Four diterpenes identified in silico were isolated from Hyptidinae and demonstrated in vitro activity against Mycobacterium tuberculosis

Abstract Plants of Hyptidinae subtribe (Lamiaceae – family), as Mesosphaerum sidifolium, are a source of bioactive molecules. In the search for new drug candidates, we perform chemical characterization of diterpenes isolated from the aerial parts of M. sidifolium was carried out with uni- and bidimensional NMR spectral data, and evaluate in silico through the construction of a predictive model followed by in vitro testing Mycobacterium tuberculosis and Mycobacterium smegmatis. Resulted in the isolation of four components: Pomiferin D (1), Salviol (2), Pomiferin E (3) and 2α-hydroxysugiol (4), as well as two phenolic compounds, rosmarinic and caffeic acids. In silico model identified 48 diterpenes likely to have biological activity against M. tuberculosis. The diterpenes isolated were tested in vitro against M. tuberculosis demonstrating MIC = 125 µM for 4 and 1, while 2 and 3 -MIC = 250 µM. These compounds did not show biological activity at these concentrations for M. smegmatis. Graphical Abstract


Introduction
The medicinal uses of plants have been documented throughout history (Casanova and Costa 2017;Koparde et al. 2019). Brazil has one of the largest vegetation covers in the world, containing six biomes and more than 12,000 species of plants. The innate abundance of natural products in Brazil puts it in a privileged position to explore active therapeutics (Barreiro and Bolzani 2009;Saraiva et al. 2015;Berlinck et al. 2017). To identify active metabolites, we can perform virtual screenings of related compound libraries to elucidate robust candidates with desired properties (Alves et al. 2017).
The Lamiaceae family is composed of 295 genera and more than 7000 species of cosmopolitan distributed herbs and shrubs (Cavalcanti et al. 2019;Mesquita et al. 2019). In Brazil, 46 genus and 524 species of this family are present (Soares et al. 2019). Lamiaceae species metabolites are known worldwide for their various uses in ethnopharmacology and nutraceuticals, but more studies are needed to characterize the active phytochemicals. (Frezza et al. 2019). Diterpenes are abundant secondary metabolites in the Lamiaceae family (Zhang et al. 2018). The Hyptidinae clade contains many abietane diterpenes, known for high rates of endemism, and is well distributed in Brazil (Bridi et al. 2021).
Among the species of Hyptidinae to be studied, Mesosphaerum sidifolium (L'H erit.), Harley & J.F.B. Pastore ¼ Hyptis umbrosa, Salzm. ex Benth. is at the forefront of interest. Popularly known in Brazil as 'bamburral' and 'Aleluia de serrote', this species is a perennial sub-shrub (da Silva and Andrade 2004;Harley and Pastore 2012). Mesosphaerum plants were commonly used in folk medicine to treat headaches, menstrual cramps, the flu, gout, and digestive problems, as well as being known as an antimicrobial and antiseptic (Bas ılio et al. 2007). While M. sidifolium (MS) leaves are used to treat nasal and auricular disorders, they also have expectorant, carminative and tonic functions (Agra et al. 2008). Previous studies have shown that MS-ethanolic extract has potent antitumor activity , as well as analgesic and antiinflammatory activity (Dos Anjos et al. 2017;Scotti et al. 2018).
Studies of other species of Lamiaceae showed potential activity against M. tuberculosis of some compounds obtained such as phenolic derivatives, diterpenes, flavonoids, phenyl propanoids and steroids (Siddiqui et al. 2012;Lirio et al. 2014;Cavalcanti et al. 2021).
In light of the antimicrobial activity of Mesosphaerum species, these extracts and similar compounds serve as promising leads to fight infections such as tuberculosis (TB). TB is an infectious disease caused primarily by M. tuberculosis, which affects the respiratory tract (Bas ılio et al. 2007;Yu et al. 2020). It was diagnosed in more than 10 million people in 2016 and is responsible for 1.45 million deaths per year. Therapeutic regimens are available to treat TB in diagnosed patients. Those with active TB should seek treatment immediately and be sure to finish their treatment regimen. Unfortunately, inadequate treatment schemes have allowed for the development and spread of resistant strains of M. tuberculosis (Giacobbo et al. 2017;Murad as et al. 2018). Multiple studies have shown that natural products, particularly diterpenes, display significant biological activity against TB compared to second line TB therapeutics (Matos et al. 2015;Silva et al. 2017;Soares et al. 2017).
The aim of our study is to use natural compounds to derive new, safer, and more efficient drug candidates to treat TB by using computational tools. Thus, this work performs a virtual screening of the Hyptidinae diterpenes through phytochemical analysis of the M. sidifolium species. That way, using the results of these screenings, we select the most promising compounds for antitubercular activity and follow with in vitro tests against M. tuberculosis.

