Prediction of SafD adhesin strong binding peptides for pilus proteins assembly suppression in the prevention of Salmonella-induced biofilm formation using virtual mutagenesis studies

Abstract

Clinical isolates of Salmonella enterica contain Saf pili that establish initial bacterial attachment with the human epithelium to form biofilms which are a common cause of several abdominal complications. Due to the rise in antibiotic-resistant strains of bacteria, an alternate strategy of inhibiting the initial bacterial contact with the epithelial layers is well-studied. Saf pili undergo a chaperone-usher pathway assembly mechanism to generate its host-recognizing functional form, SafDAA. Preventing the biogenesis of the pili by targeting the SafD and SafA proteins polymerization will prevent host recognition. In this study, virtual mutagenesis studies using the recently reported X-ray crystal structure of an N-terminal peptide co-crystallized with SafD led to the design of peptides that exhibit enhanced binding with SafD compared to its native peptide. Virtual alanine mutagenesis and protein–peptide interaction studies identified several hotspot residues. Molecular dynamics simulations and binding free energy calculations identified key pairwise interactions between the designed peptides and SafD. In addition, a library of 110 peptides that are predicted to bind strongly with SafD is prepared which can serve as an excellent resource for the discovery of novel SafD-binding peptides. This work provided new insights into the design of novel anti-virulence therapies targeting Salmonella enterica.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40203-025-00313-9.

Introduction

Microbial biofilms are a common cause of persistent infections associated with various body sites such as the skin or mucosal surfaces of the respiratory and digestive tract, oral cavity, etc. (Costerton et al. 1999). Bacterial biofilms are responsible for approximately 65% of all human infections (Wu et al. 2021). Biofilm formation usually occurs in three major stages: attachment, maturation, and dispersion (Rabin et al. 2015). Specific interactions between the proteins present on the surface of the bacteria and their cognate binding partners on the host cell surface drive the initial attachment. Bacteria have protein appendages, known as adhesins that aid in the host cell attachment. After the formation of the first layer of the biofilm, more cells are recruited from the bulk fluid and the biofilm gradually grows. After the biofilm is matured, dispersion happens, which is critical for the biofilm life cycle. Dispersion occurs either in the whole biofilm or part of it and happens due to factors such as lack of nutrients, intense competition, and outgrown population. Dispersion leads to the initiation of new biofilms at other sites (Rabin et al. 2015).

Bacteria forming biofilms exhibit high resistance to immune system activities and antibiotic treatment, thereby, making biofilm infections difficult to treat with the existing repository of antibiotics. This urges the need to better understand alternate effective therapeutic strategies. An alternate approach to anti-bacterial treatment is anti-adhesive therapy, where the first step of initial bacterial contact with the human epithelial layers is inhibited, as a therapeutic approach (Ofek et al. 2003; Asadi et al. 2019). Since agents that target the initial contact are not bactericidal, the propagation and spread of resistant strains to them are less likely (Ofek et al. 2003).

A common cause of human and animal abdominal complications such as typhoid fever and gastroenteritis is Salmonella enterica, which is a gram-negative and food-borne enteric bacterium (Zeng et al. 2017; Silva et al. 2014). According to the Centers for Disease Control and Prevention, Salmonella is one of the most common causes of foodborne intestinal infections in the United States, which results in an estimated 1.2 million human cases and $365 million in direct medical costs annually (Prevention CfDCa 2013). Salmonella enterica is subdivided into seven subspecies, namely, I, II, IIIa, IIIb, IV, VI, and VII. However, nearly 99% clinical isolates belong to the Salmonella enterica subspecies I, making it the predominant cause of Salmonella infections in humans and animals (Zeng et al. 2017). The clinical isolates of Salmonella enterica contain Saf pili, which enable bacterial microcolony formation and eventual biofilm development (Zeng et al. 2017). Based on the morphological features, the adhesins in gram-negative bacteria are classified into two major groups: fimbrial and non-fimbrial adhesins (Chatterjee et al. 2021). Earlier studies have suggested that the classical chaperone/usher family are divided into clades (Nuccio and Bäumler 2007). According to the phylogenetic tree of the fimbrial usher protein (FUP) family, six clades make up the FEP family; α-, β-, γ-, κ-, π-, and σ-fimbriae (Nuccio and Bäumler 2007).

Saf is a homopolymeric γ-clade chaperone-usher fimbriae (Nuccio and Bäumler 2007) composed of repeating major proteins SafA, periplasmic chaperone SafB, an outer membrane usher SafC, and a minor protein SafD (Zeng et al. 2017; Chatterjee et al. 2021). SafD serves as the adhesin in the Saf pili. In a recent work by Zeng et al., it has been established that Saf uses both SafA and SafD proteins to form biofilms (Zeng et al. 2017). The various Saf proteins (SafB, SafC, SafA, and SafD) undergo assembly to form the functional adhesive pilus. The nascent SafA and SafD proteins are first transported from the cytoplasm into the periplasm. The periplasmic chaperone undergoes donor strand complementation (DSC) in which it donates its strand for the correct folding of SafD or SafA proteins (Zeng et al. 2017; Remaut et al. 2006; Puorger et al. 2008). Following this, the Saf proteins undergo assembly where the N-terminal extension (Nte) of one Saf protein replaces the strand that occupies a groove in the neighboring protein. This assembly process generates a non-covalently interacting SafD–SafA_n (n is an integer) polymer (Salih et al. 2008).

After the process of the Saf proteins assembly to form the SafD–SafA_n polymer (Fig. 1a), the Saf pilus uses both the SafD and SafA proteins for host cell binding (Zeng et al. 2017). The poly-adhesive host recognition by Saf pili takes place through the following mechanism: in the first step, the SafD at the distal tip of the pili mediates the initial host recognition by binding with a host binding partner. This intimate host–bacterium association leads to the formation of bacterial colonization. Then the sequential binding of the SafA proteins happens with its host binding partners, causing stronger binding, leading to the cause of various diseases (Zeng et al. 2017). Zeng et al. showed that when SafDAA was progressively truncated into SafD, SafA, SafDA, or SafAA, very little binding was observed. Only SafDAA exhibited a strong binding signal (Zeng et al. 2017).

