Metallophiles as sources of antimycobacterial agents

Date
2023-12
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH ABSTRACT: Mycobacterial pathogens present a significant complication to disease control globally due to their resistance to numerous antibiotics. The rise in resistant strains to current chemotherapeutic treatments has prompted the search, development and implementation of new strategies to address this challenge. Harnessing the bioactivity of natural products found in the vast chemical space by using multi-disciplinary approaches has emerged as a promising way to discover new Tuberculosis drugs. This study aimed to evaluate the potential antimycobacterial activity of secondary metabolites from bacteria, fungi, and plants in-vitro and in-silico. In addition to mining for Mycobacterium tuberculosis targets, this study went further to explore other druggable targets associated with cancer in order to fully explain exhaustive in-silico bioactivity profiles. The following experiments were conducted to satisfy the aims: (i) bacteria from gold mine tailings were isolated and identified using 16S rRNA sequencing. The crude extracts from the bacteria were screened for potential activity against Mycobacterium tuberculosis (M. tb) H37Rv, Mycobacterium smegmatis MC2155, and Mycobacterium aurum A+ in-vitro. The active extracts were tentatively identified using HPLC-qTOF, GNPS, and Ms Dial. The identified compounds were virtually screened against Mycobacterium Pks13 and PknG. The natural compound that displayed high affinity was subjected to modification through multiple synthetic routes using reaction-driven enumeration. (ii) A total of 15 fungi compounds from fungi isolated from gold mine tailings were evaluated for their potential activity against M. tb PknA, PknB, PknD, and PknE proteins using extra precision molecular docking, molecular dynamics simulations, and molecular mechanics generalized born surface area (MM-GBSA) binding free energy calculation. (iii) Genomic DNA of one bacterial colony that showed activity against M. tb, was isolated and sequenced by Illumina’s NextSeq platform. The genes responsible for producing metabolites that may have antimycobacterial activity were determined using antiSMASH and PARTIC. (iv) Predictive machine learning-based quantitative structure-activity relationship models were developed with a pIC50 as the dependable variable, while features extracted from compounds found to be active against InhA were the independent variable. Another approach in developing a multitargeted SMILES-based Long Short-term Memory (LSTM) based on pIC50, and small, skewed datasets was attempted. (v) Medicinal plant species indigenous to South Africa namely Schotia brachypetala, Rauvolfia caffra, Schinus molle, Ziziphus mucronate, and Senna petersiana were evaluated for their potential antimycobacterial activity against Mycobacterium smegmatis MC2155, Mycobacterium aurum A+, and M. tb H37Rv. Although the study was specific to mycobacteria, further exploration into cytotoxic activity against MDA-MB 231 triple-negative breast cancer cells was also attempted to see if druggable targets could also be identified in eukaryotic cells as a test of the utility and robustness of the method. The constituents of the extracts possessing antimycobacterial activity were virtually screened using a rigorous Virtual Screening Workflow. The compounds exhibiting good binding, and ADME properties were returned and subjected to molecular dynamics simulations. MM-GBSA calculations were performed to evaluate the affinity of the selected compound/s to pantothenate kinase (PanK). Crude extracts from three bacterial isolates, namely Bacillus subtilis and Bacillus licheniformis, exhibited