The strains selected for this study are described in Table 1 and were isolated from Portuguese and Brazilian tuberculosis patients as part of the routine mycobacteriology laboratory services provided by Universidade NOVA de Lisboa (Lisboa, Portugal) and Universidade Federal do Rio Grande (Porto Alegre, Brazil) to the local hospitals. Since the main objective of this work was to analyze the interplay between the mutational resistance and the efflux activity, the strains were selected based on their drug resistance pattern in order to include pan-susceptible, isoniazid monoresistant, rifampicin monoresistant, ofloxacin monoresistant, poly-drug resistant (resistant to more than one first-line antituberculosis drugs other than both isoniazid and rifampicin), MDR (resistant at least to isoniazid and rifampicin), and XDR (MDR plus resistance to any fluoroquinolone and amikacin, kanamycin, or capreomycin) strains. The H37Rv ATCC27294^T reference strain was obtained from the American Type Culture Collection (Virginia, USA), and used as control. Given the retrospective nature of the work involving only anonymized bacterial isolates, informed consent was not required for this study.

First- and second-line drug susceptibility testing (DST) and MIC determination were done using the MGIT 960 system using the Epicenter V5.80A software and the TB eXIST module (Becton Dickinson, Diagnostic Systems, Sparks, MD, USA). The critical concentrations used for DST were as follows: isoniazid, 0.1 μg/ml; rifampicin, 1 μg/ml; streptomycin, 1 μg/ml; ethambutol, 5 μg/ml; amikacin, 1 μg/ml; capreomycin, 2.5 μg/ml; and ofloxacin, 1 μg/ml. The results were interpreted as follows: at the time of positivity of the proportional control [Growth units (GU) = 400] if the tubes containing the drugs were GU > 100, they were considered resistant to that concentration. If the GU of the tube containing the drugs were <100 they were considered susceptible (Springer et al., 2009; Cambau et al., 2015). In these assays, the critical concentration of the antituberculosis drugs are defined as the concentration that is required to eliminate more than 99% of the population of a control strain that never has been in contact with the drug. A strain is considered susceptible to a given drug tested at their critical concentration if, among the population, the proportion of resistant cells is <1% (critical proportion). The strain is considered resistant when the number of drug-resistant bacteria present in the drug-containing tube is >1%, when compared with the drug-free proportional growth control.

MIC determination of the antibiotics and efflux inhibitors was done within the MGIT 960 system and the growth monitored with the Epicenter V5.80A software as previously described (Machado et al., 2016). The MIC was considered as the lowest concentration with GU < 100 when the drug-free proportional control tube reached a GU = 400. The lowest concentration tested corresponded to the critical concentration of each antibiotic (susceptibility breakpoint).

An absolute growth control (undiluted) was included in every assay to monitor the normal growth of each strain. In this study, the TTD for positivity of the drug-free proportional control varied between 8 and 16 days depending on the strains. DST and MIC determination were performed at least in duplicate and the final value was given as the result of two concordant values.

qDST of the antibiotics was conducted using the MGIT 960 system and the growth monitored with the Epicenter V5.80A software and the TB eXIST module (Becton Dickinson). The test was performed and interpreted according to the protocol previously described (Cambau et al., 2015). Isoniazid was tested at 0.1, 1, 3, and 10 μg/ml, rifampicin and amikacin at 1, 4, and 20 μg/ml, capreomycin at 2.5, 5, and 25 μg/ml, and ofloxacin at 1, 2, and 10 μg/ml. Antibiotics for which the strains were resistant at the highest concentration tested by qDST, were further tested at higher concentrations by MIC determination. The interpretation of the results was done as described above. The antibiotic resistance levels were quantified as follows: isoniazid low-level resistance when resistant (R) at 0.1 and susceptible (S) at 1 μg/ml; isoniazid high-level resistance when R ≥ 1 μg/ml; rifampicin and amikacin low-level resistance when R at 4 and S at 20 μg/ml; rifampicin and amikacin high-level resistance when R ≥ 20 μg/ml; capreomycin low-level resistance when R at 5 and S at 25 μg/ml; capreomycin high-level resistance when R ≥ 25 μg/ml; ofloxacin low-level resistance when R at 1 and S at 2 μg/ml; ofloxacin high-level resistance when R ≥ 2 μg/ml (Cambau et al., 2015). The evaluation of bacterial growth/survival vs. growth inhibition was done by the comparison between the TTD, defined as the time from the start of incubation to the positivity signal of growth (Diacon et al., 2014; Bowness et al., 2015), of a given strain in the presence and absence of an inhibitor (ΔTTD). To obtain the degree of potentiation of each compound on the antibiotic activity, the ΔTTD was normalized against the control tube with no inhibitor. qDST was performed at least in duplicate and the final value was given as the result of two concordant values.

