Isoniazid pharmacokinetics/pharmacodynamics as monotherapy and in combination regimen in the hollow fiber system model of Mycobacterium kansasii

Abstract

Background.

There is limited high quality evidence to guide the optimal doses of drugs for the treatment of Mycobacterium kansasii pulmonary disease (Mkn-PD).

Methods.

We performed (1) minimum inhibitory concentration experiment, (2) isoniazid dose-response study using the hollow fiber system model (HFS-Mkn) to determine PK/PD optimized exposure, and (3) another HFS-Mkn study to determine the efficacy of high dose isoniazid (15mg/kg/day) with standard dose rifampin (10mg/kg/day) and ethambutol (15mg/kg/day). Inhibitory sigmoid maximal effect model and linear regression was used for data analysis.

Results.

MIC of the 20 clinical isolates ranged between 0.5mg/L to 32mg/L. In the HFS-Mkn, isoniazid monotherapy failed to control the bacterial growth beyond day 7. On day 7, when the maximal Mkn kill was observed, the optimal isoniazid exposure for Mkn kill was calculated as 24hr area under the concentration-time curve to the MIC of 12.41. Target attainment probability of 300mg/day dose fell below 90% above the MIC of 1mg/L. High dose isoniazid combination sterilized the HFS-Mkn in 30-days with a kill rate of −0.152±0.018 log₁₀CFU/mL/day.

Conclusion.

Despite initial kill, isoniazid monotherapy failed due to resistance emergence. Our pre-clinical model derived results suggest that higher than currently recommended 300mg/day isoniazid dose may achieve better clinical efficacy against Mkn-PD.

INTRODUCTION

Mycobacterium kansasii is one of the most clinically relevant isolated species of nontuberculous mycobacteria (NTM), and the prevalence and incidence rates of M. kansasii varies by the geographical region.¹ The clinical symptoms of M. kansasii pulmonary disease (Mkn-PD) are indistinguishable from that of M. tuberculosis and the treatment regimen (isoniazid 5mg/kg/day, rifampin 10mg/kg/day, and ethambutol 15mg/kg/day) is similar to that of M. tuberculosis except pyrazinamide is excluded in the combination regimen.² However, unlike the 6-months short course M. tuberculosis therapy, the treatment of M. kansasii pulmonary disease continues for at least 12-months.² There is limited high quality evidence to guide the optimal treatment of Mkn-PD. Our hypothesis is that the optimal dose of drugs, including isoniazid, in the combination regimens for treatment of M. kansasii will be different from those designed for M. tuberculosis and other NTM.

The numbers and global scatter of patients pose a significant challenge in performing randomized controlled trials to optimize the treatment regimens for Mkn-PD. In the absence of prospective randomized control trial for developing short-course regimens, pre-clinical models like hollow fiber model of M. kansasii (HFS-Mkn)^3-5 could potentially fill in the knowledge gap. In the present study, [1] we performed isoniazid MIC experiments with the standard laboratory strain as well as with 20 clinical isolates, [2] performed PK/PD studies using the HFS-Mkn model to determine the optimal isoniazid exposure target for M. kansasii kill followed by in silico clinical trial simulations to determine the clinical dose to achieve the exposure target and the susceptibility breakpoint above which the given clinical isoniazid dose will fail to kill M. kansasii, hence therapy failure, and [3] tested high dose isoniazid combination to compare the kill rates with the standard dose combination regimen.

METHODS

Isoniazid minimum inhibitory concentration

Standard broth micro-dilution method,⁶ with the exception of the culture medium, was used to determine the MIC of isoniazid of the standard laboratory strain (ATCC 12478) and 20 clinical isolates. Prior to each the experiment, M. kansasii stock cultures were revived and grown to logarithmic phase cultures in Middlebrook 7H9 broth supplemented with 10% oleic acid-albumin-dextrose-catalase (OADC) (hereafter termed ‘broth’). Next, the inoculum was prepared by adjusting the turbidity of the cultures to McF 0.5 standard followed by 100-fold dilution to get the inoculum with a bacterial density of ~1.5x10⁵ CFU/mL. In the next step, 180μL of the inoculum was added to each of the 96-wells that were pre-filled with 20μL of the different concentrations of isoniazid. The isoniazid concentrations ranged from 0.5mg/L to 32mg/L. The plates were sealed in a Ziplock bag to avoid evaporation. After 7-days of incubation at 37°C, cultures were inspected visually and the lowest concentration completely inhibiting the M. kansasii growth was recorded as the MIC.⁶ The experiments were performed twice with three-replicates per drug concentration.

