An In Vitro Perspective on What Individual Antimicrobials Add to Mycobacterium avium Complex Therapies

ABSTRACT

For Mycobacterium avium complex pulmonary disease (MAC-PD), current treatment regimens yield low cure rates. To obtain an evidence-based combination therapy, we assessed the in vitro activity of six drugs, namely, clarithromycin (CLR), rifampin (RIF), ethambutol (EMB), amikacin (AMK), clofazimine (CLO), and minocycline (MIN), alone and in combination, against Mycobacterium avium and studied the contributions of individual antibiotics to efficacy. The MICs of all antibiotics against M. avium ATCC 700898 were determined by broth microdilution. We performed kinetic time-kill assays of all single drugs and clinically relevant two-, three-, four-, and five-drug combinations against M. avium. Pharmacodynamic interactions of these combinations were assessed using area under the time-kill curve-derived effect size and Bliss independence. Adding a second drug yielded an average increase of the effect size (E) of 18.7% ± 32.9%, although antagonism was seen in some combinations. Adding a third drug showed a smaller increase in effect size (+12.2% ± 11.5%). The RIF-CLO-CLR (E of 102 log₁₀ CFU/ml · day), RIF-AMK-CLR (E of 101 log₁₀ CFU/ml · day), and AMK-MIN-EMB (E of 97.8 log₁₀ CFU/ml · day) regimens proved more active than the recommended RIF-EMB-CLR regimen (E of 89.1 log₁₀ CFU/ml · day). The addition of a fourth drug had little impact on effect size (+4.54% ± 3.08%). In vitro, several two- and three-drug regimens are as effective as the currently recommended regimen for MAC-PD. Adding a fourth drug to any regimen had little additional effect. In vitro, the most promising regimen would be RIF-AMK-macrolide or RIF-CLO-macrolide.

INTRODUCTION

Mycobacterium avium complex pulmonary disease (MAC-PD) is an increasingly important opportunistic infection that affects patients with underlying lung disease or local or systemic immunodeficiency (1). To treat MAC-PD, international guidelines recommend a three-drug combination therapy consisting of a macrolide (azithromycin or clarithromycin [CLR]), ethambutol (EMB), and rifampin (RIF), with amikacin (AMK) added for severe or refractory infections (2). Even with these aggressive multidrug regimens, culture conversion with prolonged treatment is attained in only 65% of patients (3, 4).

The exact contributions of the individual antimicrobial drugs to the efficacy of these regimens in vivo are controversial. In clinical trials, the addition of streptomycin to the RIF-EMB-macrolide recommended regimen yielded a short-term microbiological effect but no long-term clinical effect (5). RIF and EMB as a two-drug regimen are ineffective (6), but two recent studies have shown good outcomes of two-drug EMB-macrolide regimens without the development of macrolide resistance (7, 8). Also, recent research identified alternative antimicrobials, such as minocycline (MIN) and clofazimine (CLO), as possible additions to or replacements for antibiotics in the recommended regimen (9–11).

To study the contributions of individual antibiotics to the efficacy of these regimens in vitro, we performed single-drug time-kill assays (TKAs) of CLR, RIF, EMB, AMK, CLO, and MIN, as well as all two-, three-, four-, and five-drug combinations, and assessed activity and potential synergistic interactions against M. avium.

RESULTS

Susceptibility testing.

The MICs of the tested antimicrobials against M. avium ATCC 700898 were as follows: CLR, 1 mg/liter; EMB, 8 mg/liter; RIF, 8 mg/liter; AMK, 4 mg/liter; CLO, 1 mg/liter; MIN, 4 mg/liter.

Effect sizes for individual antimicrobials.

All effect size (E) values are reported as log₁₀ CFU per milliliter · day. All time-kill (TK) curves not shown in this article are available in Fig.S1 and S2 in the supplemental material. At 2× MIC, AMK had the highest E value (E of 82.6 log₁₀ CFU/ml · day) for single drugs, followed by RIF (E of 50.2 log₁₀ CFU/ml · day) and CLO (E of 40.4 log₁₀ CFU/ml · day); CLR had the lowest E value (E of 28.8 log₁₀ CFU/ml · day).

Effect sizes for two-drug combinations.

The TK curves for all clinically relevant combinations are shown in Fig. 1. E values for all drug combinations are presented in Table 1, and graphical representation of the increases in E values relative to the number of drugs is presented in Fig. 2.

FIG 1.

