Harnessing subtractive genomics for drug target identification in Streptococcus agalactiae serotype v (atcc baa-611 / 2603 v/r) strain: An in-silico approach

Abstract

Developing a therapeutic target for bacterial disease is challenging. In silico subtractive genomics methodology offer a promising alternative to traditional drug discovery methods. Streptococcus agalactiae infections depend on two crucial criteria: drug-resistance and the existence of virulence factors. It is essential to underline that S. agalactiae strains have emerged to be resistant to several drugs. Hence, there is a need for research on novel drugs and techniques that are potent, economical, productive, and dependable to combat S. agalactiae infections. In this study advanced computational techniques were exploited to examine potential druggable targets exclusive to this pathogen. Our study uncovered 200 non-homologous proteins in S. agalactiae serotype V (Strain ATCC BAA-611/ 2603 V/R) and identified 68 essential proteins indispensable for the bacterium’s survival. Therefore, these 68 proteins are potential targets for drug development. Subcellular localization analysis unveiled that the pathogen’s cytoplasmic membrane contained essential proteins among these vital non-homologous proteins. On the other hand, based on virulent protein predictions, six proteins were seen to be virulent. Among these, we prioritized two proteins (Sensor protein LytS and Galactosyl transferase CpsE which are exclusively found in S. agalactiae) as potential druggable targets and selected them for further structural investigation. The proteins chosen could serve as a foundation for the identification of a promising therapeutic compound that has the potential to neutralize these enzymatic proteins, thereby contributing to the reduction of risks linked to the drug-resistant S. agalactiae.

Introduction

S. agalactiae, is a gram-positive, β-haemolytic coccus, commonly known as a group B streptococcus or GBS. This encapsulated facultative anaerobe is catalase-negative and features a group B antigen in its thick peptidoglycan cell wall, with a tendency to form chains [1,2]. It is commonly found as a normal part of the body’s flora in the urogenital and lower gastrointestinal tracts, and may also be present in the oropharynx [3–7]. Based on the composition of the capsular polysaccharide, a key virulence factor, this bacterium is classified into 10 serotypes (Ia, Ib, II-IX) [1,8–10].

In 1935, Rebecca Lancefield was the first to describe the colonization of the vagina by S. agalactiae and the first description of the bacterium as a human pathogen and its role in causing invasive illness was documented in 1964 [11,12]. It is mostly responsible for multiple infections including sepsis, meningitis, soft-tissue infection, urinary tract infection, pneumonia, and premature delivery [2,13–15]. Newborns of 24–48 hours old are more susceptible (approximately 33.3%) to this pathogen compared to those of older than 48 hours (only 8%) [16]. Among adults, this bacterium has the potential to induce peripartum chorioamnionitis and bacteremia in new mothers (occurring in 0.03% of cases) as well as endocarditis and osteomyelitis in other individuals [2,17]. Individuals who are immunocompromised, elderly, or have conditions such as diabetes mellitus, alcoholism, cancer, cirrhosis, a history of stroke, or HIV/AIDS are at increased risk of Group B streptococcal infection [2,17,18]. The pathogen continues to be a primary cause of newborn sepsis, resulting in around 90,000 fatalities in early infancy and at least 57,000 stillbirths worldwide [19].

Among the 10 identified serotypes, serotype V is recognized as one of the most clinically significant strains of Streptococcus agalactiae [20,21]. Its rising importance in clinical settings can be attributed to its evolving epidemiology, increasing virulence, and the development of antibiotic resistance. This serotype poses a particular threat in both neonatal and non-neonatal infections, as well as systemic conditions such as sepsis, meningitis, and pneumonia, where it is often linked to more severe clinical outcomes compared to other serotypes. Despite global similarities in S. agalactiae colonization rates, the prevalence of specific serotypes varies significantly across different regions [22]. Evidence suggests that serotype V has become dominant in several countries, including Algeria, France, Poland, Italy, Thailand, Brazil, Portugal, and the United States, although serotype III remains the most prevalent worldwide [23–30]. Studies also show that serotype III is more common in children, while serotype V predominates in adults and the elderly, often leading to both invasive and non-invasive infections [31]. The virulence of serotype V is largely attributed to its capsular polysaccharide (CPS), which plays a crucial role in its ability to evade the host’s immune system [32]. Serotype V strains, when compared to other serotypes such as I and III, have been shown to possess distinct virulence factors, including enhanced pili formation and biofilm production, which are key to their persistence and invasiveness [33–35].

A growing concern is antibiotic resistance in Gram-positive bacteria, including Streptococcus and strains of Staphylococcus. Treatment options are complicated by mechanisms such as β-lactamase production, efflux pumps, and target site modifications that reduce susceptibility to β-lactams, macrolides, and glycopeptides [36–38]. Additionally, in GBS, resistance mechanisms are driven by genetic mutations, horizontal gene transfer, and the acquisition of resistance genes from other bacterial species, which is often compounded by the widespread use of broad-spectrum antibiotics [34,35]. Studies have revealed that Streptococcus agalactiae strains exhibit enhanced resistance to a range of drugs, including erythromycin, clindamycin, tetracycline, fluoroquinolones, ampicillin, levofloxacin, cefotaxime, chloramphenicol, and vancomycin [20,23,26,27,29,39]. In some cases, resistance to penicillin, amoxicillin, ceftazidime, and piperacillin has also been observed in S. agalactiae, despite penicillin being the recommended first-line treatment for invasive GBS infections [40–43]. Given the escalating resistance of S. agalactiae to conventional antibiotics, identifying novel drug targets for S. agalactiae serotype V is crucial for the effective prevention and management of infections.

The exceptional advancements in computational biotechnology and bioinformatics have significantly impacted drug design, leading to reduced expense and duration for traditional laboratory trials by facilitating the discovery of therapeutic candidates, structure-based drug development, and the identification of host-specific targets through genomic data analysis. A key focus has been on the subtractive genomics technique, which aims to identify proteins with therapeutic potential exclusive to pathogenic genomes while excluding similar host proteins. This methodology has successfully identified novel species-specific therapeutic targets across various pathogenic strains including [44–63].