Isoniazid monotherapy and high dose combination therapy in the HFS-Mkn

The detailed description of the hollow fiber model system (HFS)⁷ and its adaptation to perform PK/PD studies with M. kansasii (HFS-Mkn) has been published elsewhere.^{3-5, 8, 9} In the present study we mimicked isoniazid doses in slow acetylators treated with human equivalent dose of 25, 50, 150, 300, 600, and 1200mg/day, and for fast acetylators the simulated doses were human equivalent dose of 50, 150, 300 and 600mg/day. The single-nucleotide polymorphisms in the N-acetyltransferase-2 gene separate people in two phenotypic categories ¹⁰. Therefore, in the HFS-Mkn studies, we intended to mimic 4hr drug half-life for slow acetylation and 1.8hr for the fast acetylation status. Since isoniazid has low protein binding, and the circulating media (Middlebrook 7H9 with 2% dextrose) do not have any protein ¹¹, the drug concentrations achieved in the HFS-Mkn represent free (f) drug concentrations.

Briefly, 20mL logarithmic phase growth Mkn cultures (inoculum preparation as described above) were inoculated into the peripheral compartment of each of the 12 HFS-Mkn units. We used the cellulosic fiber cartridges (FiebrCell Systems Inc., MD, USA) in the experiment. Isoniazid was infused into the central compartment via a programmable syringe pump to achieve the peak concentration (C_max) at the end of 1hr infusion. To validate the concentration-time profile with each isoniazid dose, we sampled the central compartment of each HFS-Mkn unit 8 times over the 24hr period. To quantitate the changes in the bacterial burden over time with each isoniazid dose, peripheral compartment of each HFS-Mkn unit was sampled on days 3,7,10,14,21, and 28. Samples were washed twice with normal saline to eliminate drug carryover followed by resuspension of the bacterial pellet in 1mL normal saline, 10-fold serial dilution and cultured on Middlebrook 7H10 agar supplemented with 10% OADC (here in termed “agar”). The same samples were also cultured on agar supplemented with isoniazid 3xMIC concentration to estimate any drug resistant subpopulation. The cultures were incubated at 37°C for 10 days before the colony forming units per milliliter (CFU/mL) were recorded. In addition, on study day 28, the samples were also used to determine the actual change in the MIC upon prolonged isoniazid exposure in the HFS-Mkn.

Data analysis and dose finding simulations

The isoniazid pharmacokinetics in the HFS-Mkn was modeled using a one-compartment model with first-order input and elimination, using WinNonLin.¹² The observed drug concentrations were used to calculate the fC_max and fAUC_0-24 as well as isoniazid half-life in the HFS-Mkn simulating fast and slow acetylation status, C_max/MIC and AUC_0-24/MIC. The inhibitory sigmoid E_max model was used to describe the relationship between the bacterial burden and isoniazid exposures. To identify the minimal dose of isoniazid to achieve or exceed the EC₈₀ exposure, we performed Monte Carlo simulations for 10,000 virtual individuals using the population pharmacokinetic parameter estimates published previously.^{10, 13} The target attainment probability, which is how well a dose of 300mg, 450mg, 600mg, or 900mg would achieve optimal exposure in patients, at each MIC ranging from 0. 5mg/L to 32mg/L, was then calculated. GraphPad Prism (v9) was used for graphing the data.

