Rifamycin, macrolide, and ethambutol-free combinations to treat Mycobacterium avium complex lung disease

Abstract

Objective.

Poor sustained sputum culture conversion with guidelines-based therapy Mycobacterium avium complex lung disease (MAC-LD) highlight the need for new and improved multidrug combinations.

Design.

We tested eight drugs by performing MIC experiments with the ATCC700898 strain and 49 clinical isolates, followed by single and multidrug combination studies.

Results.

MIC_90s of rifabutin, rifampin, azithromycin, and ethambutol were 4, 16, 256, and 128 mg/L, respectively. Among other drugs, MIC₉₀ (mg/L) in lowest to highest rank order were for bedaquiline (0.25), tedizolid (16), tebipenem (64), and minocycline (128). As monotherapy none of the drugs had better MAC kill compared to rifampin. While minocycline alone killed 1.76±0.33 log₁₀ CFU/mL MAC, replacement of rifabutin in the standard therapy resulted in 6.27±0.28 log₁₀ CFU/mL MAC kill, showing minocycline’s synergistic effect with azithromycin and ethambutol. An example of a macrolide-rifamycin-ethambutol free oral regimen, the azithromycin-minocycline-tedizolid combination killed 4.35±0.42 log₁₀ CFU/mL MAC. Tebipenem-tedizolid-minocycline combination killed 4.82±0.42 log₁₀ CFU/mL, however, addition of bedaquiline resulted in a lower kill of 2.71±0.22 log₁₀ CFU/mL, suggesting a likely antagonism.

Conclusion.

We tested eight different drugs to identify several rifamycin, macrolide, and ethambutol-free oral multidrug combinations. The minocycline-based regimens warrant further preclinical and clinical testing.

INTRODUCTION

Among the pathogenic nontuberculous mycobacteria (NTM), ~80% of the cases of lung disease (LD) could be attributed to the slow-growing Mycobacterium avium-complex (MAC) ¹. The guidelines-based standard-of-care regime consists of a macrolide (clarithromycin (CLA) or azithromycin (AZI)), ethambutol (EMB), and a rifamycin (rifabutin (RFB) or rifampin (RIF)) ². However, the sustained sputum culture conversion (SSCC) after six months of therapy with the standard-of-care regimen is achieved in about 43-53% of patients, leaving a large proportion of patients suffering from recalcitrant disease ^{3, 4}. In other words, MAC-LD is among the most difficult to treat pulmonary infections ⁵.

Based on their own experience and patients’ response, clinicians often switch macrolides (AZI or CLA) and rifamycin (RIF or RFB). Previously, we showed that replacing one rifamycin with another may not be beneficial to prevent acquired drug resistance and better therapeutic outcomes ⁶. Using the dynamic hollow fiber model of MAC (HFS-MAC), it has also been shown that quinolones, tetracyclines, β-lactams, oxazolidinones, and several other drugs could be a better replacement for the standard-of-care drugs ^7-17. Here, we compared the kill effect of standard-of-care drugs with tedizolid (TZD), tebipenem (TBP) (without the combination of a β-lactamase inhibitor), bedaquiline (BDQ), and minocycline (MINO) alone and as three- and four-drug combinations to identify rifamycin, EMB, and macrolide-free oral drug combinations for the treatment of MAC-LD. In these exploratory preclinical studies, all experiments were performed at static drug concentrations. We also performed minimum inhibitory concentration (MIC) experiments with a collection of 49 species-identified clinical isolates of MAC to generate MIC distributions for each study drug.

MATERIALS AND METHODS

Bacteria, culture medium, and drugs

All experiments were performed with the MAC reference strain Mycobacterium avium ATCC#700898, purchased from the American Type Culture Collection (Manassas, VA, USA). We used a library of 49 MAC clinical isolates (MAC n=17, MAC subsp. Chimaera n=6, and MAC subsp. Intracellulare n=26) to generate the MIC distribution for each study drug. Bacterial stocks were prepared as published previously ^{7-14, 18, 19}. Prior to each experiment, stock cultures were grown to logarithmic phase in Middlebrook 7H9 broth supplemented with 10% oleic acid, albumin, dextrose, and catalase (OADC) for four days at 37°C (herein termed “7H9 broth”). For inoculum preparation, the turbidity of the logarithmic phase growth cultures was adjusted to McF Standard 0.5, followed by back dilution in 7H9 broth to achieve a bacterial burden of ~4.5 log₁₀ CFU/mL. Azithromycin was purchased from the University of Texas Health Science Center at Tyler campus pharmacy. Ethambutol, rifampin, rifabutin, minocycline, tedizolid, tebipenem, and bedaquiline were purchased from BOC Sciences (NY, USA).