Phytochemistry
Six compounds were isolated from the aerial parts of M. sidifolium (Supplemental material Figure S1), four of which (1-4) are abietane diterpenes first reported in this genus. Spectral data for compounds 1-3 are found in the supplemental material (Tables S1, S2 and S3, respectively). Highlighting the reassignment of chemical shifts of 13 C, as Pomiferin D (Supplementary material Figures S2-S7), Pomiferin E and 2ahydroxysugiol (Tables S1, S3 and S4, respectively).
Compound 4 was isolated in an amorphous green powder. The infrared spectrum (IR) demonstrated absorption at 3408 cm À1 , characteristic of the presence of phenolic hydroxyl. This which is corroborated by absorption at 1269 cm À1 , referring to the C-O attached to an aromatic ring (Supplementary material Figure S8). The 1 H NMR (Nuclear magnetic resonance) spectrum showed two singlets in the region of aromatic protons with chemical shifts at d H 7.88 and 6.72 ppm, attributed to H-14 and H-11, respectively. Furthermore, a doublet at d H 1.24 ppm (J ¼ 7.0 Hz) plus a septet at d H 3.18 ppm were observed and recognized as the isopropyl unit. These signals are compatible with the abietane diterpene skeleton and were attributed to H-14, H-11, 3H-16, 3H-17 and H-15, respectively (Supplementary material Table S4) (Gonzalez et al. 1989). Three singlets at d H 0.97, 1.01 and 1.22 ppm were observed, which, by comparison with the literature, were attributed to the methyls CH 3 -18, CH 3 -19 and CH 3 -20 (Supplementary  material Table S4) (Gonzalez et al. 1989). There was also a triplet of triplets at d H 4.09 ppm, attributed to a methoxy hydrogen. The 13 C NMR spectrum of Compound 4 showed 19 signals, corresponding to 20 carbon atoms. Of these, seven signals were attributed to nonhydrogenated carbon, five signals were attributed to methinic carbons, three signals were attributed to methylene carbons and four signals were attributed to methyl carbons. The presence of only two methinic sp 2 carbons implies oxygenation at C-12 and a protective effect explains the chemical shift at d C 109.45 ppm (C-11). This proposal is corroborated by the chemical shift at d C 26.75 ppm, which was attributed to C-15, under protection c of the OH group at C-12 (Supplementary material Figure S10). The signals of d H 1.01 ppm (CH 3 -19) and d C 32.56, 34.80, 48.75 and 50.33 ppm, were connected to C-18, C-4, C-5 and C-3, respectively. Furthermore, d H 0.97 ppm (CH 3 -18) and the signals d C 22.29, 34.80, 48.75 and 50.33 ppm, were identified as C-19, C-4, C-5 and C-3, respectively. The C-7 signal was reinforced by the correlation of d H 7.88 ppm (H-14) with d C 197.87 ppm (C-7) (Supplementary material Figures S11 and Figures S14 and S15).

QSAR modelling
The predictive model showed excellent performance, with a ROC curve of 0.95 for external testing and 0.95 for internal cross-validation (Supplementary material Figure  S16). The Matthews correlation coefficient (MCC) was 0.80 in the external test, revealing a robust model, in addition to excellent indices of sensitivity, specificity, accuracy and precision (Supplementary material Table S5).
We found that 48 of the 68 diterpenes analyzed displayed probable activity against M. tuberculosis, as shown in supplementary material Table S6. We also observed that most species in this subtribe occur in regions of Brazil. The four isolated diterpenes were classified as active against M. tuberculosis with sufficient confidence (Supplementary material Table S7).

Prediction of absorption, distribution, excretion and toxicity properties
Additionally, the number of violations of Lipinski's rule (LIP) were calculated for each compound. This parameter indicates whether a molecule has good oral bioavailability (absorption or permeability), determined by four physicochemical parameters: molecular mass, octanol/water partition coefficient, number of hydrogen bonding donor atoms and number of hydrogen bonding atoms (Barreiro and Fraga 2014;Piccirillo and do Amaral 2018). Lipinski concluded that if a molecule violates two or more parameters, they are very likely to be poorly permeable (Capecchi et al. 2019). As seen in Supplementary Material Table S8, the maximum number of violations found was one parameter in the molecules Salviol and Pomiferin D, indicating that the four diterpenes have a high probability of being orally bioavailable.
The risk of cytotoxicity was also assessed and revealed that the four diterpenes did not present risks of cytotoxicity in the evaluated parameters. The oral absorption rate (topological surface area) was calculated as values of % ABS > 89%, suggesting this as a highly available route (Supplementary material Table S8). With these analyses, the diterpenes isolated from M. sidifolium indicate activity against the bacterium M. tuberculosis and excellent statistics relating to bioavailability, oral absorption, and cytotoxicity. Thus, these diterpenes are compounds that may find success in treating TB, and should be examined further.