Isoniazid high dose combination therapy in the HFS-Mkn

Next, to test if high dose isoniazid can improve the efficacy of the standard combination regimen in terms of achieving faster kill rate, we performed a second HFS-Mkn study, except we used the intracellular infection model of HFS-Mkn. Inoculum was prepared as described above, and human derived THP-1 monocytes (ATCC TIB201) were infected with M. kansasii at a multiplicity of infection of 1:1 for 4hrs. After 4hr of co-culture, infected THP-1 cells were washed twice with warm RPMI-1640 by centrifugation at 1,000rpm for 5 min at room temperature to remove extracellular bacteria, and cells were resuspended in RPMI-160 supplemented with 2% fetal bovine serum (FBS). Finally, 20 mL of infected THP-1 cells were inoculated into the peripheral compartment of each of the four HFS-Mkn units. Two HFS-Mkn units were treated with an experimental regimen containing human equivalent isoniazid dose of 900mg daily plus 600mg rifampin daily plus ethambutol at 900mg daily dose. Being an intracellular experiment, the circulating media in this study was RPMI-1640 with 2% FBS. The sampling of the HFS-Mkn units for drug concentration measurement and bacterial burden was the same as the monotherapy HFS-Mkn study, except before serial dilution the THP-1 cells were stained with trypan blue for viability analysis and lysis using PBS-T (0.25% v/v) to release the intracellular bacteria.

RESULTS

The MIC of the M. kansasii laboratory strain, ATCC 12478, was 1mg/L. Table 1 list the individual MIC of the 20 clinical isolates, from the two separate experiments, with 3-replicates for each drug concentration. Therefore, 6 readouts per strains.

Table 1.

M. kansasii isoniazid MIC distribution in 20 clinical strains.

Strain ID	Isoniazid MIC (mg/L)
MRN2392724	2
MRN2739081	32
18:58688	2
MK_881	1
MK_915	0.5
MK_918	0.5
MK_860	1
MK_806	0.5
MK_978	0.5
MK_976	1
MK_925	32
MK_887	1
MK_829	32
MK_902	0.5
MK_817	0.5
MK_826	1
MK_997	8
MK_930	1
MK_1000	0.5
MK_1005	32

The concentration-time profiles of different isoniazid doses as achieved in the HFS-Mkn are shown in Figure 1. Figure 1A and 1B show the concentration-time profiles for the system mimicking the slow and fast acetylation status, respectively, and Figure 1C show the model fit for the measured versus pharmacokinetic modelled drug concentrations with a slope of 0.99 (95% Confidence Interval: 0.98 to 1.01). The isoniazid half-life in the central compartment of the HFS-Mkn was calculated as 6.03±0.36hr in systems mimicking slow acetylation versus 3.47±0.91hr for fast acetylation. The measured drug concentrations were then used to calculate the AUC_0-24/MIC, as summarized in Table 2.

Figure 1. Isoniazid pharmacokinetics in the HFS-Mkn.

Isoniazid concentration-time profiles of (A) slow acetylation and (B) fast acetylation phenotype as mimicked in the HFS-Mkn. To put the results in a clinical context, the C_max of isoniazid with 300 mg daily dose, range between 3-5 mg/L. (Cc) Regression between the pharmacokinetics modelled versus measured drug concentration, indicating good model fit (r²=0.99).

Table 2.

Isoniazid monotherapy observed concentrations in the HFS-Mkn and corresponding change in the MIC over time.

Regimen ID	Acetylation status	Cmax (mg/L)	AUC0-24/MIC	MIC (mg/L) on day 0	MIC (mg/L) on day 28
R1	Slow	0.67	5.43	1	>32
R2	Slow	1.17	9.68	1	>32
R3	Slow	2.21	20.74	1	>32
R4	Slow	4.69	41.12	1	>32
R5	Slow	8.72	78.38	1	>32
R6	Slow	17.14	166.43	1	>32
R7	Fast	0.94	5.72	1	>32
R8	Fast	1.88	11.96	1	>32
R9	Fast	3.67	23.99	1	>32
R10	Fast	8.19	47.78	1	>32
R11	Nontreated control	0	0	1	0.5*
R12	Nontreated control	0	0	1	0.5*

^*MIC of the non-treated controls was within 1-tube dilution. Hence, we considered no change in the MIC.

