Compounds 1–9 were synthesized and characterized spectroscopically using a previously reported protocol (Sammor et al. 2018).

To evaluate the accuracy of the developed assay, the performance of the standard antibiotic gentamicin (Sigma) was evaluated against the five tested bacterial strains to determine MIC values and compare them with those published by the Clinical and Laboratory Standards Institute (CLSI) (M100-S22, Vol. 32, No. 3, January 2012; Institute 2012).

Gram-negative bacteria, including E. coli (ATCC® 25922), Salmonella typhimurium (ATCC® 14028), and Klebsiella aerogenes (ATCC® 13048), and Gram-positive bacteria, including Bacillus cereus (ATCC® 13640) and Enterococcus faecalis (ATCC® 7080), were subcultured on nutrient agar and incubated overnight at 37 °C. The following day, 3–5 bacterial colonies with similar morphologies were inoculated into 10 mL of sterile Mueller-Hinton broth (MHB) and incubated overnight at 37 °C. The overnight bacterial suspensions were adjusted to a 0.5 McFarland standard (approximately 0.5–2.0 × 10⁸ CFU/mL) with sterile MHB (Bhalodia and Shukla 2011).

Antibacterial activity was evaluated using a resazurin-based microdilution assay as described previously (Drummond and Waigh 2000; Oliveira et al. 2022; Al-Akhras et al. 2026; Alqaffaf et al. 2026). Each well of a sterile 96-well plate was filled with 100 µL of MHB. The nine derivatives were dissolved in 10% DMSO and serially diluted twofold over the first 10 wells of each test row. Column 11 served as the reagent control (medium plus inoculum), and column 12 served as a negative control (sterile broth). Subsequently, 10 µL of the prepared bacterial inoculum was added to each appropriate well. All experiments were performed in triplicate. After incubation, MIC values were determined based on resazurin color changes (Fig. 2), where blue indicated growth inhibition and pink indicated active bacterial metabolism. For minimum bactericidal concentration (MBC) determination, aliquots from wells containing concentrations above the MIC were plated onto nutrient agar and incubated at 37 °C for 24 h. The MBC was defined as the lowest concentration producing no visible colony formation.

The 3D crystal structure of peptidoglycan glycosyltransferase (PDB ID: 3D3H), with a resolution of 2.31 Å (Yuan et al. 2008), was downloaded from the RCSB PDB (http://www.rcsb.org) and prepared in PyMOL 2.4.0. Protein preparation for docking involved removing non-protein components and water molecules, extracting bound ligands, correcting missing amino acids, and adding polar hydrogen atoms (Mooers 2020). For binding-site definition, a suitable grid box (20 Å) was created around the bound ligand (M4O). The 2D structures of the synthesized compounds were drawn using ChemDraw Ultra 7.0 (Cambridge Soft Corp., http://http://www.cambridgesoft.com, USA). AutoDock 4.2.1 and AutoDock Vina 1.1.2 were used for the molecular docking studies (Morris et al. 2009). Initially, the native ligand (M4O) was extracted and docked back into the binding site to validate the protocol. Then, the synthesized ligands were docked into the predefined binding site. The docking energy was tabulated, and the poses were examined using BIOVIA Discovery Studio Visualizer (version 4) (He et al. 2021).

The in silico CLC-Pred tool available within the Way2Drug web platform was used to predict the cytotoxicity of the synthesized 3,10-dihydro-2H-1,3-oxazepino[7,6-b]indole derivatives 1–9. This model evaluates the cytotoxicity of the derivatives across a large panel of human tumor-derived and normal cell lines based on structure–activity relationship analysis. The prediction output for each derivative included the probability of activity (Pa) and probability of inactivity (Pi) values for cytotoxic effects against individual human cell lines. A derivative was considered potentially cytotoxic when Pa exceeded Pi and surpassed commonly accepted probability thresholds. Derivatives exhibiting low Pa values across the tested cell lines were predicted to possess minimal cytotoxic liability. These in silico predictions were used to evaluate the preliminary safety profile of the antibacterial candidates before experimental validation.

Physicochemical and pharmacokinetic properties of the nine derivatives were predicted using the SwissADME web server. Parameters evaluated included molecular weight (MW), topological polar surface area (TPSA), and lipophilicity (Log Po/w). Predictions were generated for gastrointestinal absorption, blood–brain barrier permeability, and P-glycoprotein substrate affinity.

Cytochrome P450 (CYP) inhibition profiles were evaluated for five major isoforms, and structural alerts were screened using PAINS and Brenk filters. Predicted biological activity (Pa) for normal human cell lines (HaCaT and CNE-2) was also calculated. All in silico predictions were performed using SMILES-formatted structures, and each calculation was repeated twice to verify reproducibility.

The toxicity of 3,10-dihydro-2H-1,3-oxazepino[7,6-b]indole derivatives 1–9 was evaluated using the ProTox-3.0 web server (Banerjee et al. 2018). Predicted endpoints included carcinogenicity, cytotoxicity, hepatotoxicity, mutagenicity, and immunotoxicity according to the Globally Harmonized System toxicity classification model.

