MODS-Wayne, a Colorimetric Adaptation of the Microscopic-Observation Drug Susceptibility (MODS) Assay for Detection of Mycobacterium tuberculosis Pyrazinamide Resistance from Sputum Samples

Although pyrazinamide (PZA) is a key component of first- and second-line tuberculosis treatment regimens, there is no gold standard to determine PZA resistance. Approximately 50% of multidrug-resistant tuberculosis (MDR-TB) and over 90% of extensively drug-resistant tuberculosis (XDR-TB) strains are also PZA resistant.

in culture. Unlike traditional PZA susceptibility tests, MODS-Wayne does not require an acidic pH because the goal of the assay is to detect the production of POA from PZA metabolism and not to evaluate PZA bacterial killing. The neutral pH permits the POA to accumulate in the supernatant where it can be detected using SAF.
Each decontaminated sample was cultured in 6 wells in a 24-well plate. Nine hundred microliters of the decontaminated sample was seeded per well. Three wells corresponded to control wells, and the other 3 wells were designated for PZA addition. As the test is performed using decontaminated sample, it is possible that the number of viable bacilli, or bacillus viability, could be reduced by stress during the decontamination process; in order to increase the number of bacilli and/or to improve the viability of stressed bacteria, the plate was first incubated at 37°C with no drug. The plate reading commenced on day 5 and was then performed every 2 days until day 21, according to the standard MODS protocol (39). When at least two microcolonies were observed in the control wells, 100 l PZA (8,000 g/ml) (Sigma-Aldrich, USA) was added to one of the PZA wells (final concentration, 800 g/ml). After 3 days and 6 days, the same volume of PZA was added to one of the remaining PZA wells. Based on previous results, in order to facilitate the production of a detectable amount of POA by a Wayne reaction, after the addition of PZA, the samples were incubated for 3 days, during which time the bacilli were able to produce POA. The nonacidic pH (6.8) prevented the lethal effect of the drug and thus the inhibition of growth, and these conditions would be expected to maximize the production and accumulation of POA, the pivotal biomarker of PZA resistance. At the end of this additional period (reported as reading 1 [R1], reading 2 [R2], and reading 3 [R3]) ( Fig. 1), 100 l of 10% ferrous ammonium sulfate (Sigma, USA) was added to the corresponding control and PZA wells, and the color was registered immediately. A positive result was reported when a pink color was observed in the PZA well but not in the control well, and a negative result was reported when no color was observed in either the control well or the PZA well ( Fig. 2A and  B). Additionally, the level of growth on culture was reported in one of four categories according the degree of aggregation by microscopy, high growth, regular growth, low growth, and microcolonies, to evaluate any possible influence on the MODS-Wayne assay.

RESULTS
Agreement between the dichotomous MODS-Wayne results and the composite reference standard for each of the three drug incubation periods was estimated using the Kappa statistic in a 2-by-2 contingency table. The potential effect of bacillary load or level of growth on the MODS-Wayne results was evaluated using Fisher's exact test. For results with a positive association, a proportion test was performed to evaluate significant differences. The sensitivity and specificity (95% confidence interval [CI]) of the MODS-Wayne assay compared to those of the consensus reference test were estimated for each drug incubation period. McNemar's test was used to determine significant differences between sensitivities and specificities among MODS-Wayne results. All statistical tests were reported with 95% confidence intervals and were performed using Stata version 14.0.    (Table 2).
High and regular growth levels were associated with a higher proportion of positivity in the MODS-Wayne assay at R2 (P Ͻ 0.0001), but no significant association was observed for samples with low growth or microcolonies (P ϭ 0.56 and 0.45, respectively). Conversely, low growth level and microcolonies were associated with a higher proportion of negative MODS-Wayne results at R1 (P ϭ 0.011 and 0.0012, respectively), but no significant association was observed for samples with high and regular growth (P ϭ 0.20 and 0.14, respectively).
