The Bfx derivatives (5b and 5n) were synthesized by Medical Chemistry Laboratory of Pharmaceutical Sciences of Araraquara (Fernandes et al., 2021). These compounds were provided in their pure form and did not require any additional synthesis in our laboratory.

The minimum inhibitory concentration (MIC₉₀) of the Bfx derivatives was evaluated against Mycobacterium bovis, M. smegmatis, monoresistant, multidrug resistant (MDR), pre-extensively drug-resistant (pre-XDR), and susceptible (H37Rv) strains of Mtb following the methodology of Palomino et al. (2002) with some modifications detailed below. The compounds were serially diluted, with concentrations ranging from 0.09 to 25 μg/mL, in 96-well plates using Middlebrook 7H9 broth supplemented with 10% Oleic acid-albumin-dextrose-catalase (OADC). A suspension of the bacterial strains was added to each well, and the plates were incubated for 7 days (for H37Rv, clinical isolates, and M. bovis) or 2 days (for M. smegmatis) at 37°C in a 5% CO₂ atmosphere. Subsequently, resazurin (0.01% in sterile water) was added. Resazurin is a redox indicator that is reduced to resorufin by metabolically active bacteria, resulting in a measurable fluorescence signal that correlates with cell viability. After 24 h (for H37Rv, clinical isolates, and M. bovis) or 12 h (for M. smegmatis), fluorescence was measured at 530/590 nm using a Cytation 3 microplate reader (Biotek®).

The methodology employed in this study was based on the protocols established by Ben Hur et al. (2022), using Mueller-Hinton II broth for most of the bacteria studied. For more demanding bacteria, the medium was supplemented with lysed horse blood and β-NAD (MH-F broth) (Carkaci et al., 2017). Detailed descriptions of the methodologies and specific conditions used in the experiments can be found in the Supplementary material.

Murine macrophage J774.A1 and MRC-5 fibroblasts cells were seeded in 96-well plates (1.0 × 10⁶ cells/mL) and incubated for 24 h at 37°C, 5% CO₂ for adherence. Compounds were applied to the wells at concentrations ranging from 0.39 to 100 μg/mL. After 24 and 72 h, 30 μL of resazurin was added. Fluorescence was measured after 4 h at 530/590 nm using a Cytation 3 (Biotek®) reader. The IC₅₀ was determined as the minimum compound concentration capable of inhibiting 50% of cell viability compared to the control. The SI was calculated as IC₅₀/MIC₉₀, considering compounds with SI > 10 as safe (Amaral et al., 2024).

To evaluate the interaction between Bfx and RFP, a checkerboard assay was conducted following the methodology proposed by Berenbaum (Berenbaum, 1978).Test compounds were serially diluted in Middlebrook 7H9 broth and arranged in a 96-well plate, starting with 2x MIC for each compound. RFP was also diluted horizontally based on its MIC₉₀. The combinations of different concentrations of Bfx and RFP were incubated with Mtb (5 × 10⁵ CFU/mL) at 37°C and 5% CO₂ for 7 days. Post-incubation, 0.01% resazurin was added, and fluorescence readings were taken after 24 h to determine the Fractional Inhibitory Concentration Index (FICI). The combined effects of the compound combinations were classified according to the Fractional Inhibitory Concentration Index (FICI):

Murine macrophage J774.A1 cells (1 × 10⁶ cells/mL) were seeded in 24-well plates and incubated as previously described. After 24 h, the supernatant was discarded, and 1 mL of Mtb inoculum (2 × 10⁶ CFU/mL) was added to the wells. The plate was incubated for 2 h to facilitate infection. Subsequently, the supernatant was removed, and the cells were washed with PBS. An amikacin solution (200 μg/mL) was added to ensure the sterility of the extracellular environment. After 2 h, the wells were washed, and treatments were applied (1x, 5x, and 10x MIC₉₀ of Bfx or RFP). The cultures were then incubated for 72 h. After incubation, the cells were lysed using a Triton solution, and the supernatant was collected, serially diluted, and plated on Middlebrook 7H11 medium supplemented with 10% OADC. These plates were incubated for 20–30 days until colony-forming units (CFU) were counted (Solcia et al., 2021).

