Mycobactericidal Effects of Different Regimens Measured by Molecular Bacterial Load Assay among People Treated for Multidrug-Resistant Tuberculosis in Tanzania

Rifampin or multidrug-resistant tuberculosis (RR/MDR-TB) treatment has largely transitioned to regimens free of the injectable aminoglycoside component, despite the drug class’ purported bactericidal activity early in treatment. We tested whether Mycobacterium tuberculosis


MATERIALS AND METHODS
Patients and ethical considerations. From August 2018 to December 2019, two populations of patients with TB participated in this study. The primary target population was patients with RR/MDR-TB. A small population of patients with DS-TB was included as a control to inform the reproducibility of previous TB-MBLA study findings (14,16). Rifampin susceptibility in M. tuberculosis was confirmed using Xpert MTB/RIF (25). Moreover, all patients harbored M. tuberculosis that was deemed susceptible to fluoroquinolones and aminoglycosides by line probe assay (Hain, LifeScience, Germany), the genotype MTBDRsl version 2.0 (26). The study included all patients aged at least 18 years who were able to expectorate and provide quality early morning sputum. Quality sputum was defined by an adequate volume of .5 ml and absence of food particles. No sputum induction was done to patients who were unable to provide quality sputum. Critically ill or moribund patients, as previously defined by Robertson et al. (27), and pregnant women were excluded. Additionally, patients who interrupted treatment were excluded from the final analysis. Prior to any study procedure, all patients signed a witnessed oral or written informed consent. The study was approved by the National Institute for Medical Research (NIMR) in Tanzania (NIMR/HQ/R.8a/Vol. IX/2662). Permission to conduct the study was granted by authorities of the Kibong'oto Infectious Diseases Hospital (KIDH).
Study setting. Patients were recruited at Kibong'oto Infectious Diseases Hospital, a national center of excellence for clinical management of drug resistant (DR)-TB located in the Siha district of the Kilimanjaro region in Tanzania (25). TB-MBLA testing was performed at the National Institute for Medical Research, Mbeya Medical Research Centre branch, given that laboratory's prior experience with the assay.
Sample size determination. The numbers of patients required to determine differences in bactericidal activity over time in 4 treatment regimens were calculated as previously reported by Guo et al. (28). We assumed a Spearman correlation of 0.51 and a baseline M. tuberculosis burden of 5.5 log 10 eCFU/ml, as well as daily M. tuberculosis decline and decay rate of 0.42 and 0.05 log 10 eCFU/ml, respectively (14,16). Hence, at least 7 patients were needed per regimen to reach a power of 80% with a twosided type I error of 5%. Adjusting for least 25% of patients who were likely to be lost to follow-up, not evaluated due to negative microbiological results at baseline, and/or died, a minimum of 37 patients was desirable for sampling and analyzing at the end of 4 months of treatment.
Sputum collection, processing, and culturing of M. tuberculosis. One sample of approximately 5 ml of early morning sputum was collected from each patient for laboratory testing at day 0 (baseline) and at days 3, 7, 14, 28, 56, 84, and 112 of anti-TB treatment. Before culturing, sputum was homogenized using a sterile magnetic stirrer at room temperature for 30 min. Then, 1 ml of homogenized sputum was treated using 4 ml of 4 M guanidine thiocyanate (GTC) containing 1 M Tris-HCl (pH 7.5) and 1% (vol/vol) of b-mercaptoethanol, and was frozen at 280°C in order to preserve the M. tuberculosis RNA from degrading. The M. tuberculosis culture was performed on LJ slants from the remaining sputum samples collected at six time points from days 0, 14, 28, 56, 84, and 112 of treatment, as per previous description (29). In brief, sputum was decontaminated by N-acetyl-L-cysteine (NALC)-NaOH, and finally resuspended in 1 ml of phosphate buffer. A total of 200 ml of decontaminated sputum was inoculated into two LJ slants and incubated for up to 8 weeks to detect mycobacterial growth. Incubated LJ slants were read on a weekly basis and were deemed negative if there was no growth at week 8.
RNA extraction and RT-qPCR for TB-MBLA. M. tuberculosis quantification by TB-MBLA was performed as described by Gillespie et al. (30). In summary, M. tuberculosis RNA in 1 ml of homogenized sputum preserved in 4 ml of guanidine thiocyanate (GTC) at 280°C was extracted using the RNA pro kit (FastRNA Pro BlueKit; MP Biomedical, CA, USA) as instructed by the manufacturer. The extract was treated with DNase I enzyme (TURBO DNA-free kit; Ambion, CA, USA) to remove DNA from the dead cells. The M. tuberculosis 16S rRNA, a biomarker for viable cells, was amplified and quantified by RT-qPCR using specific primers and probes. The Cq was translated to bacterial load (estimated CFU per ml [eCFU/ ml]) using a standard curve on a Rotor gene Q 5plex platform (Qiagen). The cutoff for TB-MBLA positivity was a 30 Cq value that corresponds to 1.0 log 10 eCFU/ml, beyond which the test was considered negative (16,30).
