We used the Mtb laboratory strain H37Ra (ATCC#25177) and 20 clinical isolates (eight drug susceptible and 12 MDR-TB) provided by the South African Medical Research Council, TB Platform. The ethical approval was obtained from the Human Research Ethics Committee (UPHREC) at the Faculty of Health Sciences, University of Pretoria (Ethics Reference Number: 239/2016). Storage and culture conditions for log-phase growth Mtb cultures and transformation into semi-dormant bacteria for sterilizing activity experiments were as described in our previous publications (Gumbo et al., 2009; Srivastava et al., 2011a). Cefdinir was purchased from Sigma Aldrich (St Louis, MO, United States), and β-lactamase inhibitor, avibactam, was synthesized by the BOC Sciences (Shirley, NY, United States). Hollow fiber cartridges were purchased from FiberCell (Fredrick, MD, United States). BD BACTEC^TM MGIT^TM automated mycobacterial detection system and supplies were purchased from Becton, Dickinson and Company (NJ, United States).

We used two different methods to determine the cefdinir MIC—broth micro-dilution (CLSI, 2018), and MGIT liquid culture method (Bastian et al., 2001; Rusch-Gerdes et al., 2006; Deshpande et al., 2017b). The inoculum was prepared using the log-phase growth culture of Mtb H37Ra or the clinical isolates. The drug concentration range was 0, 1, 2, 4, 8, 16, 32, and 64 mg/L with or without combination of avibactam at a concentration of 15 mg/L, based on our previous experiments (Deshpande et al., 2017b; Srivastava et al., 2020b). Five hundred microlitres of the inoculum was added to each MGIT tube supplemented with 900 μl oleic acid, albumin, dextrose, and catalase [OADC] enrichment and 100 μl the drug; thus, a total volume of 8.5 ml. The MGIT time to positive (TTP) was recorded using the EpiCenter software (Bastian et al., 2001; Rusch-Gerdes et al., 2006). For the broth-micro dilution method, the inoculum preparation and the drug concentrations were the same as for the MGIT method, except the experiment was performed in 96-well plates. After 7 days of incubation, plates were visually examined using an inverted mirror and the concentration showing complete inhibition of the bacterial growth was recorded as the MIC. The experiments were performed twice with two replicates for each concentration.

The preparation of the inoculum and cefdinir concentration range were same as described above, except the experiment was carried out in 15 ml screw caped tubes with a total volume of 5 ml. The log-phase growth Mtb H37Ra cultures were co-incubated with different drug concentration, in replicate of three, at 37°C under 5% CO₂ and shaking conditions for 7 days. On day 7, the cultures were washed twice with normal saline to remove the carry-over drug, serially 10-fold diluted in normal saline and inoculated on Middlebrook 7H10 agar supplemented with 10% OADC (herein termed “agar”). The colony forming unit (CFU) with each concentration were recorded after 21 days of incubation at 37 C under 5% CO₂. The four-parameter inhibitory sigmoid E_max model was used to determine the relationship between the drug concentration and the bacterial burden.

In the preliminary drug screening studies (at static concentration), we found that there was no significant difference (data not shown) in cefdinir Mtb killing with or without addition of the β-lactamase inhibitor, avibactam. To confirm this, we performed an experiment with log-phase growth Mtb cultures to determine if the effect persists at the dynamic or fluctuating concentrations using the HFS-TB model of bactericidal effect (Srivastava et al., 2011a; Srivastava et al., 2011b). Twenty mL of log-phase growth Mtb H37Ra culture were inoculated into the peripheral compartment of each of eight HFS-TB units. Since percent of the time drug concentration persist above MIC (%T_MIC) is the PK/PD index linked to the β-lactam’s efficacy, the HFS-TB were treated with different cefdinir doses to achieve 50, 75, and 100%T_MIC with or without combination of avibactam at concentration of 15 mg/L. Drugs were infused into the central compartment via a computerized syringe pump over 1 h. The fresh media inflow rate (i.e., dilution) was set to mimic a 2 h cefdinir half-life (Zhang et al., 2011). The central compartment of each of the eight HFS-TB unit was sampled before drug infusion followed by at 1, 2, 3, 6, 12, 15, 18, 23.5, 25, 26, 28, 30, 36, 42, and 47.5 h after the administration of the first dose to validate the drug concentration-time profile. The peripheral compartment of each HFS-TB unit was sampled on day 0, 3, 7, 10, 14, 21, and 28 to quantify the bacterial burden. The samples were washed twice with normal saline to remove any carry over drug, serially 10-fold diluted and cultured on agar. The processed samples were also inoculated on agar supplemented with three times the cefdinir MIC to determine the proportion of the cefdinir resistant Mtb sub-population. The intent was not to determine the absolute change in the MIC. The cultures were incubated at 37°C for 21 days before CFUs were recorded. One portion of the processed sample (undiluted) was also inoculated in the MGIT tubes to record the TTP, as second pharmacodynamic measurement.