Fig. 1.

Structural analysis of the Saf proteins and their assembly, and investigation of the peptide binding site: a (Left) X-ray crystal structure of the Saf pilus (PDB ID: 5Y9H). The SafD protein is shown in red surface representation and the two SafA proteins are shown in green surface representation. (Right) Assembly of the proteins to form the SafDA-dsc complex. The Nte peptide from the neighboring SafA2 (SafA2 not shown) is shown in blue; b the structure of the SafD-dsc to show the Nte peptide (PDB ID: 5Y9G), which was used in this work

The hypothesis that led to the development of the current work is that preventing the Saf proteins from the assembly by inhibiting the interactions between the SafD and SafAA proteins will prevent host recognition. Failure to form the host-recognizing functional form of the pili will in turn prevent bacterial colonization. A similar analogy has previously been used to identify type 1 pilus assembly inhibitors that work by preventing intermolecular interactions between pilus proteins (Lo et al. et al 2014). Prevention of biogenesis of type 1 pili by targeting the pilus protein polymerization at an early assembly step using a ‘donor strand exchange’ inhibitor provided new insights into the design of novel anti-virulence therapies (Lo et al. et al 2014). The Saf-proteins assembly takes place through the interactions of the SafD protein with the N-terminal peptide (Nte) of the neighboring SafA protein (Zeng et al. 2017). Designing peptides that exhibit strong binding with the SafD protein to competitively inhibit interactions with the Nte peptide from the neighboring SafA protein can lead to the potential prevention of the Saf proteins assembly.

Although anti-adhesion therapy is well studied for a variety of biofilm-forming proteins (Ofek et al. 2003; Asadi et al. 2019), till date, no small molecules and/or peptides/peptidomimetics targeting either SafA or SafD have been designed to block Saf-mediated biofilm formation. Given a huge unmet need in this area, we, therefore, believe that developing anti-biofilm agents targeting SafA/SafD is highly significant and timely. Since no atomic level understanding of Saf pili binding and their inhibition have been reported, the structural insights that will be obtained from this study will pave the path for the future development of improved inhibitors of Salmonella-induced biofilm formation.

State-of-the-art computational modeling techniques have made significant advances in the process of drug discovery and development and have become an integral part of research in both academic as well as industrial settings (Sadybekov and Katritch 2023; Dalkas et al. 2013; Sliwoski et al. 2014). Ligand-based and structure-based modeling efforts have previously been utilized widely to identify bioactive compounds against a variety of therapeutic targets (Dings and Mayo 2007; Gruber et al. 2010; Tripathi and Bandyopadhyay 2022; Wilson and Lill 2011). In silico mutagenesis, molecular docking, molecular dynamics simulations and binding free energy calculations are some of the extensively used computational methods in the modern world drug discovery campaigns (Samanta and Doerksen 2024a, 2024b; Bromberg and Rost 2008; Khan et al. 2021; Shahab et al. 2023; Palmer et al. 2021; Geng et al. 2019). One major discovery from the field of computer-aided drug design was in the recent years by Schrodinger who leveraged their computational platform to identify multiple highly selective and structurally distinct Wee1 inhibitors. Their lead compound obtained from their computational modeling efforts is currently in clinical development with patients with advanced solid tumors (https://www.schrodinger.com/pipeline/wee1-myt1/). These inhibitors had optimized physicochemical properties. Among many more, another recent work utilized computational methods to identify novel compounds targeting a protein associated with Parkinson’s Disease, which were then tested in vitro (Li et al. 2024; Lang et al. 2015; Tan et al. 2023; Eguida et al. 2024). Computational modeling efforts have also been highly successful in peptides and peptidomimetics design for therapeutic targets (Khan et al. 2021; Shahab et al. 2023; Kuczera 2015; Mondal et al. 2022; Ochoa et al. 2022; Ciemny et al. 2018). Successful studies have been reported that have used computational modeling techniques for finding compounds that are then subjected to experimental work in the drug discovery campaigns. For instance, an in-silico compound screening technique helped identify promising compounds which were then subjected to in-vitro assays identified a first-in-class preclinical drug candidate for Huntington’s disease (Galyan et al. 2022).

In this study, we have used the recently reported X-ray crystal structure of the SafD adhesin with a co-crystallized Nte peptide, which is the N-terminal extension of SafA (PDB ID: 5Y9G) (Zeng et al. 2017) (Fig. 1b) and performed molecular modeling to design peptides that can act as potential anti-adhesives of the Saf pili by preventing Saf proteins assembly. The peptides designed in this work are predicted to show improved binding with the SafD protein than the native co-crystallized Nte peptide. Using a variety of molecular modeling approaches, such as hotspot analysis studies, protein–peptide interactions studies, virtual mutagenesis to standard amino acids as well as alanine mutagenesis, and peptide stability calculations, we have identified crucial SafD and peptide residues that can play a significant role in designing novel anti-adhesives of the SafD protein. In this work, we have used extensive all-atom molecular dynamics (MD) simulations and binding free energy calculations to obtain a detailed understanding of the key SafD–peptide pairwise interactions, which can be further modulated to enhance inhibitory peptide binding. In addition, we have prepared a library of 110 peptides that are obtained from virtual mutagenesis studies, which can serve as an excellent resource for the discovery of novel SafD peptide anti-adhesives.

