Nonclinical pharmacokinetics and relative efficacy of the first 25 novel tuberculosis drug combinations from the PAN-TB consortium: Use of the BALB/c relapsing mouse model and combination pharmacokinetics within a modeling-based framework

27 The Project to Accelerate New Treatments for Tuberculosis (PAN-TB) aims to accelerate 28 development of shorter, simpler and safer pan-TB combinations, effective for use in both 29 Drug Susceptible (DS)- and D rug Resistant (DR)- TB patients . Towards thi s aim, 30 bactericidal and sterilizing activity of 25 priority 4 -drug combinations was evaluated at 31 doses targeting clinically relevant exposures, in the BALB/c relapsing mouse model of TB. 32 The combinations comprised 8 PAN-TB drugs and candidates: bedaquiline (B), 33 pretomanid (Pa), delamanid (Del), quabodepistat (Q), sutezolid (Sut), GSK2556286 (286), 34 GSK3211830 (830) and ganfeborole (GSK3036656, (656)). Combination PK studies in 35 infected mice enabled dose selection and a population-PK approach guided dosing so that 36 compounds should achieve mean AUC 0-24 within 2-fold of their clinical target exposures 37 during the efficacy studies . All test combinations showed time-dependent bactericidal 38 activity, with six regimens reducing lung bacterial burdens below the limit of detection 39 with 8 weeks’ treatment, similar to the comparator BPaMZ (M is moxifloxacin and Z as 40 pyrazinamide). Cure/Relapse data were modelled to derive population time to cure 90% 41 mice (T90) values. Fifteen PAN-TB combinations had T90s of less than 5 months , 42 sterilizing mice faster than the standard of care for drug susceptible TB , RHZE/RH. The 43 best-performing PAN-TB combinations, BPa830Sut, BPa286Sut and BQSut286, cured 44 90% of mice in less than 3 months. These 3 top-ranked 4-drug combinations are all centered 45 on a diarylquinoline (B)/oxazolidinone (Sut) core, together with the nitroimidazole (Pa) or 46 a DprE1 inhibitor (Q) plus a novel agent such as the LeuRS inhibitor (830) or the Rv1625c 47 agonist (286). 48 49 50 Key words: PAN-TB consortium, tuberculosis, efficacy, relapse 51 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 3

52 Tuberculosis (TB) remains a global health crisis with 1.25 million deaths reported in 2023 53 [1]. Available treatments are long, complex and poorly tolerated, resulting in suboptimal 54 adherence which impact s patient outcomes. Due to the presence of significant drug-55 resistance, access to rapid drug susceptibility testing (DST) is critical to appropriate 56 treatment; the currently limited access presents another barrier to TB control [2]. 57 Development of novel pan -TB combinatio ns, defined by World Health Organization 58 (WHO) as first-line regimens that could be used without prior knowledge of a TB patient’s 59 drug-resistance profile [3], would obviate the need for DST, eliminating the time gap 60 between diagnosis and start of effective treatment. To maximize benefits, pan-TB regimens 61 should be shorter, simpler, and safer than existing treatments. Introducing such regimens 62 in high TB burden countries could result in substantial health improvements as well as 63 savings t o patients and health systems [4]. Although progress has been made towards 64 treatments that fit WHO target regimen profiles for drug -susceptible TB (DS -TB) and 65 rifampicin resistant TB (RR -TB) [3], more limited progress has been made towards 66 development of pan-TB regimens. 67 The Project to Accelerate New Treatments for Tuberculosis (PAN-TB) is a philanthropic, 68 non-profit and private sector collaboration . The Collaboration aims to accelerate 69 prioritization and development of promising pan-TB drug combinations by leveraging 70 assets, resources, technologies and e xpertise of its members (Tuberculosis Prevention | 71 PAN-TB). At the outset of the Collaboration, eight candidates and marketed drugs were 72 identified among the members as PAN-TB priorities, due to their pan-TB potential (i.e. no 73 or limited existing clinical resistance). These comprised bedaquiline (B), an ATP synthase 74 inhibitor - [5], delamanid (Del) and pretomanid (Pa), two nitroimidazole mycolic acid 75 synthesis inhibitors [6-7], quabodepistat (Q) an inhibitor of decaprenylphosphoryl -β-D-76 ribose 2' -epimerase [DprE1] [8], GSK2556286 (286) an Rv1625c agonist and, c -AMP 77 mediated cholesterol catabolism inhibitor [10], sutezolid (Sut) an RNA translation inhibitor 78 [11], GSK3211830 (830) and ganfeborole (formerly GSK3036656, listed here as 656), 79 leucyl-tRNA synthetase (LeuRS) inhibitor s [9]. These two LeuRS inhibitors are close 80 analogues with very similar PK and efficacy data. The consortium initially focused on 81 testing 830 but later shifted toward 656 as it advanced further in the clinic. 82 The Collaboration focuses on 4-drug pan-TB combinations, hypothesizing that treatment 83 with 4 drugs with novel and differing modes of action, will maximize efficacy and 84 minimize emergence of resistance. Towards prioritizing the most promising 4 -drug 85 combinations comprised of the initial 8 PAN-TB candidates, the Collaboration selected 25 86 unique regimens, avoiding inclusion of two nitroimidazoles in the same combination , for 87 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 4 nonclinical efficacy assessment. Some of the candidates had previou sly demonstrated 88 bactericidal efficacy and treatment -shortening activity when evaluated within TB drug 89 combinations in the well-established BALB/c relapsing mouse model of TB (RMM) [8, 90 12, 13, 14, 15, 16]. None had bee n evaluated within the context of these specific 25 91 combinations. 92 This work evaluated and compared the bactericidal and sterilizing efficacy of the 25 drug 93 combinations in the BALB/c RMM of TB, at exposures relevant to clinical efficacy targets. 94 An additional objective was to compare their estimated RMM time to cure parameters to 95 those of clinical benchmark regimens . Benchmark TB combinations used in the reported 96 RMM studies included the standard of care for drug -susceptible TB, Rifampicin (R), 97 Isoniazid (H), Pyrazinamide (Z), Ethambutol given as RHZE/RH and BPaMZ. BPaMZ is 98 a regimen that has been evaluated in clinical trials a nd demonstrated time to culture 99 conversion that was more rapid than RHZE/RH [17]. However, the regimen did not meet 100 the key secondary efficacy endpoint due to adverse events resulting in treatment 101 withdrawal and additionally, includes pyrazinamide and moxifloxacin, to which there is 102 significant background resistance. Therefore, BPaMZ cannot be considered as a potential 103 pan-TB regimen. We sought to identify pan-TB regimens with efficacy superior to BPaMZ 104 in this well-established mouse TB model. 105 106 107 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 5

108 Combination Pharmacokinetics Performed in Infected Mice Enabled Accurate 109 Selection of Human Equivalent Doses 110 The 25 PAN-TB test combinations are listed in Table 1. Doses for evaluation in RMM 111 efficacy studies were selected with the aim of achieving steady -state area under the 112 concentration time curve from 0 -24 hours (AUC 0-24) for each drug, across all test 113 combinations, within 2-fold of observed or targeted clinical plasma concentrations (Table 114 2). Preliminary dose selection was based on available mouse single drug plasma PK data 115 and rodent tolerability data . Doses were refined based on data generated through 5 -day 116 repeat-dose combination PK studies, conducted for each test combination, administered at 117 the preliminary doses to M.tb-infected mice. The average plasma AUC0-24 observed for 118 each drug, across the relevant 4 -drug combinations in m ice, was compared to the 119 corresponding observed or target clinical value. As a result of these comparisons, the final 120 dose for use in efficacy evaluations was decreased for Del, and increased for both 830 and 121 656, compared to the ir preliminary doses (Table 2 ). Q exposure was variable between 122 tested combinations. Although the median Q exposure in mice, at the preliminary dose, 123 was higher than the clinical target, the final dose selected was slightly higher than the 124 preliminary, to ensure exposures across all test combinations were no more than 2 -fold 125 lower than the clinical target. 126 RMM p lasma exposures were within 2 -fold of their clinical targets for most 127 combination components 128 To evaluate PK exposure achieved during RMM efficacy studies for each drug across the 129 test combinations, population PK (popPK) analysis was performed using sparse-sampled 130 drug concentration data obtained after 8 weeks’ dosing during the RMM studies, together 131 with data generated through the combination PK studies. Individual RMM drug exposure 132 distributions across all tested combinations are shown in Figures S1-S8 and Table S2. For 133 most of the tested compounds, i.e., 656, B, Del, Pa and Sut, their overall distributions of 134 RMM study drug exposures (derived plasma AUC 0-24 values) achieved after 8 weeks’ 135 dosing were within 2-fold of their clinical targets (Figure 1 and Table S2). The Q RMM 136 exposures were variable across the tested combinations: none were more than 2-fold below 137 the targeted exposure, and while the median observed Q plasma AUC0-24 was slightly more 138 than 2-fold higher than the clinical target, exposures were within 2-fold of the clinical target 139 for 8 out of 16 of the tested Q -containing combinations, namely when Q was co-140 administered with B and in the absence of Pa (Figure S7). The 286 RMM exposures were 141 slightly more than 2-fold lower than its clinical target on average (Figure 1 and Table S2). 