In this study, we aim to develop a potential therapeutic target to combat S. agalactiae serotype V through the application of sophisticated computational biotechnology and bioinformatics approaches, addressing the pressing global issue of drug resistance. Thus, in order to analyze the proteome of S. agalactiae serotype V in its entirety, the subtractive genomics technique was applied. Essential proteins for pathogen’s survival were prioritized using computational tools, followed by the removal of host homologous proteins to minimize therapeutic interference. The remaining pathogenic proteins were examined for subcellular localization and virulent capabilities, leading to the recognition of cytoplasmic membrane proteins and their virulence potential. Two promising proteins, Sensor protein LytS and Galactosyl transferase CpsE, were identified as potential therapeutic targets which are crucial for the virulence and pathogenicity of S. agalactiae serotype V. LytS regulates cell wall synthesis, stress response, and bacterial survival by maintaining cell wall integrity and mediating resistance to immune responses and antibiotics, whereas CpsE facilitates CPS biosynthesis, enhancing immune evasion by preventing phagocytosis and complement-mediated killing [32,64]. Together, these proteins contribute to the bacterium’s ability to cause severe infections, especially in neonates and pregnant women, by promoting tissue invasion and immune resistance. These enzymatic proteins underwent further protein network analysis and structural investigation, highlighting that they could potentially have the promise to serve as prospective selections for the development of vaccines or drugs that specifically target S. agalactiae serotype V.

Materials and methods

The comprehensive protocol for discovering proteins that are essential and unique to S. agalactiae serotype V (Strain ATCC BAA-611/2603V/R) for potential drug target identification was achieved by recognizing non-human homolog proteins, their essentiality in the viability of the pathogen, their involvement in important pathogen-specific metabolic pathways, their sub-cellular localization, virulence, druggability, along with structural studies is outlined in Fig 1.

Fig 1. Flowchart for identifying potential drug targets in S. agalactiae serotype V.

This shows the step-by-step process for identifying drug targets through the subtractive genomics approach.

Protein sequence retrieval

The whole proteome of S. agalactiae serotype V (Strain ATCC BAA-611/2603V/R) was retrieved from UniProt database [65–67] along with its peer-reviewed proteins in FASTA format.

Finding non-homologous proteins

To distinguish proteins that are non-homologous to the host, a BLASTp search of the NCBI database (with e-value threshold of 10⁻⁴) was conducted against Homo sapiens on the selected peer-reviewed proteins of the pathogen [68]. The non-homologous protein sequences obtained were retrieved for further analysis and the rest having significant similarities with the host were excluded.

Essential protein screening

Accordingly, the essential proteins among the non-homologous proteins were identified utilizing a BLASTp search (with a threshold e value of 10^-100) against the Database of Essential Genes and protein sequences demonstrating notable similarity with the DEG database version 15.2 were categorized as being vital for the survival of the pathogen and were selected for subsequent analysis [69–72].

Metabolic pathway exploration

The association of the chosen essential proteins in diverse metabolic pathways of the host and the pathogen was achieved by KEGG automated annotation server (KAAS) [73]. The three letter organism codes ‘sag’, ‘san’, and ‘sak’ for S. agalactiae and ‘hsa’ for H. sapiens were selected, and employing the BHH method in the KAAS server, the metabolic pathways were collected independently and compared.

Subcellular localization prediction

Subcellular location of the chosen essential proteins was initially performed by the pSORTbv3.03 tool [74,75] and cross-verified by CELLOv2.5 [76,77]. Both tools provide accurate predictions for various subcellular locations, encompassing proteins found in the cytoplasm, cytoplasmic membrane, extracellular space, cell wall, and those with unknown origins [74–77].

Virulent protein screening

Accordingly, virulent proteins from the shortlisted essential proteins were retrieved from the tool VirulentPred2.0 [78]. The tool utilizes up-to-date datasets and advanced machine learning techniques to forecast the virulence status of proteins based on their PSSM profile [78].

Determination of physiochemical parameters

The ProtParam server of ExPaSy was used to determine the physicochemical properties of the selected proteins [79].

Analysis of protein interaction networks

The potential interactions of the target proteins, with other proteins in S. agalactiae were predicted by the STRING database version 12.0 [80]. To prevent false-positive results, protein-protein interactions (PPI) studies were only considered if they had a high confidence score of 70% (0.700), choosing proteins with close interactions with at least three other proteins for further investigation.

Druggability analysis

A BLASTp search with a threshold expectation value of 10⁻⁵ was performed on the shortlisted proteins against the DrugBank 6.0 database to evaluate their druggability [81]. Freely accessible, the DrugBank 6.0 database serves as a comprehensive online resource, providing extensive data on drugs and their respective targets [81].

Prediction and assessment of three-dimensional structures

The three-dimensional structures of the target proteins were initially generated by the Swiss-Model online tool [82]. The tool uses homology-based approaches to construct 3D structures of proteins from amino acid sequences [82]. The 3D models of the selected proteins were constructed from template models that exhibited the most Global Mean Quality Estimate (GMQE) score, coverage, sequence similarity, range, and identity percentage [83–85] and was further validated by Swiss-Model structure validation (which included MolProbity scores, clash scores, and Ramachandran Plots) [86,87]. Moreover, we also utilized the SAVESv6.1 server (which included PROCHECK Ramachandran Plots and ERRAT scores) and the ProSAweb server (for Z-scores) to further verify and ensure the structural accuracy and quality of the constructed 3D structures [88–90].

Active site prediction

We employed the DoGSiteScorer to effectively pinpoint potential active sites on the modelled protein structures where specific ligands could efficiently bind to it and modify its functions [91,92]. The server effectively identifies the present active site pockets by evaluating the physicochemical properties of the protein residues.