TK curves with M. avium ATCC 700898. (a) Combinations containing recommended regimen drugs (RIF, EMB, and macrolide). (b) Combinations containing AMK. (c) Combinations containing CLO. (d) Combinations containing MIN. GC, growth control. Error bars show the standard error of the mean (SEM).

TABLE 1.

Rank of all the tested antimicrobial combinations according to their effect sizes

Overall rank	Combination	E (log10 CFU/ml · day)	Category rank for combinations of:
			1 drug	2 drugs	3 drugs	4 drugs	5 drugs
1	RIF-CLR-CLO-AMK	108				1
2	RIF-CLR-CLO-MIN	106				2
3	RIF-CLR-CLO-EMB	103				3
4	CLO-CLR-MIN-EMB	102				4
5	RIF-CLR-CLO	102			1
6	CLR-RIF-EMB-AMK-CLO	102					1
7	RIF-CLR-AMK-EMB	102				5
8	RIF-CLR-AMK	101			2
9	RIF-CLR-MIN-EMB	101				6
10	AMK-MIN-EMB	96.8			3
11	AMK-CLO	95.0		1
12	CLO-CLR-MIN	94.5			4
13	MIN-CLR-AMK	93.5			5
14	RIF-CLR-MIN	92.4			6
15	AMK-CLO-MIN	91.2			7
16	CLO-CLR-EMB-AMK	90.3				7
17	RIF-CLR-EMBa	89.1			8
18	CLO-CLR-EMB	87.8			9
19	AMK-CLR-EMB	87.6			10
20	RIF-CLR	83.2		2
21	RIF-CLO-EMB	82.8			11
22	CLO-CLR	82.6		3
23	AMK-MIN-RIF	81.6			12
24	AMK	81.6	1
25	MIN-CLR	80.0		4
26	AMK-CLR	79.6		5
27	AMK-EMB-RIF	77.0			13
28	MIN-EMB-RIF	75.4			14
29	MIN-CLO	73.0		6
30	RIF-CLO	70.5		7
31	AMK-MIN	70.5		8
32	MIN-EMB	70.0		9
33	MIN-CLR-EMB	68.2			15
34	AMK-EMB	67.6		10
35	AMK-RIF	66.6		11
36	EMB-CLR	64.6		12
37	MIN	61.2	2
38	EMB-CLO	59.7		13
39	MIN-RIF	56.6		14
40	EMB-RIF	55.5		15
41	EMB	51.6	3
42	RIF	50.2	4
43	CLO	40.4	5
44	CLR	28.8	6

^aCurrently recommended drug regimen.

FIG 2.

Changes in effect size with the addition of drugs. Combinations built from AMK (a), MIN (b), EMB (c), RIF (d), CLO (e), and CLR (f) are shown. y axis, effect size, in log₁₀ CFU per milliliter · day; x axis, number of drugs.

AMK-CLO had the highest E value of 95.0 log₁₀ CFU/ml · day among two-drug combinations, similar to that of the recommended RIF-EMB-CLR combination (E of 89.1 log₁₀ CFU/ml · day). This was followed by four other combinations, each containing CLR. However, EMB-CLR (E of 64.6 log₁₀ CFU/ml · day) ranked 12th among the 15 two-drug combinations tested.

Effect sizes for three-drug combinations.

RIF-CLR-CLO (E of 102 log₁₀ CFU/ml · day) and RIF-CLR-AMK (E of 101 log₁₀ CFU/ml · day) had the highest E values among the three-drug combinations, higher than that of the recommended regimen RIF-EMB-CLR (E of 89.1 log₁₀ CFU/ml · day). In these combinations, the addition of the third drug, CLO or AMK, increased the effect of the RIF-CLR combination by 22.8% and 21.1%, respectively. Apart from these, five other three-drug combinations had E values higher than that of the recommended RIF-EMB-CLR combination (Table 1).

Effect sizes for four- and five-drug combinations.

The top three ranked combinations were an addition of AMK, MIN, or CLO to the backbone of RIF-EMB-CLR, with E values of 108, 106, and 103 log₁₀ CFU/ml · day, respectively, which would form a four-drug regimen. However, these combinations showed modest increases in E values, compared to the standard regimen of RIF-EMB-CLR (E value increases of 21.2%, 18.9%, and 15.6%, respectively). The clinically used RIF-EMB-CLR-AMK combination had a higher E value (E of 102 log₁₀ CFU/ml · day) than the recommended three-drug regimen (E of 89.1 log₁₀ CFU/ml · day). Addition of CLO to this combination, making it the five-drug combination CLR-RIF-EMB-AMK-CLO, did not add to the E value (E of 102 log₁₀ CFU/ml · day).