Figure 2A shows the kill curves with different isoniazid doses mimicking slow acetylation status. M. kansasii in nontreated HFS-Mkn units grew from 5.41 log₁₀ CFU/mL to 8.30 log₁₀ CFU/mL in 28 days. The maximum kill below stasis (day 0 or inoculum), as 2.57 log₁₀ CFU/mL, was recorded on day 7 after which all monotherapy failed to control the bacterial growth. While there was no pre-existing isoniazid resistance in the inoculum, acquired resistance emerged in all HFS-Mkn units so much so that the entire drug susceptible population was replaced by the isoniazid resistance subpopulation on study day 28 (Figure 2B). In Figure 2C we show the kill curves with isoniazid doses depicting the fast acetylation where maximum kill below stasis was recorded on study day 3 as 0.91 log₁₀ CFU/mL. There was no significant difference in the bacterial burden between the drug treated and nontreated control HFS-Mkn units. Isoniazid resistance emergence was quick, was not significantly different between the different isoniazid doses, and entire population was replaced by the isoniazid resistant subpopulation on day 28 of the study (Figure 2D). In addition to determining the isoniazid resistant subpopulation to the 3xMIC concentration, we also tested the actual change in the Mkn isoniazid MIC on the day 28, including the nontreated controls. As shown in Supplementary Figure 1, there was no change in the Mkn isoniazid MIC in the nontreated HFS-Mkn units, whereas the MICs were recorded as >32 mg/L in all isoniazid treated systems. Therefore, it was the drug exposure that led to the emergence of acquired drug resistance to isoniazid in the HFS-Mkn, and mono-therapy failure.

Figure 2. M. kansasii kill curves in the HFS-Mkn.

(A) In the HFS-Mkn treated with isoniazid mimicking slow acetylation status, isoniazid monotherapy failed after 7 days of treatment. By day 14 the bacterial burden reached above the stasis (day 0 or inoculum). (B) Emergence of isoniazid resistance was quick and entire bacterial population was isoniazid resistant by day 28. There was virtually no difference between the total and isoniazid resistant subpopulation, 8.30 log₁₀ CFU/mL versus 8.68+0.35 log₁₀ CFU/mL. (C) All four isoniazid exposures, mimicking the fast acetylation status, failed to control the Mkn growth in the HFS-Mkn model. Similar to the slow acetylation treatment arms, there was no difference in the bacterial burden between the non-treated control and isoniazid treated systems. (D) Resistance emergence to all four drug exposures mimicking fast acetylation status was quick and entire population was drug resistance on study day 28.

Since isoniazid dose in the currently recommended combination regimen is not discriminated for the acetylation status, we combined the data for both fast and slow acetylation status in the HFS-Mkn to determine the optimal isoniazid exposure for kill, using the four-parameter inhibitory sigmoid maximal effect model (Figure 3). The relationship between the bacterial burden and isoniazid AUC_0-24/MIC, on study day 7 was described using the following equation:

$$\text{Effect}({\log }_{10}\text{CFU/mL})=[7.71\text{-}{3.81}^{\ast }({\text{AUC}}_{0\text{-}24}∕\text{MIC}{)}^2]∕[{6.32}^2+({\text{AUC}}_{0\text{-}24}∕\text{MIC}{)}^2]$$

Figure 3. Relationship between M. kansasii bacterial burden and isoniazid exposure.

The measured drug concentrations were used to calculate isoniazid AUC_0-24/MIC. On study day 7, when maximal kill was observed, inhibitory sigmoidal E_max model was used to describe the relationship between isoniazid AUC_0-24/MIC and bacterial burden in the HFS-Mkn. On study day 7, the isoniazid EC₅₀ was calculated as AUC_0-24/MIC of 6.32.

Where, E_con (bacterial burden in nontreated control) was 7.71±0.45 log₁₀ CFU/mL, E_max (maximal kill) was 3.83±0.0.59 log₁₀ CFU/mL, EC₅₀ (drug exposure required for 50% of the maximal effect) was AUC_0-24/MIC of 6.28±1.56, and H (Hill’s constant) was 2.05±1.21. EC₈₀ or the optimal isoniazid exposure for Mkn kill was calculated as AUC_0-24/MIC of 12.41.

Next, we performed in silico clinical trial simulations to calculate the isoniazid AUC_0-24 with different doses, using the input parameters (Ka=2.10 (0.39 to 28.53) h⁻¹, V=0.89 (0.73 to 1.49) L/kg, Cl=16.81 (9.28 to 77.25) Lhr⁻¹) as reported earlier,¹⁰ and target attainment probability (TAP) of these doses to achieve the exposure target, i.e., AUC_0-24/MIC of 12.41, as identified in the HFS-Mkn. The isoniazid MICs used in the simulations were obtained from our 20 clinical isolates and additional 122 Mkn isolates from a previously published study¹⁴ where the MIC₅₀ and MIC₉₀ were 1 and 8 mg/L, respectively. Figure 4 shows the TAP of 300mg, 450mg, 600mg, and 900mg per day isoniazid dose. Since patients receive the same isoniazid dose regardless of the acetylator or health status, we did not use the acetylation status in the final analysis. The 300mg/ day dose (5mg/kg/day) failed to achieve optimal exposure in 90% of the simulated patients at an MIC of 2mg/L. Even the highest simulated dose of 900mg daily (i.e., 15mg/kg/day) failed to achieve the EC₈₀ exposure at an MIC of 8mg/L, which was the MIC₉₀ among the clinical isolates.¹⁴