Identification of ligands

Given that ligand information for the modelled proteins was unavailable, the ProBis server was employed to anticipate potential ligands and their interactions [93]. A theoretical methodology is employed by this server, which uses molecular simulations to anticipate the most suitable protein-ligand combinations. These results were subsequently visualized with PyMOL v3.1.

Molecular docking studies

The AutoDock Vina in PyRx version 0.8 software was employed to dock the identified ligands with the modeled proteins serving as targets [94]. Using BIOVIA Discovery Studio version 4.5, the interactions between the ligands and the modeled proteins were then visualized.

Results

Our study aimed to identify potential drug targets to combat the S. agalactiae serotype V strain, meeting the efficacy criteria of drug targets. The efficacy criteria include the targets must be non-homologous to humans, essential for the organism, and integral to the microbe’s main metabolic processes. Furthermore, membrane potential, virulence connections, and druggability analyses were also taken into account for strengthening the selection criteria of the target proteins. The subtractive genomic analysis process is briefly presented in Table 1.

Table 1. Brief presentation of the subtractive genomic analysis of S. agalactiae serotype V.

Sl No.	Subtractive Approaches	Bioinformatics Tools and Servers Utilized	Number of Proteins
1	The whole proteome of S. agalactiae serotype V (Strain ATCC BAA-611/ 2603 V/R)	UniProt	2105
2	Peer-Reviewed Proteins of S. agalactiae serotype V (Strain ATCC BAA-611/ 2603 V/R)	UniProt	390
3	Proteins nonhomologous to H. sapiens	BLASTp (E-value 10−4)	200
4	Essential proteins	DEG server (E-value ≤ 10–100)	68
5	Essential Proteins involved only in unique metabolic pathways	KAAS at KEGG	51
6	Proteins assigned KO (KEGG Orthology) but not in any pathway	KEGG Orthology	15
7	Essential membrane proteins	PSORTb	12
8	Essential membrane proteins	CELLO	8
9	Essential virulent proteins	VirulentPred2.0	6
10	Essential membrane and virulent proteins	PSORTb, CELLO, VirulentPred2.0	2

Protein sequence retrieval

The total proteins available in the reference proteome of this strain in UniProt was 2105, among which 390 protein sequences were found to be Swiss-Prot (Peer-reviewed) proteins Fig 2. The study chose exclusively peer-reviewed sequences to avoid overrepresentation of specific proteins and computational difficulties, ensuring the accuracy and representativeness of the resulting data, thereby reducing redundancy within the proteome. For further analysis, all peer-reviewed proteins of the pathogen were collected in FASTA format, while excluding the non-reviewed proteins.

Fig 2. Total proteins of S. agalactiae serotype V.

This shows the protein counts of non-reviewed and selected peer-reviewed sequences.

Finding non-homologous proteins

Proteins that participate in various fundamental cellular systems have been identified as homologous, exhibiting analogous functions in both humans and bacteria throughout the course of evolution [95,96]. To identify non-homologous proteins, a BLASTp search of the NCBI database (with e-value threshold of 10⁻⁴) was conducted against Homo sapiens on 390 peer-reviewed proteins of the pathogen. Out of the 390 sequences, 200 were identified as non-homologous proteins, while the rest had resemblance to humans. We excluded the remaining proteins because cytotoxic reactions and adverse effects might arise from drug targets that are similar to the host genome and the aim of targeting such homologous proteins might result in detrimental consequences [97,98]. While homologous protein sequences were excluded, these 200 non-homologous proteins were chosen for further analysis.

Finding essential proteins

To identify essential proteins of the pathogen, a BLASTp search against the database of DEG (with an e-value cut off at 10^-100) was performed. Furthermore, to prevent false positive results, a manual cross-checking of the biological function of each query target protein was performed against its corresponding outcome, revealing that their functions are consistent. A total of 68 proteins (S1 Table) out of 200 were identified as crucial to the bacterium’s survival and were chosen for subsequent steps.

Metabolic pathway exploration

The KAAS server revealed that 51 out of 68 essential proteins are associated with 41 distinct pathogen-specific metabolic pathways, with no proteins involved in host pathways. Moreover, no common pathways between the host and the pathogen were observed, leading to the conclusion that the selected non-homologous proteins showed no involvement in the host pathways. The absence of shared pathways between the pathogen and host suggests distinct metabolic systems, making pathogen-specific pathways ideal targets for drug development. Focusing on these unique pathways can effectively inhibit the pathogen without harming the host, reducing toxicity and improving the safety and efficacy of treatments. Furthermore, to manage and prevent false positives, stringent filtering criteria were applied such as essential proteins (e-value cutoff of 10 ⁻ ¹⁰⁰) and non-homologous (BLASTp e-value < 10 ⁻ ⁴). Functional annotation was performed using KO identifiers, ensuring accurate pathway mapping and manual comparison of pathogen vs. host pathways further evaluated and validated the metabolic pathways. Moreover, all 68 essential proteins had KEGG Orthology (KO) identifiers (S2 Table) by the KAAS server at KEGG, except for two proteins (namely, Bis(5’-nucleosyl)-tetraphosphatase (symmetrical) and galactosyl transferase CpsE). However, it is important to note that the absence of KO annotation for these two proteins does not imply false positives as Galactosyl transferase CpsE is a well-documented enzyme involved in capsular polysaccharide biosynthesis, a key factor in bacterial virulence and immune evasion [32], whereas Bis(5’-nucleosyl)-tetraphosphatase (symmetrical) plays a critical role in nucleotide metabolism and signaling by hydrolyzing dinucleoside polyphosphates, molecules known to regulate bacterial stress responses [99]. Limited KO database coverage for these proteins likely indicates annotation gaps rather than errors in target identification, and their essential, pathogen-specific roles validate their inclusion as targets. The unique metabolic pathways are listed in Table 2.

