The R. felis proteome, comprising 1,393 sequences, was subjected to CD-HIT analysis, resulting in the exclusion of 333 paralogous sequences. To prevent cross-reactivity with human proteins, the remaining 1,060 non-paralogous sequences underwent a BLASTp analysis against the Homo sapiens proteome (78). This analysis identified 717 sequences with human homology, which were excluded, leaving 343 non-homologous proteins for further analysis.

Using DEG database, we identified 108 essential proteins in R. felis which are pivotal to the pathogen's life cycle and represent potential targets for antibacterial drug development. Additionally, using the KEGG database, we mapped 25 distinct metabolic pathways linked to these essential proteins. Identifying these pathways is critical as it provides insights into the metabolic dependencies, which could be exploited for therapeutic interventions. Supplementary Table S1 lists these specific pathways with their identifiers, while Supplementary Table S2 presents 11 unique proteins along with their corresponding metabolic pathways. To further assess their therapeutic potential, we performed a BLASTp search of these distinct vital proteins against the Drugbank database which revealed 10 proteins with substantial similarity to FDA-approved drug targets (Table 1). These proteins, which closely aligned with known FDA-approved drug targets, were selected for further investigation.

Using VFDB resource, we identified two proteins from a set of 10 being associated with the virulence of R. felis. These proteins are of particular interest as potential drug targets, especially in the context of combating drug-resistant pathogens, which present significant challenges in the treatment and may necessitate higher doses with increased risk of adverse effects (79). Targeting virulence factors is thus a promising strategy in drug development (80). Additionally, an analysis with the ARG-ANNOT V6 tool revealed that one of the virulent proteins succinate dehydrogenase, is involved in resistance mechanisms, such as drug degradation and efflux. Despite its association with resistance, succinate dehydrogenase remains a viable drug target due to its critical role in the pathogen's biology (81).

Our analysis found that 90% of proteins were cytoplasmic, with 5% in both the inner and outer membranes (Figure 2A). The identified target protein, succinate dehydrogenase, was cytoplasmic, indicating its potential as a hub protein due to its extensive interactions. Targeting succinate dehydrogenase could disrupt the function of other interacting proteins, underscoring its critical role in cellular processes (82). The protein-protein interaction (PPI) analysis showed the succinate dehydrogenase interaction with ubiquinol-cytochrome reductase (petA), NADH dehydrogenase I, chain I (nuoI), NADH dehydrogenase I, chain E (nuoE), succinyl-CoA synthase beta chain (SucC), succinyl CoA synthase alpha chain (SucD), fumarate hydratase (fumC), succinate dehydrogenase hydrophobic membrane protein (sdhD), pyruvate dehydrogenase E1 component (pdhB), succinate dehydrogenase cytochrome subunit (sdhC), and succinate dehydrogenase flavoprotein (sdhA). The PPI analysis of succinate dehydrogenase revealed a network with 11 nodes, an average node degree of 8, an average local clustering coefficient of 0.863, and a total of 44 edges (Figure 2B). The PPI enrichment p-value was calculated to be 6.51e⁻¹¹, indicating a significant interaction network, with an expected number of 14 edges. These proteins participate in various essential functions, suggesting that inhibiting succinate dehydrogenase could potentially disrupt the activity of other interacting proteins (Figure 2C) (83). This makes succinate dehydrogenase a promising candidate for a drug target.

The succinate dehydrogenase protein in R. felis presents a compelling target for drug development due to its vital role in the tricarboxylic acid (TCA) cycle and electron transport chain (ETC), both essential for ATP production (84). Succinate dehydrogenase (SDH) is a mitochondrial enzyme that plays a pivotal role in cellular metabolism by linking the tricarboxylic acid (TCA) cycle and the electron transport chain (ETC) (85). Within the TCA cycle, succinate dehydrogenase facilitates the oxidation of succinate into fumarate, while in the ETC, it reduces ubiquinone to ubiquinol, functioning as complex II. This enzyme facilitates electron transfer through its four subunits (SDHA, SDHB, SDHC, and SDHD), enabling the movement of electrons from FADH₂ to ubiquinone and subsequently to complex III, contributing to the production of adenosine triphosphate (ATP), the primary energy currency of the cell. The catalytic subunit SDHA, the largest component of the succinate dehydrogenase (SDH) complex, is responsible for oxidizing succinate into fumarate, producing FADH₂ in the process as part of the tricarboxylic acid (TCA) cycle (86). The SDHB subunit houses three iron–sulfur clusters that facilitate electron transfer from FADH₂ to the membrane-embedded subunits SDHC and SDHD. These latter subunits, situated in the inner mitochondrial membrane, form the electron transport chain's (ETC) complex II and serve as the binding and reduction site for ubiquinone (Q) to ubiquinol (QH₂) (87) (Figure 3). Therefore, its disruption can lead to impaired ATP synthesis, metabolic imbalances, and increased oxidative stress, ultimately weakening the bacterium's ability to thrive and cause infection. Given its critical function in both metabolism and virulence, succinate dehydrogenase emerges as a promising target for therapeutic interventions aimed at inhibiting the survival and proliferation of R. felis (88).