142 Notably, AUC0-24 values for 286 were lowest when co -administered with B (Figure S1). 143 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 6 Finally, 830 RMM exposures were more than 2 -fold lower than the clinical target in all 144 tested combinations (Figure S3). When comparing exposures of companion drugs between 145 Del- and Pa - containing regimens, B exposure is , at least two -times higher in the Pa -146 containing regimens than those with D el, except for the regimens of BD el656286 and 147 BPa656286, in which B exposures are similar (Figure S4). Similarly, Q exposure is more 148 than two-times higher in BPaQX regimens than those in the BD elQX regimens ( Figure 149 S7) 150 Six PAN-TB drug combinations demonstrated bactericidal activity in mice that was 151 at least as rapid as BPaMZ. 152 All 25 4-drug combinations demonstrated time -dependent bactericidal activity over the 153 course of 8 weeks’ treatment (Figures 2A and 2B). BPa286Sut and BQSut286 reduced 154 lung CFU to undetectable levels following 4 weeks of treatment, whereas BDel286Sut, 155 BDel830Sut, BPa830Sut, BPaQSut and BPaMZ required 8 weeks’ treatment. For the other 156 19 test combinations, as for the benchmark RHZE/RH, colonies were still detectable after 157 8-weeks' treatment (Figures 2A and 2B). However, 15 of these 19 reduced the mouse lung 158 bacterial burden (CFU/ lung) to a greater extent than RHZE/RH by the 8 week treatment 159 timepoint: BPa656286, BQ656286, BQ656Sut, BDelQSut, BDelQ286, BDelQ830, 160 BDel656286, PaQSut286, PaQ656Sut reduced bacterial burden by an additional 2Log10, 161 and BDelQ830, BPaQ286, B656286Sut, PaQ656286, Del656 286Sut, BPaQ656 reduced 162 CFU/ lung by an additional 1Log10 compared to RHZE/RH (Tables 3A and 3B). 163 Three PAN-TB 4-drug combinations cured 90% of mice in less than 3 months , out-164 performing RHZE/RH but not BPaMZ 165 To establish the relationship between the tested regimens and length of treatment required 166 to achieve a durable cure in mice, we quantified the proportion of mice that were culture 167 positive twelve weeks after the end of treatment (i.e. exhibiting relapse) (Tables 4A and 168 4B). No relapse events were recorded for any BPaMZ -treated mice after 6 weeks’ 169 treatment. In contrast, at least 16 weeks’ treatment were required to achieve 100% cure (no 170 relapse) for RHZE/RH . Of the PAN -TB test combinations, the best -performing were 171 BPa830Sut (100% mice cured after 8 weeks’ treatment ) and BPa286Sut, which cured all 172 mice after 10 weeks’ treatment. Six additional test combinations demonstrated 100% cure 173 after 12 weeks of treatment (Table 4A). On the other hand, 8 of the 25 test combinations 174 failed to effect cure after 12 weeks’ treatment: all mice had detectable colonies (relapse) 175 12 weeks after treatment cessation (Table 4A and B). 176 A logistic Emax model was developed based on observed cure/relapse data reported herein 177 plus a historical dataset and used to estimate time-to-cure 50% of mice (T50), and time-to-178 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 7 cure 90% of mice (T90) for each combination [23, 37]. Combinations for which 100% 179 relapse was still observed after the maximum treatment duration were excluded. The 180 developed model, using a population approach, demonstrated generally good performance 181 in predicting relapse profiles of the 4 -drug combinations, especially for comparators 182 BPaMZ and RHZE/RH for which data coming from different studies and labs were 183 available (Figure S9). Individual probability of relapse-time curves for each combination 184 by study are shown in Figure 3. The corresponding estimated population T50 and derived 185 T90 values are listed in Table 5. T90 values for the 17 analyzed novel combinations ranged 186 from 2 to 6 months, whereas the T90 for the RHZE/RH comparator was approximately 5 187 months. Fifteen combinations performed better than RHZE/RH, with shorter T90s (Table 188 5 and Figure 3). Of these, BDel286Sut, BPa656286, BQ656286, BPaQSut, BDelQ286, 189 BPaQ286 displayed a T90 of less than 4 months. The best-performing novel combinations, 190 BPa830Sut, BPa286Sut, and BQSut286, had a derived population T90 less than 3 months 191 with values of 2.11, 2.51 and 2.94 months respectively. None of the test combinations had 192 population time-to-cure parameter estimates T50 or T90 that were as low as for BPaMZ 193 (T90 of approximately one month ). When compared to population derived T90s for the 194 combinations included in the historical dataset, s ome novel PAN-TB combinations 195 appeared to have out-performed clinical reference combinations in this BALB/c mouse 196 model (Figure 4) . For example, BPa 830Sut and BPa286Sut displayed T90s similar to 197 BPaL (2.28 months) which is used for treatment of adults with Extensively Drug-resistant 198 TB (XDR-TB) or drug-intolerant or non-responsive Multi-Drug-Resistant TB (MDR-TB) 199 using a 6-month regimen [6] (Table 5 and Figure 4). 200 Comparison of T90s for combinations differing by one drug , indicate treatment-201 shortening contributions for several PAN -TB candidates, that may be context 202 dependent 203 Pairwise comparisons of T90 values were performed for 4-drug combinations differing in 204 only one component. In all cases, the presence of B reduced the T90 values and among 205 evaluable comparisons, the T90 differences were statistically significant (p < 0.05) (Table 206 6 and Table S3). For combinations differing only in the nitroimidazole (Pa versus Del), 207 several that contained Pa had a shorter population derived T90 value, compared to those 208 containing Del: Significant differences in T90s were observed for BPa830Sut vs 209 BDel830Sut (2.1 vs 5.1 months); BPaQSut vs BDelQSut (3.5 vs 4.6 months); BPa656286 210 vs BDel656286 (3.4 vs 4.3 months); and BPa286Sut vs BDel286Sut (2.5 vs 3.2 months) . 211 (Table 6 and Table S3). However, exposures of B and Q are about 2-times higher in Pa 212 containing regimens than those of in the D el-containing regimens. Therefore, this 213 difference between Pa and D el may be partly explained by the different drug -drug 214 interactions in this mouse model. For certain 4-drug combinations population T90 values 215 did not differ when Q replaced a nitroimidazole : no significant differences or a low level 216 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 8 of confidence in the difference were observed between T90s for BQSut286 vs BDel286Sut 217 (2.9 vs 3.2 months) or BQ 656286 vs BPa656286 (3.4 vs 3.4 months) or BQSut286 vs 218 BPa286Sut ( 2.9 vs 2.5 months ). On the other hand, r eplacing Del with Q significantly 219 reduced the T90 for BDel656286 vs BQ656286 (4.3 vs 3.4 months) (Table 6 and Table 220 S3). 221 The presence of 286 had a variable impact on 4-drug combination T90 depending on the 222 companion drugs. A reduction in T90 (faster cure) was observed when 286 replaced Q in 223 all tested regimens (BDel286Sut vs BDelQSut (3.2 vs 4.6 months) or BPa286Sut vs 224 BPaQSut (2.5 vs 3.5 months )), and when 286 replaced 830, 656 or Sut in some tested 225 regimens: (BDel286Sut vs BDel830Sut (3.2 vs 5.1 months); BQ656286 vs BQ656Sut (3.4 226 vs 4.0 months) ; BDelQ286 vs BDelQSut (3.6 vs 4.6 months) and BDel656286 vs 227 BDel830Sut (4.3 vs 5.1 months)). Finally, the presence of Sut significantly reduced 228 (improved) the T90 when it replaced 656 (BPaQSut vs BPaQ656 ( 3.5 vs 4.4 months) or 229 BQSut286 vs BQ656286 ( 2.9 vs 3.4 months) . As for B, Sut was present in all 3 best -230 performing 4-drug combinations that displayed the shortest T90s: BPa830Sut, BPa286Sut 231 and BQSut286. 232 In addition to these T90 pairwise comparisons , AUC0-24 ratios (AUCratio) were calculated 233 and compared for each of the 3 common drugs between pairs of combinations that differed 234 by only one drug. Overall, when only one drug differed, exposure ratios for the common 235 drugs in paired combinations were variable -, although some significant ratios were 236 observed (p < 0.001) very few are more than 2-fold. (Table 6 and Table S3). 237