Figure 4. MIC distribution and probability of target attainment of different isoniazid clinical doses.

The probability of isoniazid 300mg daily dose to achieve the optimal exposure fell below 90% at an MIC of 2mg/L, whereas the 450mg and 600mg daily dose could achieve the EC₈₀ exposure up to an MIC of 4mg/L. The highest simulated dose of 900mg daily failed to achieve the EC₈₀ exposure at an MIC of 8mg/L, which was the MIC₉₀ among the clinical isolates.¹⁴

Since the monotherapy HFS-Mkn results suggest that that higher than standard daily dose may be needed for better treatment outcome, we tested a high dose isoniazid (900mg/day) combination regimen in the HFS-Mkn. In the combination therapy study, the fC_max of isoniazid, rifampin and ethambutol measured in the HFS-Mkn units were 14.62±2.68 mg/L, 1.25±0.032 mg/L, and 7.05±0.61 mg/L, respectively (Supplementary Figure 2A). The number of viable THP-1 cells in the HFS-Mkn units treated with high dose isoniazid combination are shown in Supplementary Figure 2B. In Figure 5 we show that the bacteria in the nontreated control grew from 4.32± log₁₀ CFU/mL on day 0 (inoculum) to 6.07 log₁₀ CFU/mL in 30 days. Whereas the bacterial burden in the high dose isoniazid regimen treated HFS-Mkn units was below the limit of detection (0.69 log₁₀ CFU/mL) in 30 days, hence sterilization of the systems. The kill rate with the high dose isoniazid combination was calculated as −0.15±0.02 log₁₀ CFU/mL/day. There was no emergence of drug resistance to any of the study drug treated with the high dose isoniazid combination.

Figure 5. High dose isoniazid combination and killing efficacy in the HFS-Mkn.

Isoniazid 900mg daily dose in combination with standard dose rifampin and ethambutol sterilized the HFS-Mkn units in 30 days, unlike the standard dose combination that earlier shown failed to control the bacterial growth after day 21 in the same model.⁸

DISCUSSION

Nontuberculous mycobacteria come in many different forms, with M. kansasii presenting more with a tuberculosis-like picture. M. kansasii is a common enough problem to warrant public health attention, however, lacks information from randomized trials to identify the optimal treatment regimen and optimal duration of therapy.¹⁵ Indeed, M. kansasii is classified as a rare disease by the US Genetics and Rare Diseases Information Center [https://rarediseases.info.nih.gov/diseases]. The currently recommended regimen, based on multi-society guidelines, is a combination (similar to that used for tuberculosis) of daily isoniazid (300mg/day), rifampin (600 mg/day), and ethambutol (15mg/kg/day) where the patients should be treated for 12-months² The guidelines categorized the level of evidence for the currently recommended regimen as having the lowest evidence level by GRADE (Grading of Recommendations, Assessment, Development and Evaluations) criteria. This means that more work needs to be done to optimize current regimen and identify new regimens that could be administered for shorter duration of 6-months or less and identify a strategy that could overcome the paucity of randomized clinical trials.

Acknowledging that treatment of M. kansasii pulmonary disease is always a combination therapy, to achieve better therapy outcome we propose that the activity of each drug alone as monotherapy should be determined before proceeding to test the various drug combinations. In the present study, we show that irrespective of the acetylation status, isoniazid was able to kill M. kansasii. However, resistance emergence to monotherapy was faster in the HFS-Mkn mimicking the fast acetylation as it would result in lower drug exposure for the same clinical dose. The isoniazid clinical dose for treatment of M. kansasii pulmonary disease is adopted from that recommended for tuberculosis disease caused by M. tuberculosis where the current recommendations from the World Health Organization are to treat patients with an isoniazid dose of 5mg/kg, irrespective of the acetylator status. However, a recent study by Sundell et al¹⁶ describes a novel model-based isoniazid dosing strategy accounting for variability in the pharmacokinetics and metabolism affect by the genetic polymorphism. Such model-based dosing strategies could inform on dose selection, potentially reduce the risk of isoniazid toxicity and result in improved treatment outcome.