Statistical analysis. Data were recorded in a clinical case report form, entered, and cleaned before statistical analysis. Patients who completed 8 treatment visits and had positive pretreatment TB-MBLA results were analyzed and visualized in R softward version 4.0.2 (http://www.R-project.org). Continuous variables, such as age, body mass index (BMI) in kg/m 2 , and time to TB-MBLA negativity were described by the median with the 25 th and 75 th interquartile range (IQR), and were compared across different regimens using a Kruskal-Wallis test. Accordingly, proportions for HIV status, gender, cavitary lung disease on chest, and previous TB treatment were compared across different regimens using a chi-square or Fisher's exact test. Using baseline bacterial load, chest cavity, HIV, silicosis, and gender as fixed effects, the rate of M. tuberculosis killing (log 10 eCFU/ml) was fitted on quartic polynomial nonlinear mixed effects (NLME) for repeated measures as previously described (31,32). Individual patients were accounted for random effect. A model was reliably selected if it had low Akaike information criterion but high intraclass correlation coefficient ( Table 1). The mean difference in M. tuberculosis load, due to two different regimens received by patients at the end of 4 months of treatment, was compared using one-way analysis of variance (ANOVA) and Tukey's test for repeated measures (33). An injectable regimen without bedaquiline was used as a reference regimen. The median time to TB-MBLA and culture conversion to negative was estimated using the Kaplan-Meier method, and was compared across different regimens using a log rank test (34). Cox proportional hazards regression models were used to estimate the hazard ratios (HR) for M. tuberculosis killing, and was adjusted for the effects of HIV, baseline bacillary load, cavitary disease, silicosis, gender, history of treatment for drug-sensitive TB, and clearance rate. We computed an overall mean M. tuberculosis load of 4.0 log 10 eCFU/ml, and it was used to categorize a patient's bacterial load as "high bacterial load" versus "low bacterial load" depending on whether patients had detectable M. tuberculosis above or below this mean, respectively. A P value of ,0.05 was considered significant. A 95% confidence interval (CI) of the mean clearance rate and HR was included.

RESULTS
Population. Of 59 patients enrolled, 37 patients produced a total of 296 serial sputa for final analysis. Reasons for exclusion and patient's distribution are outlined in Fig. 1. Among 296 serial sputa analyzed, 104 sputa came from 13 patients who received an injectable but bedaquiline-free regimen, 72 sputa were from 9 patients who received an injectable bedaquiline-containing regimen, 64 from 8 patients who received an alloral bedaquiline-based regimen, and 56 sputa from 7 patients who were treated for drug-sensitive TB with conventional RHZE. Clinical and demographic parameters are presented in Table 2. Twenty-seven (73%) out of 37 patients were male. Their median (IQR) age was 37 (32 to 49) years. Patients who received standard RHZE treatment were younger than those who received RR/MDR-TB treatment regimens (P = 0.038). Also, 11 (30%) patients were living with HIV with a CD4 T cell count of 208 (95% CI 144 to 272) cells/ml. More patients with HIV received an all-oral than injectable-based treatment regimen (P = 0.001).
Mycobactericidal activities of different regimens over time. The M. tuberculosis load measured by TB-MBLA and culturing in Fig. 2 Table 3, P , 0.001). An injectable bedaquiline-containing regimen had the highest mean M. tuberculosis killing rate, followed by an all-oral bedaquiline-based regimen compared to the injectable but bedaquiline-free reference regimen (Table 3, P = 0.019). Kanamycin-containing regimens in Fig. 3 had rapid bactericidal activity at day 14, but this was not translated into long-term bactericidal effect (P , 0.001). An all-oral bedaquiline-based regimen had a sharp decline after day 28.