The protein 3D structure provided by AlphaFold2, includes some regions of disorder, which are indicated by low pLDDT values (89). In the resulting model, most residues exhibit very high confidence scores (pLDDT > 90), indicating the model's accuracy (Figure 4A). The dark green areas in the predicted aligned error plot signify high accuracy, while the light green areas suggest higher error rates (Figure 4B) (90). The structure was validated using the PROCHECK and ERRAT servers which revealed good quality of three-dimensional models (91). Ramachandran plot categorized residues into three regions: favored (86.3%), allowed (12.3%), and disallowed (0.4%) (Figure 4C). Additionally, the ERRAT server evaluated the model's quality based on statistical interactions among non-bonded atoms of different types. ERRAT scores around 92.7% indicate high-resolution structures as shown in Figure 4D.

The functional activity of a protein depends on the binding of a ligand to its active site (92). Identifying this binding site is crucial for understanding the protein's role and for drug discovery (93). MOE utilized the 3D template structure, which shares similarities with known ligand-binding sites, to identify potential binding pockets through geometric analysis (94). Multiple active sites were found, with the site having the highest energy profile and key functional residues selected (Figure 5A). Predicting protein ligands is challenging, as similar protein folds do not always imply similar functions (95). Conversely, proteins with different folds can exhibit similar biochemical properties, emphasizing that the binding site is crucial for protein function (96). The ProBiS revealed high sequence and structural homology with rhodoquinol-fumarate reductase complexed with Flavin-adenine dinucleotide (FAD) (PDB ID 3VR8) (Figure 5B). A lower binding energy, such as that observed here, suggests a stronger binding affinity, which often correlates with effective inhibition of enzyme activity.

Various bioinformatics tools are used for molecular docking in drug discovery (97). MOE module provides detailed graphical representations and ranks receptor-ligand binding affinities using the S-score, a measure of binding free energy (in kcal/mol), where a lower score indicates a more favorable binding (98). Docking of Flavin-adenine dinucleotide (FAD) to succinate dehydrogenase revealed five distinct conformations. Conformation 1, with a high binding energy of −8.47 kcal/mol, was selected for further analysis due to its superior binding affinity compared to the lowest score of −7.50 kcal/mol. This indicates that conformation 1 represents the most stable and potentially effective binding mode for FAD, supporting its role in the protein's function (Figure 6, Table 2) (99).

A comprehensive in silico screening of the 18,000-compound ZINC library was conducted against the active site of succinate dehydrogenase using an intensive docking approach. This process generated numerous docked conformations, ranked according to their docking scores. Compounds exhibiting lower binding affinities than the reference inhibitor, FAD with a binding energy threshold of −8.47 kcal/mol, were eliminated from further consideration as potential hit candidates. The screening identified over 11,000 compounds with binding energies surpassing that of FAD, ranging between −6.32 kcal/mol and −4.42 kcal/mol (Figure 7A, indicated in dark red). These results suggest that the inhibition of succinate dehydrogenase by these molecules could represent promising leads, as the favorable binding energies imply that these compounds form stable interactions with succinate dehydrogenase, which could effectively inhibit its function (100). Compounds demonstrating higher binding affinities than the FAD inhibitor were prioritized for further investigation due to their potent inhibitory potential against succinate dehydrogenase (Figure 7B). From this group, six compounds emerged as viable therapeutic candidates against succinate dehydrogenase in R. felis: ZINC67847806, ZINC67982856, ZINC67974679, ZINC67895371, ZINC05668040, and ZINC05670149 which were used for further analysis.

The selected compounds were subjected to post-docking interaction analysis to gain further insights into the pharmacological characteristics and binding dynamics of succinate dehydrogenase. Multiple interactions between each ligand and the receptor were observed during the molecular docking analysis. The docking rank order, based on binding scores, is as follows: ZINC67974679 > ZINC67895371 > ZINC67982856 > ZINC67847806 > ZINC05670149 > ZINC05668040. In the docking study, ZINC67974679 demonstrated a binding energy of −6.54 kcal/mol. The compound's five-aromatic ring facilitated the formation of one pi interaction with HIS118 achieved binding energy of −0.8 kcal/mol, and two H-donor interactions with TYR184 and ASP128, with bond lengths of 3.71 and 3.07 Å, and interaction energy (E) of −0.9 and −6.3 kcal/mol, respectively (Figure 8A). ZINC67895371 achieved a docking score of −6.40 kcal/mol. The six and five aromatic rings of this compound established a pi-hydrogen bond with MET119 and ASP189, characterized by a bond length of 3.78 Å, 3.95 and an energy of −0.6 and −0.7 kcal/mol (Figure 8B). ZINC67982856 was highly compatible with the succinate dehydrogenase binding site, displaying a binding score of −6.02 kcal/mol. The six aromatic rings engaged in one H-donor and a pi-H interaction with TYR184 and MET119, with bond lengths ranging from 3.64 to 4.43 Å and interaction energies between −0.7 and −0.9 kcal/mol (Figure 8C). ZINC67847806 formed one hydrogen bond as a donor with TYR184, resulting in a binding energy of −6.00 kcal/mol with bond lengths ranging from 3.97 to 3.75 Å and interaction energies between −0.7 and −0.8 kcal/mol (Figure 8D). Conversely, ZINC05670149 initiated two hydrogen bonds as donors from TYR184 and HIS118. This compound demonstrated a binding score of −5.27 kcal/mol (Figure 8E). Finally, ZINC05668040 docked with succinate dehydrogenase, producing a binding energy of −4.73 kcal/mol. The compound's aromatic ring mediated a single pi-hydrogen bond with the MET119, and three hydrogen bonds as acceptors (HIS131, PHE132), and one hydrogen bond as a donor with ASP128 residue (Figure 8F). Compounds with lower docking scores, such as ZINC67974679 and ZINC67895371, are expected to show increased stability within the binding site and thus greater inhibitory potential, supporting their prioritization for further investigation. The lower binding energies observed in these interactions suggest that these compounds form stable complexes with succinate dehydrogenase, potentially disrupting the enzyme's activity effectively (101). This correlation between docking score and inhibitory potential underscores these candidates' potential efficacy. The detailed binding interactions within succinate dehydrogenase active site for the selected compounds is presented in Table 3.