Results

Protein–peptide interaction analysis

In this work, we have designed peptides that exhibit strong binding with the SafD protein of the Saf pili to competitively inhibit interactions with the Nte peptide from the neighboring SafA protein that can potentially prevent SafD–SafAA assembly. The SafA1_Nte has a total of 15 amino acids. Several intermolecular interactions are observed between the SafA1_Nte peptide (denoted as “P”) and the two neighboring β-sheets contributing from the SafD protein (denoted as “B.1” and “B.2”) (Fig. 2a, with the pairwise interactions noted at the bottom of the figure). The key structural insights obtained from the crystal structure of the SafD–Nte peptide complex served as the starting point in modulating the SafD and the peptide interactions to design blocking peptides that are predicted to bind strongly with SafD.

Fig. 2.

Investigation of the peptide binding site and hotspot residues prediction: a Calculated interaction energies of the SafD residues at the peptide binding site (in kcal/mol) calculated using Schrödinger BioLuminate. The SafD residue IDs corresponding to the row number (x-axis) are summarized in the table on the right; b hotspot residues identification at the SafD–peptide interacting interface, calculated using ANCHOR tool. A list of the hotspot residues and their contribution to binding free energy upon mutation is shown (in kcal/mol). Yellow and gray represent the hotspot residues in the native co-crystallized peptide and in SafD, respectively. The chain names of the peptide and the SafD protein are “P” and “A”, respectively

The peptide binding pocket on the SafD protein was analyzed using Schrödinger and the calculated interaction energies for the binding pocket residues were calculated (Fig. 2a). The residues from the SafD B.1- and B.2-sheets that formed the peptide binding pocket range 34–45 (row numbers 18–29) and 134–150 (row numbers 1–17), respectively (Fig. 2a). It was seen that all SafD residues in the peptide binding pocket exhibited negative calculated interaction energy values, with residues at positions 150, 137, and 41 exhibiting the worst calculated interaction energies. In addition, SafD hotspot residues prediction was performed using the ANCHOR Tool. Interestingly the three residues at positions 150, 137, and 41 that are predicted to show the worst calculated interaction energies using Schrödinger BioLuminate were also not identified as hotspot residues in the ANCHOR Tool alanine mutagenesis studies (Right Fig. 2a vs. Right Fig. 2b; “P” is the peptide chain name and “A” is the SafD chain name). Therefore, both the ANCHOR Tool and Schrödinger calculated interaction energies calculations predicted consistent SafD residues that are important for ligand/peptide binding.

To obtain an exhaustive list of the various protein–peptide interactions that are seen in the SafD–Nte peptide, in this work, a variety of molecular modeling approaches to investigate the protein–peptide interaction interface, were used. First, we have utilized the power of TCL (Tool Command Language), a string-based scripting language, to obtain the intermolecular interactions at the SafD–peptide interface. The interacting residues within 5 Å of the SafD–native peptide interface, are shown in Fig. 3a. In addition to the in-house TCL scripts to investigate the interacting interface, the Protein Interaction Analysis tool in Schrödinger was used to identify the pairwise interactions between SafD and the co-crystallized N-terminal peptide (Fig. 3b, Table 1 and Table S1).

Fig. 3.

Structural analysis of the SafD–native peptide (Pep0) complex: a Protein–Pep0 interaction analysis conducted using in-house Tool Command Language. (Top panel) Structure of Pep0. (Bottom panel) The Pep0 is shown in yellow NewCartoon representation and the two SafD β-sheets (B.1 and B.2-sheets) are shown in white NewCartoon representation. Intermolecular hydrogen bonding interactions are shown in cyan dashed lines. In both images, the polar, non-polar, acidic, and basic residues are shown in green, white, red, and blue licorice, respectively; b protein–Pep0 interaction analysis conducted using the Protein Interaction Analysis tool in Schrödinger; c–d molecular dynamics simulation analysis: (c) (Top panel) RMSD (in Å) of the heavy atoms of Pep0 during 1 μs MD simulation. (Bottom Panel) The total number of specific contacts SafD makes with Pep0 over the course of the MD simulation trajectory; d two-dimensional representation of SafD–Pep0 interactions as obtained from the 1 μs MD simulation trajectory. Interactions that occurred for ≥ 30% of the simulation time in the MD trajectory are shown

Table 1.

Pairwise intermolecular interactions between SafD (chain name: A) and the native co-crystallized peptide (chain name: P)

SafD residues	Peptide residues	Distance (Å)	Specific interactions	#HB	#Salt bridges	SafD residues	Peptide residues	Distance (Å)	Specific interactions	#HB	#Salt bridges
A: Ala 147	P: Gln 168	2.1	1 × hb to P: Gln 168	1	0	A: Leu 143	P: Val 172	1.9	2 × hb to P: Val 172	2	0
A: Phe 145	P: Lys 170	2.4	2 × hb to P: Lys 170	2	0	A: Trp 141	P: Ile 174	1.9	2 × hb to P: Ile 174	2	0
A: Asp 144	P: Ser 171	2.1	1 × hb to P: Ser 171	1	0	A: Asp 139	P: Phe 176	2.1	1 × hb to P: Phe 176	1	0
A: Ala 45	P: Ser 177	1.9	1 × hb to P: Ser 177	1	0	A: Thr 35	P: Glu 167	2.0	1 × hb to P: Glu 167	1	0
A: Phe 43	P: Ser 177	2.3	1 × hb to P: Ser 177	1	0	A: Lys 32	P: Glu 167	2.4	1 × salt bridge to P: Glu 167	0	1

The inter-residues distances, the type of interaction, and the number of hydrogen bonding, and salt bridge interactions are summarized in the table

The native co-crystallized peptide complexed with SafD was then subjected to MD simulations. The root-mean-square deviation (RMSD) of the native peptide showed that it was highly stable at the SafD binding interface during the MD simulation (Fig. 3c). The peptide exhibited several contacts with the SafD protein during the course of the simulation which resulted in the stability of the peptide binding at the SafD interface. The pairwise intermolecular interactions between SafD and the native peptide that were maintained ≥ 30% of the MD simulation time are shown in Fig. 3d. In addition to hydrogen bonding interactions between the peptide backbone and the neighboring SafD residues, the sidechains of the native peptide residues also exhibited several intermolecular interactions with SafD (such as with residues SafD.D144, SafD.T142, and SafD.F43).