Comparison of mean relative effect sizes.

Adding a second drug led to a mean increase of 18.7% ± 31.9%, and addition of a third drug led to a further increase of 12.2% ± 11.5%; with the addition of a fourth drug, the mean increase was 4.54% ± 3.08% (Fig. 3; also see Table S1).

FIG 3.

Changes in mean relative effect size of two-, three-, and four-drug combinations. Whiskers represent 95% confidence intervals.

Bliss independence analyses.

The calculated Bliss independence-derived ΔE (ΔE_BI) values for the two- and three-drug combination TKAs are presented in Table 2.Among the two-drug combination, CLO-CLR against M. avium was the most synergistic combination, with the highest ΔE_BI of 57.5%; the most antagonistic combination was MIN-RIF, with an ΔE_BI of −27.5%. Among the three-drug combinations, CLR-RIF-CLO was most synergistic, with an ΔE_BI of 35.5%, and antagonism was seen for RIF-EMB-AMK, with an ΔE_BI of −20.7%. The sigmoidal E_max curves for the six antibiotics RIF, EMB, CLA, AMK, MIN, and CLO against M. avium, along with their E_max, E₅₀, and Hill slope values, are shown in Fig. S3. Detailed Bliss independence-derived data are presented in Tables S2 and S3.

TABLE 2.

Bliss independence-derived ΔE values for two- and three-drug combinations

Combination	ΔEBI (%)
Two-drug combinations
CLO-CLR	57.5
RIF-CLR	31.4
EMB-CLR	21.8
MIN-CLR	20.1
MIN-CLO	6.12
AMK-CLO	5.79
CLO-RIF	2.70
EMB-MIN	−1.22
EMB-CLO	−3.72
AMK-CLR	−8.94
EMB-RIF	−24.8
MIN-AMK	−25.0
EMB-AMK	−26.7
AMK-RIF	−27.4
MIN-RIF	−27.5
Three-drug combinations
CLR-RIF-CLO	35.5
CLR-EMB-CLO	34.9
CLR-CLO-MIN	32.5
CLR-AMK-EMB	22.1
CLR-RIF-EMB	12.0
CLR-RIF-MIN	11.9
CLR-RIF-AMK	6.37
EMB-RIF-CLO	1.07
MIN-EMB-AMK	−1.43
CLR-MIN-AMK	−2.93
CLR-EMB-MIN	−6.13
MIN-AMK-CLO	−6.27
MIN-RIF-EMB	−12.4
MIN-RIF-AMK	−16.8
RIF-EMB-AMK	−20.7

DISCUSSION

Here, we show that the recommended regimen for MAC-PD combines the killing capacity of RIF, EMB, and CLR with extensive synergy, but other combinations may also merit further preclinical and possibly clinical evaluation.

TK dose-response experiments with individual antibiotics against M. avium have been performed previously, with kill curves comparable to ours, with the exception of that for CLR, for which we found lower effectivity overall, which may be because of the different reference strain we used (11–14). The efficacy of the single antimicrobials and combinations are similar to the activity and ranking by Yajko et al. in a 7-day TKA investigating peak serum concentrations for each drug with two CFU measurements and a murine macrophage infection model (15). Those investigators also found a rifamycin-CLR-AMK combination to be especially effective. They found the recommended three-drug regimen to be more effective than we did, acknowledging they used the in vitro more active rifabutin instead of RIF. Generally, however, they too found a trend of diminishing return with adding more drugs (15).

Multidrug combinations against MAC have been evaluated in various mouse models. Ji et al. studied single drugs and three-drug combinations based on CLR in beige mice infected with M. avium (16). They recorded very similar efficacies for CLR-EMB-MIN and CLR-EMB-RIF combinations, comparable to only slight differences in effect sizes observed in our data. In this study and similar experiments by Inderlied et al., AMK alone was most effective in vivo, similar to our in vitro observations (16, 17). Lazard et al. found no added benefit by combining MIN with CLR in mice (18), which we did see in vitro. Klemens et al. also did not see added benefit of combining CLR with AMK, EMB, or RIF in vivo but did observe increased activity of CLR combined with CLO in M. avium treatment (19). Ji et al. also reported significant decreases in CFU counts with combinations containing CLR and CLO (16). Similarly, Lanoix et al. observed increased activity of CLR when combined with CLO, similar to our in vitro study, but they noted no substantial increase in activity with a further addition of RIF to a CLR-CLO combination, which contradicts our findings (20).