Indole-fused oxazepines such as 3,10-dihydro-2H-1,3-oxazepino[7,6-b]indole have been reported to exhibit antibacterial activity by disrupting bacterial cell wall assembly, mainly by interfering with peptidoglycan biosynthesis (Akhmetova et al. 2022). Within this pathway, glycosyltransferase enzymes play a major role in cell wall synthesis by catalyzing the polymerization of glycan chains from lipid-linked precursors (Lipid II), establishing the backbone of the bacterial cell wall (Breukink and de Kruijff 2006; Egan et al. 2020). The fusion of the indole nucleus with a seven-membered oxazepine ring produces a distinct structure that can mimic or compete with natural bacterial substrates. This structural mimicry allows binding to penicillin-binding proteins (PBPs), which possess glycosyltransferase activity, thus preventing cross-linking and polymerization of the peptidoglycan layer (Sauvage et al. 2008; Sauvage and Terrak 2016). Inhibiting these enzymes causes osmotic instability and ultimately bacterial cell lysis.

Previous computational docking and in vitro studies indicate that the nitrogen and oxygen atoms within the oxazepine ring facilitate hydrogen bonding with the active sites of glycosyltransferase domains, particularly PBP1 and PBP2 (Onoabedje et al. 2016). In addition to cell wall inhibition, synthetic indole derivatives have been shown to disrupt membrane potential and reduce ATP pools, thereby reducing respiratory metabolism (Chandal et al. 2024). Furthermore, these derivatives interfere with exopolysaccharide production by inhibiting bacterial glucosyltransferases, reducing biofilm formation and enhancing susceptibility to host defenses (Teo and Roper 2015).

To ensure mammalian safety, toxicological evaluation of indole-fused scaffolds can be performed using in silico computational modeling. Quantitative structure–toxicity relationship analysis and SwissADME profiling can predict hepatotoxicity and mutagenicity, ensuring that derivatives fall within the “drug-like” space defined by Lipinski’s Rule of Five (Mo et al. 2024). Importantly, indole moieties are associated with antioxidant and cytoprotective activities, potentially mitigating oxidative stress in mammalian tissues by scavenging reactive oxygen species (Zeng et al. 2024).

To investigate the antibacterial potential of 3,10-dihydro-2H-1,3-oxazepino[7,6-b]indole derivatives, nine derivatives of this scaffold, with variations in the position and nature of the substituents on the aromatic ring, were tested against five Gram-positive and Gram-negative bacterial strains. These derivatives 1–9 (Fig. 1) were synthesized as previously described by Sammor et al. (2018).

Microdilution plates were evaluated for inhibition of bacterial growth by the tested derivatives; the results are shown in Fig. 2. The absence of bacterial growth and the change in resazurin color in the tested well indicate the antimicrobial activity of the synthesized derivatives.

As shown in Table 1, the validated assay findings were highly compatible with CLSI or EUCAST recommendations.

A significant increase in the bacterial count was detected after serial dilution of the tested derivatives, and no resazurin color change was detected at the MIC, as shown in Table 2, indicating the antibacterial activity of the derivatives.

The MICs and MBCs of derivatives 1–9 were evaluated (Table 2). All derivatives showed broad-spectrum activity in the concentration range of 0.0048–0.620 mg/mL. The lowest MIC among the tested derivatives was 0.0048 mg/mL for derivative 6 against S. typhimurium, which is lower than that of gentamicin.

These derivatives showed substantial variation in their ability to inhibit K. aerogenes. Derivative 2 showed low potency (MIC = 0.620 mg/mL), whereas derivatives 8 and 9 showed the greatest inhibitory effect (MIC = 0.08 mg/mL). All derivatives showed antibacterial activity against E. coli, with derivative 4 having the highest potency (MIC = 0.0136 mg/mL). In contrast, derivatives 7 and 9 showed no activity against S. typhimurium. Furthermore, derivatives 3 and 6 showed no inhibitory effects against E. faecalis.

The derivatives showed a weak correlation between their structure and potency, as shown by derivatives 2, 6, and 8 with a benzyl substituent at R2, whereas a greater inhibitory effect on B. subtilis was reported by Olgen et al. (2008) with the same benzyl substituent. In addition, derivatives 1, 5, and 9 have similar R1 and R2 substituents but still vary in potency. The variation in antibacterial activity observed in this study could be attributed to differences in chemical structure (Fig. 1), modifications and structural differences in the cell wall, and pathogen drug resistance due to genetic factors.

Derivatives 1, 4, 6, 8, and 9 displayed high potency against both Gram-negative and Gram-positive bacteria and could be considered potential broad-spectrum antibacterial agents for use individually or in combination with other antimicrobial agents. They could also have potential use as disinfectants.

The MIC and MBC values were analyzed statistically. No-inhibition (NI) values were treated as missing observations, not as zero. Because MIC and MBC data are dilution-based and positively skewed, geometric means and nonparametric statistics are more appropriate than arithmetic means alone.

The descriptive statistics by microorganism, shown in Table 3, indicate that the lowest overall geometric-mean MICs were observed against S. typhimurium and E. coli, whereas E. faecalis showed the highest geometric-mean MIC. The same pattern was generally observed for MBC values.