Mtb cultures (5 mL at 10⁵ CFU/mL) were prepared and divided into four groups: one treated with Bfx, one with isoniazid (INH), another with ethambutol (EMB), and an untreated group, following the described conditions. After incubation, the cells were washed with PBS and fixed overnight with Karnovsky fixative solution (containing 2% glutaraldehyde, 2% paraformaldehyde, and 3% sucrose in 0.1 M sodium cacodylate buffer, pH 7.2). The cells were then washed with 0.1 M sodium cacodylate buffer, pH 7.2, postfixed with 2% osmium tetroxide for 30 min, and washed with distilled water. Dehydration was carried out in an increasing series of acetone (50–100%), followed by critical point drying using CPD 030 (BALZERS, Geneva, IL, United States) and SCD 500 metallization (LEICA, Wetzlar, Hesse, Germany), before being analyzed using a JSM 7001F scanning electron microscope (15 kV) (Jeol—Tokyo, Japan) (Caleffi-Ferracioli et al., 2016).

To select Mtb strains resistant to the most active Bfx compounds, a bacterial inoculum at 10⁵ CFU/mL was seeded on Middlebrook 7H11 agar plates supplemented with OADC and various concentrations of compound 5n (0.78 to 25 μg/mL). Plates were incubated at 37°C in a 5% CO₂ atmosphere until bacterial growth was observed. A colony that grew at concentrations above the MIC₉₀, as determined by the REMA method, was identified as a resistant mutant. DNA was then extracted from the resistant colony using glass beads. The colony was vortexed with TE buffer and beads, followed by supernatant transfer, precipitation with sodium acetate and ethanol, and incubation at −20°C. After centrifugation, the supernatant was discarded, and the pellet was washed with 70% ethanol, dried, and resuspended in milliQ water. The DNA concentration and quality were assessed using a Nanodrop spectrophotometer and stored at −20°C. Whole Genome Sequencing was conducted at SeqCenter (Pittsburgh, PA, United States) using Illumina HiSeq technology with a coverage depth greater than 100. Variant calling was performed with Breseq software, comparing the sequences to the reference Mtb H37Rv genome (Deatherage and Barrick, 2014).

The 2D structure of compound 5n was designed using ChemDraw Professional 17.0, and its 3D analog was generated in Chem3D (Cousins, 2005). The molecular geometry was energy optimized using the auto-optimization tool in Avogadro software, applying a universal force field (Hanwell et al., 2012). The modeling of the Rv1855c protein was performed using the I-TASSER web server (Yang and Zhang, 2015).¹ To determine the druggability of the Rv1855c protein pockets and their binding affinity with compound 5n, PockDrug was used (Hussein et al., 2015),² selecting pockets with a probability greater than 95% and a minimum proximity of 5.5 Å (Primo et al., 2023). Molecular docking analysis was conducted with AutoDock Vina (Trott and Olson, 2010; Eberhardt et al., 2021). Finally, ligand-receptor interactions were analyzed and visualized using Discovery Studio Visualizer software (Pawar and Rohane, 2021).

Female BALB/c mice (7–8 weeks old) were housed under controlled conditions and stratified into groups: Infection Control (IC), Post-Infection Control (PIC), RFP, 5b, and 5n, each in triplicate. Mice were infected intranasally with 20 μL of Mtb H37Rv suspension (1.5 × 10⁵ CFU/mouse). Fourteen days post-infection, the IC group was sacrificed to measure the bacterial load present in their lungs at the beginning of the experimental treatment. Treatment groups received daily oral suspensions of 5b or 5n at 200 mg/kg/mouse, and RFP at 15 mg/kg/mouse. Control groups received saline solution. The PIC group was sacrificed on day 42 to compare how the infection evolves over time without active treatment. After completing the experimental timeline, lung tissues were collected, homogenized, and cultured on Middlebrook 7H11 agar to determine the bacterial load, with incubation at 37°C and 5% CO₂ for up to 60 days to count CFU (Mourik et al., 2018).

The results presented here are expressed as the mean ± standard deviation (SD) of three experimental analyses. Statistical significance analyses were performed using GraphPad Prism 5.01 (GraphPad Software Inc., La Jolla, CA) through a one-way Analysis of Variance (ANOVA) followed by Dunnett’s post hoc test. Statistical significance was considered for p-values less than 0.05.