Accuracy measures for MODS-Wayne. Results from each reading day were compared to the composite reference standard independently. Sensitivities (true resistant percentage) in R1, R2, and R3 were similar, ranging from 92.7% to 97.6% (P Ͼ 0.05). However, the specificities showed a broader range, increasing from 57.9% to 99.3% (Table 5) and with a statistically significant difference upon comparison (P Ͻ 0.05).
As the MODS-Wayne and the classic Wayne assays obey the same principles, they were directly compared. At day 16, the sensitivity of MODS-Wayne was 93.5% and specificity was 93.8%, with an agreement and kappa index of 93.8% and 0.79, respectively (P ϭ 0.001). (Data from other time points not shown.) Ten isolates were PZA resistant according to the MODS-Wayne assay but PZA susceptible according to the classic Wayne test. When their genotypes were analyzed, nine isolates had mutations associated with PZA resistance (Q10P, Q10R, H51R, and H57R), and one strain was wild type.

DISCUSSION
This study shows for the first time that an adaptation of the MODS assay combined with the Wayne test (MODS-Wayne) can accurately determine M. tuberculosis resistance to PZA directly from AFB smear-positive sputum samples and can achieve this in a relatively short period of time compared to other PZA DSTs. In addition, it is inexpensive, simple to perform, and requires minimal training of laboratory staff. The MODS-Wayne assay is based upon the principle that POA is only produced by PZA-susceptible bacilli, and therefore, detection of POA in the supernatant of a MODS culture through visual observation of a Wayne colorimetric reaction functions as an accurate proxy for PZA sensitivity. In contrast with previous efforts to modify MODS for PZA resistance testing, in this adaptation, after sputum samples are decontaminated, the stressed M. tuberculosis bacilli are grown at a nonharmful neutral pH, in the absence of any drug until mycobacterial growth is observed. PZA is only added after this time. The neutral  pH is maintained throughout. This allows amplification of the bacterial population, thereby maximizing the release of POA into the extracellular environment while preventing bacterial killing. As MODS-Wayne only evaluates the presence of POA in the supernatant, there is no need to satisfy the culture conditions, including an acidic pH, necessary to demonstrate PZA's antibacterial activity. The advantage of using a neutral pH is that acidity inhibits the growth of M. tuberculosis and therefore reduces the test's specificity (44). MODS-Wayne overcomes other limitations of the traditional DSTs for PZA, in that there is no requirement for an isolate to be grown first (indirect test), or for a large inoculum size (3 to 5 mg) in order to avoid false-negative results (30). MODS-Wayne reported a median turnaround time of 18.5 days (range, 14 to 29 days) for PZA-susceptible isolates and 21 days (range, 14 to 25 days) for PZA-resistant isolates for results in R3. Although the MODS-Wayne turnaround time was higher than that reported for direct MGIT-PZA (range, 11 to 16 days), Demers et al. (45) reported that 41% (163/398) of the direct MGIT-PZA results were uninterpretable mostly due to growth failure, and most of the indirect tests (i.e., classic Wayne test) require primary isolation of the isolate first, which means that from the time when the specimen was collected until the susceptibility results were obtained, the turnaround time ranged from 18 to 95 days. Total time from sample to results is in fact much shorter for MODS-Wayne, which is a direct test.
The use of a composite standard where no good single standard exists has been described (12). Using our composite reference standard (MGIT, classic Wayne test, and pncA sequencing), we found that around half of the selected MDR-TB strains were PZA resistant (30/57 samples [52.6%]), but few of the selected pansusceptible strains were PZA resistant (4/101 samples [4.0%]). These findings are in agreement with rates reported by other studies (3,6,20).