For the tested compound–organism combinations, the MBC was approximately five times the MIC. The median MBC/MIC ratio was 5.0 for every microorganism (Table 4).

This consistent ratio indicates that the compounds generally required approximately fivefold higher concentrations to produce a bactericidal effect than to inhibit visible growth. The ratio was 2.0 for gentamicin in most cases. Table 5 ranks the tested compound–organism activity by the geometric mean across the available microorganisms.

Among the synthesized derivatives, compound 6 had the lowest geometric-mean MIC and MBC, followed by compound 4. Compound 7 was the least active derivative, followed by compound 2. Compared with gentamicin, compound 6 required approximately 3.7-fold higher concentrations on a geometric-mean MIC basis and approximately 9.2-fold higher concentrations on an MBC basis. The other derivatives were substantially less potent than gentamicin.

Limited literature on oxazepine-containing antibacterial derivatives underscores the novelty of this scaffold. Although a recent review of indole–imidazole hybrids and indole C2 derivatives bearing 1,3,4-oxadiazole thioether moieties reported antibacterial activity against Gram-positive and Gram-negative strains, it largely relied on zone-of-inhibition or single-point MIC screening against standard strains (Bellapukonda et al. 2026).

The redocked ligand showed a binding pose similar to that of the PDB-bound native ligand (M4O), with approximately the same ring and chain orientations and some variability in angles. The redocked ligand interacted with the protein by forming five hydrogen bonds: OCG interacted with Glu83 (2.5 Å), Ser116 (2.6 Å), and Gln121 (2.5 Å); Ser116 interacted with OBG (2.4 Å); and O1E interacted with Lys124 (3.3 Å). The redocked ligand also interacted within the pocket via hydrophobic interactions with Glu83, Ala97, Arg100, Gly114, and Lys201. These interactions were similar to those of the native ligand (Fig. 3).

Similarly, all synthesized ligands were docked into the binding site and were found to dock well. The binding energy values for the poses of all synthesized derivatives are presented in Table 6.

Based on the antibacterial screening results, derivative 6 was identified as the most active derivative, and its binding pose is shown in Fig. 4. Derivative 6 showed good interaction within the protein pocket, where the benzene and pyridine rings, in addition to the indole heterocycle, interacted via hydrophobic interactions with Lys62, Tyr65, and Leu159.

From the antibacterial screening results, compound 6 was found to be the most active compound; its binding pose is presented in Fig. 4. This compound showed good interaction within the protein pocket: the benzene and pyridine rings, in addition to the indole heterocycle of the ligand, interacted via hydrophobic interactions with the amino acids Lys62, Tyr65, and Leu159.

The in silico predictive results obtained using CLC-Pred suggest that all tested derivatives 1–9 exhibit low-probability activity across the human cell-line panel. No derivative exceeded the 0.5 Pa threshold, the standard benchmark for likely biological toxicity.

The toxicological predictions indicated a favorable safety profile for all derivatives. The mean Pa values for all derivatives ranged from 0.219 to 0.253, indicating that chemical modifications did not introduce significant mammalian toxicity. Comparison between cell types revealed that normal cell lines showed lower average activity than tumor-derived lines, reinforcing selectivity.

Comparison between cell types revealed that normal cell lines (e.g., skin and lung fibroblasts) showed significantly lower average activity (Pa = 0.151) than tumor-derived lines (Pa = 0.249) (Table 7). This distinction reinforces the candidates’ selectivity. The use of in silico prediction at this stage is considered a validation step for selecting compounds with a high therapeutic index before proceeding to in vitro and in vivo testing (Lounkine et al. 2012).

The in silico cytotoxicity screening using CLC-Pred indicates that derivatives 1–9 possess a favorable preliminary safety profile. All derivatives exhibited a low probability of cytotoxic activity across the evaluated human cell lines, reflected by consistently low Pa values. Notably, derivatives assessed against a wider range of cell lines maintained comparable predictions, supporting the robustness of the safety assessment.

The absence of predicted cytotoxicity suggests that the oxazepino–indole scaffold and its substituent variations do not significantly interact with cellular pathways responsible for mammalian cell damage. From a drug-development perspective, this is particularly advantageous for antibacterial candidates, as it implies selective activity toward bacterial targets while minimizing the risk of host-cell toxicity.

Overall, the CLC-Pred results support the continued evaluation of this derivative series and justify further experimental validation using in vitro cytotoxicity assays to confirm the predicted safety margins.

The SwissADME analysis (Table 8) shows that the tested derivatives had molecular weights ranging from 449.46 to 624.87 g/mol. Most derivatives displayed high gastrointestinal absorption. Derivative 5 showed the most favorable drug-likeness profile, with a Log P value of 2.50 and zero violations of Lipinski’s rules.

The QSAT analysis suggests that all nine derivatives are safe at the cellular and genomic levels. All derivatives were predicted to be inactive for cytotoxicity, mutagenicity, and carcinogenicity (Table 8). This indicates that the observed antibacterial activity is not due to nonspecific host toxicity.