Genomic DNA was isolated using the QIAamp DNA mini kit (QIAGEN, GmbH, Hilden, Germany) according to the manufacturer's instructions. The most common mutations in inhA regulatory region, and katG, gyrA, and rrs genes were studied using the Genotype MTBDRplus and MTBDRsl (Hain Lifescience GmbH, Nehren, Germany) according to the manufacturer's instructions. Genomic analysis of the complete inhA, katG, and tlyA genes and eis promoter region was performed by PCR amplification and DNA sequencing using the primers described elsewhere (Feuerriegel et al., 2009; Machado et al., 2013; Perdigão et al., 2013). The annealing temperatures were 60°C for inhA, 62°C for katG, 56°C for eis, and 55°C for tlyA.

Real-time detection of efflux activity was performed with a semi-automated fluorometric method using a Rotor-Gene 3000 thermocycler (Corbett Research, Sydney, Australia) (Paixão et al., 2009; Viveiros et al., 2010; Rodrigues et al., 2015). The assays were performed as previously described (Machado et al., 2012). M. tuberculosis strains were grown in 10 ml of Middlebrook 7H9 (Difco, Madrid, Spain) supplemented with 10% OADC (oleic acid, albumin, dextrose, catalase; Becton and Dickinson) with 0.05% Tween 80 at 37°C until an OD_{600 nm}(OD₆₀₀) of 0.8. The cells were collected by centrifugation at 2940 × g for 3 min, the supernatant discarded, the pellet washed in PBS and centrifuged again in the same conditions.

Strains were grown in MGIT tubes (mycobacteria growth indicator tubes; Becton Dickinson) for the MGIT 960 system in the presence of half MIC of each antibiotic as follows: MtbPT3 was exposed to isoniazid; MtbPT5 was exposed to rifampicin; MtbBR1 was exposed to ofloxacin; H37Rv, MtbPT7, and MtbPT11 were exposed to isoniazid or rifampicin. Total RNA was extracted using a GTC/Trizol method as previously described (Machado et al., 2016). The relative expression level of the efflux pump genes mmpL7, p55, efpA, mmr, Rv1258c, and Rv2459 was analyzed by RT-qPCR. The primer sets and sequences of oligonucleotides used are described elsewhere (Machado et al., 2012). RT-qPCR assay was performed in a Rotor-Gene 3000 thermocycler and followed the protocol recommended for use with the QuantiTect SYBR Green RT-PCR Kit (QIAGEN) with the following amplification program: reverse transcription for 30 min at 50°C; initial activation step for 15 min at 95°C; 35 cycles of denaturation at 94°C for 30 s, annealing at 52°C for 30 s and extension at 72°C for 30 s; a final extension step at 72°C for 5 min; and an additional step at 50°C for 15 s followed by melt analysis (50–99°C). The determination of the relative mRNA expression level was performed using the comparative quantification cycle (Cq) method (Livak and Schmittgen, 2001). The relative expression of each target gene was determined by comparison of the relative quantity of the respective mRNA in the presence of the antibiotic to the non-exposed condition. Each strain was assayed in triplicate using total RNA obtained from three independent cultures. A relative expression level equal to 1 indicated that the expression level was identical to the unexposed strain. Genes showing expression levels above one were considered to be overexpressed. Genes showing expression levels above two were considered to be significantly overexpressed. Normalization of the mRNA levels was done using the M. tuberculosis 16S rDNA as internal control for each experiment and presented as the mean-fold change (±SD) compared with the control.

