of 32 years (range 1-87 years). Table 1 provides an overview of relevant clinical, CSF, central nervous system (CNS) imaging and extra-neural findings. Two culture-confirmed TBM patients had normal CSF protein and sugar values, with leukocyte counts of 7/μ l and 233/μ l (lymphocytes 45% and 85%) respectively. Both had chest radiographs suggestive of pulmonary TB. In total, only 93 (26.6%) patients had a chest radiograph findings suggestive of TB.
Of 27 positive CSF M. tuberculosis cultures, 25 (92.6%) were recovered on L-J solid media and subjected to DST. The majority of these patients 19 (76%) had chest radiograph findings suggestive of TB; median age 26 years (range 3-82 years); 68.0% (17/28) were female. Among positive isolates nearly half (12/25; 48.0%) demonstrated phenotypic drug resistance. The demographic profile and clinical features of TBM patients with and without phenotypic drug resistance are summarized in Table 2. Hospital outcomes were comparable; 2/12 (16.7%) cases with drug resistant disease (mono-S, mono-K) died in hospital and 3/13 (23.1%) of those with drug susceptible disease; 3/12 cases with drug resistant disease had moxifloxacin added to the treatment regimen. All 3 MDR cases had a good outcome. Case #18 received moxifloxacin and meropenem in addition to the standard TB treatment, while cases #11 and #19 received standard TB treatment with cefoperazone or meropenem and vancomycin; no second-line injectables were used.
Drug resistance patterns observed in TBM patients infected with different M. tuberculosis genotypes are reflected in Table 3. Beijing was the dominant genotype (20/25; 80.0%) identified and was associated with high rates of drug resistance (45%; 9/20). This included all 3 MDR strains; 1 with pre-XDR (MDR with additional resistance to either fluoroquinolones or second-line injectable agents). Figure 2 provides more genotypic detail and assesses the phylogenetic relationship of the various M. tuberculosis genotypes identified. There was no evidence of genotype clustering suggestive of common source exposure or laboratory contamination. Table 4 summarizes all the mutations detected in genes associated with drug resistance with corresponding MIC values. All strains with phenotypic isoniazid resistance had mutations in katG (S315T, R463L) and high MIC values (> 32.0 μ g/ml). Most (3/4) strains with phenotypic rifampin resistance had mutations in rpoB, all of these mutations occurred in 81-base pair RRDR which could be detected by Gene Xpert MTB/RIF ® . A single case with phenotypic low-level mono-rifampin resistance was not detected by Gene Xpert MTB/RIF ® ; no mutation could be identified on full sequencing of the rpoB gene. Pan-resistance to all injectable drugs was observed in strains with mutations in rpsL, gidB and rrs-KAN. No phenotypic fluoroquinolone resistance was detected, despite the presence of non-synonymous mutations in the gyrA gene.      32 . These isolates were recovered from a select group of referred patients and were not necessarily representative of strains transmitted within the community. However, with high rates of drug resistant TB among pulmonary TB cases it is expected that similar rates will be observed among TBM cases if these strains are readily transmitted. In fact, the high rates of drug resistance found among M. tuberculosis strains isolated from TBM patients referred from all over Shaanxi province, suggest that these strains are widely dispersed and readily transmitted within the wider community.
There is now convincing evidence of epidemic spread of drug-resistant M. tuberculosis strains in multiple settings. A study of new TBM cases in south western China 33 reported an even higher rate (18%) of MDR-TBM than what we detected. As in previous studies 34,35 , we found the clinical features of drug-resistant and -susceptible TBM to be similar, which emphasizes the importance of routine DST or to identify drug-resistant TBM cases. The current study did not identify any TBM cases in children < 3 years of age, which is the age group at highest risk of TBM following M. tuberculosis infection 36 . However, most children would have been sent to specialized pediatric hospitals that did not contribute to study recruitment. In general, pediatric TB case rates are relatively low in Shaanxi province, which may be a consequence of good BCG coverage, unique population demographics with limited exposure of young children or missed diagnoses. It is estimated that 76% of infants are BCG vaccinated  at birth in China 37 . Data from Shaan'xi Province are limited, but a review of BCG vaccination and scar rates conducted from 1992 to 2001 suggested excellent BCG coverage (97%) 38 . The good outcomes achieved in our TBM cohort may be related to early treatment initiation, non-severe disease at presentation, excellent treatment adherence and the absence of HIV co-infection. The fact that all 3 MDR-TBM patients improved is unexpected, given that MDR-TBM is usually associated with a poor prognosis 39 . All 3 cases were immunocompetent, were diagnosed early, and received additional treatment with antibiotics that has known activity against M. tuberculosis 40 . Typically the CSF in TBM reflects a pleocytosis with raised protein (> 1mg/dl) and reduced sugar 17 . However, a number of patients did not meet these criteria, including two cases with culture-confirmed TBM. Given the ancillary evidence of TBM, these specimens are unlikely to have been contaminated. Similar observations have been made in India 41 , where only 80% of TBM patients had the classic CSF triad of reduced sugar, raised protein (>1mg/dl) and pleocytosis (total leucocyte count of >20 cells/ml). This emphasizes the need to consider TBM even in patients with initial mild clinical symptoms and minor CSF abnormalities. The fact that a high percentage of patients with culture-confirmed drug resistant TBM was female and had chest radiograph abnormalities has not been observed before and is difficult to explain mechanistically.