Both the classic and the quantitative Wayne assays (8, 28) evaluate PZase activity using PZA concentrations ranging from 100 g/ml to 400 g/ml (21). In this study, we used a PZA concentration of 800 g/ml in order to obtain a maximum signal without compromising specificity. This concentration was established in a previous study optimizing the assay, as it enabled us to most accurately discriminate isolates with weak-positive Wayne activity (PZA susceptible) from those with negative Wayne activity (PZA resistant) (our unpublished data). Compared to the classic Wayne test, MODS-Wayne showed an agreement and kappa index of 93.8% and 0.79, respectively. Ten isolates showed discrepant results, i.e., were PZA susceptible according to the classic Wayne assay and PZA resistant according to MODS-Wayne; nine isolates clearly have genuinely PZA-resistant phenotypes because they have mutations in pncA that are close to the active site (Q10P and Q10R) or affect directly the metal-binding site (H51R and H57R), both of which are associated with PZA resistance (7). Additionally, these isolates reported PZA resistance by MGIT-PZA. This discrepancy may be attributable to misinterpretation during the reading of the classic Wayne assay, as difficulty in the correct interpretation of weak-positive samples has been reported as a limitation of this test (8,29,30). On the contrary, the single wild-type strain that was falsely reported as PZA resistant by the MODS-Wayne may have been be due to the lack of PZA addition during the incubation period. This isolate was reported as PZA susceptible by MGIT-PZA.
Although MODS-Wayne does not evaluate pncA mutations, it had 96.9% agreement and a kappa index of 0.90 with pncA sequencing. One disadvantage of molecular methods is that interpretation of the results requires an understanding of the roles of pncA gene mutations and PZase activity (4). Not all mutations in the putative promoter or the pncA gene have been reported, and not all reported mutations are associated with PZA resistance. Nevertheless, MODS-Wayne is a phenotypic test that indirectly evaluates pncA mutations (i.e., mutations that affect pncA expression or impair PZase activity) by detecting POA, the enzymatic product of the PZase. In this case, a mutation database is not required to interpret the results. Six of 7 isolates harboring K48T maintained PZase activity and were reported as positive by MODS-Wayne and PZA susceptible by MGIT-PZA. K48T is an example of a mutation that is not associated with PZA resistance. More studies are necessary to understand the real effect of some mutations that do not impair PZase activity on PZA susceptibility.
As a phenotypic test, MODS-Wayne depends on the growth of M. tuberculosis in culture and its capacity to produce a detectable amount of POA. Our results showed that the level of growth influenced the time to positivity on MODS-Wayne, and there was also a trend in association of the bacillary load on smear microscopy with time to results, although this did not reach statistical significance. We hypothesized that a low number of PZA-susceptible bacilli can lead to a negative result in the early reading days, which can be considered false PZA resistance. This could be mitigated by using a longer duration of culture prior to the addition of PZA to the assay, which would increase the number of PZA-susceptible bacilli and give a positive MODS-Wayne result.
Our data show that the accuracy of MODS-Wayne (92% sensitivity and 99.3% specificity) is comparable to that of other tests for PZA resistance detection (MGIT, classic Wayne, and pncA sequencing). The MODS-Wayne assay and MGIT-PZA had an agreement and kappa index of 95.3% and 0.86, respectively. The sensitivity and specificity of the MODS-Wayne assay were optimal in R3 (92.7% and 99.3%, respectively); although sensitivity was slightly higher in R2 (97.6%), our assumption is that longer culture duration may permit a more accurate discrimination of the PZA-resistant from the PZA-susceptible isolates. This will need to be confirmed in future studies.
Finally, it is important to highlight that the MODS-Wayne assay lends itself to incorporation into the standard MODS assay. This would generate a single unique test that would be able to diagnose TB, MDR-TB, XDR-TB, and PZA resistance from the same sputum sample with high accuracy and in a relatively short time.
In conclusion, here, we describe the MODS-Wayne test, a modification of the standard MODS method that is able to determine PZA susceptibility with high accuracy and in a relatively short time. Although DST for most first-and second-line TB drugs is available, PZA susceptibility testing is not routinely performed despite widespread use of the drug because of well-described difficulties in testing its activity in vitro. MODS-Wayne overcomes these difficulties and also has the advantages of low cost and being easy to perform and interpret. Thus, the MODS-Wayne assay offers the real possibility of introducing routine PZA susceptibility testing in resource-constrained and/or high-TB-burden settings. This would represent a major advance in the treatment of TB patients worldwide.