Mycobacterial Interspersed Repetitive Unit—Variable Number of Tandem Repeats (MIRU-VNTR) genotyping was performed by multiplex PCR amplification of 24-loci MIRU-VNTR, as described previously (Supply et al., 2006). Spoligotyping was performed as per Kamerbeek et al. (1997). The genotype of these strains was analyzed using the MIRU-VNTRplus web application (Allix-Béguec et al., 2008; Weniger et al., 2010) and the SITVITWEB database (Demay et al., 2012).

Data analysis was carried out using the Student's t-test. A ^*P < 0.05 was considered statistically significant and highly significant when ^**P < 0.01 and ^***P < 0.001 (two-tailed tested).

Firstly, to establish a correlation between the phenotypic drug resistance level and the presence of resistance associated mutations, we searched for mutations in the genes associated with drug resistance to the antibiotics studied (Table 2).

The three pan-susceptible strains showed no mutation in any of the genes evaluated. Resistance to isoniazid was associated with mutations in the katG and inhA genes which are in accordance with published data (Cambau et al., 2015; Domínguez et al., 2016). MtbPT3 showed the S315T mutation in katG and MtbPT4 presented the C-15T substitution in the inhA promoter region. Among the nine MDR strains, five harbored mutations in katG, of which four had the S315T mutation and one has a rare mutation at codon 735. The remaining four strains harbored double mutations in inhA. The double mutations C-15T/S94A were detected in strains MtbPT7 and MtbPT9 and the pair C-15T/I194T was detected in MtbPT10 and MtbPT11.

In respect to rifampicin resistance, all MDR strains and the two rifampicin monoresistant strains showed the mutation S531L in rpoB. This is the most common mutation found in clinical isolates resistant to rifampicin (Cambau et al., 2015; Domínguez et al., 2016). Strain MtbBR2 carries the mutation D516Y in rpoB but it was found susceptible to rifampicin. Although was not our aim to evaluate susceptible strains, MtbBR2 was initially found to be resistant to rifampicin at 1 μg/ml (critical concentration). After re-testing MtbBR2 for rifampicin resistance, it was found susceptible, although with a MIC close to the breakpoint, and was reclassified as poly-drug resistant. The role of the mutation D516Y in rifampicin resistance is not clear. It has been reported by others to be associated with high-level rifampicin resistance (Cambau et al., 2015), low-level resistance (Williams et al., 1998; Zaczek et al., 2009), as well it has been found in rifampicin susceptible strains (Somoskovi et al., 2006; Andres et al., 2014), therefore a non-consensual genotype associated to resistance important to be included and analyzed in the context of this study.

Regarding second-line drugs, the MDR strains MtbBR6 and MtbPT8 were additionally resistant to one second line-injectable drug, and MtbBR2 was resistant to ofloxacin. MtbBR6 and MtbPT8 harbored a mutation in rrs, A1401G, and in strain MtbBR2 was found the A90V mutation in gyrA. MtbPT9, MtbPT10, and MtbPT11 were classified as XDR. Strains MtbPT10 and MtbPT11 harbored the rrs A1401G. This is the most common mutation found in clinical isolates resistant to second-line aminoglycosides (Cambau et al., 2015; Domínguez et al., 2016). For MtbPT9, no mutation could be found in the rrs gene. The resistance to amikacin was assigned to a mutation in eis promoter and the resistance to capreomycin was attributed to the insertion of a GT base pair at position 755/756 of the tlyA gene. The mutation G-10A in the eis promoter region was previously showed to be associated with resistance to kanamycin (Zaunbrecher et al., 2009) and with low-level resistance to amikacin (Perdigão et al., 2013). The insertion of a GT base pair at position 755/766 of tlyA was previously showed to be associated with capreomycin resistance (Perdigão et al., 2013). Resistance to ofloxacin of these three strains was associated with common mutations in gyrA at codons 91 and 94 (Table 2; Cambau et al., 2015; Domínguez et al., 2016). MtbBR1 exhibited monoresistance to ofloxacin associated with a mutation in gyrA, D94N.