Minimum inhibitory concentration

The broth micro-dilution method was used to determine the MIC of each of the eight study drugs against the reference strain and 49 clinical isolates, as described by the Clinical Laboratory Standards Institute, EUCAST, and our recent publications ^20-22. The concentration range of each drug is listed in Table 1. Experiments were performed with two replicates per drug concentration. MIC was defined as the lowest drug concentration that completely inhibited visible microbial growth in the wells and was read by three independent observers. When a trailing effect was present, drug concentration associated with 80% growth inhibition, determined by visual inspection of pellet size compared to the nontreated controls, was determined as the MIC ^{23, 24}.

Table 1. Drug and concentrations used in the experiments.

The concentration range for each drugs (in alphabetical order) used in the MIC experiments and resulting MIC of the reference strain were as listed below. Also, listed are the drug concentrations used in the monotherapy and multi-drug combination studies, selected from previously published ^{6, 7, 13, 14, 18, 22, 25} and unpublished studies.

Drug	Concentration range for MIC experiments (mg/L)	MIC (mg/L) ATCC#700898	EC80 concentration (mg/L) used in the experiments
Azithromycin	0.03 to 16	8	15.14
Bedaquiline	16 to 256	0.015	0.467
Ethambutol	0.25 to 16	8	7.16
Minocycline	1 to 64	1	5.23
Rifabutin	4 to 128	0.3	0.123
Rifampin	0.25 to 128	0.3	0.206
Tebipenem	0.0075 to 0.25	0.5	0.525
Tedizolid	1 to 16	8	1.48

Static concentration monotherapy and multidrug combination studies.

We performed a series of experiments with all eight drugs using EC₈₀ or the drug concentration mediating 80% of maximal effect or optimal concentration, selected from previous experiments ^{6, 7, 13, 14, 18, 22, 25} and unpublished data for tebipenem (without combination of β-lactamase inhibitor) and bedaquiline. The drug concentrations (EC₈₀) are listed in Table 1. Inoculum preparation was as described above. In the monotherapy experiments, the total volume was 5 mL in 15 mL screw-cap tubes, where 4.5 mL of the inoculum was added to each tube, followed by the addition of 0.5 mL of the drug at 10X concentration. For the multidrug combination experiments, the procedure remained the same, except the inoculum volume was adjusted by adding 7H9 broth to keep the initial bacterial burden the same across different drug-treated arms. After 7 days of incubation at 37°C, 1 mL culture from each of the 15 mL tubes was collected in sterile 1.8 mL centrifuge tubes. Cultures were washed twice in normal saline to remove the carry-over drug, resuspended in 1 ml normal saline, and 10-fold serially diluted before inoculation on Middlebrook 7H10 agar supplemented with 10% OADC (herein termed “agar”). Colony-forming units (CFU) were recorded after 10 days of incubation at 37°C and used in subsequent data analysis.

Data analysis

MIC data were recorded using MS-EXCEL and used to calculate the MIC₅₀ and MIC₉₀ among the 49 clinical isolates. GraphPad Prism (v10.4.2) was used for graphing of the data and to perform ordinary one-way analysis of variance to compare the MAC kill of drugs as monotherapy and 3- and 4-drug combinations. Linear regression was used to calculate the kill rate with each drug combination.

RESULTS

MIC of the ATCC strain against the eight study drugs is listed in Table 1. MIC distribution of the 49 clinical isolates (MAC n=17, MAC subsp. Chimaera n=6, and MAC subsp. Intracellulare n=26) is shown in Table 2. MIC₅₀ and MIC₉₀ of each of the eight study drugs are also listed in Table 2. The results of the monotherapy studies, showing MAC kill with each drug compared to the nontreated controls on study day 7, in increasing order, are shown in Figure 1. The lowest MAC kill was observed with tebipenem as 1.07 ± 0.37 log₁₀ CFU/mL, whereas rifabutin killed 2.50 ± 0.52 log₁₀ CFU/mL of MAC. Next, to compare the kill effect as a single drug, we used rifampin as the comparator in the one-way analysis of variance; the results are shown in Supplementary Table 1. Given the shows overlapping 95% Confidence Intervals (95% CI) that cross zero, none of the drugs had better MAC kill compared to rifampin.