activity against M. tb H37Rv, Mycobacterium smegmatis MC2155, and Mycobacterium aurum A+. The classes of secondary metabolites identified in this study are known to possess antibacterial activity. Virtual screening of the secondary metabolites against PknG and Pks13, returned cyclo-(L-Pro-4-OH-L-Leu) and vazabitide A with pre-MD MM-GBSA values of -42.81 kcal/mol and -47.62 kcal/mol, respectively. The modification of vazabitide A yielded a compound with a higher affinity of -85.80 kcal/mol to the Pks13, binding as revealed by the post-MD MM-GBSA. SAMN36381076 was assigned to be B. licheniformis whole genome analysis. The genome length of B. licheniformis SAMN36381076 was estimated to be 4.213156 Mb, with a G+C content of 46.08%, comprising 58 contigs and exhibiting an N50 length of 165,033 bp. The biosynthetic gene clusters identified included fengycin, butirosin A, butirosin B, schizokinen, pulcherriminic acid, bacillibactin, bacillibactin E, bacillibactin F, lichenicidin VK21 A1, Lichenicidin VK21 A2, and thermoactinoamide A. These gene clusters are known for producing secondary metabolites with antimicrobial activity. Furthermore, The B. licheniformis SAMN36381076 possesses genes that encode for six diverse antibiotic resistance mechanisms, with efflux pumps as the predominant mechanism of resistance. Metabolic analysis of B. licheniformis SAMN36381076 showed that the presence of genes involved carbohydrate degradation and assembly processes, oxyanion biogeochemical cycling, and nitrogen cycling. In-silico evaluation of fungi compounds against ser/thr kinases showed the lowest ΔGBind values of aurovertin D against PknA (-50.9 kcal/mol), aurovertin D against PknB (-50.7 kcal/mol), verticillin A against PknD (-36.8 kcal/mol), and roquefortine C against PknE (-53.4 kcal/mol). Molecular dynamics simulation showed that the PknD-verticillin A exhibited the highest stability. Furthermore, the post-MD MMGBSA ΔGBind showed that verticillin A has a high affinity for PknD -53.67 kcal/mol. The results indicated that verticillin A is a potential hit compound that can befurther optimized and modified to develop a potent antimycobacterial inhibitor. The classical machine learning models developed from logistic regression and multi-layer perceptron were identified to have significant performance metrics on the InhA dataset. The results (- R2) from the multitarget Long Short Term Memory (LSTM) model indicated the need for hyperparameter tuning. However, further external validation of the two classification models is needed. In this study, the bioactive compounds present in R. caffra and S. molle showed average activity against M. tb H37Rv (MIC 0.25-0.125 mg/mL). Norajmaline with a docking score of -7.47 kcal/mol, and pre-MM-GBSA of -37.64 kcal/mol was returned from the rigorous virtual screening. Molecular dynamics simulation and post-MD MM-GBSA revealed the stable binding of norajmaline to PanK (-58. 73 kcal/mol). Results from the Flow cytometry analysis of treated MDA-MB 231 cells revealed that the dichloromethane extracts from S. petersiana, Z. mucronate, and ethyl acetate extracts from R. caffra and S. molle induced higher levels of apoptosis than the control cisplatin. In conclusion, this study serves as a starting point for the in-silico discovery of potent antimycobacterial compounds from metallophiles (fungi and bacteria) and plants. Virtual screening accelerates the drug discovery process by identifying compounds that may possess activity, thus they can be modified to increase potency. The incorporation of a large dataset of compounds comprising different biological conditions but with the same endpoint can be used to develop robust models with exceptional generalization capabilities.