Median time to M. tuberculosis killing. There was strong positive correlation in time to sputum conversion between TB-MBLA and culture (r = 0.46 [95% CI 0.36 to 0.55]; P , 0.001). The overall median time to sputum TB-MBLA conversion to negative was 56 (IQR 28 to 84) days. The median times to TB-MBLA conversion to negative were 28, 42, and 84 days among patients on injectable bedaquiline, an all-oral bedaquiline- based regimen, and injectable but bedaquiline-free regimens, respectively. Irrespective of treatment regimen, 92% (34/37) of patients had negative culture results compared to 65% (24/37) of negative TB-MBLA at day 56 (P = 0.037). The number of patients who converted to sputum negative by culture and TB-MBLA per treatment regimen is shown in Fig. 4. Among 13 patients who received the injectable but bedaquiline-free regimen, 2 and 7 of them remained culture and TB-MBLA positive, respectively, whereas all 8 patients who received injectable bedaquiline-containing regimens had negative LJ culture and TB-MBLA at day 56 ( Fig. 4A to D). Favorably, all patients on injectable bedaquiline for MDR-TB and standard RHZE regimen for DS-TB had negative  TB-MBLA at days 56 and 84, respectively. Compared to 31% (4/13) of patients who received an injectable but bedaquiline-free regimen, only 11% (1/9) of those who received an all-oral bedaquiline-containing regimen remained positive TB-MBLA but negative LJ culture at day 112 of treatment ( Fig. 4A and B versus Fig.4E and F; P = 0.283).

DISCUSSION
This study shows for the first time to our knowledge that the killing rates of M. tuberculosis in patients treated for RR/MDR-TB, as well as those with concomitant TB with silicosis, varies with treatment regimens. As measured by TB-MBLA, M. tuberculosis decreased significantly over time on treatment, and this kinetic correlated with what was observed using LJ culture medium. For decades, culture has been used as a routine microbiological tool for monitoring drug-resistant TB treatment response (15,35), but in many settings with endemic TB, culture is unavailable or limited to specialized centers. Importantly, culture results can take up to 8 weeks from the time of sputum collection, which delays patient care if a treatment decision is made based on a result from a specimen collected 2 months earlier. Given the continued decentralization of RR/MDR-TB services in Tanzania and elsewhere, monitoring treatment response in laboratories capable of performing qPCR, such as with Xpert MTB/RIF, will allow lab-   oratory assays to impact treatment decisions closer to the point-of-care. Therefore, this study in RR/MDR-TB complements the growing evidence supporting the application of TB-MBLA in routine clinical management (14,16,36). Interestingly, our findings suggest that bactericidal activity at day 14 may not be a suitable predictor of the long-term efficacy of a regimen, particularly when that regimen contains bedaquiline. In this cohort at day 14, more than 75% of people had a positive TB-MBLA and more than half had a positive culture result. However, between 14 and 56 days we observed substantial M. tuberculosis killing in those treated with a bedaquiline-containing regimens, suggesting that evaluation of bactericidal activity be performed later, such as at day 56, for modern RR/MDR-TB regimens. Using culture, one previous phase 2b clinical trial reported high bactericidal activity of a bedaquilinecontaining regimen in patients with DS-and RR/MDR-TB (37). However, detectable M. tuberculosis beyond day 56 in our study supports this trial's argument that day 56 is unreliable as an indicator of a regimen's ability to either predict long-term treatment outcomes or shorten treatment duration (37). This further raises the question of whether TB-MBLA may in fact be a superior predictor to culture.
Another important finding from this study of TB-MBLA is that M. tuberculosis killing kinetics were regimen dependent. Overall, there was rapid and prominent killing of M. tuberculosis at day 14 for patients who received kanamycin regardless of receipt of bedaquiline. However, superior activity of kanamycin-containing regimens at day 14 had no long-term bactericidal effect. As a result, 3 patients on the injectable but bedaquiline-free regimen remained positive by TB-MBLA but negative by LJ culture after 4 months of treatment. On the other hand, patients who received an all-oral bedaquiline-containing regimen achieved these rates of killing at or after 1 month of treatment. This observation concurs with previous reports that the bactericidal activity of bedaquiline in MDR-TB is delayed at the beginning, but accelerates later in therapy (38). Usually, recovery of M. tuberculosis by TB-MBLA correlates better with Mycobacteria Growth Indicator Tube (MGIT) liquid culture than with LJ solid culture, which may partially explain the discrepancy between the two tests at month 4 of treatment (16). This argument supports previous findings that culturing M. tuberculosis on LJ recovers a lower yield than in MGIT liquid culture (39). Nonetheless, our findings, as measured by TB-MBLA, fit with the pharmacodynamical understanding that kanamycin and other aminoglycoside/polypeptides that are active against mycobacteria will primarily exert their effect against those extracellular organisms that are rapidly dividing, and these may be more abundant early in the treatment course (40,41).