It is important that a drug can kill different metabolic populations of Mtb (Mitchison, 1979). Also, the efficacy determined using standard drug-suscetible laboratory strain may not be same against MDR-TB strains. Therefore, we performed sterilizing activity studies of cefdinir using a MDR-TB clinical strain (16D). The isoniazid and rifampin phenotypic resistance in this clinical starin was also confirmed by whole genome sequencing, using the methods described previously (Srivastava et al., 2006; Deshpande et al., 2017b; Srivastava et al., 2017). The whole genome sequencing showed presence of drug resistance associated mutation in katG (Ser315Thr), rpoB (Ser450Leu, Tyr564His), embB (Met306Val), pncA (Val139Gly), gidB (Leu16Arg, Ser100Phe) and gyrA (Clu21Gln, Ser95Thr, Gly247Ser, Gly668Asp) and ponA1 (Pro631Ser) genes of Mtb.

Prior to the experiment, 4 day old log-phase growth cultures were transformed into semi-dormant bacilli (SDB) growing under acidic condition (pH 5.8). The detailed method of transformation to SDB has been published elsewhere (Gumbo et al., 2009; Srivastava et al., 2011a). Next, we examined 14 different doses of cefdinir, in a combined dose-effect and dose-frationation study design to achieve 0, 8, 16, 25, 32, 42, 50, 64, 84, and 100%T_MIC with either once daily or twice daily dosing schedule. The number of HFS-TB units was 16, including two non-treated control systems. The study was performed without addition of avibactam. The sampling of the peripheral compartment to validate the drug concentration-time profile and of the central compartment to determine the total as well as the drug resistant sub-population was performed as described above.

Avibactam was measured using a previously validated method (Deshpande et al., 2017a; Deshpande et al., 2017b; Srivastava et al., 2020b). We developed an LC-MS/MS methods for measurement of cefdinir. Briefly, Cefdinir and ceftazidime-d5 (internal standard, IS) were purchased from Sigma (St Louis, MO, United States) and Toronto Research Chemicals (Toronto, Canada), respectively. LC-MS/MS analysis was performed using Waters Acquity UPLC coupled with Waters Xevo TQ mass spectrometer. Data was collected using MassLynx version 4.1 SCN810 software. Separation was achieved by injecting 2 μl of sample on a Waters Acquity UPLC HSS T3 column (50 × 2.1 mm; 1.8 μm) using a binary gradient. Stock solutions of the standard and IS were prepared in 80:20 methanol:water at a concentration of 1 mg/ml. Calibration curve, low- and high-quality control samples (LQC and HQC) were prepared by diluting the stock solution in blank medium. Samples were diluted 1:20 with IS solution in 0.1% aqueous formic acid (FA). Solvents for UPLC were: (A) 0.1% aqueous FA, and (B) 0.1% FA in methanol. Flow rate was 0.2 ml/min; total run time was 6 min. Compounds were detected using positive ESI in MRM mode. The transitions used were m/z 396–227 (cefdinir), and m/z 552–468 (ceftazidime-d5). The between day percentage coefficient of variation (%CV) for analysis of low and high (brackets) quality controls were 3% (1%). The inter- and intra-day %CV were 4 and 2%. The lower limit of quantitation was 0.01 µg/ml.

