Isonicotinic acid N-oxide, from isoniazid biotransformation by Aspergillus niger, as an InhA inhibitor antituberculous agent against multiple and extensively resistant strains supported by in silico docking and ADME prediction

Abstract Biotransformation of isoniazid produced isonicotinic acid (1), isonicotinic acid N-oxide (2), and isonicotinamide (3) which were isolated by column chromatography using silica gel and Sephadex LH 20 and elucidated using various spectroscopies. This is the first report for isolation of 2 . Antituberculosis activity was evaluated against Mycobacterium tuberculosis strains: drug sensitive (DS), multiple drug resistant (MDR) and extensively drug resistant (XDR). 1-3 and isoniazid showed MICs of 63.49, 0.22, 15.98 and 0.88 µM, respectively, against the DS strain. For the MDR strain, 2 and 3 exhibited MICs of 28.06 and > 1000 µM, respectively, while 1 was inactive. Moreover, 2 had an MIC of 56.19 µM against XDR strain, while 1 and 3 were inactive. Docking simulation using enoyl ACP reductase (InhA) revealed favorable protein-ligand interactions. In silico study of pharmacokinetics and hepatotoxicity predicted 1-3 to have good oral bioavailability and 2 to have a lower hepatoxicity probability than isoniazid. Graphical Abstract


Introduction
Isoniazid is one of the most common antituberculosis drugs. It binds to enoyl ACP reductase (InhA) enzyme by forming a covalent adduct with the NAD þ cofactor of the enzyme and thus inhibits the synthesis of lipids leading to prevention of the cell wall synthesis and development (Vilch eze and Jacobs 2007). First, isoniazid was recommended as a monotherapy then, it was used in combination with pyrazinamide, rifampin or both to combat antimicrobial resistance (Baselt 2014). Multiple drug resistant (MDR) M. tuberculosis is resistant to at least rifampin and isoniazid while extensively drug resistant (XDR) strain is resistant to fluoroquinolones and one or more secondline injections (Reller et al. 2000). A continuous search for new drugs is a demand to combat the global problem of the antibiotic resistance. Herein, we conducted fungal bioconversion of isoniazid and screened the isolated metabolites for antituberculosis activity using DS, MDR and XDR strains. In vitro assay for evaluation of InhA inhibition by the isolated compounds was carried out. Moreover, in silico molecular docking to InhA and ADME prediction studies were carried out.

Structure elucidation
Isoniazid transformation produced 1 and 2 by Aspergillus niger NRRL 328 and 3 by A. niger AUMC 4156. The structures were elucidated (Tables S1, S2 and Figures S1-S13). The spectral data of 1 and 3 were in agreement with those for isonicotinic acid and isonicotinamide, respectively (Koczo n et al. 2003, de Souza et al. 2010. Superimposed IR spectra of 3 and an authentic isonicotinamide showed a typical fingerprint. 1 HNMR spectrum of 2 showed two doublets at d H 8.12 (J ¼ 6.6 Hz) and 7.72 (J ¼ 6.6 Hz) for H-2/H-6 and H-3/H-5, respectively. The signals for hydrazide group were absent. DEPTQ-135 spectrum showed two signals at d C 138.2 and 126.7 for C-2/C-6 and C-3/C-5, respectively. Two quaternary carbons at d C 166.3 and 138.1 were assigned to a carboxylic acid and C-4, respectively. The signal of C-2/C-6 and their protons were deshielded compared to those of 1 which proposed N-substitution with an electronegative group. IR spectrum exhibited a band at 1398 cm À1 for N-O functionality. ESI-MS spectrum of 2 showed an [M-H]ion at m/z 138.0, which is 16 Da extra to that of 1. These spectral data were found in agreement with the published data for the synthetic isonicotinic acid N-oxide (Marandi et al. 2012). This is the first report for its isolation and characterization from a natural source.