238 This translational research utilized the well-established BALB/c TB RMM within a 239 modelling-based framework to evaluate and compare bactericidal and sterilizing efficacy 240 of 25 novel 4 -drug combinations of 8 PAN-TB candidate drugs, administered at dose s 241 targeting clinically-relevant exposures. The absolute and relative time -dependent 242 bactericidal activities of the RHZE/RH and BPaMZ regimens , as observed here, were 243 consistent with previous data published by Evotec and others [18, 19, 20]. Time-dependent 244 sterilizing activity, evaluated by modelling BALB/c RMM data, has been estimated using 245 varying methodologies by several groups. Despite these differing approaches, the derived 246 population T90s for RHZE/RH and BPaMZ in the present study are consistent with 247 published values [21, 22, 23]. We identified 15 novel PAN-TB 4-drug combinations - i.e. 248 with relevance to the WHO pan -TB Target Regimen Profile - with shorter time to cure 249 (population derived T90) in this model than the current standard of care for DS-TB, 250 RHZE/RH. 251 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 9 The best -performing combinations , BPa830Sut, BPa286Sut, and BQSut286 , achieved 252 T90s of less than 3 months in this BALB/c TB model and for BPa830Sut and BPa286Sut, 253 T90 values were similar to BPaL. This means that if developable and safe, these 2 regimens 254 could represent alternatives to BPaL with a pan-TB potential. 255 None of the PAN -TB combinations cured mice as rapidly as the benchmark regimen 256 BPaMZ, which demonstrated potential to cure patients in less time than RHZE/RH in 257 clinical trials. However, it is promising that several novel 4 -drug combinations with pan-258 TB potential cured BALB/c mice faster than RHZE/RH and at least as rapidly as BPaL. 259 Further, both BPa286Sut and BQSut286 demonstrated rapid bactericidal activity, similar 260 to that of BPaMZ. Rapidly bactericidal regimens may be beneficial by quickly reducing 261 symptoms and infectiousness of TB patients, positively influencing patient and population 262 level outcomes. The observation that BPa286Sut and BQSut286 did not match the cure rate 263 of BPaMZ even after demonstrating similar bactericidal activity is interesting and deserves 264 further attention. First, it underscores what has always been known in TB therapy that rapid 265 elimination of replicating organisms does not necessarily lead to rapid sterilization and 266 lasting cure, even for regimens that contain highly efficacious and long half-life agents like 267 B. Second, it points to a certain sterilizing activity for M and Z, the two agents in BPaMZ, 268 and demands a concerted effort in the field to find agents with similar sterilizing potential. 269 All 10 PAN-TB combinations that demonstrated T90s of less than 4 months contained B - 270 including the 5 with the most rapid bactericidal activity . In contrast , most B-free 271 combinations didn’t clear lung bacteria after 8 weeks’ treatment or cure mice after 12 272 weeks of treatment. Both these findings highlight the potentially significant contribution 273 of B and possibly other diarylquinolines to pan-TB treatments. This is in line with previous 274 work demonstrating the importance of this ATP-synthase inhibitor class in the TB 275 treatment backbone [13, 19, 24 ]. The introduction of B has significantly improv ed 276 treatment outcomes and survival among patients with MDR/XDR-TB. However, since its 277 implementation in 2012, reports of B resistance have emerged [26]. The identification of 278 potent B-free regimens is therefore of interest . The novel B -free 4 -drug combination 279 DelQSut286 demonstrated a slightly shorter T90 than RHZE/RH (by 0.4 months) in this 280 study, offering a qualified hope for B-free regimen for the future. 281 These results also confirmed the contribution of protein synthesis inhibitors to the 282 performance of novel regimens, with the oxazolidinone Sut present in all 3 combinations 283 with T90s of less than 3 months. Indeed, interest in Sut as a potential novel TB drug led to 284 its evaluation as part of regimens in the Phase2b SUDOCU trial [ 39] and the PAN -TB 285 Gates MRI-TBD06-21-trial (Int J Tuberc Lung Dis 2025: 29 (11 Suppl 1): S1 – S963). 286 Further, the novel drug candidate 286, which inhibits cholesterol metabolism [3] - a new 287 mode of action for potential TB drugs - is included in 2 of the top 3. The sterilizing activity 288 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 10 observed here for BPa286Sut is similar to published data for BPa286L, where full 289 sterilization of mice was achieved within 3 months [12]. L, like Sut, is an oxazolidinone 290 drug and has demonstrated treatment-shortening activity in mice [16, 17] . Notably, in the 291 study of Nuermberger et al [12], undetectable levels of mouse lung CFU were reached after 292 2 months of treatment with BPa286L as opposed to 4 weeks for BPa286Sut in our study. 293 This difference in bactericidal kinetics may be related to a more potent bactericidal activity 294 of Sut, compared to L, as shown in the study of Tasneen et al [24]. In addition to Sut, the 295 oxaborole leucyl tRNA synthetase inhibitor 830 was present in one of the top 3 regimens, 296 demonstrating the potential utility of this novel mechanism of action which also inhibits 297 protein synthesis, for the treatment of TB. 298 This is the first time preclinical studies have compared head-to-head bactericidal and 299 sterilizing activity for drug combinations including the two nitroimidazole drugs Pa and 300 Del, at doses confirmed to be relevant to their respective clinical exposures. In these 301 studies, Pa was dose d at 40 mg/kg and Del at 2 mg/kg, lower doses than used in some 302 previously published BALB/c RMM studies [13, 14, 27]. Overall, and especially when 303 considering T90s, the Pa-containing combinations tested here performed better than those 304 containing Del. However, exposures of B and Q are about 2-times higher in Pa containing 305 regimens than those of in the D el-containing regimens. Since B is a known potent 306 sterilizing drug, this difference between Pa - and Del-containing regimens may be partly 307 explained by the different drug -drug interactions in this mouse model. Whether similar 308 drug-drug interactions occur in humans will need to be carefully evaluated. In previous 309 work conducted to compare their bactericidal activity as monotherapies in two mouse 310 models, Del at 2.5 mg/kg demonstrated similar efficacy to Pa at 20 mg/kg and 30 mg/kg 311 [22]. Our results are consistent with this outcome, taking into account the higher dose and 312 presumably exposure of Pa used in our studies . Interestingly, the difference observed in 313 sterilizing activity between Pa-containing and Del-containing matched combinations was 314 sometimes but not always observed when comparing their relative bactericidal activity. For 315 example, whereas BPa286Sut and BDel286Sut reduced lung CFU to undetectable levels at 316 the end of 4 and 8 weeks ’ treatment respectively, with corresponding T90s of 2.5 vs 3.2 317 months, this correlation between bactericidal and sterilizing activities was not observed for 318 BPa830Sut and BDel830Sut. Both regimens reduced bacterial levels to undetectable levels 319 with 8 weeks’ treatment, whereas there is a 3-month difference in T90 between the two 320 combinations (T90s of 2.1 and 5.1 months respectively) . Further work is needed to 321 understand the drivers of sterilizing versus bactericidal activity for these combinations and 322 to identify the PK and/or PD factors driving the differentiation in behavior between these 323 two sets of nitroimidazole-containing combinations. Some possible contributors are greater 324 post-antibiotic effect ( PAE) for Pa than for Del [28], differentiated PD interactions with 325 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 11 companion drugs for Pa versus Del, and differing impacts on exposures of highly 326 efficacious companion drugs for the two nitroimidazoles. 327 328 Our pop-PK analysis indicated that differences in T90s were associated with significantly 329 lower AUC0-24 values for B or Q, or both, in Del-containing combinations compared to the 330 Pa-containing equivalents while exposure was in the 2x range of clinical target for almost 331 all the comparisons . B exposure in BDelQ830 is the lowest compared to all other B-332 containing combinations and significantly lower (more than 2-fold) compared to BPaQ656. 333 (Table 6, Figure S4, Table S3). In other hand, this analysis indicated that differences in 334 differences in time to cure were associated with a significant lower AUC0-24 values for Q 335 in all B-free combinations compared to the B- containing combinations (Table 6, Table 336 S3). More generally, numerical results, included in Table S3, suggest variable relationships 337 between drug exposure variations and T90 changes. Although some statistically significant 338 AUCratio were observed, most of them are not clinically relevant as they are within the 2-339 fold range of clinical target , it is difficult to identify associations between drug exposure 340 and sterilization activity, as significant T90 reductions occur together with either increased 341 or reduced AUCratio (i.e., >2 or <.0.5, respectively) across paired combinations where one 342 drug differs. Overall, this exploratory analysis highlights that observed exposure-T90 343 associations must be treated with caution , in the absence of established PK -PD 344 relationships for each compound within the context of these specific 4-drug combinations. 345 Further caution should be exercised when considering translation: any PK drug -drug 346 interaction occurring which may be responsible for observed changes in AUC ratios when 347 one drug is substituted for another, may not translat e from mouse to human considering 348 metabolic and/or transporter mechanisms that influence drug (parent and/or metabolite) 349 disposition may differ. 