Mono-drug resistance against streptomycin and other injectable drugs has also been found in other settings 42 . It may be explained by discrepant adherence to oral and injectable drugs during TB retreatment, or prolonged courses of an injectable agent given for other indications. MDR strains that includes streptomycin resistance are common among patients failing first-line treatment in Mongolia, indicating likely transmission of these strains given the absence of previous streptomycin exposure 43 . Mono-resistance against para-aminosalicylic acid and prothionamide is difficult to explain, given the poor potency of these drugs and guidance to only use them as part of an MDR-TB regimen, but have been reported in previous Asian studies 44,45 . Inappropriate antibiotic use is a major challenge in China 46 and all of Asia and since access to second-line TB drugs is poorly regulated exposure to these drugs may have occurred outside well supervised programs.
The 25-loci MIRU used in our study showed good discriminatory power, even among Beijing strains in which strain resolution is poor with standard 24-loci MIRU typing 47 . The strain population structure was similar to those observed among pulmonary TB patients in China 48 without any genotype clustering suggestive of common source exposure or laboratory contamination. The position of an atypical Beijing strain in the middle of the Beijing phylogenetic tree is unexpected, suggesting that MIRU-25 strain resolution remains sub-optimal despite HGDI optimization. There is a possibility that DRs may be subject to cycling convergent evolution that confuse the strain specific signal; especially within highly monomorphic Beiing lineage strains 49 . TBM caused by Beijing genotype strains has been associated with a higher risk of drug resistance 50 , similar to findings in patients with pulmonary TB 51 . A review of osteo-articular TB cases diagnosed in Beijing, China, identified any drug resistance in more than half of the isolates (62/113; 54.9%) ; MDR was present in 15.0% (17/113). The vast majority of strains were of the Beijing lineage; 80.5% SIT 1 52 .
Rifampin mono-resistance has been reported with at relatively high frequency among new cases in Turkey 53 ; rising rates of rifampin mono-resistance has also been reported in parts of South Africa 54 . We detected a single case of rifampin mono-resistance that tested negative on Gene-Xpert MTB/RIF ® , but acknowledge that phenotypic DST can be inaccurate close to the cutoff levels defining resistance 8,55 . All three MDR strains in our study tested positive for rifampin resistance by Gene-Xpert MTB/RIF ® . The fact that rpsL (K43R) was associated with high-level streptomycin resistance, but was also detected in 8% (2/24) of streptomycin-sensitive strains, opens the possibility of multiple gene interactions defining a resistant phenotype and demonstrates the hazards of deducing drug resistance phenotypes from simplistic mutation analyses 56,57 . The only other mutation found in the rpsL gene was a silent mutation in codon 39 (T39T), which has not been associated with streptomycin resistance 58 . Similarly, gidB gene mutations (E92D and A205A) were not associated with drug resistance in our study, but gidB L16R mutations have been associated with resistance in Latin American-Mediterranean lineage strains 58 . The KatG (S315T) mutation is the most common mutation causing high-level isoniazid resistance, while the KatG (R463L) variation is thought to be a polymorphism 59 . Numerous gyrA mutations have been associated with fluroquinolone resistance, but the S95T gyrA mutation that we observed was present in all the TBM strains collected and not associated with drug resistance 26 .
A major study limitation was the small number of culture-confirmed TBM cases; unfortunately this is the reality in all TBM studies 60,61 , especially those enrolling HIV-uninfected TBM patients 33,62,63 . Culture yield may have been further compromised by delays in processing time and the common use of empiric antibiotics before a CSF sample is collected 64 . However, the study cohort represents one of the largest HIV-uninfected TBM cohorts to date. Our study was also limited by its geographic restriction, but findings should be relevant to greater western China and the surrounding region. Clinical, drug resistance and genotypic data from TBM patients in different parts of China would be highly informative. This may be achieved by a nationally coordinated TBM surveillance project to follow-up on our findings.
In conclusion, high rates of drug resistance were found among culture-confirmed TBM cases in Shaanxi Province, China. Findings suggest primary transmission of drug-resistant Beijing strains within the community, which will require accurate identification and treatment of drug resistant pulmonary TB cases to limit transmission. Routine DST is also indicated to optimize the treatment of TBM cases, since clinical features do not differentiate drug resistant and drug susceptible cases.
MGIT 960 MIC. e L511P, S531L, L533P with Escherichia coli numbering (all within the 81 base pair RRDR). f Detected in all the isolates, both drug resistant and drug susceptible; *compared to H37Rv.