We performed pharmacokinetic modeling, with priors from literature, using the measured drug concentrations in the HFS-TB (Shimada et al., 1989; Srivastava et al., 2011a; Srivastava et al., 2017). We used two different software’s, ADAPT (D'argenio and Schumitzky, 1997) and Phoenix WinNonlin 8.1 (Certara USA, Inc., MO, United States) for pharmacokinetic modeling to compare the results. The measured drug concentration in each HFS-TB unit was used to calculate the ratio of peak to MIC (C_max/MIC), 0–24 h area under the concentration-time curve to MIC (AUC_0–24/MIC), and the %T_MIC for each cefdinir doses. Drug concentration and bacterial response relationships were examined using the inhibitory sigmoid E_max model for microbial kill, and one-way analysis of variance to compare the dosing schedule was performed in GraphPad Prism v 8.0 (La Jolla, CA, United States).

Since the circulating media in the HFS-TB was Middlebrook 7H9 both with 10% dextrose (i.e., no protein present binding), the measured drug concentrations represent the free or available drug in the HFS-TB units. The calculated C_max/MIC with three cefdinir doses were 4.46, 16.34, 82.88; %T_MIC were calculated as 50, 60, and 100; and the corresponding AUC_0–24/MIC were 31.16, 98.17, and 454.7. As shown in Figure 3A, difference in the TTP, recorded in the HFS-TB units treated with different cefdinir exposures in the presence or absence of avibactam was not statistically significant (p > 0.05). The kill curves with each cefdinir exposure, with or without avibactam, shown in Figure 3B, was not significantly different, similar to the TTP results. All three cefdinir exposures killed drug susceptible Mtb H37Ra, though the extents of kill varied in a dose dependent manner (2.11 vs 3.67 vs 6.95 log₁₀ CFU/ml, respectively in 28 days).

Since in the HFS-TB study performed with the log-phase growth cultures there was no significant different in the bacterial burden in the systems treated with or without combination of avibactam, this set of experiment was performed with cefdinir alone. Figures 4A,B show the PK modeled predicted and observed cefdinir concentrations in each HFS-TB unit, treated with once or twice daily dosing schedule. In the HFS-TB, the cefdinir clearance was calculated as 0.368 (95% CI: 0.366–0.369) L/h, volume of distribution of 0.248 (95% CI: 0.229–0.267) L, and half-life of 4.679 (95% CI: 4.33–5.038) h.

The extent of MDR-TB bacterial kill with different cefdinir exposure varied in a dose dependent manner. On day 28, the maximal kill (E_max) compare to the non-treated control with cefdinir exposure of 100%T_MIC exposure was 4.02 log₁₀ CFU/ml. There was no difference in the bacterial burden in systems treated with cefdinir once daily or twice daily, for the same AUC_0–24/MIC exposure (p > 0.05). Next, we determined the relationship between the drug exposure (C_max/MIC, AUC_0–24/MIC or %T_MIC) and bacterial burden at each sampling time-point using the inhibitory Sigmoid E_max model. The Akaike Information Criteria score (AIC) (Akaike, 1974) was used to select the PK/PD parameter (with lowest AIC score) associated with cefdinir microbial kill. Table 2 show the AIC scores for TTP and CFU/ml on each sampling day for each PK/PD index. We found that, on study day 28, for both TTP (Figures 5A–C) and log₁₀ CFU/ml (Figures 5D–F), AUC_0-24/MIC had lower AIC score compare to %T_MIC or C_max/MIC. Therefore, we determined that in the HFS-TB, AUC_0–24/MIC was the PK/PD index linked to the cefdinir’s sterilizing efficacy against Mtb. Using the CFU/ml readouts, the EC₅₀ was calculated as an AUC_0–24/MIC of 104.1 with an H of 0.80. The EC₈₀ was calculated an AUC_0–24/MIC of 578.86. There was no cefdinir resistance recorded on agar supplemented with three times MIC.