The Bfx compounds demonstrated significant potential in inhibiting monoresistant strains to INH, MDR, and Pre-XDR, as well as the drug-sensitive Mtb strain H37Rv, as shown in Table 1. Compound 5n exhibited high efficacy against most clinical isolates, except for the MDR strain CF48, where it reached an inhibitory concentration of 22.38 ± 9.37 μM. However, it stood out for its ability to eliminate Pre-XDR strains at concentrations as low as 0.28 μM, outperforming all the commercial drugs tested, including RFP and LNZ, which showed significantly higher values.

On the other hand, compound 5b not only demonstrated potent activity against Mtb H37Rv (0.70 ± 0.26 μM) but was also effective against multiple MDR and Pre-XDR strains, such as CF165 (0.24 ± 0.00 μM), CF171 (0.25 ± 0.01 μM), and CF162 (0.26 ± 0.00 μM). These low inhibitory concentrations suggest that the structural modifications in compound 5b may enhance its ability to evade the specific resistance mechanisms presented by these strains. It is important to note that strains CF165 and CF171 are resistant to INH and RFP; CF61 and CF162 are resistant to amikacin; CF48 and CF105 are resistant to both amikacin and linezolid; and CF81 and CF169 are resistant to RFP, INH, and a fluoroquinolone.

Compounds 5n and 5b were evaluated against a variety of Gram-negative, Gram-positive bacteria, and fungi, including Candida albicans. According to the data presented in Table 2, both compounds showed a lack of significant activity against all the bacterial and fungal strains tested, with MIC₉₀ values exceeding 200 μM for all species analyzed. In contrast, commercial drugs such as RFP exhibited notable activities, particularly against Staphylococcus epidermidis (0.001 μM) and Staphylococcus saprophyticus (0.009 μM).

These results indicate that compounds 5n and 5b do not have a broad spectrum of action. On the contrary, when comparing these data with the previous results obtained against Mtb, where both compounds demonstrated high efficacy, it can be inferred that 5n and 5b might act as narrow-spectrum antimicrobial agents. This suggests that the mechanisms of action of 5n and 5b may be highly specialized for Mtb, avoiding disruption of the gut microbiota and potentially reducing adverse effects associated with broad-spectrum treatments. This narrow focus could also minimize the risk of cross-resistance with other classes of antibiotics, a key factor in combating the growing threat of antimicrobial resistance.

Compound 5n showed an IC₅₀ greater than 230.24 μM in murine macrophages J774.A1 and 221.52 μM in human fibroblasts MRC-5. These values indicate a low cytotoxicity of the compound toward healthy cells. The selectivity indices (SI) greater than 2,400 for both cell lines reflect a remarkable ability of the compound to inhibit Mtb H37Rv (MIC₉₀ of 0.09 ± 0.04 μM) without significantly affecting healthy cells (Table 3). This indicates that compound 5n has a wide therapeutic margin.

Compound 5b, on the other hand, presented an IC₅₀ of 40.83 ± 10.72 μM in J774.A1 and 57.07 ± 48.15 μM in MRC-5. Although these values reflect higher cytotoxicity compared to 5n, the calculated selectivity indices of 57.14 for J774.A1 and 81.43 for MRC-5 still indicate an acceptable therapeutic margin. The antimicrobial activity of compound 5b, with an MIC₉₀ of 0.7 ± 0.26 μM against Mtb H37Rv, suggests that despite the relatively higher toxicity, it remains a viable compound for the study in the inhibition of Mtb, provided that doses are carefully considered to minimize adverse effects.

To explore the combined effectiveness of these compounds with RFP, a checkerboard assay was conducted to evaluate the synergy between Bfx and RFP against Mtb H37Rv. The results showed FICI values of 0.083 for compound 5n and 0.11 for compound 5b, indicating significant synergy (Supplementary Table S1). This suggests that the combination of Bfx with RFP could enhance treatment efficacy, allowing for the use of lower concentrations of both compounds, potentially contributing to a reduction in overall treatment toxicity.

Mtb was treated with compounds EMB, INH, and 5n, as well as an untreated control, all at concentrations equivalent to 1x MIC (Figure 2). The purpose of these experiments was to evaluate the morphological effects induced by these compounds on Mtb using scanning electron microscopy (SEM). The obtained images allowed for a comparison of the structural integrity of untreated mycobacteria with the morphological alterations caused by the treatments, thus providing a clearer view of the efficacy and possible mechanisms of action of each compound.