Determination of minimum inhibition concentration (MIC)
Antituberculosis activity of isoniazid and 1-3, was evaluated using microplate alamar blue assay method (Table S3). First, the compounds were tested against DS strain M. tuberculosis ATCC 25177/H37Ra. MICs of 1, 2, 3 and isoniazid were 63.49, 0.22, 15.98 and 0.88 mM, respectively. Secondly, screening against MDR M. tuberculosis ATCC 35822 was investigated. While 1 was inactive, MICs of 2 and 3 were 28.06 and > 1000 mM, respectively. Lastly, activity against XDR clinical isolate was tested and 2 showed an MIC of 56.19 mM while 1 and 3 were inactive. These results implied that 2 was four times more active than isoniazid against DS strain, more active than 3 against MDR strain and the only active compound against the tested XDR strain which could be related to the N-oxide functionality. Thus, the potential mechanism for 2 as antituberculous agent was investigated further for inhibition of InhA enzyme.

Enzyme assay for inhibition of M. tuberculosis InhA by 2
Enzyme assay revealed that 2 is more potent than isoniazid with IC 50 of 0.20 ± 1.20 mM and 0.70 ± 0.58 mM, respectively (Table S4 and Figure S14). The inhibition was dose dependent in concordance with previously reported data for isoniazid (Manning et al. 2015).

In silico molecular docking
Molecular docking of isoniazid and 1-3 to InhA of M. tuberculosis (PDB ID: 2X23) was carried out using Molegro virtual docker software. Isoniazid and 1-3 showed binding energies of À71.38, À68.92, À73.44 and À67.06 kcal/mol, respectively (Table S5). The amino acids involved in the hydrogen bonding interaction of isoniazid and 1-3 with InhA protein are presented in Table S6 and Figure S15. Hydrogen bonding interaction with Tyr158 was found in the docking poses of isoniazid and 1-3. The distance between the pyridine ring of isoniazid or 1-3 and pyridine ring of NAD þ was around 4 Å, which ensures k-k stacking interaction that further stabilizes the inhibitor's conformation. This was in concordance with the reported data for isoniazid (Mardianingrum et al. 2021).

Prediction of drug likeness and pharmacokinetic ADME
Using SwissADME tool, 1-3 and isoniazid were predicted to be drug-like with high probability of good oral bioavailability according to Lipinski, Veber and Egan rules with zero deviation (Table S7). Only 1 could permeate blood brain barrier ( Figure S16). In addition, PreADMET tool (Table S8) indicated good absorption and distribution of the compounds (Doucet et al. 2017). Percentage of plasma protein binding (%PPB) of 2 and isoniazid were less than 90% (2.96 and 1.61, respectively). This indicated a good probability of well distribution in the body (Lee et al. 2003).

Prediction of the toxicity of compound 2
Hepatotoxicity caused by isoniazid treatment is a manifestation of chronic toxicity (Badrinath and John 2018). Hepatotoxicity of 2 was predicted in comparison to isoniazid using four different prediction webservers ProTox-II (Banerjee et al. 2018), ADVERPred (Ivanov et al. 2018), HepatoPred-EL (Ai et al. 2018) and ADMETlab (Xiong et al. 2021). For 2, the hepatotoxicity probability was lower than that of isoniazid. It was 0.52, 0.43, 0.66 and 0.50, respectively, compared to isoniazid 0.94, 0.95, 0.77 and 0.78, respectively using the prediction servers in the same order mentioned above (Table S9). The LD 50 for 2 predicted using ProToxII webserver was 3123 mg/kg, while for isoniazid was 133 mg/kg indicating a potential safer dose range. Future in vivo studies for full pharmacokinetic parameters of 2 will be conducted.

General (supporting information)
3.2. Microorganisms, culture and fermentation procedures (supporting information) 3.3. Isolation of the compounds (supporting information) 3.4. Antituberculosis activity (supporting information) 3.5. Spectroscopic data (supporting information) 3.6. Molecular docking study (supporting information) 3.7. Prediction of drug likeness and ADME (supporting information) 3.8. Prediction of the toxicity of compound 2 (supporting information)

Conclusion
The results demonstrated that 1-3 possessed an anti-tuberculosis activity against DS strain. More interestingly, 2 exhibited a profound activity against MDR and XDR strains. Docking to InhA indicated that 2 showed energy score of interaction nearly the same as isoniazid and better than 1 and 3. All compounds have a potential of good pharmacological parameters based on the drug-likeness. PreADMET evaluation predicted good permeability and distribution.