Table 2. Metabolic pathways unique to S. agalactiae serotype V.

Sl. No	Unique Pathways	Total Proteins	Pathway ID	Proteins Involved
1	Pentose phosphate pathway	1	00030	DEOB_STRA5
2	Amino sugar and nucleotide sugar metabolism	2	00520	MURA_STRA5, MURB_STRA5
3	Pyruvate metabolism	2	00620	ACKA_STRA5, CAPP_STRA5
4	Propanoate metabolism	1	00640	ACKA_STRA5
5	Oxidative phosphorylation	2	00190	ATPD_STRA5, PPAC_STRA5
6	Photosynthesis	1	00195	ATPD_STRA5
7	Carbon fixation by Calvin cycle	1	00710	CAPP_STRA5
8	Other carbon fixation pathways	2	00720	CAPP_STRA5, ACKA_STRA5
9	Methane metabolism	2	00680	CAPP_STRA5, ACKA_STRA5
10	Fatty acid biosynthesis	2	00061	FABH_STRA5, FABZ_STRA5
11	Glycerolipid metabolism	2	00561	PLSX_STRA5, PLSY_STRA5
12	Glycerophospholipid metabolism	1	00564	PLSY_STRA5
13	Purine metabolism	1	00230	DEOB_STRA5
14	Pyrimidine metabolism	3	00240	PYRH_STRA5, KCY_STRA5, KTHY_STRA5
15	Glycine, serine and threonine metabolism	1	00260	KHSE_STRA5
16	Cysteine and methionine metabolism	1	00270	KHSE_STRA5
17	Phenylalanine, tyrosine and tryptophan biosynthesis	4	00400	AROB_STRA5, AROD_STRA5, AROA_STRA5, AROC_STRA5
18	Taurine and hypotaurine metabolism	1	00430	ACKA_STRA5
19	D-Amino acid metabolism	4	00470	ALR_STRA5, DDL_STRA5, MURI_STRA5, MURD_STRA5
20	Peptidoglycan biosynthesis	7	00550	MURA_STRA5, MURB_STRA5, MURC_STRA5, MURD_STRA5, DDL_STRA5, UPPP_STRA5, MURG_STRA5
21	Teichoic acid biosynthesis	1	00552	UPPP_STRA5
22	Thiamine metabolism	1	00730	RSGA_STRA5
23	Nicotinate and nicotinamide metabolism	1	00760	NADE_STRA5
24	Pantothenate and CoA biosynthesis	1	00770	COAD_STRA5
25	Biotin metabolism	1	00780	FABZ_STRA5
26	RNA polymerase	1	03020	RPOA_STRA5
27	Ribosome	3	03010	RS3_STRA5, RS4_STRA5, RL10_STRA5
28	Aminoacyl-tRNA biosynthesis	2	00970	SYGA_STRA5, SYGB_STRA5
29	Protein export	1	03060	SECA_STRA5
30	RNA degradation	1	03018	RNY_STRA5
31	DNA replication	1	03030	RNH3_STRA5
32	Nucleotide excision repair	1	03420	UVRC_STRA5
33	Mismatch repair	1	03430	EX7L_STRA5
34	Homologous recombination	5	03440	RECF_STRA5, RECO_STRA5, RECR_STRA5, RECA_STRA5, RUVA_STRA5
35	ABC transporters	1	02010	PSTS1_STRA5
36	Bacterial secretion system	1	03070	SECA_STRA5
37	Two-component system	3	02020	PSTS1_STRA5, DNAA_STRA5, LYTS_STRA5
38	Cell cycle – Caulobacter	2	04112	DNAA_STRA5, MURG_STRA5
39	Quorum sensing	1	02024	SECA_STRA5
40	Tuberculosis	1	05152	PSTS1_STRA5
41	Vancomycin resistance	3	01502	DDL_STRA5, ALR_STRA5, MURG_STRA5

Subcellular localization prediction

The pSORTbv3.0.3 tool predicted eight cytoplasmic membrane proteins, 51 cytoplasmic proteins, and nine proteins with uncertain or unknown subcellular localization from the shortlisted essential proteins. Subsequently, CELLOv2.5 predicted 12 membrane proteins, 54 proteins as cytoplasmic, and two extracellular proteins. However, both CELLOv2.5 as well as pSORTbv3.03 did not exhibit any proteins belonging in the cell wall region (S3 Table). The predicted results of the subcellular localization of essential proteins of S. agalactiae are presented in Fig 3. Out of the 8 and 12 cytoplasmic membrane proteins predicted by pSORTbv3.03 and CELLOv2.5, respectively, five proteins were predicted by both tools (Table 3).

Fig 3. Essential, non-homologous proteins and their subcellular localization by pSORTbv3.03 and CELLOv2.5.

This shows the protein counts of sequences predicted by subcellular localization prediction.

Table 3. List of cytoplasmic membrane proteins identified by both pSORTbv3.03 and CELLOv2.5.

Sl No.	Gene name	Proteins	UniProt ID	pSORTb v3.03	CELLO v2.5
1	plsY	Glycerol-3-phosphate acyltransferase	P67167	Membrane	Membrane
2	ftsK	DNA translocase FtsK	Q8CX05	Membrane	Membrane
3	uppP	Undecaprenyl-diphosphatase	Q8E260	Membrane	Membrane
4	lytS	Sensor protein LytS/ Sensor histidine kinase LytS	Q8E218	Membrane	Membrane
5	cpsE	Galactosyl transferase CpsE	Q9AFI0	Membrane	Membrane

Prediction of virulent proteins

Virulent factors in pathogens facilitate bacterial adhesion, colonization, invasion, and disease pathogenesis [100,101]. The identification of virulent proteins, among 68 essential non homologous proteins, using VirulentPred2.0 revealed six proteins that are listed in Table 4. Inhibition of these proteins could significantly impact the pathogen’s functionality within the host organism. Among the six predicted proteins, LytS and CpsE were found to be both virulent and present in cytoplasmic membrane which implies that these two proteins are potential drug targets Table 4 (Bold and Italic).