Summing up, we have selected a representative and diverse sample of clinical strains of M. tuberculosis from two geographical distinct areas harboring the most representative gene mutations in known drug targets, accounting for clinical resistance to the most important first- and second-line antituberculosis drugs. As per the most commonly accepted dogma about the mechanisms accounting for drug resistance in M. tuberculosis, the sole cause of the resistance levels noticed in these strains are expected to be derived exclusively from the diminished affinity of the mutated genes toward the respective antibiotic (Zhang and Yew, 2009; Domínguez et al., 2016).

Secondly, we quantified the drug resistance levels of the strains toward the antibiotics in study. For this purpose, the MICs of the antibiotics and efflux inhibitors were determined for each strain and the results are presented in Table 3. The MICs of the efflux inhibitors were found to be quite homogenous among the seven groups of strains. Conversely, the MICs of the antibiotics showed that for the same mutation different resistance levels were present. Therefore, we hypothesize that the different resistance levels presented for the same genotype are due to different levels of active efflux of the antibiotics. The existence of active efflux was evaluated indirectly through the reduction of the resistance levels of the antibiotics in the presence of the efflux inhibitors verapamil, thioridazine and chlorpromazine by qDST and MIC determination (Table 4).

Thirdly, to confirm the existence of active efflux systems on these strains we have evaluated their ability to efflux ethidium bromide, a fluorescent efflux substrate, by real-time fluorometry (Paixão et al., 2009; Viveiros et al., 2010; Machado et al., 2012). To perform these assays, one representative of each drug resistant group was selected: MtbPT3, monoresistant to isoniazid; MtbPT5, monoresistant to rifampicin; MtbBR1, monoresistant to ofloxacin; MtbPT7, multidrug resistant, and MtbPT11, extensively drug resistant (Figure 1). M. tuberculosis H37Rv and MtbPT1 were included as representatives of drug susceptible strains. The strains were incubated with ethidium bromide at the equilibrium concentration determined for each strain and verapamil at half MIC (see Table 3 for MIC values) for 1 h. After this period, verapamil was washed out from the solution, the cells were assayed with glucose, as energy source, and without glucose, and their efflux capacity was measured by fluorometry (see Section Materials and Methods for details). As can be observed in Figure 1, efflux of ethidium bromide was detected in all these strains (colored curve), although with different efficiencies, which was inhibited in presence of verapamil at half MIC (black dashed line). It was also noted that the efflux of ethidium bromide is not affected by the presence of glucose (compare gray curve with the colored curve on each graph) which shows that the external energization of the cells is not necessary to guarantee an optimal efflux activity in M. tuberculosis, a common requirement for efficient efflux in Gram-negative bacteria (Paixão et al., 2009; Viveiros et al., 2010). The assays performed with the susceptible strains further confirmed that efflux activity in M. tuberculosis is an intrinsic characteristic of susceptible and drug resistance strains (Machado et al., 2012). The different degrees of efflux activity observed, e.g., MtbBR1 presented increased efflux activity and MtbPT1 presented reduced efflux activity compared with the other strains, may be related to the different environmental conditions to which these strains were subject in the clinical setting. To further confirm that the efflux activity of these strains could be inhibited in the presence of the efflux inhibitors, we performed ethidium bromide accumulation assays in presence of verapamil, thioridazine, and chlorpromazine (Figure 2) and determined the RFF index (Table 5). The RFF index is a measure of how efficient is the inhibitor in promoting intracellular accumulation of ethidium bromide. The results showed that the three inhibitors were capable to increase the ethidium bromide accumulation in all the tested strains (Table 5). The most efficient efflux inhibitor in these assays was verapamil (6/7), followed by thioridazine (2/7), and chlorpromazine (1/7). The results demonstrated that these compounds are able to promote intracellular accumulation of ethidium bromide on M. tuberculosis susceptible, monoresistant and MDR/XDR strains, clearly putting in evidence that active efflux is inhibited by these compounds.