Identifying protein–protein interaction hotspots by mutational analysis and designing novel peptide analogs

Hotspot identification using ANCHOR tool

Alanine mutagenesis was performed using the ANCHOR Tool and the residues’ contribution to the binding free energy were calculated (Meireles et al. 2010). Change in the binding energy calculated on the crystallized SafD–peptide complex (PDB ID: 5Y9G) indicated that residues E167, Q168, K170, S171, V172, I174, V175, F176, and S177 of the native peptide and residues T35, L37, V39, F43, R44, T140, Y141, L143, and F145 of SafD can act as hotspot residues. These hotspot residues are predicted to play a significant role in the SafD–peptide interactions and can be targeted for small molecule pilus assembly suppression design (Fig. 2b).

Residue scanning using BioLuminate and molecular dynamics simulations

The residues in the native co-crystallized Nte peptide were mutated to standard amino acids and the residue scanning and affinity maturation functionalities implemented in the BioLuminate module in Schrödinger (Zhu et al. 2014; Salam et al. 2014; Beard et al. 2013; Schrödinger Release 2021-2,2021) were used to calculate the stability of the designed novel peptides and the change in binding affinity ∆∆G(bind) between SafD and the mutant peptides, as explained under Materials and Methods. A total of 110 stable mutant peptides were generated that showed favorable binding affinity with SafD (Fig. 4). Of the 110 mutant peptides, 81 peptides showed a ∆∆G (bind) of < − 100 kcal/mol (Table S2). In this work, we have considered the top 10 mutant peptides with respect to binding affinity and performed extensive all-atom MD simulations to investigate the nature of SafD–mutant peptide interactions in detail (Table 2). The MD simulations were followed by binding free energy calculations conducted on the frames extracted from the MD simulation trajectories to obtain an accurate estimation of the peptide binding.

Fig. 4.

The generated 110 SafD binding peptides as obtained from virtual mutagenesis studies using Residue Scanning: a ∆∆G(bind) denotes the change in the binding affinity due to the mutations (in kcal/mol) of the 110 SafD binding peptides; b ∆∆G(stability) denotes the change in the stability of the peptides due to the mutations (in kcal/mol) of the 110 SafD binding peptides

Table 2.

The top 10 SafD binding peptides from the generated peptide library, with respect to their ∆∆G (bind), are summarized. The residue positions of the peptides (chain name: P), their residue names before and after mutation, and the ∆∆G (bind) values (in kcal/mol) are listed

Peptide #	Residue	Original	Mutant	∆∆G (bind) (kcal/mol)
Pep1	P:164 P:168 P:169 P:170 P:171 P:172	PRO GLN GLN LYS SER VAL	TRP PHE ARG TRP ARG MET	− 189.04
Pep2	P:164 P:168 P:169 P:170 P:171 P:177	PRO GLN GLN LYS SER SER	TRP PHE ARG TRP ARG TYR	− 188.28
Pep3	P:164 P:168 P:169 P:170 P:171 P:177	PRO GLN GLN LYS SER SER	TRP PHE ARG TRP ARG MET	− 186.8
Pep4	P:164 P:168 P:169 P:170 P:171 P:173	PRO GLN GLN LYS SER ASP	TRP PHE ARG TRP ARG THR	− 185.02
Pep5	P:164 P:167 P:168 P:169 P:170 P:171	PRO GLU GLN GLN LYS SER	TRP GLN PHE ARG TRP ARG	− 184.55
Pep6	P:164 P:165 P:168 P:169 P:170 P:171	PRO ASN GLN GLN LYS SER	TRP LEU PHE ARG TRP ARG	− 183.85
Pep7	P:164 P:168 P:169 P:170 P:171 P:173	PRO GLN GLN LYS SER ASP	TRP PHE ARG TRP ARG ASN	− 183.82
Pep8	P:164 P:168 P:169 P:170 P:171 P:173	PRO GLN GLN LYS SER ASP	TRP PHE ARG TRP ARG LEU	− 183.49
Pep9	P:164 P:168 P:169 P:170 P:171	PRO GLN GLN LYS SER	TRP PHE ARG TRP ARG	− 183.33
Pep10	P:164 P:168 P:169 P:170 P:171 P:178	PRO GLN GLN LYS SER SER	TRP PHE ARG TRP ARG LEU	− 183.09

Mutant peptide1. The structure of the mutant peptide showing the best binding affinity with SafD is shown in Fig. 5. The proposed peptide is a hexa-mutant peptide; P164W/Q168F/Q169R/K170W/S171R/V172M (denoted as 164.168.169.170.171.172.WFRWRM or Pep1). During MD simulation, this mutant peptide when bound to SafD showed lower fluctuations than the native co-crystallized peptide, as could be seen from the peptide RMSD (Fig. 5a vs. 3c). Similar to the bound native peptide, 164.168.169.170.171.WFRWRM also exhibited several hydrogen bonding interactions with the neighboring SafD residues (Fig. 5c). The sidechain of the mutated residue, Pep1.Q169R, showed an electrostatic interaction with neighboring SafD.D144. Such an electrostatic interaction, which was missing in the native peptide, helped in enhancing the binding of the mutant peptide. Additionally, a hydrogen bonding interaction between the sidechain of Pep1.K170W and SafD.T35 was also observed in the mutant peptide (Fig. 5c). The intermolecular interaction between the sidechain of Pep1.S177 and SafD.F43 was preserved both in the native as well as in the mutant peptide, Pep1.

Fig. 5.