Table 2.

Minimum inhibitory concentration distribution of the study drugs among the 49 MAC clinical isolates.

Organism (n=49)	RFB	RIF	AZM	ETB	BDQ	MINO	TBP	TDZ
Mycobacterium avium	0.03	0.5	64	>64	0.0075	0.25	2	2
Mycobacterium avium	4	16	256	>64	0.015	128	2	8
Mycobacterium avium	2	16	32	8	0.0075	64	4	8
Mycobacterium avium	0.25	8	64	8	0.015	16	4	4
Mycobacterium avium	0.06	0.25	32	>64	ND	8	4	1
Mycobacterium avium	0.03	4	>256	8	0.0075	16	1	16
Mycobacterium avium	0.125	16	32	64	0.015	16	4	16
Mycobacterium avium	0.03	4	>256	64	0.06	4	2	4
Mycobacterium avium	0.25	8	16	4	0.015	16	4	4
Mycobacterium avium	0.25	8	128	8	0.0075	64	4	8
Mycobacterium avium	0.125	4	64	4	0.015	16	2	8
Mycobacterium avium	0.5	>16	>256	8	0.015	32	16	8
Mycobacterium avium	0.25	4	64	32	0.06	16	64	16
Mycobacterium avium	0.125	4	32	32	0.015	16	16	8
Mycobacterium avium	0.5	>16	128	32	0.03	64	16	16
Mycobacterium avium	0.03	2	32	8	0.015	4	16	8
Mycobacterium avium	0.03	4	32	8	0.0075	32	2	4
Mycobacterium avium subsp. chimaera	0.25	8	64	8	0.015	64	8	8
Mycobacterium avium subsp. chimaera	0.25	16	256	8	0.03	64	4	16
Mycobacterium avium subsp. chimaera	1	>16	128	4	0.015	64	4	4
Mycobacterium avium subsp. chimaera	0.5	16	128	4	0.03	128	16	8
Mycobacterium avium subsp. chimaera	0.03	1	32	8	0.015	1	2	8
Mycobacterium avium subsp. chimaera	0.25	>16	128	16	0.03	32	16	16
Mycobacterium avium subsp. intracellulare	0.125	8	64	4	0.015	32	2	8
Mycobacterium avium subsp. intracellulare	0.25	>16	64	4	0.0075	64	4	8
Mycobacterium avium subsp. intracellulare	0.25	16	64	4	0.0075	64	4	8
Mycobacterium avium subsp. intracellulare	0.25	>16	64	4	0.0075	64	16	8
Mycobacterium avium subsp. intracellulare	0.25	16	32	4	0.015	128	4	4
Mycobacterium avium subsp. intracellulare	0.125	16	128	4	0.0075	32	2	4
Mycobacterium avium subsp. intracellulare	0.25	16	128	4	0.015	64	8	16
Mycobacterium avium subsp. intracellulare	16	>16	>256	128	0.25	>128	64	8
Mycobacterium avium subsp. intracellulare	16	>16	>256	128	0.25	>128	64	8
Mycobacterium avium subsp. intracellulare	16	>16	>256	128	0.25	>128	64	8
Mycobacterium avium subsp. intracellulare	16	>16	>256	128	0.25	>128	64	8
Mycobacterium avium subsp. intracellulare	16	>16	>256	128	0.25	>128	>64	8
Mycobacterium avium subsp. intracellulare	0.125	2	32	8	0.015	64	4	4
Mycobacterium avium subsp. intracellulare	0.06	2	128	16	0.0075	32	2	8
Mycobacterium avium subsp. intracellulare	0.125	4	64	4	0.0075	32	2	2
Mycobacterium avium subsp. intracellulare	0.25	16	64	4	0.015	128	4	8
Mycobacterium avium subsp. intracellulare	0.03	2	>256	8	0.0075	64	4	16
Mycobacterium avium subsp. intracellulare	0.06	8	128	8	0.0075	32	1	8
Mycobacterium avium subsp. intracellulare	0.03	4	64	16	0.0075	4	1	8
Mycobacterium avium subsp. intracellulare	0.125	8	32	4	0.015	64	8	8
Mycobacterium avium subsp. intracellulare	0.25	8	128	8	0.015	64	64	8
Mycobacterium avium subsp. intracellulare	0.25	4	64	4	0.03	64	64	4
Mycobacterium avium subsp. intracellulare	0.25	4	128	4	0.015	64	16	8
Mycobacterium avium subsp. intracellulare	0.5	16	128	4	0.015	64	16	16
Mycobacterium avium subsp. intracellulare	0.25	8	64	>128	0.015	64	1	4
Mycobacterium avium subsp. intracellulare	0.25	16	128	16	0.015	64	16	8
MIC Range (mg/L)	0.03 to 16	0.25 to >16	16 to >256	4 to >128	0.0075 to 0.03	0.25 to >128	1 to >64	1 to 16
MIC50	0.25	8	64	8	0.015	64	4	8
MIC90	4	16	256	128	0.25	128	64	16