AFRIKAANSE OPSOMMING: Mikobakteriese patogene bied 'n beduidende komplikasie tot siektebeheer wêreldwyd as gevolg van hul weerstand teen talle antibiotika. Die toename in weerstandige stamme teen huidige chemoterapeutiese behandelings het aanleiding gegee tot die soektog, ontwikkeling en implementering van nuwe strategieë om hierdie uitdaging aan te spreek. Die benutting van die bioaktiwiteit van natuurlike produkte in die chemiese ruimte deur die gebruik van multidissiplinêre benadering het na vore gekom as 'n belowende manier om nuwe tuberkulose middels te ontdek. Die doel van hierdie studie was om die potensiële antimikobakteriële aktiwiteit van sekondêre metaboliete van bakterieë, swamme en plante te evalueer. Hierdie studie het ook die antikankeraktiwiteit van die plante geëvalueer. Die volgende eksperimente is uitgevoer om die doelwitte te vervul: (i) isolasie en identifikasie van bakterieë uit goudmyn uitskot deur gebruik te maak van 16S rRNA volgorde bepaling. Die ru-ekstrakte van die bakterieë is gesif vir potensiële aktiwiteit teen Mycobacterium tuberculosis (M. tb) H37Rv, Mycobacterium smegmatis MC 2 155 en Mycobacterium aurum A+ in vitro . Die aktiewe ekstrakte is voorlopig geïdentifiseer deur gebruik te maak van HPLC- qTOF , GNPS en Ms Dial. Die geïdentifiseerde verbindings is virtueel gesif teen Mycobacterium Pks13 en PknG. Die natuurlike verbinding wat hoë affiniteit getoon het, is onderworpe aan modifikasie deur die gebruik van verskeie sintetiese roetes deur reaksie-gedrewe konfigurasies. (ii) 'n Totaal van 15 swam verbindings van swamme van goudmyn uitskot is geëvalueer vir hul potensiële aktiwiteit teen M. tb PknA-, PknB-, PknD- en PknE-proteïene deur ekstra presiesheid molekulêre koppeling, molekulêre dinamika-simulasies en molekulêre meganika veralgemeende gebore oppervlak area (MM-GBSA) binding vry energie berekeninge. (iii) Genomiese DNA van een bakterie wat aktiwiteit teen M. tb getoon het , is geïsoleer en georden deur Illumina se NextSeq platform. Die gene wat verantwoordelik is vir die vervaardiging van metaboliete wat antimikobakteriële aktiwiteit kan hê, is bepaal deur gebruik te maak van antiSMASH en PARTIC. (iv) Voorspellende masjienleer-gebaseerde kwantitatiewe struktuur-aktiwiteit verwantskap modelle is ontwikkel met 'n pIC 50 as die betroubare veranderlike, terwyl kenmerke uittreksels uit verbindings wat aktief teen InhA was, is as die onafhanklike veranderlike gebruik. Nog 'n benadering in die ontwikkeling van 'n multi-geteikende SMILES-gebaseerde lang korttermyn geheue (LSTM) gebaseer op pIC 50, en klein, skewe datastelle was probeer. (v) Medisinale plantspesies inheems aan Suid-Afrika, naamlik Schotia brachypetala , Rauvolfia caffra , Schinus molle , Ziziphus mucronate, en Senna petersiana is geëvalueer vir hul potensiële antimikobakteriese aktiwiteit teen Mycobacterium smegmatis MC 2 155, Mycobacterium aurum A+ en Mycobacterium tuberculosis H37Rv, en sitotoksiese aktiwiteit teen MDA-MB 231 trippel-negatiewe bors kankerselle . Die bestanddele van die ekstrakte wat antimikobakteriese aktiwiteit het, is virtueel gesif deur gebruik te maak van 'n streng virtuele siftings werkvloei. Die verbindings wat goeie binding en ADME-eienskappe getoon het, is teruggestuur en aan molekulêre dinamika-simulasies onderwerp. MM-GBSA berekeninge is uitgevoer om die affiniteit van die geselekteerde verbindings vir pantotenaat kinase ( PanK) te evalueer. Ru-ekstrakte van drie bakteriese isolate, naamlik Bacillus subtilis en Bacillus licheniformis, het aktiwiteit getoon teen M. tb H37Rv, Mycobacterium smegmatis MC 2 155 en Mycobacterium aurum A+. Dit is bekend dat die sekondêre metaboliete klasse wat in hierdie studie geïdentifiseer is, antibakteriese aktiwiteit het. Virtuele