350 351 Similar to B-free combinations, nitroimidazole-free combinations are of interest for their 352 utility against nitroimidazole -resistant TB which may emerge in the future [30]. We 353 evaluated the sterilizing activity of the DprE1 inhibitor Q, within multiple 4 -drug 354 combinations, for the first time . When either Pa or Del was replaced by Q, these 355 nitroimidazole-free combinations performed as well as the corresponding Pa - containing 356 combinations (i.e. BQSut286; T90 < 3 months) suggesting a similar sterilizing contribution 357 of this DprE1 inhibitor compared to nitroimidazoles, in these particular combinations. The 358 recently reported evaluation of Q together with B and Del in a Phase II clinical trial 359 supports the positive performance of Q within nitroimidazole-containing drug 360 combinations [29]. Our mouse study suggest s Q might be useful in nitroimidazole-free 361 regimens 362 363 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 12 The datasets and modeling approaches used in this work allowed us to compare efficacy of 364 combinations across studies and to historical data, to compare exposures achi eved in PD 365 studies to clinical targets and to explore comparative contributions of drugs using PK and 366 PD data. However, further investigation of PK/PD for each agent within the combinations 367 of interest, and direct comparisons of PK and PD for 4 -drug combinations versus their 3-368 drug components, are needed to better understand the contributions of each agent and, in 369 some cases, their active metabolites, to overall combination efficacy, exposure-response 370 relationships and potential PK or PD drug-drug interactions. 371 372 These studies constitute the first step in PAN -TB's nonclinical strategy to prioritize and 373 characterize novel drug combinations to support decision-making and progression through 374 development. The use of this well-established mouse model together with careful dose 375 selection and exposure evaluation, allowed us to assess and compare bactericidal and 376 sterilizing activity of large numbers of drug combinations at relevant exposures in mice as 377 a first step towards prioritization and characterization. The population model ling 378 approaches used for both T90 derivation and PK allowed us to rank combinations with 379 historically tested combinations, across studies and understand the relationship between 380 duration of treatment and cure for each tested combinations while reducing the overall 381 number of mice required . Although the T90 rankings from these BALB/c data appeared 382 generally consistent with clinical results, the model has limitations, including pathology 383 that does not include all lesion types seen in TB patients (e.g those exhibiting caseous 384 necrosis and cavitation). 385 386 For this reason, the extent to which findings from BALB/c RMM studies translate to the 387 clinic is hard to predict without further clinical data for comparison to mouse -tested 388 combinations or BALB/c mouse RMM data for regimens recently evaluated in the clinic 389 (e.g Rifapentine (P)HMZ) [31]. Notably, the recently completed Gates MRI -TBD06-201 390 PAN-TB trial, which evaluated the PAN -TB investigational regimens DBQS and PBQS, 391 concluded that although both regimens demonstrated strong bactericidal activity by the end 392 of treatment when administered for 4 months neither, showed sufficient evidence for being 393 able to treat TB in 3 months or less based on 2- and 3-month sputum culture conversion 394 rates and TB recurrences after treatment completion . When assessed by 12 -month post-395 randomization unfavorable outcome rates, the 4-month PBQS regimen performed similarly 396 to 6-month HRZE (Int J Tuberc Lung Dis 2025: 29 (11 Suppl 1): S1 – S963). 397 398 These findings are consistent with the data reported here, where BPaQS ut and BDelQSut 399 T90s are 3.48 and 4.6 months, respectively compared with RHZE/RH which cured 90% of 400 mice in 5.09 months. 401 402 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 13 403 For combinations of interest identified through these studies, as well as those prioritized 404 based on subsequent BALB/c RMM studies, PAN -TB intends to generate significant 405 additional nonclinical data for use as translational model ling inputs for prediction of 406 clinical performance and to support decision-making. Some examples include nonclinical 407 lesion PK studies, activity studies in caseum (using an ex vivo assay), and evaluation of 408 T90s in mouse models featuring caseous necrotic lesions (the Kramnik RMM). Evaluation 409 of p erformance of combinations against mouse models where the infecting strain is 410 resistant to a key drug class will also be important. Finally, combination efficacy studies 411 will be conducted in mouse models to better understand the individual contributions of 412 each agent, to confirm the ir pharmacological value and to inform future regimen design. 413 Promising data has been recently reported for both newer diarylquinolines (e.g. 414 sorfequiline (TBAJ-876) [25]) and oxazolidinones (e.g. TBD09 –[32]) as well as for 415 TBD11, a compound from a different chemical series with a similar mechanism of action 416 as 286, suggest potential to improve on the high -performing first generation PAN -TB 417 regimens by introducing these next generation compounds. Accordingly, the PAN -TB 418 collaboration has introduced sorfequiline, TBD09 and TBD11 to its candidate pool and in 419 addition to seeking a better understanding of the top performing regimens reported here, is 420 conducting further BALB/c RMM studies using th e above workflow, to evaluate and 421 compare their bactericidal and sterilizing activities at exposures likely to be targeted in the 422 clinic in potential next generation regimens. The most important outcome of this first study 423 from the PAN -TB consortium has been the successful assembly of novel drugs from 424 different partner organizations and designing and testing drug combinations in a consistent 425 format that will highlight the most promising regimens ready for direct clinical evaluation, 426 thus reducing the time to delivery of these needed interventions to patients. 427 428

429 Animals and ethics 430 All mouse experiments were carried out at the Evotec France SAS animal facility. This 431 facility is accredited by the French Ministry of Agriculture and by the Association for 432 Assessment and Accreditation of Laboratory Animal Care International (AAALAC). All 433 studies were performed under the European Communities Council Directive 434 (2010/063/EU) for the care and use of laboratory animals and approved by local Ethical 435 Committee CEPAL: CE 029 and authorized by the French Ministry of Education, 436 Advanced Studies , and Research. Six-week-old female BALB/cJRj mice from Janvier 437 Laboratories were group housed in bioconfined cages (Isocage, Tecniplast®) under a 12h 438 light:12h dark with free access to filtered water and a standard rodent diet (AO4C, Safe, 439 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 14 France). An ambient temperature of 22 ± 2°C, a relative humidity of 55 ± 10 % , and a 440 negative pressure of -20Pa were maintained throughout the study. All mice were allowed 441 to acclimatize to their new environment for at least 5 days prior to the start of the study. 442 Drug Formulations and Dosing Strategies 443 Drugs were acquired or provided by PAN -TB consortium members , and formulations 444 prepared for dosing as follows at the selected doses (Table 2). B (J&J IM) was formulated 445 in 20% 2 -hydroxypropy-β-cyclodextrin; Pa (TB Alliance) was formulated in 10% 446 Hydroxy-propyl-beta-cyclodextrin and 2% soy lecithin for dosing at 100mg/kg for the 447 BPaMZ control group, and at 40 mg/kg for the PAN -TB 4-drug combinations. M (LTK 448 Laboratories) and Z (Sigma) were co-formulated in water for dosing at 100mg/kg and 150 449 mg/kg, respectively. R (Sigma) was prepared in water for dosing at 10 mg/kg; H (Sigma), 450 Z (Sigma) and E (Sigma) were co-formulated in water for dosing at 10, 150 and 100mg/kg, 451 respectively. Del (Otsuka) and Q (Otsuka) were formulated in 5% Arabic gum. 286 (GSK), 452 656 (GSK) and 830 (GSK) were formulated in 1% methylcellulose solution. Sut (TB 453 Alliance) was formulated in 0.5% methylcellulose and 5% PEG200 . For each PAN-TB 454 combination, each drug was dosed individually with an interval of 2 hours between each 455 drug. The order of dosing for each drug in each combination was based on the half-life of 456 each drug where drugs with longest half -lives were administered first and drugs with 457 shorter half-lives were given later in the day. The drug dosing order for each combination 458 is indicated in the name of the combination i.e. the drug listed first was dosed first, and so 459 on. For the BPaMZ comparator, B was administered first, followed by Pa and finally MZ, 460 dosed as a co-formulation. For the RHZE/RH comparator, R was dosed individually for 461 the first 8 weeks, followed by co-formulated HZE. For the following 8 weeks, R and H 462 were dosed individually. The same 2h interval was applied between 2 administrations for 463 the comparators as for the test combinations. 