The recommended three-drug therapy consisting of a macrolide, RIF, and EMB achieves a treatment success rate of around 65%, as shown in a meta-analysis by Diel et al. (4). Notably, the effectiveness of the standard regimen relies on the macrolide, as RIF and EMB mainly suppress macrolide resistance (21). Recent evidence suggests that EMB is the key drug to prevent macrolide resistance, as two-drug EMB-macrolide regimens are effective in mild MAC-PD and do not lead to macrolide resistance (7, 8). Replacing the less active drugs in this regimen was attempted by Roussel and Igual with a macrolide-MIN-CLO combination in 22 patients, with a success rate of 63%, on par with that of the recommended three-drug regimen (9). Field and Cowie replaced RIF with CLO in a clinical trial with 30 patients initially (22), and a long-term follow-up study of this regimen with 107 patients was published 13 years later by Jarand et al. (10). Overall, they found a 95% culture conversion rate with 49% relapse, thus having a considerably higher culture conversion rate (10). This aligns well with our in vitro data in which these three combinations have similar rankings and are synergistic according to Bliss independence analyses.

Severe MAC-PD cases are treated with a four-drug regimen, with the addition of AMK (2). The added benefit of aminoglycosides with the recommended three-drug regimen was shown in a randomized clinical trial in which the group receiving additional streptomycin had a significantly increased culture conversion rate (71% versus 51%) (23). Similarly, the recent CONVERT trial (ClinicalTrials registration number NCT02344004) compared a guideline-based regimen with and without added AMK liposomal inhaled suspension (ALIS) in patients with refractory MAC-PD and found a significantly improved culture conversion rate in patients treated with ALIS (29% versus 9%), emphasizing the added benefit of aminoglycosides (29). Four-drug regimens were more effective in vitro too, including the addition of AMK, but the differences between the individual regimens were marginal in our experiments.

Replacing drugs in the recommended regimen and exploiting possible synergistic interactions found in vitro might enhance the bactericidal properties of the recommended regimen beyond the action of the macrolide. The PERC trial (EudraCT registration number 2015-003786-28) aims to replace RIF with CLO in a three-drug regimen, which we do find to be effective and synergistic in vitro. Other possibilities, based on our data, would be RIF-AMK-macrolide or RIF-CLO-macrolide combinations; a variant of the latter, with rifabutin instead of RIF, is currently being tested in a phase 3 clinical trial (ClinicalTrials registration number NCT04616924).

Some limitations in this study should be considered. First, all of our assays use fixed antibiotic concentrations, without the addition of fresh antibiotics throughout the experiment. This might have led to an underestimation of drug performance because of drug degradation (25). Dynamic models with the addition of fresh drugs to mimic human pharmacokinetics, such as the hollow-fiber model, eliminate this issue. Second, we performed our assays at 2× MIC. The high drug load might mask more subtle drug interactions observable at 1× MIC, especially in assays with three or four drugs, but might emphasize drug interactions in two-drug assays. Because we did not perform all possible three-drug assays, Bliss independence calculations for four-drug assays could not be interpreted. Therefore, we cannot assess synergism in these assays and can solely compare effect sizes. Also, the TKA focuses on the killing capacity of antimicrobials, whereas the RIF-EMB combination is mostly inactive in vitro and in vivo but prevents the emergence of macrolide resistance in the RIF-EMB-macrolide regimen (6, 24); such roles may not be captured in TKAs. The TKA records killing of planktonic mycobacteria, and the activity of some antibiotics may be significantly enhanced or diminished inside macrophages, where the mycobacteria reside during infection (11, 15). Similarly, we seem to have selected concentrations close to the E_max of RIF and not the E₀, making the estimation of its effect size less secure and possibly overemphasizing the role of RIF in these assays. Lastly, we tested only M. avium ATCC 700898 in these assays. Clinical strains that may have been preexposed to antibiotics may be more resistant to treatment or have deviating growth patterns. The data at hand might therefore overestimate the effects of multidrug combinations.

In conclusion, we show that, in vitro, several two- and three-drug regimens are as effective as the currently recommended regimen for MAC-PD but also that expanding the drug regimen for MAC-PD beyond three drugs leads to diminishing returns. The most promising candidate three-drug regimens for upcoming trials would be RIF-CLO-macrolide or RIF-AMK-macrolide.

MATERIALS AND METHODS

Bacterial strains and antimicrobials.