Table 4. Essential, non-homologous proteins predicted to be virulent by VirulentPred2.0.

Sl No.	Gene name	Proteins	UniProt ID	KO ID	VirulentPred2.0
1	hslO	33 kDa chaperonin	P64402	K04083	Virulent
2	ruvA	Holliday junction branch migration complex subunit RuvA	Q8DWW6	K03550	Virulent
3	pstS	Phosphate-binding protein PstS	Q8DZV4	K02040	Virulent
4	lytS	Sensor protein LytS	Q8E218	K07704	Virulent
5	recO	DNA repair protein RecO	Q8E2G8	K03584	Virulent
6	cpsE	Galactosyl transferase CpsE	Q9AFI0	–	Virulent

Physiochemical property analysis

The physiochemical properties of the selected two proteins, namely Sensor protein LytS and Galactosyl transferase CpsE, were analyzed by ProtParam tool. The analysis showed that both proteins are slightly hydrophobic, basic, stable, and highly thermostable in nature. It is important to note that Galactosyl transferase CpsE is more basic in nature than Sensor Protein LytS. The detailed results of physiochemical property analysis of two proteins are presented in Table 5.

Table 5. ProtParam analysis of Sensor protein LytS and Galactosyl transferase CpsE.

Physicochemical Properties	Sensor protein LytS	Galactosyl transferase CpsE
Amino acids	581	449
Molecular weight	64548.62	52364.13
Theoretical pI	7.71	9.34
Instability Index	35.59	33.46
Aliphatic Index	107.01	103.94
Grand average of hydropathicity (GRAVY)	0.075	0.059
Estimated half-life in Mammalian reticulocytes, in-vitro	30 hours	30 hours
Estimated half-life in Yeast in-vivo	>20 hours	>20 hours
Estimated half-life in Escherichia coli, in-vivo	>10 hours	>10 hours

Analysis of protein interaction networks

The significant association of these two chosen proteins with other proteins in the pathogen was identified using the STRING database, and the amino acid sequences of Sensor protein LytS and Galactosyl transferase CpsE was uploaded to the server. The Sensor protein LytS developed 4 PPI networks (Fig 4A) depicted as lytS in red node. LytS had 4 edges, 3 anticipated edges, 4 nodes, with a 2.0 average node degree. Protein-Protein Interaction enrichment p-value was 0.331, with 0.833 average local clustering coefficient. Moreover, it interacted with two neighbouring two-component system response regulator proteins (lytR and SAG1016) and one neighbouring conserved hypothetical protein (SAG0184). On the other hand, Galactosyl transferase CpsE had 20 PPI networks (Fig 4B) represented as cpsE in red node. There were around 20 nodes, 125 edges, with an average node degree of 12.5, a clustering coefficient of 0.888, 21 predicted edges, and a PPI enrichment p-value of more than 1.0e-16. Furthermore, it interacted with numerous neighbouring proteins mostly involved in the assembly, regulation and biosynthesis of the GBS capsular polysaccharide (cpsD, cpsC, cpsB, cpsL, cpsA, cpsJ, cpsG, cpsH, cpsM, cpsO, cpsN, cpsK, cpsF) along with different other neighbouring important enzymes (neuC, neuD, neuA, neuB, etc).

Fig 4. Prediction of protein-protein interactions of the two proteins.

The chosen proteins are represented in red nodes. Each node represents all the proteins produced by a single, protein-coding gene locus. Empty nodes represent proteins of unknown 3D structure and filled nodes indicate that a 3D structure is known or predicted. Edges represent protein-protein associations. (A) Sensor protein LytS protein. (B) Galactosyl transferase CpsE protein.

Druggability analysis

Binding of a small molecule, such as a drug or vaccine, to a protein can induce a functional change, thereby converting the protein into a “druggable target.” This transformation typically alters the protein’s activity, resulting in therapeutic effects that enhance the health or condition of the host organism [102]. Therefore, a BLASTp analysis was performed on the two selected proteins using an e-value threshold of 10⁻⁵, against the DrugBank 6.0 database. The results revealed that both LytS and CpsE were recognized as potential novel drug targets, as no known compounds targeting these proteins were found within the DrugBank database.

Prediction and assessment of three-dimensional structures

Understanding protein functions, ligand interactions, and dynamic behaviors relies heavily on the prediction of protein tertiary structures [103]. Thus, the online server Swiss-Model generated fifty initial templates for each protein, and the templates with the most GMQE score (LytS = 0.89, CpsE = 0.83), coverage (LytS = 1.00, CpsE = 1.00), sequence similarity (LytS = 0.55, CpsE = 0.61), range (LytS = 1–581, CpsE = 1–449) and identity percentage (LytS = 81.9, CpsE = 99.55) were chosen to construct the 3D models. Moreover, the predicted structures of Sensor protein LytS and Galactosyl transferase CpsE had 0.80 and 1.06 MolProbity scores, clash scores of 0.65 and 0.93, with 97.58% and 95.97% of residues in the favored regions of the Ramachandran plot respectively, according to the Swiss-Model structure validation (Fig 5). In addition, PROCHECK analysis revealed that the residues in the favored region of the Ramachandran plot of LytS and CpsE was 92.1% and 93.3%, respectively (Fig 5). The ERRAT scores of 98.025 for LytS and 93.253 for CpsE (Fig 6) further confirmed the good quality of the structures. Furthermore, the Z-scores of the proteins were −9.1 for LytS and −6.29 for CpsE (Fig 6). The accuracy and quality of the constructed 3D structures for the selected proteins were confirmed, suggesting that the predicted models conform to the typical structural range observed in NMR and X-ray crystallography data.

Fig 5. Determination of the three-dimensional structures.