Molecular docking
To identify a possible mechanism of action of these diterpenes in M. tuberculosis, a study of molecular docking was carried out. In docking, the interactions and energies of interaction between the molecule and target active site are analyzed and compared with the inhibitors of these proteins and drugs that are already used in the treatment of the disease. An aminotransferase (PDB ID 5YHV) and an oxidoreductase (5VRN) and their respective inhibitors were obtained from the Protein Data Bank. The diterpenes had better interaction energy than some control drugs and had energy similar to or better than the inhibitors in the two proteins, as can be seen in supplementary material Figure S17.
The drug terizidone showed the best coupling energy in the two proteins, aminotransferase, and oxidoreductase (InhA). For the aminotransferase enzyme, the diterpenes had lower binding energy than the cocrystallized inhibitor, but Pomiferin D stands out, as the nest diterpene binding energy (-96.08 kJ.mol À1 ) and presented energy like the cocrystallized inhibitor (-96.70 kJ.mol À1 ). However, all diterpenes showed better interaction energies than the prothionamide, ethionamide, cycloserine and pyrazinamide controls (Supplementary material Figure S17). Pomiferin D showed interactions with residues similar to terizidone control and/or complexed inhibitor; among the hydrogen bonds it is possible to highlight Ser100, Arg242 and Ser101, in addition to the steric interactions with the residues Ser100, Tyr125, Asn174, Lys234 and Tyr128. Some of these residues mentioned are also found in the compounds 2ahydroxysugiol, Salviol and Pomiferin E, and we hypothesize that these similarities contribute to the good anchoring energies presented by all the analysed compounds, reinforcing the possible antitubercular potential and mechanism of action of the diterpenes of this study in relation to interaction with aminotransferase (Supplementary material Figure S18 and Table S9).
For the InhA enzyme, Pomiferin E (-99.30 kJ.mol À1 ) presented the best coupling value compared with the other diterpenes; however, all diterpenes analysed showed better energies than the controls, except for terizidone (-108.71 kJ.mol À1 ) (Supplementary material Figure S19). Pomiferin E did not show hydrogen bonds with the protein, but the steric bonds gave a better coupling to this compound than the other diterpenes. The steric interaction with the Leu218 residue in the protein is present in terizidone and all the diterpenes analyzed, however, it is not found in the prothionamide control (-56.43 kJ.mol À1 ) or the complex inhibitor (Supplementary material Figure S19 and Table S9).
We analyzed of the possible metabolites of these diterpenes, their toxicities in humans and the ADME properties. Only metabolite 35 of Salviol (Supplementary  material Table S10) presented a very high risk of toxicity on the reproductive system and the liver, with a score above 85 (Supplementary material Figures S20 and S21). All compounds showed high probability to be absorbed in the gastrointestinal tract and all were predicted like inhibitor the CYP2D6 (Supplementary material Table S11).

Antitubercular biological activity
The molecules had their biological activities tested in vitro against M. tuberculosis H37Ra, whose data can be found in Table 1. The compounds 2a-hydroxysugiol and Pomiferin D showed greater activity against M. tuberculosis with Minimum Inhibitory Concentration (MIC) 125 mM, while Salviol and Pomiferin E showed MIC 250 mM. Thus, the result of the biological test validated the antitubercular activity prediction model used, in which all the diterpenes tested were promising against M. tuberculosis, corroborating the in silico study. However, these compounds did not at these concentrations show biological activity for M. smegmatis.
Studies reported by Soares et al. demonstrate that some kaurane and trachylobane diterpenes have antituberculosis activity, with MIC values similar than the secondchoice antituberculosis drug Cycloserine (122.4-498 lM) . Thus, we suggest that the diterpenes identified in this study also have similar potency to Cycloserine.

Instrumentation and general procedures
The detailed description is provided in the supplementary material.

Plant material
The detailed description is provided in the supplementary material. aMIC values reported here were observed in two independent experiments or the highest value found among three independent tests. bINH, isoniazid; RIF, rifampicin; MOX, moxifloxacin.

Isolation of chemical constituents
The detailed description is provided in the supplementary material.

Quantitative structure-activity relationship (QSAR) modelling
The detailed description is provided in the supplementary material.
3.5. Predicting absorption, distribution, excretion, toxicity properties, hepatic metabolites and ADME approach The detailed description is provided in the supplementary material.

Molecular docking
The detailed description is provided in the supplementary material.

Mycobacterial susceptibility assay
The detailed description is provided in the supplementary material.

Conclusions
We performed phytochemical characterization of the M. sidifolium plant through the isolation of Pomiferin D, Salviol, Pomiferin E, 2a-hydroxysugiol, rosmarinic acid and caffeic acid. Isolated abietane diterpenes are herein reported for the first time in the Mesosphaerum genus, and the 13 C signal assignment was done unambiguously, for Pomiferin D, Pomiferin E and 2a-hydroxysugiol. In silico analysis that includes virtual screening of antitubercular activity and ADMET prediction suggested that 48 of the 68 Hyptidinae diterpenes were likely to be active against M. tuberculosis. Of the four diterpenes analyzed in this research, Salviol showed the highest potential for activity accompanied by high risk for reproductive and/or renal toxicity, which is causing high concern regarding its use. However, all four diterpenes were tested in vitro against M. tuberculosis and demonstrated antitubercular activity at very high concentrations: 2ahydroxysugiol and Pomiferin D with MIC ¼ 125 mM, while Salviol and Pomiferin E showed MIC ¼ 250 mM. As predicted, these compounds did not show biological activity at these concentrations for M. smegmatis.

Disclosure statement
No potential conflict of interest was reported by the authors.