In the untreated control (Figures 2A,B), the mycobacteria exhibited a typical bacillary morphology, with intact and well-defined cellular structures, without any signs of structural damage. These images served as a reference to assess the effects of the antimicrobial compounds.

Treatment with EMB at 1x MIC (Figures 2C,D) showed damaged cell surfaces and the presence of cellular debris, suggesting a bactericidal effect characterized by disruption of cell wall integrity. The disruption in bacillary morphology, along with the disorganization of the cell surface, indicates that EMB exerts its action through interference with the synthesis of the Mtb cell wall, resulting in the loss of bacterial viability.

Treatment with INH at 1x MIC (Figures 2E,F) resulted in the appearance of amorphous structures and cell lysis. The treated mycobacteria showed a complete loss of morphology and the formation of disintegrated cellular debris, consistent with a potent lytic effect. This result supports the known mechanism of action of INH, which inhibits the synthesis of mycolic acids essential for the structure and function of the Mtb cell wall.

Finally, treatment with compound 5n at 1x MIC (Figures 2G,H) also caused morphological damage to Mtb. Mycobacteria treated with 5n showed distorted filamentous structures and considerable damage to the integrity of the bacilli. These observations suggest that 5n induces a unique mechanism of action that may involve the disruption of critical components of the Mtb cell wall or plasma membrane, resulting in deformation and possible cell death.

To evaluate the mechanism of action of 5n, we aimed to isolate a Bfx-resistant mutant. 7H11 agar plates with varying concentrations of the compound were prepared, and we successfully isolated a colony capable of growing at a concentration of 13.5 μM. This colony appeared on the solid medium 40 days after inoculation, was collected, and re-inoculated onto a plate containing 13.5 μM of compound 5n. New colonies were obtained; some were used for genetic material extraction for sequencing, while others were cultured for storage. We considered these colonies to be 5n-resistant mutants due to their growth on a medium containing more than 150 times the previously determined MIC of 0.09 μM. The genetic material was subjected to whole-genome sequencing and variant analysis using Breseq (Deatherage and Barrick, 2014). Comparative analysis was conducted between (a) the reference genome of Mtb H37Rv available at https://www.ncbi.nlm.nih.gov/datasets/genome/GCF_000195955.2/ (reference strain); (b) the Mtb H37Rv genome from our laboratory used as a control (control strain); and (c) the genome of the 5n-resistant mutant obtained in the previous procedure (5n Mutant). This analysis revealed four single nucleotide polymorphisms (SNPs) among the mutated genes in the control strain and the 5n Mutant.

In the control strain, mutations were identified in the PPE protein family genes (PPE13 and PPE18 at two distinct positions) and in Rv1855c. It is important to note that the genetic alterations identified in these genes did not induce phenotypic changes in their final protein products, as the SNPs were classified as silent mutations. These alterations did not impact the final amino acid sequence during protein translation. However, in the Rv1855c gene, a single base pair deletion was observed, resulting in a frameshift mutation in a hypothetical oxidoreductase. This genomic alteration observed in the laboratory control strain H37Rv did not hinder the in vitro growth of the inoculum; on the contrary, it allowed for the optimization of mycobacterial growth. We hypothesize that Bfx 5n exerted a “restorative function” by reversing the mutation in the control strain, as the strain needed to protect itself and reduce nitrosative stress. However, to understand the 5n-receptor interactions, we decided to perform complementary in silico analyses, as the results observed in the genomic sequencing data analysis are not yet sufficient to confirm the true target and mechanism of action of Bfx 5n but shed light on potential resistance mechanisms generated by Mtb against this compound.

To support the genetic sequencing results and the potential mechanisms of action, in silico studies were conducted (Figure 3). The 2D structure of compound 5n was converted to 3D using Chem3D software, and Avogadro was used to relax and optimize the compound (Figure 3A). I-TASSER provided three structural models of the protein encoded by the Rv1855c gene (Rv1855c Protein) with C-scores of 0.69, −3.65, and − 5. Based on the manufacturer’s guidelines, we selected the structure with a C-score of 0.69 (Figure 3B). Distances, classifications, types, and quantities of interactions were visualized using Discovery Studio Visualizer. 2D and 3D images of the ligand-receptor interactions for each of the tested pockets were obtained (Figure 3C).