464 Relapsing Mouse Model 465 The 25 PAN -TB combinations were evaluated across two RMM studies. The first 12 466 combinations were tested in Study 1 (named Wave 2022 in the C-PATH APEX database: 467 https://c-path.org/tools-platforms/tb-apex) and the second 13 combinations in Study 2 468 (named Wave 2023 in the C-PATH APEX database: https://c-path.org/tools-platforms/tb-469 apex). BPaMZ and RHZE/RH comparators were included as controls in each Study. Both 470 studies included 4 to 6 mice per group, per time point, allocated following a similar 471 approach to that reported previously [23] and described in Table S1. M.tb H37Rv stock 472 solution was prepared at exponential growth phase in 7H9 medium / 10% OADC (oleic 473 acid-albumin dextrose-catalase) / 15% glycerol. At Day -14, female BALB/c mice were 474 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 15 anaesthetized with 2.5% isoflurane in 97.5% oxygen and were intranasally infected with 475 50µL M.tb H37Rv at an inoculum level of 4.5 Log 10 CFU/mouse.. Treatment started 14 476 days post infection, designated Day 0. All drug combinations except RHZE/RH were dosed 477 once daily by oral gavage for 5 days a week (omitting the weekend) for 2, 4, 6, 8, 10, or 12 478 weeks. RHZE/RH was dosed for 2, 4, 6, 8, 10, 12, 14 or 16 weeks where RHZE was given 479 for the first 8 weeks followed by RH only for the following 8 weeks. For analysis of mouse 480 lung bacterial burden at the end of each designated treatment period, mice were sacrificed 481 24h post last dosing. For relapse assessment, mice were sacrificed 12 weeks after the end 482 of each designated treatment period. Following sacrifice, lungs were collected , weighed 483 and lung samples homogenized and plated undiluted, or serially diluted on 7H11-OADC + 484 0.4% activated charcoal. Plates were incubated at 37°C for 6 weeks for CFU quantification. 485 Undetectable level was considered when no colonies were observed in all plates for the 486 sample. 487 488 Pharmacokinetics 489 Combination PK studies were conducted , before the RMM studies, in female BALB/c 490 mice, intranasally infected with M.tb H37Rv in the same manner as for the RMM studies 491 (see RMM section of Materials and Methods) . Beginning 14 days post-infection (Day 0) 492 24 mice per group were treated with each test combination via oral gavage either once or 493 once daily for 5 days. The 5-day dosing period was selected based on the short terminal 494 half-lives of the drugs ( Table S3), suggesting at least 80% of the s teady-state is reached 495 after 5 days treatment duration (5 days > 5 times of T1/2). The only exception is the 496 bedaquiline M2 (desmethyl) metabolite (B -M2), which exhibits a longer half -life. It is 497 assumed that the late sparse PK samples in the RMM study (Week 8) provide enough 498 information to correctly estimate the PK parameters for this entity. Doses were selected 499 based on historical data from PAN-TB members. The order and spacing of administrations 500 were performed as for the RMM studies (see Dose Formulations and Dosing Strategies) . 501 Blood samples were collected from the tail vein on days 1 (after single dosing) and 5 (after 502 once daily repeat administration) at 0.5, 2.5, 5.5, 7.5, 10, 12, 14, 24, 36, 29, and 31 hours 503 post administration of the first drug. Additionally, for each combination, plasma and blood 504 intra-cardiac concentrations were assessed on Day 1 and Day 5 at 2.5, 7.5, 14, and 31 hours 505 post-first compound dosing. During RMM studies, on the 5 th day of the 8 th week of 506 treatment, blood was collected from the tail vein at 0.5, 2.5, 5.5, 7.5, 9 and 24 hours post -507 first compound dosing. In all cases, after blood processing, each compound, including B -508 M2, and the active (sulfoxide) metabolite of sutezolid (PNU -101603) were quantified by 509 liquid chromatography tandem mass spectrometry (LC -MS/MS). Bioanalytical methods, 510 samples handling, as well as MS conditions are described in supplemental data. All PK 511 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 16 assessments were conducted in blood. Blood and plasma concentrations determined 512 through the Combination PK studies were used to calculate mouse geometric mean Blood-513 to-Plasma (BP) ratios ( Table 2 ) for each drug . These values were used to facilitate 514 translation of blood to plasma exposure values for head -to-head comparison s between 515 observed mouse and clinical target exposures (see Population PK Modelling). 516 Population PK Modelling 517 For each compound tested across the 25 PAN-TB combinations, a population 518 pharmacokinetic (popPK) model was developed to describe the mouse PK profile. After 519 identifying each specific model, post-hoc estimates were used to calculate total exposures, 520 i.e. AUC0-24, accounting for the variability observed across animals in terms of summary 521 statistics (median, 5 th, 25th, 75th and 95th percentiles). Blood concentration data collected 522 during the RMMs were pooled with those collected in the Combination PK studies to 523 ensure a sufficient number of observations required for robust model development and 524 parameters estimation. More details about model selection criteria are provided in the 525 supplemental material (see PopPK model verification and quality criteria). For B and Sut 526 their main active metabolites , B-M2 and PNU-101603 respectively, were included in the 527 model description to derive the overall exposure. As the parent drug and corresponding 528 metabolite contribute to efficacy in mouse and human at different relative levels [33, 34, 529 35, 36], the active moiety approach was adopted to compare the relevance of preclinical 530 exposure values to clinical exposures, considering the plasma AUC 0-24/MIC90 ratios 531 (AUC24MIC) in both human and mouse: 532 𝐴𝑈𝐶24𝑀𝐼𝐶 = (𝐴𝑈𝐶24 𝑀𝐼𝐶 ) 𝑝𝑎𝑟𝑒𝑛𝑡 + (𝐴𝑈𝐶24 𝑀𝐼𝐶 ) 𝑚𝑒𝑡𝑎𝑏𝑜𝑙𝑖𝑡𝑒 533 To enable comparisons of mouse exposures against clinical plasma targets, simulated blood 534 concentrations were converted to plasma values using the compound -specific geometric 535 mean Blood-to-Plasma (BP) ratio (Table 2). The area under the curve, during the 24 hours 536 post drug administration (AUC0-24, ngh/mL) was determined, assuming a dense simulation 537 time grid (i.e., 0.1 h sampling frequency) which enabled robust determination of the total 538 exposure in plasma. PopPK modelling was performed with Monolix Suite® (Build 2024R1, 539 Lixoft) running on a Windows 11, 64 -bit operating system. All models were identified 540 using a Stochastic Approximation Expectation-Maximization ( SAEM) algorithm. All 541 analyses for the extrapolation of exposure metrics were performed using R Statistical 542 Software (v4.4.2; R Core Team 2024). 543 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 17 Logistic Emax Model 544 A logistic Emax model was applied, as previously described [23] using data obtained from 545 Study 1 and Study 2 plus a historical dataset. The historical dataset consisted of data 546 utilized in previous modelling efforts [37] as well as other literature data [19, 38, 39] and 547 unpublished historical data generated at Evotec . The Evotec internal studies assessed 548 efficacy of RHZ/RH, BPaL and PaMZ dosed 5/7 or 7/7 days per week in a BALB/c RMM 549 and data from these studies is available via the C-PATH APEX platform (https://c-550 path.org/tools-platforms/tb-apex). 551 Data were pooled from the same treatment group collected from different studies (i.e. 552 within the historical and present datasets). For each combination, i and study k, γ and T50 553 parameters were estimated and subsequently used to calculate the related T90, according 554 to the following formula: 555 𝑇90𝑖 = 10 𝑙𝑜𝑔10(𝑇50𝑖)+( 1 𝛾′𝑖 𝑙𝑜𝑔10( 90 100−90)) (Population) 556 𝑇90𝑖𝑘 = 10 𝑙𝑜𝑔10(𝑇50𝑖𝑘)+( 1 𝛾′𝑖 𝑙𝑜𝑔10( 90 100−90)) (Individual) 557 Population parameters T50 and T90 enabled characterization of the average time to 50% 558 or 90% cure respectively, for a particular combination across studies, independent of study-559 related factors (e.g., starting inoculum), while individual parameters (e.g. T50ik and T90ik) 560 provide the average time to 50% or 90% cure for each specific study. To limit the number 561 of parameters to be estimated and overcome identifiability issues , c ombinations in the 562 historical dataset that share the same drugs but were administered at different dose levels 563 or with differing dose schedules were parametrized with different T50 parameters, 564 assuming the same γ value (i.e. steepness of the relapse -time curve). To estimate model 565 precision and support comparison s of the combinations, confidence intervals (CI) were 566 computed for T90s. Since T90 was derived from T50 and γ estimates, the delta method 567 [41] was employed to compute an approximation of its standard error (SE), assuming that 568 correlation between T50 and γ is negligible (assessed by the non -parametric Spearman 569 correlation test being T50 and γ not normally distributed, with p -value lower than 0.05 570 considered statistically significant) . Once T90 SE had been obtained, 2.5 th and 97.5 th 571 percentiles (i.e. CI lower limit and C Iupper limit , respectively) were computed for each 572 combination by assuming normally distributed data, as follows: 573 𝑇90 𝐶𝐼𝑙𝑜𝑤𝑒𝑟 𝑙𝑖𝑚𝑖𝑡 = 𝑇90 − 𝑆𝐸 ∗ 1.96 574 𝑇90 𝐶𝐼𝑢𝑝𝑝𝑒𝑟 𝑙𝑖𝑚𝑖𝑡 = 𝑇90 + 𝑆𝐸 ∗ 1.96 575 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 18 The logistic Emax model was developed using NONMEM® 7.5.1 and data handling was 576 performed using SAS® version 9.4 and R Statistical Software (v4.4.2; R Core Team 2024). 577 Statistical Analysis 578 Statistical analysis was conducted to evaluate differences between a set of pairwise drug 579 combinations that differed by only one component. The analysis considered both overall 580 T90 results and exposure levels (AUC0-24) of the drugs common to the combinations. 581 Specifically, T90 results were compared using a Z -test, assuming that T90 values are 582 normally distributed, and their standard errors are known. AUC comparisons were 583 performed using a t -test with unequal variances on log -transformed data. No formal 584 correction for multiple comparisons was applied, as the analyses were exploratory . P-585 values below 0.05 were considered statistically significant. 586