(Rifampin, RIF; Rifabutin, RFB; Azithromycin, AZI; Tedizolid, TZD; Tebipenem, TBP; Bedaquiline, BDQ; Minocycline, MINO.

Figure 1. MAC kill compared to nontreated control after coincubation with the EC₈₀ concentration of each drug.

The number above each bar is the mean of the three replicates. All eight drugs killed MAC. The drugs are shown in the figure with increasing kill from left to right. The lowest MAC kill was observed with tebipenem (without any beta-lactamase inhibitor). The extent of MAC killing with rifampin and rifabutin static concentrations was not statistically different, confirming previous findings using a dynamic hollow fiber model of MAC (HFS-MAC) ⁶. Azithromycin, bedaquiline, tedizolid, and minocycline MAC kill were also not statistically different from rifampin or rifabutin.

The results for RIF, TZD, MINO, TBP, and BDQ as monotherapy and six three-drug EC₈₀ combinations with rifampin are shown in Supplementary Figure 1A-F. Highest MAC kill was observed with RIF-MINO-TBP combination (5.14 ± 0.08 log₁₀ CFU/mL) followed by RIF-TZD-TBP (3.94 ± 0.03 log₁₀ CFU/mL), RIF-MINO-BDQ (3.77 ± 0.15 log₁₀ CFU/mL), RIF-MINO-TZD (3.75 ± 0.01 log₁₀ CFU/mL), RIF-TBP-BDQ (3.26 ± 0.05 log₁₀ CFU/mL), and RIF-TZD-BDQ (3.20 ± 0.02 log₁₀ CFU/mL).

Figure 2A-C shows the results of two new rifamycin-free oral regimens for MAC. The kill of MAC with RFB was 2.50±0.52 log₁₀ CFU/mL, with AZI was 2.10±0.32 log₁₀ CFU/mL, 1.62±0.12 log₁₀ CFU/mL with EMB, and 3.29±0.28 log₁₀ CFU/mL for the 3-drug combination of RFB-AZI-EMB (Figure 2A). In comparison, as shown in Figure 2B, while the MAC kill with MINO alone was only 1.76±0.33 log₁₀ CFU/mL, replacement of RFB with MINO resulted in an impressive 6.27±0.28 log₁₀ CFU/mL MAC kill with AZI-EMB-MINO combination, showing MINO’s synergistic effect. We took another step and replaced both RFB and EMB (Figure 2C) and observed AZI-MINO-TZD killed 4.35±0.42 log₁₀ CFU/mL MAC kill. We also compared the kill slope of these regimens. The kill slope with AZI-EMB-MINO combination was −0.87 (95% CI: −0.89 to −0.85) log₁₀ CFU/mL/day, significantly different from the AZI-TZD-MINO −0.60 (95% CI: −0.67 to −0.53) log₁₀ CFU/mL/day and for standard-of-care RFB-AZI-EMB combination was −0.44 (95% CI: −0.47 to −0.42) log₁₀ CFU/mL/day. Thus, replacement of RFB and EMB resulted in two MINO-based regimens with improved MAC kill.

Figure 2. Minocycline-based oral regimens and MAC kill comparison with the rifamycin-macrolide-ethambutol standard-of-care regimen.