(A) Three-dimensional structure of Sensor protein LytS with Ramachandran plots generated by (B) Swiss-Model Structure Validation and (C) PROCHECK. (D) Three-dimensional structure of Galactosyl transferase CpsE with Ramachandran plots generated by (B) Swiss-Model Structure Validation and (C) PROCHECK.

Fig 6. Structural quality assessment of the two target protein structures.

(A) Overall quality assessment, (B) local quality assessment, and (C) ERRAT scores of the Sensor protein LytS structure. (D) Overall quality assessment, (E) local quality assessment, and (F) ERRAT scores of the Galactosyl transferase CpsE structure. Panels (A) and (D) display Z-scores representing the compatibility of the predicted 3D structures with known protein models of similar sizes. Panels (B) and (E) show per-residue knowledge-based energy profiles highlighting potential structural inaccuracies. Panels (C) and (F) present ERRAT scores evaluating model reliability based on non-bonded atom-atom interactions, where higher scores indicate better quality.

Active site analysis

After modeling the protein structures, it was essential to identify a suitable binding interface to facilitate ligand binding. The DoGSiteScorer tool subsequently identified 19 potential binding pockets for the Sensor protein LytS, and the site exhibiting the highest drug score of 0.81 was favored. In a similar manner, 19 potential binding pockets were identified for Galactosyl transferase CpsE, and the pocket with the highest drug score of 0.81 was favored for additional analysis. S4 Table provides detailed information on the amino acid residues within the selected binding pockets of both LytS and CpsE, while the binding sites of both proteins are visually represented in Fig 7.

Fig 7. Active sites of the two target proteins.

The active sites of the two target proteins, (A) Sensor protein LytS and (B) Galactosyl transferase CpsE, are highlighted in yellow, their corresponding amino acid residues are magnified in the inset, and their protein backbones are depicted as blue ribbons.

Identification of ligands

Protein binding site identification and their respective ligands plays a critical role in drug discovery and pharmaceutical research. These sites are crucial both structurally and functionally, serving as regions where different drug molecules bind to induce the desired biological response [93]. Exploiting the ProBis server, the ligand 128, officially named as Spiro(2,4,6-trinitrobenzene[1,2A]-2O’,3O’-methylene-adenine-triphosphate with the IUPAC designation (3aR,4R,6R,6aR)-4-(6-aminopurin-9-yl)-N-hydroxy-6-[[hydroxy-[hydroxy(phosphonooxy)phosphoryl]oxyphosphoryl]oxymethyl]-3’,5’-dinitrospiro[3a,4,6,6a-tetrahydrofuro[3,4-d][1,3]dioxole-2,4’-cyclohexa-2,5-diene]-1’-imine oxide, was identified in connection with the Sensor protein LytS. The ligand 128 was attained from a template with the PDB ID: 1I5D (from Thermotoga maritima). Whereas for Galactosyl transferase CpsE, the ligand HEM, formally named as Protoporphyrin-IX-containing-Fe with the IUPAC designation 3-[18-(2-carboxyethyl)-8,13-bis(ethenyl)-3,7,12,17-tetramethyl-23H-porphyrin-21-id-2-yl]propanoate;iron(2+), was attained from a template with the PDB ID: 2WDQ (from Escherichia coli). The structural representations of the two identified ligands are presented in Fig 8.

Fig 8. Ligands identified for the two selected drug targets.

(A) Spiro(2,4,6-trinitrobenzene[1,2A]-2O’,3O’-methylene-adenine-triphosphate (Ligand 128) with assigned IUPAC name (3aR,4R,6R,6aR)-4-(6-aminopurin-9-yl)-N-hydroxy-6-[[hydroxy-[hydroxy(phosphonooxy)phosphoryl]oxyphosphoryl]oxymethyl]-3’,5’-dinitrospiro[3a,4,6,6a-tetrahydrofuro[3,4-d][1,3]dioxole-2,4’-cyclohexa-2,5-diene]-1’-imine oxide for Sensor protein LytS (B) Protoporphyrin-IX-containing-Fe (HEM) with IUPAC name 3-[18-(2-carboxyethyl)-8,13-bis(ethenyl)-3,7,12,17-tetramethyl-23H-porphyrin-21-id-2-yl]propanoate;iron(2+) for Galactosyl transferase CpsE. The nitrogen atoms of the ligands are depicted in blue, oxygen atoms in red, phosphorus atoms in orange, iron atoms in purple, and carbon atoms in white.

Molecular docking studies

The ligand exhibiting the lowest docking score in molecular docking studies is regarded as the most effective, as it signifies a stronger and more stable binding affinity with the target protein, thereby making it a prime candidate for further exploration in drug design and therapeutic development [104]. The interactions between the modeled structures of LytS and CpsE and their respective ligands were investigated using molecular docking performed with AutoDock Vina in PyRx v0.8. Ligand 128 demonstrated a stronger binding affinity with LytS, yielding a binding energy of –9.1 kcal/mol, whereas ligand HEM interacted with CpsE yielding a binding energy of –8.9 kcal/mol. As depicted in Fig 9, ligand 128 established hydrogen bonds with the amino acid residues His479, Asn537, Arg541, Gln470, Thr531, Arg478, and Asn381 of LytS. In comparison, ligand HEM interacted with CpsE through hydrogen bonds with the residues Ser352 and Leu369.

Fig 9. Illustration of the interactions between ligands and their corresponding drug targets through molecular docking.

(A) Three-dimensional and two-dimensional interaction profiles of ligand 128 with the sensor protein LytS. (B) Three-dimensional and two-dimensional interaction profiles of ligand HEM with galactosyltransferase CpsE. In the three-dimensional representations, amino acid residues of the target proteins are shown as pink sticks, while interacting ligands are depicted in green, blue, and red. In the two-dimensional illustrations, dotted lines colored green, blue, pink, purple, and red indicate various types of bonds formed between the target proteins and their respective ligands.