587 This work was supported by the Gates Foundation [ INV-008993]. The conclusions and 588 opinions expressed in this work are those of the author(s) alone and shall not be attributed 589 to the Foundation. Under the grant conditions of the Foundation, a Creative Commons 590 Attribution 4.0 License has already been assigned to the Author Accepted Manuscript 591 version that might arise from this submission. Please note works submitted as a preprint 592 have not undergone a peer review process. 593 We would like to thank the Evotec BSL3 In Vivo Pharmacology, Bioanalytical and 594 Pharmacometrics teams for their valuable support and collaboration. We are especially 595 grateful to Augusto Celon and Andrea Boscolo Panzin for their insightful input and 596 contributions to the analytical aspects of this work, as well as Jeanne Jaen, Fanny Deglave 597 and Eric Erdocain for expert bioanalytical work. 598 Conflict of interest statement 599 C.A.-P. is a full-time employee and potential stockholder of Johnson &Johnson (previously 600 Janssen Pharmaceutica). 601 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 19 602

603 [1] World Health Organization, Global tuberculosis report 2024 604 [2] Cox H, Dickson-Hall L, Ndjeka N, Van't Hoog A, Grant A, Cobelens F, Stevens W, 605 Nicol M . 2017. Delays and loss to follow -up before treatment of drug -resistant 606 tuberculosis following implementation of Xpert MTB/RIF in South Africa: a 607 retrospective cohort study. PLoS Med. 14(2):e1002238. doi: 608 10.1371/journal.pmed.1002238. 609 [3] Lienhardt C, Dooley KE, Nahid P, Wells C, Ryckman TS, Kendall EA, Davies G, 610 Brigden G, Churchyard G, Cirillo DM, Di Meco E, Gopinath R, Mitnick C, Scott 611 C, Amanullah F, Bansbach C, Boeree M, Campbell M, Conradie F, Crook A, Daley 612 CL, Dheda K, Diacon A, Gebhard A, Hanna D, Heinrich N, Hesseling A, Holtzman 613 D, Jachym M, Kim P, Lange C, McKenna L, Meintjes G, Ndjeka N, Nhung NV, 614 Nyang'wa BT, Paton NI, Rao R, Rich M, Savic R, Schoeman I, Makokotlela BS, 615 Spigelman M, Sun E, Svensson E, Tisile P, Varaine F, Vernon A, Diul MY, 616 Kasaeva T, Zignol M, Gegia M, Mirzayev F, Schumacher SG. 2024. Target 617 regimen profiles for tuberculosis treatment. Bull World Health Organ. 102(8):600-618 607. doi: 10.2471/BLT.24.291881. Epub 2024 May 28. 619 [4] WHO Target regimen profiles for tuberculosis treatment. Nov 2, 620 2023. https://www.who.int/publications-detail-redirect/9789240081512 [DOI] 621 [PMC free article] [PubMed] 622 [5] Hards K, Robson JR, Berney M, Shaw L, Bald D, Koul A, Andries K, Cook GM. 2015. 623 Bactericidal mode of action of bedaquiline. J Antimicrob Chemother. 70(7):2028 -624 37. doi: 10.1093/jac/dkv054. Epub 2015 Mar 8. PMID: 25754998 625 [6] Conradie F, Bagdasaryan TR, Borisov S, Howell P, Mikiashvili L, Ngubane N, 626 Samoilova A, Skornykova S, Tudor E, Variava E, Yablonskiy P, Everitt D, Wills 627 GH, Sun E, Olugbosi M, Egizi E, Li M, Holsta A, Timm J, Bateson A, Crook AM, 628 Fabiane SM, Hunt R, McHugh TD, Tweed CD, Foraida S, Mendel CM, Spigelman 629 M; ZeNix Trial Team . 2022 . Bedaquiline-Pretomanid-Linezolid Regimens for 630 Drug-Resistant Tuberculosis. N Engl J Med 387(9): 810 -823. 631 10.1056/NEJMoa2119430 632 [7] Khoshnood S, Taki E, Sadeghifard N, Kaviar VH, Haddadi MH, Farshadzadeh Z, 633 Kouhsari E, Goudarzi M, Heidary M. 2021. Mechanism of Action, Resistance, 634 Synergism, and Clinical Implications of Delamanid Against Multidrug -Resistant 635 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 20 Mycobacterium tuberculosis . Front Microbiol 12: 717045. 636 10.3389/fmicb.2021.717045 637 [8] Hariguchi N, Chen X, Hayashi Y, Kawano Y, Fujiwara M, Matsuba M, Shimizu H, 638 Ohba Y, Nakamura I, Kitamoto R, Shinohara T, Uematsu Y, Ishikawa S, Itotani M, 639 Haraguchi Y, Takemura I, Matsumoto M. 2020. OPC-167832, a Novel Carbostyril 640 Derivative with Potent Antituberculosis Activity as a DprE1 Inhibitor. Antimicrob 641 Agents Chemother 64(6). 10.1128/aac.02020-19 642 [9] Tenero D, Derimanov G, Carlton A, Tonkyn J, Davies M, Cozens S, Gresham S, 643 Gaudion A, Puri A, Muliaditan M, Rullas -Trincado J, Mendoza -Losana A, 644 Skingsley A, Barros-Aguirre D. 2019. First-Time-in-Human Study and Prediction 645 of Early Bactericidal Activity for GSK3036656, a Potent Leucyl-tRNA Synthetase 646 Inhibitor for Tuberculosis Treatment. Antimicrob Agents Chemother 63(8). 647 10.1128/aac.00240-19 648 [10] Brown KL, Wilburn KM, Montague CR, Grigg JC, Sanz O, Pérez -Herrán E, Barros 649 D, Ballell L, VanderVen BC, Eltis LD . 2023. Cyclic AMP-Mediated Inhibition of 650 Cholesterol Catabolism in Mycobacterium tuberculosis by the Novel Drug 651 Candidate GSK2556286. Antimicrob Agents Chemother 67(1): e0129422. 652 10.1128/aac.01294-22 653 [11] Bruinenberg P, Nedelman J, Yang TJ, Pappas F, Everitt D . 2022. Single Ascending-654 Dose Study To Evaluate the Safety, Tolerability, and Pharmacokinetics of 655 Sutezolid in Healthy Adult Subjects. Antimicrob Agents Chemother 66(4): 656 e0210821. 10.1128/aac.02108-21 657 [12] Nuermberger EL, Martínez -Martínez MS, Sanz O, Urones B, Esquivias J, Soni H, 658 Tasneen R, Tyagi S, Li SY, Converse PJ, Boshoff HI, Robertson GT, Besra GS, 659 Abrahams KA, Upton AM, Mdluli K, Boyle GW, Turner S, Fotouhi N, Cammack 660 NC, Siles JM, Alonso M, Escribano J, Lelievre J, Rullas-Trincado J, Pérez-Herrán 661 E, Bates RH, Maher -Edwards G, Barros D, Ballell L, Jiménez E. 2022. 662 GSK2556286 Is a Novel Antitubercular Drug Candidate Effective In Vivo with the 663 Potential To Shorten Tuberculosis Treatment. Antimicrob Agents Chemother. 2022 664 Jun 21;66(6):e0013222. doi: 10.1128/aac.00132-22. Epub 2022 May 24. 665 [13] Tasneen R, Garcia A, Converse PJ, Zimmerman MD, Dartois V, Kurbatova E, Vernon 666 AA, Carr W, Stout JE, Dooley KE, Nuermberger EL. 2022. Novel Regimens of 667 Bedaquiline-Pyrazinamide Combined with Moxifloxacin, Rifabutin, Delamanid 668 and/or OPC -167832 in Murine Tuberculosis Models. Antimicrob Agents 669 Chemother. 66(4):e0239821. doi: 10.1128/aac.02398-21. 670 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 21 [14] Walter K, Te Brake LHM, Lemm AK, Hoelscher M, Svensson EM, Hölscher C, 671 Heinrich N. 2024. Investigating the treatment shortening potential of a combination 672 of bedaquiline, delamanid and moxifloxacin with and without sutezolid, in a murine 673 tuberculosis model with confirmed drug exposures. J Antimicrob Chemother. 674 1;79(10):2607-2610. doi: 10.1093/jac/dkae266. 675 [15] Williams K, Minkowski A, Amoabeng O, Peloquin CA, Taylor D, Andries K, Wallis 676 RS, Mdluli KE, Nuermberger EL. 2012. Sterilizing activities of novel combinations 677 lacking first- and second-line drugs in a murine model of tuberculosis. Antimicrob 678 Agents Chemother. 56(6):3114-20. doi: 10.1128/AAC.00384-12. 679 [16] Lanoix JP, Betoudji F, Nuermberger E. 2014. Novel regimens identified in mice for 680 treatment of latent tuberculosis infection in contacts of patients with multidrug -681 resistant tuberculosis. Antimicrob Agents Chemother. 2014;58(4):2316 -21. doi: 682 10.1128/AAC.02658-13. 683 [17] Cevik M, Thompson LC, Upton C, Rolla VC, Malahleha M, Mmbaga B, Ngubane N, 684 Abu Bakar Z, Rassool M, Variava E, Dawson R, Staples S, Lalloo U, Louw C, 685 Conradie F, Eristavi M, Samoilova A, Skornyakov SN, Ntinginya NE, Haraka F, 686 Praygod G, Mayanja -Kizza H, Caoili J, Balanag V, Dalcolmo MP, McHugh T, 687 Hunt R, Solanki P, Bateson A, Crook AM, Fabiane S, Timm J, Sun E, Spigelman 688 M, Sloan DJ, Gillespie SH; SimpliciTB Consortium . 2024. Bedaquiline -689 pretomanid-moxifloxacin-pyrazinamide for drug -sensitive and drug -resistant 690 pulmonary tuberculosis treatment: a phase 2c, open -label, multicentre, partially 691 randomised controlled trial. The Lancet Infectious Diseases 24(9): 1003 -1014. 692 10.1016/S1473-3099(24)00223-8 693 [18] Li SY, Irwin SM, Converse PJ, Mdluli KE, Lenaerts AJ, Nuermberger EL. 2015. 694 Evaluation of moxifloxacin-containing regimens in pathologically distinct murine 695 tuberculosis models. Antimicrob Agents Chemother. 59(7):4026 -30. doi: 696 10.1128/AAC.00105-15. 697 [19] Li SY, Tasneen R, Tyagi S, Soni H, Converse PJ, Mdluli K, Nuermberger EL . 2017. 698 Bactericidal and Sterilizing Activity of a Novel Regimen with Bedaquiline, 699 Pretomanid, Moxifloxacin, and Pyrazinamide in a Murine Model of Tuberculosis. 700 Antimicrob Agents Chemother 61(9). 10.1128/aac.00913-17 701 [20] Zhang M, Li SY, Rosenthal IM, Almeida DV, Ahmad Z, Converse PJ, Peloquin CA, 702 Nuermberger EL, Grosset JH. 2011. Treatment of tuberculosis with rifamycin -703 containing regimens in immune -deficient mice. Am J Respir Crit Care Med. 704 183(9):1254-61. 705 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 22 [21] Mourik BC, Svensson RJ, de Knegt GJ, Bax HI, Verbon A, Simonsson USH, de 706 Steenwinkel JEM. 2018. Improving treatment outcome assessment in a mouse 707 tuberculosis model. Nature/Scientific report 8:5714. 708 [22] Mudde SE, Ayoun Alsoud R, van der Meijden A, Upton AM, Lotlikar MU, Simonsson 709 USH, Bax HI, de Steenwinkel JEM. 2022. Predictive Modeling to Study the 710 Treatment-Shortening Potential of Novel Tuberculosis Drug Regimens, Toward 711 Bundling of Preclinical Data. J Infect Dis 225(11):1876-1885. 