To further corroborate these findings we analyzed the transcriptional profile of putative efflux pumps described in the literature as being associated with antibiotic resistance phenotypes of M. tuberculosis strains (Viveiros et al., 2012; Black et al., 2014; da Silva et al., 2016). RT-qPCR was done for six genes previously shown to be overexpressed in response to antibiotic exposure (Machado et al., 2012, 2016). The results showed the occurrence of significant changes in the expression levels in all strains upon exposure to the antibiotics except in the susceptible H37Rv strain (Figure 3). Concerning the monoresistant strains, in MtbPT3 exposed to isoniazid only efpA was found overexpressed. In strain MtbPT5 exposed to rifampicin, from the six genes tested, four were found overexpressed namely mmpL7, mmr, p55, and efpA. The ofloxacin resistant strain MtbBR1 when exposed to ofloxacin showed significant overexpression all the genes tested (≥two-fold expression). The MDR and XDR strains were exposed to rifampicin or isoniazid and similar patterns were observed. Three to six genes were overexpressed upon exposure to isoniazid or rifampicin. P55 showed the highest level of expression on both strains and for both antibiotics. Overall, EfpA was overexpressed in all strains independently of the antibiotic to which it was exposed. Contrary, Rv2459 was found overexpressed only in the ofloxacin monoresistant strain exposed to ofloxacin. For the remaining efflux pumps analyzed we could not associate an antibiotic to the overexpression of a particular efflux pump. We have observed a general pattern of efflux pump gene expression upon the exposure to the antibiotics.

If efflux really contributes to the overall drug resistance level seen above, the delay in the growth of each drug resistant strain, due to the stress imposed by the combination of an antibiotic plus an efflux inhibitor at subinhibitory concentrations, will render them more susceptible to the effect of the antibiotic. To test this hypothesis, we performed qDST for selected M. tuberculosis strains and investigated the TTD of growth in the presence and absence of the efflux inhibitors to attain clinical correlation (Machado et al., 2016).

The results showed a delay between the TTD of the tube containing the antibiotic plus an efflux inhibitor and the tube containing only the antibiotic ranging from 1 to 63 days depending on the antibiotic concentration and the drug combination (Supplementary Tables 2–4). Rifampicin at 1 μg/ml in combination with the efflux inhibitors showed a delay in the TTD of strain MtbPT11 of 48 h in the presence of thioridazine, corresponding to a potentiation of rifampicin activity of 63.23% when compared with rifampicin only. Similarly, in the presence of verapamil or chlorpromazine, the delay was 72 h corresponding to a potentiation of rifampicin activity of 95.81 and 96.77%, respectively. Noteworthy, this strain became susceptible to rifampicin at 20 μg/ml in the presence of verapamil (Figure 4) and chlorpromazine, with a delay in the TTD of 386 and 195 h, respectively (Supplementary Table 2). In Figure 5 is presented the qDST of isoniazid for the same strain and it can be observed the increasing of the TTD at each concentration of isoniazid in the presence of half MIC of verapamil. These results showed that the antibiotic activity is clearly potentiated by the efflux inhibitors as demonstrated by the delay on the TTD of growth although, as seen previously, they do not reach susceptibility levels due to the antibiotic-target mutations they carry.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.