PockDrug identified 17 pharmacodynamic pockets (Supplementary Table S2), of which only those with a probability greater than 95% of being pharmacodynamic sites were selected. Specifically, pockets 9, 12, and 16 were chosen due to their high pharmacodynamic probabilities of 99, 98, and 95%, respectively. The selection of docking models obtained from AutoDock Vina was determined using three main criteria: ligand affinity for the receptors, the number of simultaneous interactions between the amino acids in each pharmacological pocket and the ligand, and the specific type of bonds formed. AutoDock Vina generated nine models for each molecular docking performed between the ligand and the selected receptor pockets (Supplementary Table S3). 5n bound strongly and stably to pharmacodynamic pockets 9, 12, and 15 of the Rv1855c protein of Mtb, with binding energies of −7.6, −8.1, and − 8.2 kcal/mol, respectively. The distances, rankings, types, and quantities of interactions were visualized using Discovery Studio Visualizer. 2D and 3D images of the ligand-receptor interactions were obtained for each of the tested pockets (Figure 3C).

Compound 5n exhibited a variety of interactions with different pockets of the Rv1855c protein (Supplementary Table S4). In pocket 9, seven specific interactions were observed, notably a hydrogen bond with TYR87, characterized by its strength and stability, with a distance of 3.326 Å. The other six interactions in this pocket were of the hydrophobic pi-alkyl type with residues such as ARG88, LEU92, VAL18, PRO58, LEU60, and again ARG88, demonstrating the compound’s tendency to engage in nonpolar interactions. In pocket 12, compound 5n produced eight interactions, including a conventional hydrogen bond with GLY184, with a distance of 2.59085 Å, and a carbon-hydrogen bond with GLN8, with a distance of 3.52464 Å. The remaining six interactions were hydrophobic, highlighting multiple pi-pi T-shaped interactions with the TRP116 residue, suggesting significant interaction with aromatic rings. Finally, in pocket 15, eight interactions were recorded, where compound 5n formed a conventional hydrogen bond with GLY184 and a pi-donor hydrogen bond with GLY182, with distances of 2.52025 Å and 3.19934 Å, respectively. Additionally, six hydrophobic interactions were observed, with pi-pi T-shaped interactions with TRP116 being particularly notable, emphasizing the relevance of the compound’s spatial configuration and its ability to integrate into the hydrophobic environment of the protein pocket.

In intramacrophagic studies, the bacterial load within macrophages was evaluated following treatment with RFP at 10 times its MIC, resulting in a 0.7 log₁₀ inhibition. Compound 5n showed a 0.9 log₁₀ inhibition, while compound 5b achieved a bacterial growth reduction of 2.3 log₁₀. These results highlight the efficacy of 5n and 5b, compared to RFP, demonstrated greater potency against intracellular Mtb H37Rv (Figure 4A).

In the in vivo analyses, Balb/C mice were infected with Mtb H37Rv and administered the corresponding treatments, following a protocol detailed in the methodology (Figure 4B). During the four-week treatment, no adverse changes in behavior or animal loss were observed, suggesting good tolerability of the compounds. Initially, 1.5 × 10⁵ CFU per animal were inoculated, and 14 days later, the infected control group (IC) presented a bacterial count in the lungs of 1.4 × 10⁵ CFU (4.2 log₁₀). At the end of the 4 weeks, the mean bacterial concentration in the lungs of the mice in the PIC group was 1.5 × 10³ CFU (3.0 log₁₀), representing a 1.22 log₁₀ reduction attributable to the activation of the mice’s immune system, capable of reducing the bacterial load in lung tissue. When comparing the effects of the treatment with the compounds, a 1.23 log₁₀ reduction was observed with RFP, a 1.45 log₁₀ reduction with 5b, and a 3.0 log₁₀ reduction with 5n. This difference in the efficacy of 5n and 5b may be related to their molecular structures, where the presence of a pyridine in 5n could be contributing to its higher bactericidal activity compared to the sulfide group present in 5b (Patel et al., 2020; Marinescu and Popa, 2022; Villamizar-Mogotocoro et al., 2020).