MD simulation and structural analysis of the 1 μs trajectory of SafD–Pep1 complex: a (Top panel) RMSD (in Å) of the heavy atoms of Pep1. (Bottom Panel) The total number of specific contacts SafD makes with Pep1 over the course of the trajectory; b three-dimensional structure of the designed Pep1 and its neighboring SafD residues in the binding pocket. The mutated and the non-mutated residues of the peptide are shown in green and yellow, respectively; c two-dimensional representation of SafD–Pep1 interactions as obtained from the 1 μs MD simulation trajectory. Interactions that occurred for ≥ 30% of the simulation time in the MD trajectory are shown. The mutated residues are marked in green

Mutant peptides 2 and 3. To investigate the effect of this conserved pairwise interaction (Pep.S177–SafD.F43), residue S177 was mutated to TYR (denoted as Pep2). This yielded the mutant peptide that exhibited the second-best binding affinity with SafD (Table 2 and Fig. 6). It was seen that the mutated residue S177Y faces away from residue SafD.F43, thereby disrupting the previously observed hydrogen bonding interaction with the backbone of SafD.F43. This led to a slight reduction of the binding affinity of Pep2 compared to Pep1. Another mutation was performed on S177 (S177M) to further validate the role of residue S177 in the interaction with the SafD adhesin (leading to Pep3). Pep3 showed the third-best binding affinity value with SafD (Table 2 and Fig. 7). The sidechain of S177M did not preserve the hydrogen bonding interaction with the backbone of SafD.F43, leading to slightly reduced SafD binding affinity, further highlighting the importance of residue S177 in the interaction with SafD. Pep3 exhibited an electrostatic interaction between E167 with neighboring SafD.K146 that was maintained for 30% of the MD simulation time. Residue Q169R exhibited an electrostatic and a hydrogen bonding interaction with the neighboring SafD.D144, which were maintained for 73% and 74% of the MD simulation time, respectively. The hydrogen bonding interaction between D173 and SafD.T142 was conserved for Pep1, Pep2, Pep3, Pep5, Pep6, Pep7, Pep9 and Pep10. Residue SafD.W141 also exhibited polar interactions with the mutant peptides for all 10 mutant peptides investigated here.

Fig. 6.

MD simulation and structural analysis of the 1 μs trajectory of SafD–Pep2 complex: a (Top panel) RMSD (in Å) of the heavy atoms of Pep2. (Bottom Panel) The total number of specific contacts SafD makes with Pep2 over the course of the trajectory; b three-dimensional structure of the designed Pep2 and its neighboring SafD residues in the binding pocket. The mutated and the non-mutated residues of the peptide are shown in green and yellow, respectively; c two-dimensional representation of SafD–Pep2 interactions as obtained from the 1 μs MD simulation trajectory. Interactions that occurred for ≥ 30% of the simulation time in the MD trajectory are shown. The mutated residues are marked in green

Fig. 7.

MD simulation and structural analysis of the 1 μs trajectory of SafD–Pep3 complex: a (Top panel) RMSD (in Å) of the heavy atoms of Pep3. (Bottom Panel) The total number of specific contacts SafD makes with Pep3 over the course of the trajectory; b three-dimensional structure of the designed Pep3 and its neighboring SafD residues in the binding pocket. The mutated and the non-mutated residues of the peptide are shown in green and yellow, respectively; c two-dimensional representation of SafD–Pep3 interactions as obtained from the 1 μs MD simulation trajectory. Interactions that occurred for ≥ 30% of the simulation time in the MD trajectory are shown. The mutated residues are marked in green

Mutant peptide 4. Next, the mutant peptide that exhibited the fourth-best binding affinity with SafD was studied (Pep4; Fig. 8), in which the role of residue D173 in the binding with SafD was investigated. Residue D173 was seen to exhibit hydrogen bonding interactions with SafD.T142 that were maintained for ~ 30–50% of the simulation time (for Pep1, Pep2, and Pep3; Figs. 5, 6, and 7). Mutating D173 to THR disrupted this interaction with SafD, leading to slightly reduced binding affinity (Fig. 8). Overall, Pep4 exhibited a lower number of total contacts with SafD than Pep1, Pep2, or Pep3 (Fig. 8a vs Figs. 5a, 6a, and 7a).

Fig. 8.

MD simulation and structural analysis of the 1 μs trajectory of SafD–Pep4 complex: a (Top panel) RMSD (in Å) of the heavy atoms of Pep4. (Bottom Panel) The total number of specific contacts SafD makes with Pep4 over the course of the trajectory; b three-dimensional structure of the designed Pep4 and its neighboring SafD residues in the binding pocket. The mutated and the non-mutated residues of the peptide are shown in green and yellow, respectively; c two-dimensional representation of SafD–Pep4 interactions as obtained from the 1 μs MD simulation trajectory. Interactions that occurred for ≥ 30% of the simulation time in the MD trajectory are shown. The mutated residues are marked in green

Mutant peptide 5. The fifth strong SafD-binding mutant peptide (Pep5, Fig. 9b) was subjected to MD simulation, and protein–peptide interactions studies were conducted. In Pep5, the importance of residue E167 was investigated and so mutated to GLN. E167Q exhibited a hydrogen bonding interaction with SafD.A149 that was maintained 96% of the simulation time (Fig. 9c). Although this might seem to be an improvement in terms of binding interaction, the mutation of E167 to GLN disrupted the electrostatic interaction that was earlier observed between E167 and SafD.K146 (in Pep3; c.f. Figure 7) that was maintained for 30% of the simulation time. Pep5 was seen to be stable during the simulation and showed very little fluctuations and maintained a similar total number of specific contacts with SafD as the native co-crystallized peptide (Fig. 9a).

Fig. 9.