A pivotal aspect of drug discovery is the examination of ADMET properties, as it significantly reduces time and expenses during clinical trials (102). The pharmacokinetic evaluation revealed key insights into the drug-likeness, ADME characteristics, and blood-brain barrier (BBB) penetration of the selected compounds (103). The Lipinski Rule of Five (RO5) served as the foundation for determining drug-likeness, stipulating that drug-like molecules should possess a molecular weight under 500 Daltons, with a maximum of 10 hydrogen bond acceptors and <5 hydrogen bond donors and a logP (lipophilicity) value below 5 (104). Following these criteria, compounds ZINC20115475, ZINC02688148, and ZINC04259566 satisfied the RO5, whereas ZINC95543764, ZINC04232055, and ZINC04231816 exhibited a single violation each. Nonetheless, all six compounds were determined to lack BBB permeability, which evaluates the compound's ability to cross the protective blood-brain barrier to reach the central nervous system. The compounds adhering to drug-likeness criteria progressed to the subsequent phase of the study. For preliminary ADME property estimation, the pkCSM tool was utilized (105). This analysis encompassed solubility in pure water (mg/L), gastrointestinal absorption (HIA), which indicates the compound's potential for oral absorption in the intestine, permeability, inhibition of liver enzymes such as CYP 2C19, CYP 2C9, CYP 2D6, and CYP 3A4, as well as Caco-2 cell permeability, a model used to predict intestinal drug absorption and permeability through cell monolayers (106). The water solubility of these compounds ranged from −2 to −5, with ZINC02688148 exhibiting high solubility. Caco-2 permeability values spanned from −0.025 to 0.94, with ZINC04259566 demonstrating high cell permeability and ZINC95543764 the lowest. The high permeability observed for ZINC04259566 aligns with studies suggesting that compounds with permeability values exceeding 0.5 tend to exhibit favorable intestinal absorption profiles, making them promising candidates for oral drug delivery (77). Additionally, all compounds displayed good potent HIA permeability, leading to the recommendation of ZINC67982856 for further experimental validation (Table 4). Moreover, an Ames mutagenicity test conducted using the pkCSM tool assessed the potential toxicity of these compounds, including maximum tolerated dose (human), minnow toxicity, T. pyriformis toxicity, oral rat acute toxicity (LD50), hepatotoxicity, and skin sensitization (107). The use of T. pyriformis as a model organism in toxicity studies is due to its sensitivity to chemical compounds, providing valuable insights into aquatic toxicity (106). All compounds except ZINC20115475 yielded a negative Ames test, indicating a lack of mutagenic potential. The LD50 values ranged from 2.8 to 4.3 mol/kg, with ZINC02688148 exhibiting the highest value, indicating lower acute oral toxicity in rats. These results are consistent with other studies where compounds with higher LD50 values demonstrated greater safety margins, particularly for therapeutic agents intended for prolonged use (108). The data for oral rat acute toxicity (LD50), hepatotoxicity, and skin sensitization were predicted computationally using the pkCSM tool, which evaluates toxicity based on chemical structure and quantitative structure-activity relationships (QSAR). Given that a chemical's toxicity is often predicted based on its molecular structure, ZINC20115475 was excluded from further studies due to its positive result in the Ames test, suggesting possible mutagenicity. While ZINC02688148, ZINC04259566, ZINC04232055, and ZINC04231816 were found to be hepatotoxic, none of the compounds induced skin sensitization. The hepatotoxic compounds were not considered for future investigation, as previous study indicate the dose-dependent liver toxicity profiles and mitigate potential risks (109). T. pyriformis exhibited maximum tolerance to ZINC20115475 (0.367 log μg/L), whereas the remaining compounds were less tolerated. Detailed information can be found in Table 5.