sifting van die verbindings teen PknG en Pks13, siklo-(L-Pro-4-OH-L-Leu) en vazabitied A met pre-MD MM-GBSA het teruggekeer met waardes van -42.81 kcal/mol en -47.62 kcal/mol, onderskeidelik. Die modifikasie van Vazabitide A tot die Pks13 binding het gelei tot 'n verbinding met 'n hoër affiniteit van -85.80 kcal/mol soos geopenbaar deur die post-MD MM-GBSA. B. licheniformis SAMN36381076 is geïdentifiseer deur heelgenoom analise. Die genoom lengte van B. licheniformis SAMN36381076 is geskat op 4,213156 Mb, met 'n G+C-inhoud van 46,08%, wat 58 kontigs bevat en 'n N50-lengte van 165,033 bp vertoon. Die biosintetiese geen groepe wat geïdentifiseer is, sluit in fengysien, butirosien A , butirosien B, skisokienen, pulcherriminensuur, bacillibactin , bacillibactin E , bacillibactin F, lichenicidin VK21 A1 , Lichenicidin VK21 A2, en thermoactinoclusters. Hieride geen groepe is bekend vir produksie van metaboliete met antimikrobakteriële aktiwiteit. Verder beskik die B. licheniformis SAMN36381076 oor gene wat kodeer vir ses diverse antibiotika weerstand meganismes, met uitvloei pompe as die oorheersende meganisme van weerstand. Metaboliese analise van B. licheniformis SAMN36381076 het getoon dat die teenwoordigheid van gene koolhidraat afbraak- en samestellings prosesse, oksianion biogeochemiese siklusse en stikstofsiklusse ingesluit het. In-silico evaluering van swam verbindings teen ser/ threokinases het die laagste ΔG- bindingswaardes van aurovertien D teen PknA (-50.9 kcal/mol), aurovertien D teen PknB (-50.7 kcal/mol), vertisillien A teen PknD (-36.8 kcal/mol) getoon, en roquefortine C teen PknE (-53,4 kcal/mol). Molekulêre dinamika-simulasie het getoon dat die PknD-vertisillien A die hoogste stabiliteit getoon het. Verder het die post-MD MMGBSA ΔG Bind getoon dat vertisillien A 'n hoë affiniteit vir PknD -53.67 kcal/mol het. Die resultate het aangedui dat vertisillien A 'n potensiële tref verbinding is wat verder geoptimaliseer en aangepas kan word om 'n kragtige antimikobakteriële inhibeerder te ontwikkel. In hierdie studie het die bioaktiewe verbindings teenwoordig in R. caffra en S. molle gemiddelde aktiwiteit teen M. tb H37Rv (MIC 0.25-0.125 mg/mL) getoon. Norajmaline met 'n dok telling van -7.47 kcal/mol, en pre-MM-GBSA van -37.64 kcal/mol is teruggekeer van die streng virtuelesifting. Molekulêre dinamika-simulasie en post-MD MM-GBSA het die stabiele binding van norajmaline aan PanK (-58. 73 kcal/mol) geopenbaar. Resultate van die vloeisitometrie-analise van behandelde MDA-MB 231-selle het aan die lig gebring dat die dichloormetaan ekstrakte van S.petersiana , Z. mucronate en etielasetaat ekstrakte van R. caffra en S. molle het hoër apoptosevlakke geïnduseer as die kontrole-cisplatien. Laastens, die klassieke masjienleer modelle wat uit logistiese regressie en multi-laag perseptron ontwikkel is, is geïdentifiseer om beduidende prestasiemaatstawwe op die InhA-datastel te hê. Die resultate (-R 2 ) van die multiteiken-LSTM-model het die behoefte aan hiper parameter verfynning aangedui. Verdere eksterne validering van die twee klassifikasie modelle is egter nodig. Ter afsluiting, hierdie studie dien as 'n beginpunt vir die in-silico ontdekking van kragtige antimikobakteriële teenmiddels van metallofiele (swamme en bakterieë) en plante. Virtuele sifting versnel die geneesmiddel ontdekkings proses deur verbindings te identifiseer wat aktiwiteit kan hê, en aangepas kan word om hulle kragtigheid te verhoog. Die inkorporering van 'n groot datastel van verbindings wat verskillende biologiese toestande bevat maar met dieselfde eindpunt kan gebruik word om robuuste modelle met buitengewone veralgemenings vermoëns te ontwikkel.
Description
Thesis (PhD)--Stellenbosch University, 2023.
Keywords
Citation