712 [23] Sordello S, Brock L, Tagliavini A, Federico D, Boulenc X, Pergher M, Huc Claustre 713 E, Metcalf D, Walter N, Robertson G, Clary J, Berg A, Mdluli K, Hermann D, 714 Flood D, Upton A. 2026. A modeling-based framework to evaluate forgiveness of 715 tuberculosis treatment in a BALB/c relapsing mouse model . Antimicrob Agents 716 Chemother Antimicrob Agents Chemother. 15:e0110925. doi: 10.1128/aac.01109-717 25. Online ahead of print.PMID: 41537590 718 [24] Tasneen R, Betoudji F, Tyagi S, Li SY, Williams K, Converse PJ, Dartois V, Yang T, 719 Mendel CM, Mdluli KE, Nuermberger EL. 2015 . Contribution of Oxazolidinones 720 to the Efficacy of Novel Regimens Containing Bedaquiline and Pretomanid in a 721 Mouse Model of Tuberculosis. Antimicrob Agents Chemother. 60(1):270 -7. doi: 722 10.1128/AAC.01691-15.[25] Li SY, Converse PJ, Betoudji F, Lee J, Mdluli K, 723 Upton A, Fotouhi N, Nuermberger EL. 2023. Next -Generation Diarylquinolines 724 Improve Sterilizing Activity of Regimens with Pretomanid and the Novel 725 Oxazolidinone TBI -223 in a Mouse Tuberculosis Model. Antimicrob Agents 726 Chemother. 67(4):e0003523. doi: 10.1128/aac.00035-23. 727 [26] Nimmo C, Millard J, van Dorp L, Brien K, Moodley S, Wolf A, Grant AD, Padayatchi 728 N, Pym AS, Balloux F, O'Donnell M. 2020. Population -level emergence of 729 bedaquiline and clofazimine resistance -associated variants among patients with 730 drug-resistant tuberculosis in southern Africa: a phenotypic and phylogenetic 731 analysis. Lancet Microbe. 2020 Aug;1(4):e165 -e174. doi: 10.1016/S2666 -732 5247(20)30031-8. PMID: 32803174 733 [27] Pieterman ED, Keutzer L, van der Meijden A, van den Berg S, Wang H, Zimmerman 734 MD, Simonsson USH, Bax HI, de Steenwinkel JEM. 2021. Superior Efficacy of a 735 Bedaquiline, Delamanid, and Linezolid Combination Regimen in a Mouse 736 Tuberculosis Model. J Infect Dis. 224(6):1039-1047. doi: 10.1093/infdis 737 [28] Matthew J Reichlen , Emmanuel Musisi , Samuel T Tabor , Holly Nielsen, Ashley M 738 Gerwing , Firat Kaya, Matthew Zimmerman, Martin I Voskuil, Gregory T 739 Robertson, Nicholas D Walter. 2025. Same class, different activity: Delamanid and 740 pretomanid have comparable bactericidal activity but pretomanid potently inhibits 741 Mycobacterium tuberculosis ribosomal rRNA synthesis. bioRxiv [Preprint]. 2025 742 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 23 Dec 30:2025.12.30.697025. doi: 10.64898/2025.12.30.697025.PMID: 41509415 743 PMCID: PMC12776329 DOI: 10.64898/2025.12.30.697025 744 [29] Dawson R, Diacon AH, Takuva S, Liu Y, Zheng B, Karwe V, Hafkin J. Quabodepistat 745 in combination with delamanid and bedaquiline in participants with drug -746 susceptible pulmonary tuberculosis: protocol for a multicenter, phase 2b/c, open -747 label, randomized, dose -finding trial to evaluate safety and efficacy. Trials. 2024 748 Jan 19;25(1):70. doi: 10.1186/s13063-024-07912-5. PMID: 38243296 749 [30] Lee BM, Harold LK, Almeida DV, Afriat -Jurnou L, Aung HL, Forde BM, Hards K, 750 Pidot SJ, Ahmed FH, Mohamed AE, Taylor MC, West NP, Stinear TP, Greening 751 C, Beatson SA, Nuermberger EL, Cook GM, Jackson CJ. 2020 Predicting 752 nitroimidazole antibiotic resistance mutations in Mycobacterium tuberculosis with 753 protein engineering. PLoS Pathog. 2020 Feb 7;16(2):e1008287. doi: 754 10.1371/journal.ppat.1008287. eCollection 2020 Feb. PMID: 32032366 755 [31] Dorman SE, Nahid P, Kurbatova EV, Phillips PPJ, Bryant K, Dooley KE, Engle M, 756 Goldberg SV, Phan HTT, Hakim J, Johnson JL, Lourens M, Martinson NA, 757 Muzanyi G, Narunsky K, Nerette S, Nguyen NV, Pham TH, Pierre S, Purfield AE, 758 Samaneka W, Savic RM, Sanne I, Scott NA, Shenje J, Sizemore E, Vernon A, Waja 759 Z, Weiner M, Swindells S, Chaisson RE, Tuberculosis Trials Consortium. 2021. 760 Four-month rifapentine regimens with or without moxifloxacin for tuberculosis. N 761 Engl J Med 384:1705–1718. 10.1056/NEJMoa2033400. 762 [32] Crowley BM, Boshoff HI, Boving A, Tan VY, Zhu J, Hoyt F, Miller RR, Ehrhart J, 763 Boyce CW, Young K, Nantermet PG, Su J, Yang L, Painter RE, Corcoran EB, Hoar 764 JL, Oh S, Holtzman DL, Levi M, Anderson A, Otieno MA, Zimmerman M, Kaya 765 F, Massoudi LM, Ramey ME, Bauman AA, Lenaerts AJ, Roberston GT, Dartois 766 V, Wells CD, Barry CE 3rd, Olsen DB. 2026. Discovery and development of a new 767 oxazolidinone with reduced toxicity for the treatment of tuberculosis. Nat Med. 768 2026 Jan 13. doi: 10.1038/s41591 -025-04164-x. Online ahead of print. PMID: 769 41530381. 770 [33] Li, W., Vazvaei-Smith, F., Dear, G., Boer, J., Cuyckens, F., Fraier, D., Liang, Y., Lu, 771 D., Mangus, H., Moliner, P., Pedersen, M. L., Romeo, A. A., Spracklin, D. K., 772 Wagner, D. S., Winter, S., & Xu, X. S. (2024). Metabolite Bioanalysis in Drug 773 Development: Recommendations from the IQ Consortium Metabolite Bioanalysis 774 Working Group. Clinical pharmacology and therapeutics, 115(5), 939 –953. 775 https://doi.org/10.1002/cpt.3144 776 777 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 24 [34] MacLeod, A. K., Coquelin, K. S., Huertas, L., Simeons, F. R. C., Riley, J., Casado, 778 P., Guijarro, L., Casanueva, R., Frame, L., Pinto, E. G., Ferguson, L., Duncan, C., 779 Mutter, N., Shishikura, Y., Henderson, C. J., Cebrian, D., Wolf, C. R., & Read, K. 780 D. (2024). Acceleration of infectious disease drug discovery and development 781 using a humanized model of drug metabolism. Proceedings of the National 782 Academy of Sciences of the United States of America, 121(7), e2315069121. 783 https://doi.org/10.1073/pnas.2315069121 784 785 [35] Rouan, M. C., Lounis, N., Gevers, T., Dillen, L., Gilissen, R., Raoof, A., & Andries, 786 K. (2012). Pharmacokinetics and pharmacodynamics of TMC207 and its N -787 desmethyl metabolite in a murine model of tuberculosis. Antimicrobial agents and 788 chemotherapy, 56(3), 1444–1451. https://doi.org/10.1128/AAC.00720-11 789 790 [36] Wallis, R. S., Dawson, R., Friedrich, S. O., Venter, A., Paige, D., Zhu, T., Silvia, A., 791 Gobey, J., Ellery, C., Zhang, Y., Eisenach, K., Miller, P., & Diacon, A. H. (2014). 792 Mycobactericidal activity of sutezolid (PNU-100480) in sputum (EBA) and blood 793 (WBA) of patients with pulmonary tuberculosis. PloS one, 9(4), e94462. 794 https://doi.org/10.1371/journal.pone.0094462 795 [37] Berg A, Clary J, Hanna D, Nuermberger E, Lenaerts A, Ammerman N, Ramey M, 796 Hartley D, Hermann D. 2022. Model -Based Meta-Analysis of Relapsing Mouse 797 Model Studies from the Critical Path to Tuberculosis Drug Regimens Initiative 798 Database. Antimicrob Agents Chemother. 2022 Mar 15;66(3):e0179321. doi: 799 10.1128/AAC.01793-21. Epub 2022 Jan 31. 800 [38] Xu J, Li SY, Almeida DV, Tasneen R, Barnes -Boyle K, Converse PJ, Upton AM, 801 Mdluli K, Fotouhi N, Nuermberger EL. 2019. Contribution of Pretomanid to Novel 802 Regimens Containing Bedaquiline with either Linezolid or Moxifloxacin and 803 Pyrazinamide in Murine Models of Tuberculosis. Antimicrob Agents Chemother . 804 2019 Apr 25;63(5):e00021 -19. doi: 10.1128/AAC.00021 -19. Print 2019 805 May.PMID: 30833432 806 [39] Walter ND, Born SEM, Robertson GT, Reichlen M, Dide-Agossou C, Ektnitphong 807 VA, Rossmassler K, Ramey ME, Bauman AA, Ozols V, Bearrows SC, Schoolnik 808 G, Dolganov G, Garcia B, Musisi E, Worodria W, Huang L, Davis JL, Nguyen NV, 809 Nguyen HV, Nguyen ATV, Phan H, Wilusz C, Podell BK, Sanoussi ND, de Jong 810 BC, Merle CS, Affolabi D, McIlleron H, Garcia -Cremades M, Maidji E, Eshun -811 Wilson F, Aguilar-Rodriguez B, Karthikeyan D, Mdluli K, Bansbach C, Lenaerts 812 AJ, Savic RM, Nahid P, Vásquez JJ, Voskuil MI. 2021. Mycobacterium 813 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 25 tuberculosis precursor rRNA indicates treatment -shortening activity of drugs and 814 regimens. Nature Communications 12(1):2899. 815 816 [40] Heinrich N, Manyama C, Koele SE, Mpagama S, Mhimbira F, Sebe M, Wallis RS, 817 Ntinginya N, Liyoyo A, Huglin B, Minja LT, Wagnerberger L, Stoycheva K, 818 Zumba T, Noreña I, Peter DD, Makkan H, Sloan DJ, Brake LT, Schildkraut J, 819 Aarnoutse RE, McHugh TD, Wildner L, Boeree M, Aldana BH, Phillips PPJ, 820 Hoelscher M, Svensson EM; PanACEA consortium. 2025 . Sutezolid in 821 combination with bedaquiline, delamanid, and moxifloxacin for pulmonary 822 tuberculosis (PanACEA -SUDOCU-01): a prospective, open -label, randomised, 823 phase 2b dose -finding trial. Lancet Infect Dis. 2025 Nov;25(11):1208 -1218. doi: 824 10.1016/S1473-3099(25)00213-0. Epub 2025 Jul 8. 825 826 [41] Casella G, Wu R, Wu SS, Weidman ST; National Research Council (US) Board on 827 Mathematical Sciences and Their Applications . 2002. Making Sense of 828 Complexity: Summary of the Workshop on Dynamical Modeling of Complex 829 Biomedical Systems.Washington (DC): National Academies Press (US); 2002. 830 PMID: 25057601 831 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 26 832 FIGURE LEGENDS 833 Figure 1 Distribution of plasma AUC0-24 (AUC0-24/MIC for B and Sut) from popPK 834 analysis versus clinical target. 835 836 Figure 2 Lung bacterial burden at end of treatment in Study 1 (A) and Study 2 (B). 837 Whole lung CFU of BALB/c mice, intranasally infected with M.tb H37Rv, after different 838 durations of oral treatment with 4-drug combinations, dosed 5/7. Treatments were initiated 839 2-weeks post infection. 840 Figure 3 Probability of relapse with treatment duration for BALB/c mice treated with 841 combinations tested in Study 1 and Study 2 RMM studies (ordered by individual T90 842 in the legend). 