MD simulation and structural analysis of the 1 μs trajectory of SafD–Pep5 complex: a (Top panel) RMSD (in Å) of the heavy atoms of Pep5. (Bottom Panel) The total number of specific contacts SafD makes with Pep5 over the course of the trajectory; b three-dimensional structure of the designed Pep5 and its neighboring SafD residues in the binding pocket. The mutated and the non-mutated residues of the peptide are shown in green and yellow, respectively; c two-dimensional representation of SafD–Pep5 interactions as obtained from the 1 μs MD simulation trajectory. Interactions that occurred for ≥ 30% of the simulation time in the MD trajectory are shown. The mutated residues are marked in green

Mutant peptide 6. Although residue N165 in nine of the studied peptides, (Pep1–Pep5 and Pep7–Pep10), did not show any polar interactions with the neighboring SafD residues, to investigate the effect of solvent on N165, the residue was mutated to an amino acid with a hydrophobic sidechain, LEU, forming Pep6 (Fig. 10). A slight deterioration of the binding affinity compared to Pep1–Pep5 was observed (Table 2). Mutation of ASN to LEU resulted in a slight change in the solvent exposure of the residue at position 165, as evident from Fig. 9c.

Fig. 10.

MD simulation and structural analysis of the 1 μs trajectory of SafD–Pep6 complex: a (Top panel) RMSD (in Å) of the heavy atoms of Pep6. (Bottom Panel) The total number of specific contacts SafD makes with Pep6 over the course of the trajectory; b three-dimensional structure of the designed Pep6 and its neighboring SafD residues in the binding pocket. The mutated and the non-mutated residues of the peptide are shown in green and yellow, respectively; c two-dimensional representation of SafD–Pep6 interactions as obtained from the 1 μs MD simulation trajectory. Interactions that occurred for ≥ 30% of the simulation time in the MD trajectory are shown. The mutated residues are marked in green

Mutant peptides 7 and 8. A network of hydrogen bonding interactions between D173 and SafD.T143 was heavily conserved amongst all mutant peptides. Mutation at position 173 to THR disrupted this network of hydrogen bonding interactions (Fig. 8). In Pep7, the change from D173 to ASN was found to keep one of the two hydrogen bonding interactions with SafD.T143 intact which was maintained for 53% of the simulation time (Fig. 11). It is important to note that Pep6–Pep10 showed very little difference in ∆Affinity (–183.09 to–183.85 kcal/mol). Mutation of the residue at position 173 to a hydrophobic residue, LEU, (Pep8) disrupted all hydrogen bonding interactions with neighboring residues of SafD, which is an expected outcome. This finding is similar to what was observed for the mutant peptide Pep4 (with D173T) (Fig. 12).

Fig. 11.

MD simulation and structural analysis of the 1 μs trajectory of SafD–Pep7 complex: a (Top panel) RMSD (in Å) of the heavy atoms of Pep7. (Bottom Panel) The total number of specific contacts SafD makes with Pep7 over the course of the trajectory; b three-dimensional structure of the designed Pep7 and its neighboring SafD residues in the binding pocket. The mutated and the non-mutated residues of the peptide are shown in green and yellow, respectively; c two-dimensional representation of SafD–Pep7 interactions as obtained from the 1 μs MD simulation trajectory. Interactions that occurred for ≥ 30% of the simulation time in the MD trajectory are shown. The mutated residues are marked in green

Fig. 12.

MD simulation and structural analysis of the 1 μs trajectory of SafD–Pep8 complex: a (Top panel) RMSD (in Å) of the heavy atoms of Pep8. (Bottom Panel) The total number of specific contacts SafD makes with Pep8 over the course of the trajectory; b three-dimensional structure of the designed Pep8 and its neighboring SafD residues in the binding pocket. The mutated and the non-mutated residues of the peptide are shown in green and yellow, respectively; c two-dimensional representation of SafD–Pep8 interactions as obtained from the 1 μs MD simulation trajectory. Interactions that occurred for ≥ 30% of the simulation time in the MD trajectory are shown. The mutated residues are marked in green

Mutant peptides 9 and 10. Pep9 (∆affinity =–183.33 kcal/mol), a penta-mutant peptide, showed slightly higher fluctuations in the RMSD during MD simulation (Fig. 13). Overall Pep9 exhibited similar polar intermolecular interactions with the neighboring SafD residues as all mutant peptides investigated in this work. Pep10 (∆∆G(bind) affinity =–183.09 kcal/mol) is a hexa-mutant peptide, exhibiting the point mutations observed in Pep9 along with a 6th point mutation, D178L, (Fig. 14). Many of the peptides investigated in this work exhibited several water-mediated hydrogen bonding interactions.

Fig. 13.

MD simulation and structural analysis of the 1 μs trajectory of SafD–Pep9 complex: a (Top panel) RMSD (in Å) of the heavy atoms of Pep9. (Bottom Panel) The total number of specific contacts SafD makes with Pep9 over the course of the trajectory; b three-dimensional structure of the designed Pep9 and its neighboring SafD residues in the binding pocket. The mutated and the non-mutated residues of the peptide are shown in green and yellow, respectively; c two-dimensional representation of SafD–Pep9 interactions as obtained from the 1 μs MD simulation trajectory. Interactions that occurred for ≥ 30% of the simulation time in the MD trajectory are shown. The mutated residues are marked in green

Fig. 14.

MD simulation and structural analysis of the 1 μs trajectory of SafD–Pep10 complex: a (Top panel) RMSD (in Å) of the heavy atoms of Pep10. (Bottom Panel) The total number of specific contacts SafD makes with Pep10 over the course of the trajectory; b three-dimensional structure of the designed Pep10 and its neighboring SafD residues in the binding pocket. The mutated and the non-mutated residues of the peptide are shown in green and yellow, respectively; c two-dimensional representation of SafD–Pep10 interactions as obtained from the 1 μs MD simulation trajectory. Interactions that occurred for ≥ 30% of the simulation time in the MD trajectory are shown. The mutated residues are marked in green

It is important to note that all 10 peptides investigated in this work exhibited lower RMSD fluctuations than the native co-crystalized peptide.