The M. avium ATCC 700898 (American Type Culture Collection [ATCC], Manassas, VA, USA) reference strain was used for all TKAs. The strain was preserved at −80°C in Trypticase soy broth with 40% glycerol and was freshly cultured before each assay. AMK, RIF, EMB, CLR, MIN, and CLO were obtained from Sigma-Aldrich (Zwijndrecht, The Netherlands). All of these drugs were dissolved in water except for CLR, CLO, and RIF, which were dissolved in acetone (CLR) or dimethyl sulfoxide (DMSO) (CLO and RIF).

Susceptibility testing.

The MIC for each individual antibiotic against M. avium ATCC 700898 was determined according to Clinical and Laboratory Standards Institute (CLSI) recommendations (26) except for CLO, for which we determined MICs in Middlebrook 7H9 (M7H9) broth (BD Bioscience, Erembodegem, Belgium) instead of cation-adjusted Mueller-Hinton broth (CAMHB) (BD Bioscience) because of its poor solubility in CAMHB (12).

TKAs for single antimicrobials.

TKAs for the individual antimicrobials were performed using previously published methods (12). Bottles containing 10 ml of CAMHB (or M7H9 broth for CLO) with 0.05% Tween 80 and 10% oleic acid-albumin-dextrose-catalase (OADC) growth supplement (Becton, Dickinson, and Company) were inoculated with a 1:100 dilution of a freshly prepared broth culture of M. avium ATCC 700898 with 0.5 McFarland standard turbidity. The six antibiotics used in the dose-response kinetic TKAs had concentrations ranging from 0.25× MIC to 32× MIC. On days 0, 1, 2, 3, 4, 5, 7, 10, and 14, the bacterial population of each bottle was quantified using the triplicate spot-plating method on M7H10 (BD Bioscience) agar plates (12). Plates were incubated at 37°C for 7 days, and the CFU per milliliters were counted.

TKAs for antimicrobial combinations.

Combinations of two, three, four, and five of the aforementioned antimicrobials were subjected to TKAs, following the same protocol as described above. For all assays that included CLO alone or in a drug combination, M7H9 broth was used; otherwise, CAMHB was used. All drugs were used at a 2× MIC concentration for a clear antimicrobial response. Table 2 presents a full overview of the drug combinations assessed.

Response surface analysis.

Response surface analysis to assess interactions according to Bliss independence was performed as described previously (27). The area under the curve (AUC) was calculated from the log CFU versus time plots for days 0 to 14 using the trapezoidal rule after averaging the results from the two replicates and normalizing the results to the baseline colony count. The effect was then calculated according to the following formula: effect_x = AUC_{growth control} – AUC_x, where x is any given curve other than the growth control. We assumed that CAMHB and M7H9 broth had the same growth capacity with similar antibiotic efficacies. To assess potential interactions in the TK experiments, we calculated the expected effect for a combination under Bliss independence (E_comb,BI) to be compared with the observed effect (E_comb,obs).

Bliss independence analysis was performed for two combinations as described previously (28), using the following formula: E_comb,BI = E_max,A × [E_B/E_max,A + E_A/E_max,A − (E_A/E_max,A) × (E_B/E_max,A)], where E_B is for the drug with the lower E value, E_A is for the drug with the higher E value, and E_max,A is the greatest maximum effect. The E_max values of the antibiotics were determined by fitting a sigmoidal E_max model to the concentration-response single-drug data using the ordinary least-squares method. For three-drug interactions, we expanded the formula by multiplying it by the effect size of the third drug (E_C), giving E_ABC,BI = (E_A/E_max,A + E_B/E_max,A + E_C/E_max,A) – (E_A × E_B/E_max,A^2) – (E_A × E_C/E_max,A^2) – (E_B × E_C/E_max,A^2) – (E_A × E_B × E_C/E_max,A^3) × E_max,A, where E_A is for the antibiotic with the greatest maximum effect. All reported effect sizes are rounded to three significant digits.

The difference between the observed effect and the expected effect was quantified as a percent difference, relative to the expected effect, as follows: ΔE_BI = (E_comb,obs – E_comb,BI)/E_comb,BI. We defined a ΔE_BI of 0% ± 10% as no interaction; anything less than −10% was defined as antagonistic, and anything more than 10% was defined as synergistic (28). To calculate the percentage of the relative effect that is the effect of adding more drugs to a standing regimen with a greater effect size, the following formula was used: [(E_comb,obs/E_A) – 1] × 100.

Statistics.

Calculations were performed using GraphPad Prism version 5.03 (GraphPad Software Inc., La Jolla, CA, USA) or R version 3.1.2 (R Foundation for Statistical Computing, Vienna, Austria) (https://www.r-project.org). Values are reported with the standard error of the mean (SEM).