843 Sterilization curves indicating the probability of relapse over treatment time, constructed 844 by fitting observed relapse data (Table 4) to an Emax model developed using a large 845 historical RMM dataset. Observed relapse data are indicated for each test regimen using 846 open symbols and crosses. The time to 50% cure/relapse is estimated from these curves, 847 and the time to 90% cure (i.e. 10% relapse) is derived utilizing time to 50% cure estimates 848 together with steepness of the curve (gamma) as explained in Methods. 849 Figure 4 Ranking of population T90 values with 95% confidence intervals for all 850 combinations (historical dataset + Study 1 and Study 2 RMM combinations). 851 852 853 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 27 854 TABLES 855 856 Delamanid-containing combinations Pretomanid-containing combinations Nitroimidazole-free combinations 1. B Del Q 830 11. B Pa Q 656 21. B Q 656 286 2. B Del Q 286 12. B Pa Q 286 22. B Q 656 Sut 3. B Del Q Sut 13. B Pa Q Sut 23. B Q Sut 286 4. B Del 656 286 14. B Pa 656 286 24. Q 656 286 Sut 5. B Del 830 Sut 15. B Pa 830 Sut 25. B 656 286 Sut 6. B Del 286 Sut 16. B Pa 286 Sut 7. Del Q 656 286 17. Pa Q 656 286 8. Del Q 656 Sut 18. Pa Q 656 Sut 9. Del Q Sut 286 19. Pa Q Sut 286 10. Del 656 286 Sut 20. Pa 656 286 Sut Total: 10 Total: 10 Total: 5 Table 1 PAN-TB 4-drug Combinations Combinations in blue font were tested in Study 1. Combinations in black font were tested in Study 2. 830 was replaced by 656 in Study 2. 857 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 28 858 Compounds (Blood to plasma ratio) Target human plasma AUC0-24 (µg*h/mL)

for target human plasma AUC0-24 Initial Dose (mg/kg) Mouse/ human plasma AUC0-24 ratio Dose selected for RMM studies (mg/kg) Bedaquiline (0.60) 23 NDA 204384 25 1.02 25 Bedaquiline active metabolite BDQ-M2 (0.83) 6 - - Pretomanid (1.71) 52 NDA 212862 40 1.35 40 Delamanid (0.48) 5.9 Provided by Otsuka 6 3.4 2 Sutezolid (1.57) 7.1 (Wallis et al., 2014) 100 1.1 100 Sutezolid active metabolite PNU-101603 (1.01) 36.8 - - Quabodepistat (0.64) 2.5 Dawson R et al 2025 9 3.58 14 GSK2556286 [286] (1.04) 22.5 (predicted) provided by GSK 35 1.12 35 GSK3211830 [830] (2.09) 6.5 (predicted) provided by GSK 2 0.2 5 GSK3036656 [656] (2.04) 6.5 (based on clinical exposure) (Tenero et al 2019) 4 0.5 8 Table 2 Doses Used in Mouse Studies to reach Clinical Target Plasma Exposures 859 860 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 29 A Study 1 861 Mean + SD Log10 CFU/ lung (positive culture mice/total number of mice) Groups Day -13 (1 dpi) Day 0 end of 2W Treatment end of 4W Treatment end of 8W Treatment Untreated 5.0 ± 0.1 (5/5) 7.2 ± 0.2 (5/5) NA NA NA B Pa 286 Sut NA NA 4.5 ± 0.2 (5/5) 0.0 ± 0.0 (0/5) 0.0 ± 0.0 (0/6) B Q Sut 286 NA NA 4.6 ± 0.3 (5/5) 0.0 ± 0.0 (0/5) 0.0 ± 0.0 (0/6) B Pa M Z NA NA 4.3 ± 0.2 (5/5) 0.5 ± 0.5 (2/4) 0.0 ± 0.0 (0/5) B Pa 830 Sut NA NA 3.7 ± 0.7 (5/5) 1.3 ± 0.9 (3/4) 0.0 ± 0.0 (0/6) B Pa Q Sut NA NA 4.4 ± 0.2 (5/5) 1.4 ± 0.9 (4/5) 0.0 ± 0.0 (0/6) B Del 286 Sut NA NA 4.9 ± 0.9 (5/5) 2.1 ± 1.4 (3/4) 0.0 ± 0.0 (0/6) B Del 830 Sut NA NA 4.9 ± 0.4 (5/5) 3.2 ± 0.3 (5/5) 0.0 ± 0.0 (0/6) Pa Q Sut 286 NA NA 4.6 ± 0.6 (5/5) 1.4 ± 1.6 (2/4) 0.4 ± 0.8 (1/3) B Del Q 286 NA NA 5.6 ± 0.1 (5/5) 3.6 ± 0.5 (4/4) 0.4 ± 0.6 (1/3) B Del Q Sut NA NA 4.9 ± 0.6 (5/5) 4.3 ± 0.8 (5/5) 0.4 ± 0.5 (3/6) B Del Q 830 NA NA 5.8 ± 0.3 (5/5) 4.0 ± 0.6 (5/5) 1.0 ± 1.1 (3/6) B Pa Q 286 NA NA 5.5 ± 0.2 (5/5) 3.4 ± 0.3 (5/5) 1.4 ± 0.4 (6/6) Del Q Sut 286 NA NA 5.2 ± 0.5 (5/5) 3.7 ± 0.6 (4/4) 2.4 ± 0.8 (6/6) R H Z E / R H NA NA 6.0 ± 0.4 (5/5) 4.2 ± 0.3 (5/5) 2.2 ± 0.2 (5/5) 862 863 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 30 B Study 2 864 Mean + SD Log10 CFU/ lung (positive culture mice/total number of mice) Groups Day -13 (1 dpi) Day 0 end of 2W Treatment end of 4W Treatment end of 8W Treatment Untreated 4.6 ± 0.3 (5/5) 7.4 ± 0.1 (5/5) NA NA NA B Pa M Z NA NA 4.8 ± 0.3 (3/3) 0.4 ± 0.3 (2/3) 0.0 ± 0.0 (0/3) B Pa 656 286 NA NA 3.8 ± 0.4 (5/5) 0.4 ± 0.8 (1/4) 0.1 ± 0.2 (1/6) B Q 656 Sut NA NA 4.1 ± 0.1 (3/3) 2.4 ± 0.3 (3/3) 0.4 ± 0.8 (1/4) Pa Q 656 Sut NA NA 4.6 ± 0.3 (4/4) 2.2 ± 0.7 (4/4) 0.6 ± 0.9 (2/5) B Q 656 286 NA NA 4.3 ± 0.1 (3/3) 2.7 ± 0.5 (4/4) 0.4 ± 0.7 (2/4) B Del 656 286 NA NA 5.6 ± 0.1 (2/2) 2.2 ± 0.3 (3/3) 0.6 ± 0.7 (3/5) B 656 286 Sut NA NA 3.9 ± 0.5 (3/3) 2.6 ± 0.0 (2/2) 0.8 ± 1.1 (1/2) Pa Q 656 286 NA NA 5.7 ± 0.1 (4/4) 1.8 ± 0.4 (4/4) 1.3 ± 1.2 (3/5) Del 656 286 Sut NA NA 4.3 ± 1.2 (5/5) 3.5 ± 0.2 (5/5) 0.6 ± 0.6 (4/6) Pa 656 286 Sut NA NA ND 2.5 ± 0.2 (4/4) 1.6 ± 0.0 (1/1) B Pa Q 656 NA NA 4.8 ± 0.1 (5/5) 3.9 ± 0.2 (5/5) 1.5 ± 1.0 (3/3) Q 656 286 Sut NA NA 2.5 ± 0.7 (4/4) 2.5 ± 0.9 (5/5) 2.2 ± 0.6 (6/6) Del Q 656 Sut NA NA 4.0 ± 1.4 (5/5) 3.5 ± 0.3 (4/4) 2.0 ± 0.6 (5/5) R H Z E NA NA 5.4 ± 0.2 (5/5) 3.9 + 0.1 (5/5) 2.4 + 0.3 (6/6) Del Q 656 286 NA NA 6.1 ±0.4 (5/5) 5.5 ±0.2 (5/5) 4.4 ±0.1 (6/6) Table 3: Lung bacterial burden prior and at the end of treatment in Study 1 (A) and Study 865 2 (B) after different durations of oral treatment with 4-drug combinations. 866 867 868 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 31 869 870 A _ Study 1 871 COMBINATION 2W TX + 12WR 4W TX + 12WR 6W TX + 12WR 8W TX + 12WR 10W TX + 12WR 12W TX + 12WR 16W TX + 12WR B Pa M Z 100% (5/5) 16% (1/6) 0% (0/6) 0% (0/5) - 0% (0/4) - B Pa 830 Sut - 100% (4/4) 100% (4/4) 0% (0/6) 0% (0/5) 0% (0/6) - B Pa 286 Sut - 100% (3/3) 75% (3/4) 33% (2/6) 0% (0/6) 0% (0/6) - B Q Sut 286 - 100% (4/4) 75% (3/4) 33% (2/6) 20% (1/5) 0% (0/6) - B Pa Q Sut - 100% (4/4) 75% (3/4) 66% (4/6) 16% (1/6) 0% (0/6) - B Del 286 Sut - 100% (4/4) 100% (4/4) 100% (6/6) 33% (2/6) 0% (0/6) - B Del Q 286 - 100% (4/4) 100% (4/4) 66% (4/6) 50% (3/6) 0% (0/6) - B Pa Q 286 - 100% (4/4) 100% (4/4) 100% (6/6) 16% (1/6) 20% (1/5) - B Del 830 Sut - 100% (4/4) 100% (4/4) 100% (4/4) 83% (5/6) 60% (3/5) - B Del Q Sut - 100% (3/3) 100% (4/4) 100% (2/2) 100% (6/6) 83% (5/6) - Del Q Sut 286 - 100% (4/4) 100% (4/4) 100% (6/6) 100% (6/6) 83% (5/6) - 2R H Z E /2R H - - - 100% (6/6) 100% (6/6) 83% (5/6) 0% (0/6) Pa Q Sut 286 - 100% (4/4) 100% (4/4) 83% (5/6) 83% (5/6) 100% (6/6) - B Del Q 830 - 100% (4/4) 100% (4/4) 100% (6/6) 100% (5/5) 100% (5/5) - 872 873 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 32 874 875 B _ Study 2 876 COMBINATION 2W TX + 12WR 4W TX + 12WR 6W TX + 12WR 8W TX + 12WR 10W TX + 12WR 12W TX + 12WR 16W TX + 12WR B Pa M Z 100% (4/4) 40% (2/5) 0% (0/5) 0% (0/5) - - - B Pa 656 286 - 100% (4/4) 100% (4/4) 83% (5/6) 33% (2/6) 0% (0/6) - B Q 656 286 - 100% (3/3) 100% (3/3) 83% (5/6) 33% (2/6) 0% (0/6) - B 656 286 Sut - 100% (3/3) 66% (2/3) 80% (4/5) 17% (1/6) 50% (3/6) - Pa Q 656 Sut - 100% (4/4) 100% (4/4) 100% (6/6) 80% (4/5) 67% (2/3) - B Del 656 286 - 100% (4/4) 100% (4/4) 100% (6/6) 80% (4/5) 33% (2/6) - B Q 656 Sut - 100% (4/4) 100% (4/4) 100% (5/5) 100% (5/5) 33% (2/6) - 2R H Z E/2R H - - - 100% (5/5) 83% (5/6) 50% (3/6) 17% (1/6) B Pa Q 656 - 100% (4/4) 100% (4/4) 100% (6/6) 100% (6/6) 66% (4/6) - Pa Q 656 286 - 100% (4/4) 100% (4/4) 100% (6/6) 100% (6/6) 100% (6/6) - Pa 656 286 Sut - 100% (4/4) 100% (4/4) 100% (6/6) 100% (5/5) 100% (6/6) - Q 656 286 Sut - 100% (4/4) 100% (4/4) 100% (6/6) 100% (6/6) 100% (5/5) - Del 656 286 Sut - 100% (3/3) 100% (4/4) 100% (6/6) 100% (6/6) 100% (5/5) - Del Q 656 SUT - 100% (4/4) 100% (4/4) 100% (6/6) 100% (6/6) 100% (6/6) - Del Q 656 286 - 100% (4/4) 100% (4/4) 100% (6/6) 100% (6/6) 100% (6/6) - Table 4 Percentage of relapsing mice, following different treatment durations with the 877 tested drug combinations in study 1 (A) and in Study 2 (B). 878 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 33 879 STUDY Combinations Estimated Population T50 (Months) Derived Population T90 (Months) Both B Pa M Z 0.86 1.06 Study 1 B Pa 830 Sut 1.95 2.11 Study 1 B Pa 286 Sut 2.04 2.51 Study 1 B Q Sut 286 2.11 2.94 Study 1 B Del 286 Sut 2.97 3.22 Study 2 B Pa 656 286 2.85 3.38 Study 2 B Q 656 286 2.85 3.39 Study 1 B Pa Q Sut 2.32 3.48 Study 1 B Del Q 286 2.77 3.56 Study 1 B Pa Q 286 2.92 4.00 Study 2 B Q 656 Sut 3.74 4.04 Study 2 B Del 656 286 3.60 4.28 Study 2 B Pa Q 656 3.99 4.40 Study 1 Del Q Sut 286 4.13 4.60 Study 1 B Del Q Sut 4.13 4.60 Study 1 B Del 830 Sut 3.93 5.08 Both R H Z E / R H 4.15 5.09 Study 2 Pa Q 656 Sut 4.08 5.35 Study 2 B 656 286 Sut 2.92 6.07 Study 1 Pa Q Sut 286, B Del Q 830 NC NC Study 2 Q 656 286 Sut, Del Q 656 Sut, Pa Q 656 286, Del 656 286 Sut, Del Q 656 286, Pa 656 286 Sut NC NC Table 5 Estimated population T50 and derived population T90 values for combinations 880 tested in Study1 and Study 2 RMM studies. 881 NC, Not computable for combinations which showed 100% relapse at the last treatment 882 time point. 883 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 34 884 885 Table 6 Statistical comparison of T90 and AUCratio between combinations differing by one 886 drug. 887 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint 35 Green (Delta T90 0) indicate statistically significant differences (p 2; AUC 0-24 889 DrugCombination1 > 2×AUC0-24 DrugCombination2) and red (AUC ratio <0.5; AUC 0-24 DrugCombination1 3 months). 892 NS means “Not significant” (p > 0.05). NC means “Not Computable” T90 value. 893 830 and 656 are considered as equivalent compounds. 894 895 896 .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint .CC-BY 4.0 International licenseavailable under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (whichthis version posted May 7, 2026. ; https://doi.org/10.64898/2026.05.05.722941doi: bioRxiv preprint