Binding free energy calculations using prime-MM-GBSA

Binding free energy calculations using the molecular mechanics with generalized Born surface area (MM-GBSA) method were performed on frames extracted from each SafD–peptide complex MD simulation trajectory as explained under Materials and Methods. All studied peptides showed negative average binding free energies (Fig. 15a). As evident from Fig. 15a, the 10 designed mutant peptides exhibited improved SafD binding energies than the native co-crystallized peptide. Figure 15b summarizes the affinity maturation results. The affinity maturation LOGO plot signifies the dominant mutant form of each peptide residue to show enhanced SafD binding. As discussed earlier in the virtual mutagenesis studies, residue S177 is critical to SafD binding.

Fig. 15.

Binding free energies and LOGO plot summarizing the favorable peptide residue mutations: a Average binding free energies (in kcal/mol) from MD simulation trajectory for the SafD–peptide complexes calculated using Prime MM-GBSA; b LOGO plot from the affinity maturation results

Generated mutant peptides database

The 3D structures of the mutant peptides complexed with SafD are supplied with the manuscript (in compressed MAEGZ file format). A total of 110 mutant peptides that are predicted to show favorable binding with the SafD adhesin are prepared using the virtual mutagenesis methodology adopted in this study. All 110 exhibited negative binding affinity with SafD. Of the 110 mutant peptides, 81 peptides showed a ∆ binding affinity of <–100 kcal/mol.

Discussion

It is interesting to note that all 110 mutant peptides that exhibited negative ∆∆G (bind), favored/tolerated at least one mutation at the terminal position (residue 164) and at position 169, indicating that the two amino acids played a critical role in SafD binding, through polar and/or hydrophobic interactions with the binding pocket. Additional structural investigation showed that the mutation of residues P164 and Q169 to amino acids with larger side chains (for instance, to TRP (Fig. 16a and b) or to ARG (Fig. 16c and d), respectively) allowed better fitting into the SafD B.2-sheet amino acids that constitutes the binding pocket (Fig. 16). From the hotspot prediction analysis studies, the SafD residues at positions 141 and 143 were seen to act as hotspot residues crucial for SafD–peptide interactions. This was in agreement with the extensive all-atom 1 μs MD simulations where W141 and L143 showed to form polar interactions with all 10 mutant peptides. SafD.W141 and SafD.L143 formed hydrogen bonding interactions with the backbone of the peptide residues 174 and 172, hence mutation of peptide residues 174 and 172 did not affect these hydrogen bonding interactions.

Fig. 16.

Structural analysis of the native and the mutant peptides; (a and c) The B.2-sheet of the SafD protein and residues P164 and Q169 of the native peptide (Pep0) in Surface representation (C gray); (b and d) The B.2-sheet of the SafD protein and mutated residues P164W and Q169R of the designed mutant peptide in Surface representation (C green), showing that the mutated residues fit better into the binding pocket than the native peptide residues

In recent years, computational modeling techniques have made significant advances in the process of drug discovery and development and have become an integral part of research in both academic as well as industrial settings (Sadybekov and Katritch 2023; Dalkas et al. 2013; Sliwoski et al. 2014; Kliger 2010; Maurya et al. 2019; Jennings and Tennant 2006). In this study, we have used virtual screening methods to mutate a native peptide and obtained peptide hits that exhibited enhanced predicted binding to SafD along with significant stability. These peptides could be further subjected to be developed for therapeutic purposes as potential alternate approach towards anti-bacterial treatment.

Peptide stability is essential to ensure that they maintain their functional structure and are capable to perform their biological roles effectively (Wang et al. 2022; Al Musaimi et al. 2022; Maher and Brayden 2021). In our virtual screening protocol, only mutations that generated stable peptides (negative values of ∆∆G) were allowed– mutant peptides that exhibited positive ∆∆G (stability) values were not retained. Therefore, all peptides obtained as the result of virtual mutagenesis screening in this work exhibited significant predicted stability. We have shortlisted the top 10 peptides for further study based on their improved predicted binding affinity to SafD, compared to the native peptide. These top 10 peptides also exhibited significant stability (∆∆G (stability) values ranging ~  − 350 to − 100 kcal/mol).

The designed peptides that exhibited enhanced binding with the SafD adhesin are predicted to prevent the Saf proteins from assembly to form the host-recognizing functional form of the pili, by competitively inhibiting the intermolecular interactions between SafD and SafAA proteins. Advanced and detailed molecular modeling studies on the prepared library of designed peptides will help identify additional important SafD–peptide interactions that could be further modulated to enhance SafD–peptide binding and design high binding affinity novel peptides. These peptides could be further studied using in vitro studies and in vivo validation. A point of concern with therapeutic peptides is their pharmacokinetic properties. Machine-learning-based in-silico ADMET prediction of therapeutic peptides could be performed. The molecular modeling studies conducted here is the first of its kind in Saf pili binding peptide design study.

Materials and methods

Generation of the 3D structure of SafD adhesion

The X-ray crystal structure of the SafD-dsc was used as a starting point of this study (PDB ID: 5Y9G) (Zeng et al. 2017). The protein was prepared using the Protein Preparation Wizard in Maestro at pH 7.0 ± 2.0. All existing water molecules from the protein PDB were removed. Energy minimization of the protein structure was performed using the OPLS4 forcefield (Lu et al. 2021). The structure was then split into SafD and the N-terminal peptide using Maestro and chain names of “A” and “P” were assigned to the SafD adhesin and the Nte peptide, respectively.

Protein–peptide interaction studies

The pairwise interactions between the SafD adhesin and the peptides were investigated using two methods: (a) using in-house Tool Command Language (TCL) scripts and visualizing in VMD. The interacting residues within 5 Å of the SafD–peptide interface, were calculated; (b) using the Protein Interaction Analysis tool in Schrödinger.