Discussion

Developing new therapeutic drugs and vaccines are challenging. Advancements in computational research, sequence-based technologies, and the availability of diverse pathogens’ genomes and proteomics data have made it easier to perform the task. Subtractive genomics approach is invaluable for drug discovery because it isolates pathogen-specific proteins, minimizing potential harm to the host by targeting only the pathogen’s essential proteins. This method reduces the risk of off-target effects, ensuring therapeutic specificity. Non-homologous proteins differ from host proteins in that they do not share sequence similarity or functional characteristics, making them ideal candidates for drug targeting and vaccine development without interfering with the host’s own cellular machinery. In silico subtractive genome techniques are promising in identifying specific genes and proteins in various organisms including Clostridium botulinum [105], Mycoplasma pneumoniae [55], Streptococcus pneumonia [106], Campylobacter jejuni [107], Salmonella typhi [50], Legionella pneumophila [63], Arabidopsis thaliana [108], Meningococcus B [109], Eubacterium notadum [110], Fusobacterium nucleatum [111], Salmonella enterica subsp. Poona [112], Treponema pallidum [113], Staphylococcus aureus N315 [114], Acinetobacter baumanii [115,116], Bartonella bacilliformis [117], Bordetella pertussis [118], Serratia marcescens [119], and Staphylococcus aureus [120].

In this study, we employed a subtractive genomics-based computational approach to screen the entire proteome of S. agalactiae serotype V (ATCC BAA-611/ 2603 V/R) for the identification of potential drug targets. Afterwards, 200 proteins from the set of peer-reviewed proteins of the pathogen were recognized to have no similarities to the host H. sapiens proteins. In some studies, human homologous proteins (up to 40% similarity) were selected to identify potential drug targets [121], based on the idea that low similarity in sequence would produce a slightly different protein structure from human host and it would be safely considered as a drug target [121,122]. This could be the case, when no potential drug targets are found in the non-homologous sequence pool. Later on, we utilized the DEG database version 15.2, that successfully lead to the selection of essential non-similar proteins out of the non-homologous sequences (S1 Table). The promising candidates for organism-specific drug targets are proteins that are vital for the pathogen’s survival and do not share homology with the host genome [72,98,115,117,123].

Subsequently, KEGG KAAS analysis revealed 51 essential non-homologous proteins were involved in 41 unique pathways of the pathogen (Table 2). These unique pathways including two-component system, quorum sensing, peptidoglycan biosynthesis, and various amino acid metabolism systems such as alanine, aspartate, glutamate, cysteine, methionine, tyrosine, histidine metabolism, etc., are essential for the survival of S. agalactiae and when interrupted would cause the bacterium to not function properly [54,107]. Environmental stress can increase gene expression and metabolic processes in bacteria, potentially leading to the development of resistant pathogens with distinct metabolic processes [62,120]. Thus, these essential proteins involved in such unique pathways alone could be excellent targets for drugs and vaccines [62,120].

In addition, protein localization is crucial in drug development, as it governs the formulation and design of novel drugs and vaccines, as demonstrated in studies including Acinetobacter baumanii [46,115], Bartonella bacilliformis [117], Bordetella pertussis [118], Streptococcus pneumoniae [57,124], Salmonella typhi [48], Mycoplasma genitalium [125], Neisseria meningitides [60], Staphylococcus aureus [62,120], and Treponema pallidum [113]. In several studies, cytoplasmic proteins were selected as drug targets for their availability and involvement in metabolic pathways crucial to the survival of those pathogens [44,58,125,126]. However, in our study we emphasized on the cytoplasmic membrane proteins in view of the fact that they can further be utilized as a target for vaccine development, as membrane proteins are considered to be the best criteria for vaccines [106,127,128]. Moreover, it was seen that membrane proteins are essential for integral cellular signal detection, transduction, and various other biological processes [129]. They act as diffusion barriers for ions, water, transport systems, and nutrients and when disrupted or degraded, can lead the bacteria to cease functioning and can serve as excellent druggable targets [129]. It is important to note that a significant number of druggable targets were derived from membrane proteins [45,127–135].

In addition, the identification of unique virulent factors of S. agalactiae would represent a substantial contribution, given that these factors are crucial in the control or degradation of the host immune system [136]. In Helicobacter pylori, virulence factors like VacA, a vacuolating cytotoxin, are associated with the development of gastric diseases, with specific genotypes linked to ulcerogenic and non-ulcerogenic strains [137]. Hence, from the set of essential proteins, six virulent proteins (Table 4) were predicted by the tool VirulentPred2.0, implying that inhibition of these proteins could render the pathogen non-virulent and significantly impact the pathogen’s functionality within the host organism [47,49,50,59,105,111]. As a result, we prioritized two proteins, namely sensor protein LytS and galactosyl transferase CpsE, that had the potential to be utilized as promising drug candidates because of their virulent nature, cytoplasmic membrane characteristics and their involvement in essential pathogen-specific pathways of S. agalactiae.

The sensor protein LytS, also known as sensor histidine kinase LytS, encoded by the gene lytS, is a part of the two-component regulatory system LytR/LytS of S. agalactiae. Two-component regulatory systems are key contributors to bacterial responses to environmental changes, regulating gene expression in response to stimuli, which enables bacteria to perceive, respond, and cope with stressful conditions [138,139]. LytS, self-phosphorylates its cytoplasmic domain by transferring the phosphate group upon detecting specific signals or stress conditions, activating the transcriptional response regulator LytR. The phosphorylated LytR then binds to DNA, regulating the expression of target genes [138,139]. The LytSR two-component system, consequently, influences virulence factors like biofilm formation, resistance to host antimicrobial peptides, and bacterial autolysis, enhancing the survival of S. agalactiae in the human host [22,138,139]. In essence, LytS acts as a molecular switch, initiating a signaling cascade that influences gene expression, behavior, and virulence, enhancing bacterial adaptation and making it a key player in pathogenesis. Disruption of LytS would impair the bacterial ability to respond to environmental stresses, reduce cell wall integrity, and prevent the activation of virulence factors regulated by the two-component system.