The shorter overall time to sputum conversion to negative, as measured by TB-MBLA and conventional culture, for all patients who received bedaquiline regardless of kanamycin further supports the argument that bedaquiline should be a cornerstone of regimens designed to shorten the duration of MDR-TB treatment (42). The conventional injectable but bedaquiline-free regimen has been in practice for decades, even though more than 40% of patients treated with this regimen had unfavorable outcomes in settings where TB is endemic (43). Aminoglycosides such as kanamycin are no longer part of the current MDR-TB treatment regimens, not due to lack of bactericidal activity, as our data would suggest the contrary in the early treatment period, but rather because of the significant toxicity and patient intolerance that leads to treatment interruption (24,44). From a microbiological perspective alone, as demonstrated in this study and others, such as Mpagama et al. (45), and also in a more patient-centered approach, our results demonstrate the potential importance of finding more tolerable substitutes for kanamycin that can match the early bactericidal effect.
The main strength of this study is that we utilized TB-MBLA to model killing rates among patients with RR/MDR-TB and those with TB/silicosis. We have shown that patients with TB/silicosis had slower M. tuberculosis killing rates by TB-MBLA compared to those with TB and without silicosis. This low rate of killing could partially be attributed to the underlying pulmonary pathophysiology, which can include progressive massive fibrosis (46, 47) and a blunted local host immune response to M. tuberculosis infection (46). We observed a similarly lower rate of M. tuberculosis killing among patients with RR/MDR-TB who had high initial bacterial load, which is consistent with previous studies of TB-MBLA kinetics from patients with drug-sensitive TB (14,16,36). In this study, approximately 1 and 4 out of 10 patients had, respectively, positive LJ culture and positive TB-MBLA at day 56. This supports the previous argument that TB-MBLA is more sensitive compared to agar-based Loewenstein-Jensen culture, in which the M. tuberculosis population gets lost due to contamination at later time points (18). Limitations of the study include the timing of endpoints, which were limited to 4 months, such that predicting long-term treatment success was beyond the scope of this study. Nevertheless, modeling M. tuberculosis killing for 4 months as we accomplished here has been used as a marker for treatment failure and relapse in several observational studies (35,48), and exceeds the duration of monitoring used in other trials of RR/MDR-TB regimens that have employed conventional culture-based techniques (37). Additionally, this study had no control over the treatment regimens prescribed. However, given the feasibility of TB-MBLA and the comparability of this study's findings to prior studies with TB-MBLA in drug-susceptible TB (16), we plan to apply TB-MBLA systematically within an ongoing operational research protocol for injectable-free RR/MDR-TB treatment in Tanzania that employs standardized regimens over various treatment durations. Lastly, because of the small number of patients per treatment regimen, these findings should be cautiously inferred to other RR/MDR-TB populations. Nevertheless, a longitudinal cohort design in this study allowed control of variabilities between patients, as well as intrapatient tracking of each regimen's bactericidal activities over time (28,49).
In conclusion, patients who received bedaquiline-containing regimens exhibited higher M. tuberculosis killing rates and had shorter time to sputum TB-MBLA and culture conversion to negative. While both kanamycin-containing regimens had superior bactericidal activity during the first 2 weeks of RR/MDR-TB treatment, the addition of bedaquiline allowed for improved killing after 1 month of therapy. Together, these findings provide insight into formulating optimal all-oral bedaquiline-containing regimens with the best potential to shorten duration of MDR-TB treatment (37,44,50). Given that TB-MBLA does not require laboratory procedures associated with culture and the prolonged time to receive a culture-based result, we envision it can be used to make regimen adjustments in the presence of anti-TB drug susceptibility testing results.

ACKNOWLEDGMENTS
This study received financial support from the EDCTP2 program supported by the European Union project (grant number TMA2016SF-1463-REMODELTZ) and DELTAS Africa Initiative (Afrique One-ASPIRE/DEL-15-008). The Afrique One-ASPIRE is funded by a consortium of donors, including the African Academy of Sciences, Alliance for Accelerating Excellence in Science in Africa, the New Partnership for Africa's Development Planning and Coordinating Agency, the Wellcome Trust (107753/A/15/Z), and the UK Government. The funding bodies had no role in the conceptualization, methodology, data interpretation, or writing of the manuscript.
We acknowledge Batuli Mono, Taji Mnzava, Joseph Kachala, and Bibie Said of KIDH for their assistance with recruitment and data collection from study participants. We also thank Elisha S. Juma and Sarapia P. Malya of KIDH, and Emmanuel Sichone and Joseph John of NIMR Mbeya for assisting with laboratory work. In addition, we also acknowledge the KIDH administration for granting permission to conduct this study.
We have no conflicts of interest to declare.