Unlike M. tuberculosis, there are no clinical studies with M. kansasii and isoniazid to determine the PK/PD exposure target and clinical dose finding. Humans are the best model for a human disease; however, it would be unethical to perform clinical studies without prior knowledge of drug efficacy and potency. Therefore, preclinical models are used to fill the knowledge gap. The drug-induced bacterial killing and resistance emergence is often determined using the in vitro systems, where HFS-Mkn provides an opportunity to study such relationship at dynamic drug concentrations in the same system.

In the present study, using the HFS-Mkn model and standard laboratory strain, we identified isoniazid AUC_0-24/MIC of 12.41 as the target drug exposure. The in silico clinical trial simulation indicate that the currently recommended 300mg daily dose will fail to achieve this exposure target in patients as soon as the MIC of the infecting strain reaches to 2mg/L. Even the highest simulated dose of 900mg daily was not able to achieve the EC₈₀ exposure at an MIC of 8mg/L, which was the MIC₉₀ among the clinical isolates. However, treatment of Mkn pulmonary disease is always a combination therapy where the drugs may protect each other. Therefore, it is a possibility that higher than 300mg/day dose may result in faster sputum culture conversion in patients where the isoniazid MIC of the infecting strain are higher than 2mg/L. Indeed, high dose isoniazid (15mg/kg/day or 900mg/day) in combination of rifampin and ethambutol was able to sterilize the HFS-Mkn in 28 days at a kill rate of −0.15±−0.02 log₁₀ CFU/mL/day. Elsewhere we have shown that, in the same HFS-Mkn model, the standard dose combination regimen failed to kill the entire bacterial population in 28 days and the kill rate was −0.038±0.038 log₁₀ CFU/mL/day, ~4-fold slower than the isoniazid high dose combination ⁸. Thus, in theory, the high dose isoniazid could potentially improve the efficacy of the standard drug combination. However, we advise caution when applying the results form in vitro and in silico models to patients, as clinical validation of our findings is still needed.

Our study has limitations. In the presented HFS-Mkn study, while we examined the 1200mg/day dose for slow acetylators, the same dose was not tested for the fast acetylation status. Due to the NAT2 polymorphism differences, slow acetylators will achieve higher drug exposure, hence increased risk of toxicity. Therefore, the 1200mg/day dose should have been tested for the fast acetylators. Next, the achieved isoniazid half-life in the HFS-Mkn was longer than the intended. However, we always report the free drug concentrations as measured in the samples collected from the hollow fiber systems and use to describe the drug-bacterial burden relationship. Therefore, while the drug half-life was longer than intended, it does not change the overall study outcome. In the HFS-Mkn, isoniazid monotherapy failed due to the acquired drug resistance. While we showed that the MIC of the laboratory strain changed from 1mg/L to >32mg/L after 28 days exposure to isoniazid, with the current data we could not establish the mechanism of isoniazid resistance in HFS-Mkn. We hypothesize that the change in the MIC and resistance emergence to isoniazid in the HFS-Mkn could be due to the efflux pumps, as we have shown elsewhere for M. tuberculosis.¹⁷. Finally, the experiments were performed with only one laboratory strain of Mkn to identify the isoniazid exposure target. This target could change for clinical isolates with different MIC.

To conclude, isoniazid exposure target for the M. kansasii kill, in the HFS-Mkn, was identified as the AUC_0-24/MIC of 12.56, and 15mg/kg/day dose in combination with other efficacious drugs could result in better therapy outcome for M. kansasii pulmonary disease. Clinical studies are warranted to determine the performance of the high dose isoniazid and validation of the HFS-Mkn findings.

HIGHLIGHTS.

Similar to the clinical observations, in the pre-clinical model isoniazid at currently prescribed 300mg/day dose failed to kill M. kansasii due to emergence of drug resistance. Higher isoniazid dose is required for better therapy outcome of M. kansasii pulmonary disease.