Conversely, galactosyl transferase CpsE, encoded by the gene cpsE, is a key enzyme in the biosynthesis of capsular polysaccharide and played a pivotal role in the assemblage of capsular polysaccharide in group B streptococci [32,140]. The polysaccharide capsule, consisting of a series of repeating monosaccharides units including glucose, galactose, and N-acetylglucosamine, represents a significant virulence factor in pathogens belonging to the Streptococcus genus and plays a significant role in enabling these bacteria to evade the innate immune response by providing protection against phagocytosis, opsonization, along with the complement system [32,141]. CpsE catalyzes the transfer of a galactose residue from UDP-galactose (UDP-Gal) to an undecaprenyl phosphate acceptor, initiating the formation of the CPS oligosaccharide repeating unit, which occurs at the cytoplasmic face of the bacterial cell wall [32]. CpsE is part of a larger biosynthetic complex, which includes other enzymes like CpsJ, CpsK, and CpsA, each playing distinct roles in capsular polymerization, sialylation, and insertion into the bacterial cell wall [32]. Moreover, CPS biosynthesis influences biofilm formation, further enhancing GBS pathogenicity during infection. Thus, inhibiting CpsE weakens the protective capsule, exposing the bacteria to phagocytosis and immune detection, thereby increasing sensitivity to treatment and immune responses.

The STRING server afterwards revealed that the two selected proteins could function as core proteins associating with three or more neighboring proteins of S. agalactiae. As a result, repressing these proteins can inhibit the proper functioning of other related proteins [56,58,63,119]. ProtParam analysis predicted that both proteins were hydrophobic, basic, stable, and thermostable in nature, with LytS having 581 amino acids and CpsE having 449 amino acids in length (Table 5). Furthermore, druggability analysis revealed both proteins to be novel drug targets.

In addition, the study effectively predicted, analyzed, and evaluated the 3D structures of the two chosen proteins. It was seen that the residues in the Ramachandran favoured regions of both proteins were more than 85% and the Molprobity score was between the expected range of −4–2 [142], thereby confirming that the above values and calculations satisfied the structural validation criteria and showed that the predicted structures were of high quality as reported in various other studies [44,49,50,56,58,59,61,63,108,110,112,113,119,120,143,144]. Moreover, through the utilization of the ProBis server, ligand 128 was identified as a potential binder for LytS, while HEM was found to interact with CpsE. The binding sites of both proteins were analyzed through the DoGSiteScorer tool and visualized using PyMOL 3.1. Subsequent molecular docking studies revealed that both ligands exhibited notable binding affinities. Specifically, ligand 128 demonstrated a binding affinity of –9.1 kcal/mol with LytS, while HEM displayed a binding affinity of –8.9 kcal/mol with CpsE. These findings underscore the significant binding interactions between the ligands and their targets, suggesting their potential utility in the modulation of these proteins for therapeutic applications. As a result, we successfully identified and analyzed new proteins that show great promise as potential therapeutic targets. The potential of these proteins was prudently determined pertaining to their fundamental contribution to play an essential part in the survival of the pathogen and their feasibility in combating S. agalactiae serotype V (Strain ATCC BAA-611/2603V/R). As far as we are aware, this serotype has never been the subject of a subtractive genomics study before, and we believe it will provide a promising new strategy for preventing the spread of the antibiotic-resistant bacteria. Furthermore, to evaluate the potential for cross-species drug targeting, we performed a BLASTp-based sequence homology analysis of S. agalactiae serotype V proteins LytS and CpsE against a panel of clinically relevant bacterial pathogens. These proteins exhibited a range of 40% to 98% sequence similarity to various sensor histidine kinases and sugar transferases across multiple species, including Bacillus anthracis, Streptococcus pneumoniae, Staphylococcus aureus, and Mycobacterium tuberculosis, among others (S5 Table). High sequence identity (>70%), as observed with the LytS homolog in S. mutans, suggests conserved structural domains and potential drug-binding sites, supporting the feasibility of multitarget antimicrobial development. Proteins with moderate similarity (40–60%), such as homologs in S. pneumoniae, B. subtilis, and C. botulinum, may still serve as viable targets, pending structural validation of conserved ligand-binding pockets. In contrast, homologs with <40% identity are less likely to be druggable unless functionally conserved domains are retained. However, it requires further structural and functional investigations which is not in the scope of the current article.

Conclusion

The development of new drugs has been significantly expedited by the utilization of bioinformatics tools to extract and analyze genome and proteome sequences from diverse pathogens across multiple databases. The subtractive genomics methodology is impressively capable of effectively resolving challenges which are prevalent in traditional drug discovery methodologies. Using subtractive genomics approach, we have successfully identified two proteins, namely LytS and CpsE, that have the potential to be used as drug targets. While this research establishes a strong computational foundation in early-stage drug discovery, it also opens avenues for further in vitro and in vivo studies that are crucial for translating these computational insights into clinically viable therapies, which is beyond the scope of the current work. Nonetheless, an extensive pipeline for drug target identification has been developed in this study, which has the potential to facilitate and aid future experimental research to functionally characterize these targets in vivo and validate their druggability through high-throughput screening and preclinical evaluation. This approach holds promise for the creation of alternative therapies, offering a valuable framework for future drug discovery.

Supporting information

S1 Table. All DEG Essential Proteins.

(XLSX)

S2 Table. List of KO ID.

(XLSX)

S3 Table. All Subcellular Localizations.

(XLSX)

S4 Table. Residues present in the active sites of Sensor protein LytS.

(XLSX)

S5 Table. Galactosyl transferase CpsE.

(XLSX)

Data Availability

All relevant data are within the manuscript and its Supporting information files.

Funding Statement

The author(s) received no specific funding for this work.