Predicting hotspots at the SafD–peptide interface using ANCHOR tool

To identify anchor residues, the ANCHOR tool, which uses ALA substitution for each residue and calculates the contribution to the binding free energy of each residue side-chain, was used (Salih et al. 2008). The SafD–Nte peptide complex (PDB ID: 5Y9G) was submitted to the ANCHOR tool web server.

Residue scanning and affinity maturation

We propose that the hotspot residues at the SafD–peptide complex interface will act as starting points for the designing of peptide inhibitors of the Saf pili. Therefore, to identify these hotspot residues, the Residue Scanning and the Protein affinity maturation techniques in the BioLuminate program of Schrödinger were used (Zhu et al. 2014; Salam et al. 2014; Beard et al. 2013; Schrödinger Release 2021-2,2021). Residues P164–S178 of the N-terminal peptide in the SafD-peptide complex were mutated to the known standard amino acids and the effect of the mutations, such as the change in the binding affinity of each mutated peptide to the SafD adhesin, changes in the stability of the peptide, etc., were calculated. The binding affinity values are calculated in implicit solvent and a negative value of the binding affinity change indicates stronger binding of the mutant peptide to SafD than the native co-crystallized N-terminal peptide. Side-chain prediction with backbone minimization was allowed for each mutant peptide. Simultaneous mutations of the native peptide residues to the standard amino acids were allowed and the maximum number of simultaneous deviations from the input was set to 6. The Monte Carlo optimization method was used and the maximum number of steps for optimization was set to 2000.

Binding affinity prediction method

The change in the binding affinity between the binding partners upon mutation is denoted by ∆∆G (bind) (Beard et al. 2013; Damuka et al. 2020; Bhachoo and Beuming 2017; Ferrari et al. 2024; Aguilera-Pesantes et al. 2017). The following thermodynamic cycle was used to calculate the change in the binding affinity of the peptide due to the mutation, where ΔG₃ and ΔG₄ were calculated with Prime MM-GBSA (using an implicit solvation model) (Schrödinger Release 2021-2,2021):

R represents the receptor (SafD), L represents the native co-crystallized peptide and L’ represents the mutated peptide. R + L and R + L’ denote the unbound SafD and the peptide. R.L and R.L’ denote the bound SafD–peptide complex (Schrödinger Release 2021-2,2021). The change in the binding affinity is given by:

1

Peptide stability prediction method

The following thermodynamic cycle was used to calculate the stability of the peptide due to the mutation, where ΔG₃ and ΔG₄ were calculated with Prime MM-GBSA (using an implicit solvation model) (Schrödinger Release 2021-2,2021):

L(u) is the unfolded native co-crystallized peptide, L(f) is the folded native co-crystallized peptide, L’(u) is the unfolded mutant peptide and L’(f) is the folded mutant peptide (Schrödinger Release 2021-2,2021). The change in the stability is given by:

2

A negative value of the stability indicates that the mutant peptide was more stable than the native co-crystallized Nte peptide.

Molecular dynamics simulations

The SafD–peptide complexes were subjected to all-atom MD simulations, using the OPLS4 force field in Desmond, Schrödinger (Bowers et al. 2006; Schrödinger Release 2021-2). Each protein–peptide complex was solvated with a 12 Å TIP3P water buffer and neutralized with Na⁺ and Cl^– ions. The default Desmond system-relaxation protocol was used for equilibration, followed by a 1 μs production run in the NPT ensemble at 300 K (Samanta and Doerksen 2023). The non-bonded interactions were kept at a cut off of 9 Å. The Nosé–Hoover chain thermostat and Martyna–Tobias–Klein barostat were used for the production run. A timestep of 2 fs was used for each MD simulation.

Prime MM-GBSA calculations

Prime Molecular Mechanics Generalized Born/Surface Area (MM-GBSA) binding free energy calculations were performed on the MD simulations trajectories of the SafD–peptide complexes. 100 snapshots per MD simulation were extracted from the 1 μs MD simulation trajectories. The thermal_mmgbsa script by Prime Schrödinger was used to compute the binding free energies of the ligands on frames extracted from the trajectory at an interval of 10 ns. Using this script, water molecules from the frames were removed and the VSGB continuum solvation model for water was used with a dielectric constant set to 80. The average of the MM-GBSA ∆G binding energy throughout the MD simulation time of each SafD–peptide complex was further investigated. The widely used MM-GBSA method was used to calculate the enthalpic contributions to the binding energies. The MM-GBSA method neglects the entropic change of protein–ligand binding and is widely used to rank ligands rather than for calculating absolute binding-free energies (Mishra and Koča 2018; Sun et al. 2018; Samanta et al. 2023).

Generation of the mutant peptides database

Peptides that exhibited negative binding affinity to the SafD adhesin after virtual mutagenesis using Schrödinger Residue Scanning were retained.

Conclusions

This work uses a variety of molecular modeling techniques to identify and investigate SafD–peptide interactions and to design peptides that are predicted to bind strongly with the SafD adhesin. A library of 110 peptides that are predicted to exhibit strong SafD binding is prepared which can serve as an excellent resource for the discovery of novel SafD binding peptides. This work helped provide new insights into the design of novel anti-virulence therapies targeting Salmonella enterica.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization, PS; methodology, PS; validation, PS; formal analysis, PS and SG; investigation, PS; data curation, P.S.; writing—original draft preparation, P.S.; writing—review and editing, P.S. and S.G.; visualization, P.S.; supervision, P.S.; project administration, P.S. All authors reviewed the manuscript.

Funding

This research received no external funding.

Data availability statement

The data presented in this study are available in the article, its supplementary material document, and the library of the protein–designed peptide complexes along with their 3D coordinates in a compressed MAEGZ format file is provided with this article.

Declarations

Institutional review board statement

Not applicable.

Informed consent

Not applicable.

Competing interests

The authors declare no competing interests.