(A) Rifabutin showed the best kill among the three standard-of-care drugs, and the combination killed MAC more than each drug alone. (B). Minocycline monotherapy MC kill was lower (statistically not significant) compared to rifabutin. However, the replacement of rifabutin with minocycline resulted in a two-fold higher MAC kill, showing synergy with azithromycin and ethambutol. (C). Both Rifabutin and ethambutol were replaced with minocycline and tedizolid. This new combination was also more effective in killing MAC compared to the standard-of-care regimen. (Rifampin, RIF; Rifabutin, RFB; Azithromycin, AZI; Tedizolid, TZD; Tebipenem, TBP; Bedaquiline, BDQ; Minocycline, MINO.)

Next, we examined the bacterial kill with rifamycin, EMB, and macrolide-free oral drug combinations. Supplementary Figure 2A-D shows the MAC kill with each drug (TZD, MINO, TBP, and BDQ) as monotherapy and three-drug EC₈₀ combinations. The TZD-MINO-TBP combination showed 4.43 ± 0.63 log₁₀ CFU/mL MAC kill compared to the 3.26 ± 0.45 log₁₀ CFU/mL with TZD-TBP-BDQ, 3.31 ± 0.03 log₁₀ CFU/mL with MINO-TBP-BDQ, and 3.19 ± 0.22 log₁₀ CFU/mL with TZD-MINO-BDQ combination.

However, we observed that the addition of bedaquiline, to examine the possibility of four-drug combinations, resulted in lower bacterial kill. Initially, there was 4.35±0.42 log₁₀ CFU/mL MAC kill observed with AZI-MINO-TZD combination (Figure 3A), but bedaquiline combination killed only 3.09±0.21 log₁₀ CFU/mL MAC (Figure 3B). Similarly, the 4.82±0.42 log₁₀ CFU/mL kill with TBP-TZD-MINO combination (Figure 4C) reduced to 2.71±0.22 log₁₀ CFU/mL when BDQ was added (Figure 4D). The likely antagonistic BDQ effect was also observed in the kill slopes, which were −0.60 (95% CI: −0.67 to −0.53) log₁₀ CFU/mL/day for AZI-MINO-TZD versus −0.42 (95% CI: −0.44 to −0.39) log₁₀ CFU/mL/day with AZI-MINO-TZD-BDQ combination. The kill slope of the TBP-TZD-MINO combination changed from −0.66 (95% CI: −0.73 to −0.60) to −0.36 (95% CI: −0.39 to −0.33) log₁₀ CFU/mL/day when BDQ was added to this combination. This observation of drug combination-dependent synergy and antagonism has been shown previously in studies describing two- and three-month short-course pan-TB regimens ²⁶.

Figure 3. Bedaquiline and likely antagonism.

Bedaquiline was as effective as other study drugs, however, the Panel (A) and (B) show that synergy and antagonism are not only concentration-dependent, but also which drugs are used in combination. These results also highlight the importance of in vitro drug testing for synergy and antagonism, otherwise expensive and time-consuming in in vivo models. The number above the bars represents the mean kill compared to nontreated controls on study day 7. (Rifampin, RIF; Rifabutin, RFB; Azithromycin, AZI; Tedizolid, TZD; Tebipenem, TBP; Bedaquiline, BDQ; Minocycline, MINO.)

Overall, as summarized in Supplementary Table 2 and ranked in Supplementary Figure 3, replacement of rifamycin, EMB, and macrolide in the standard-of-care regimens resulted in several oral drug combinations showing improved MAC kill and faster kill slopes.

DISCUSSION

Treatment of MAC-LD is complex, and treatment failure as well as relapse are common, even when following the multi-society guidelines and recommendations for drug and regimen selection ^{2, 27}. The adherence to the guideline-based treatment approach by the clinicians is low ²⁸, and switching of drugs within the class and between antibiotic classes is not uncommon. Here we present data describing several oral drug combinations replacing rifamycin, macrolide, and ethambutol. These new multi-drug combinations need further examination in the dynamic preclinical models, such as the HFS-MAC,^{6, 9, 13-16, 18, 25, 29-32} following the recent announcement from the USA Food and Drug Administration.³³

In the present study, we tested eight drugs for activity against MAC. First, we generated MIC distributions using a collection of 49 clinical isolates. Among the standard-of-care drugs, rifabutin MIC₉₀ (4 mg/L) was 4-fold lower compared to rifampin (16 mg/L), and azithromycin and ethambutol MIC₉₀ were very high, at 256 mg/L and 128 mg/L, respectively. Among other drugs, MIC₉₀ in lowest to highest rank order were for bedaquiline (0.25 mg/L), tedizolid (16 mg/L), tebipenem (without a beta-lactamase inhibitor; 64 mg/L), and minocycline (128 mg/L). Thus, if MIC is considered the sole arbitrator of a drug’s efficacy, rifabutin, bedaquiline, and tedizolid would be the drugs to study further to develop new combination regimens. However, subsequent experiments tell a different story.

Second, in the static EC₈₀ concentration time-kill experiments, the maximum MAC kill with rifabutin and rifampin was not statistically different, confirming the results of a previous HFS-MAC study, which found that switching one rifamycin with another may not provide added clinical benefits.⁶ The MAC kill with other study drugs was also not different from rifabutin or rifampin. Third, while rifampin monotherapy killed only 2.02 ±0.12 log₁₀ CFU/mL MAC, different combinations, excluding macrolide and ethambutol, resulted in additivity and higher kill, ranging between 3.19 ± 0.02 log₁₀ CFU/mL with rifampin-tedizolid-bedaquiline combination to 5.14 ± 0.08 log₁₀ CFU/mL MAC kill with the rifampin-minocycline-tedizolid combination. These data suggest that rifampin needs other companion drugs instead of macrolides and ethambutol to improve the treatment of MAC-LD.

Fourth, we identified three rifamycin-macrolide-ethambutol free oral three-drug combinations consisting of tedizolid, minocycline, bedaquiline, and tebipenem showing a similar extent of MAC kill in static concentration experiment as with the standard-of-care regimen that killed 3.29 ± 0.28 log₁₀ CFU/mL MAC. In addition, the tedizolid-minocycline-tebipenem combination showed a 1.53 ± 0.14 log₁₀ CFU/mL higher MAC kill compared to the standard-of-care regimen.

Fifth, we show that bedaquiline needs careful consideration when added to multidrug regimes for MAC. For example, the MAC kill with the azithromycin-tedizolid-minocycline combination was 4.35 ± 0.28 log₁₀ CFU/mL. However, the addition of bedaquiline to create a four-drug combination resulted in 1.26 ± 0.21 log₁₀ CFU/mL lower kill. Similar bedaquiline effect was was also noted when added to the tedizolid-minocycline-tebipenem combination. While our studies were not designed to systematically analyze the synergy and antagonism between drugs, we propose further experiments with bedaquiline are warranted if it has to be advanced for the treatment of MAC-LD.

Sixth, the MIC of the drugs does not always correlate with the clinical response. Therefore, to show as a proof-of-concept that MIC alone should not be the decision-making tool for use of a drug for MAC-LD, we compared the MAC kill with the standard-of-care RFB-AZI-EMB combination by replacing rifabutin with minocycline (MIC₉₀ = 128 mg/L). Interestingly, this rifamycin-free MINO-based regimen showed an impressive 2.99 ± 0.01 log₁₀ CFU/mL higher kill. Another rifamycin and EMB-free combination of AZI-TZD-MINO also showed 1.06 ± 0.14 log₁₀ CFU/mL higher kill compared to the standard regimen. Thus, these MINO-based regimens, despite high MICs, warrant investigation for clinical use.

Our study has a limitation in that all experiments were performed at static concentrations, which could potentially overestimate a drug’s efficacy. However, previously we tested several of the study drugs at dynamic concentration using human-like drug pharmacokinetics and showed a potential role in the treatment of MAC-LD.^{7, 13, 14, 18, 25} Thus, for screening studies, the static concentration experiments, including MIC testing, should be interpreted carefully.

To conclude, we tested eight different drugs comparing the activity against MAC with rifamycin, azithromycin, and ethambutol, identified several rifamycin-free oral combinations. We propose to test the new minocycline-based regimens in dynamic preclinical models and explore therapeutic potential in clinical settings.

Supplementary Material

ADDITIONAL INFORMATION: The Supplementary Figure and Supplementary Table are available online under “Supplementary Data”.

DATA AVAILABILITY STATEMENT.

Upon a reasonable request, the raw data for the results presented in the manuscript is available with the corresponding author.