Interaction and simulation studies suggest the possible molecular targets of intrinsically disordered amyloidogenic antimicrobial peptides in Acinetobacter baumannii

Abstract Acinetobacter baumannii is one of the causing agents of nosocomial infections. A wide range of antibiotics fails to work against these pathogens. Hence, there is an urgent requirement to develop other therapeutics to solve this problem. Antimicrobial peptides (AMPs) are a diverse group of naturally occurring peptides that have the ability to kill diverse groups of microorganisms. The major challenge of using AMPs as therapeutics is their unstable nature and the fact that most of their molecular targets are still unknown. In this study, we have selected intrinsically disordered and amyloidogenic AMPs, showing activity against A. baumannii, that is, Bactenecin, Cath BF, Citropin 1.1, DP7, NA-CATH, Tachyplesin, and WAM-1. To identify the probable target of these AMPs in A. baumannii, calculation of docking score, binding energy, dissociation constant, and molecular dynamics analysis was performed with selected seventeen possible molecular targets. The result showed that the most probable molecular targets of most of the intrinsically disordered amyloidogenic AMPs were UDP-N-acetylenol-pyruvoyl-glucosamine reductase (MurB), followed by 33–36 kDa outer membrane protein (Omp 33–36), UDP-N-acetylmuramoyl-l-alanyl-d-glutamate-2,6-diaminopimelate ligase (MurE), and porin Subfamily Protein (PorinSubF). Further, molecular dynamics analysis concluded that the target of antimicrobial peptide Bactenecin is MurB of A. baumannii, and identified other molecular targets of selected AMPs. Additionally, the oligomerization capacity of the selected AMPs was also investigated, and it was shown that the selected AMPs form oligomeric states, and interact with their molecular targets in that state. Experimental validation using purified AMPs and molecular targets needs to be done to confirm the interaction. Communicated by Ramaswamy H. Sarma


Introduction
Acinetobacter baumannii is a Gram's negative nosocomial pathogen and intrinsically resistant to many antibiotics.It has become a prevalent hospital-acquired infection and is part of the ESKAPE group of pathogens.Bacteria adapt to protect themselves against different antibiotics via diverse mechanisms (Kaushik, Sharma, et al., 2022;Kaushik, Tiwari, Joshi et al., 2022;Kaushik, Tiwari, Tiwari et al., 2022;Roy et al., 2018;Sharma et al., 2021;Tiwari, 2019;Tiwari et al., 2012;2019;Tiwari & Moganty, 2014;Verma et al., 2021).The bacteria's defences have also become exceedingly complex and intricate, but we also don't have any new drugs to control the emerging A. baumannii infections.Antimicrobial therapeutic drug discovery has risen in the last few years due to the widespread multidrug resistance to antibiotics.Many alternative therapies have been developed to prevent or treat antimicrobial resistance, such as antibodies, phage therapy, lysins, antimicrobial peptides, etc. (O'Neill, 2016).Antimicrobial peptides are one of the antimicrobial agents having a lot of potential for the anti-Acinetobacter therapeutic development (Domalaon et al., 2016).Antimicrobial peptides are produced by the host's innate immune system with immunomodulatory activity.These are low molecular weight peptides possessing various activities against pathogens.In the last few years, numerous studies have been performed to study AMPs and their ability to kill Acinetobacter along with its drug-resistant strains (Jung et al., 2021).
Despite their various advantages, antimicrobial peptides have some restrictions, including stability, production costs, cytotoxicity, and availability, which may limit their clinical use.To overcome the limitations related to the stability and denaturation of AMPs in cells, we investigated whether antibacterial AMPs against A. baumannii have intrinsically disordered states and can form amyloids.Amyloids are the most stable form of any peptide, and using intrinsically disordered peptides, we would be able to avoid their denaturation (Almeida & Brito, 2020).The antibacterial activity of cationic intrinsically disordered antimicrobial peptides depends on peptide chain length, net charge, lipidation and environmental condition rather than ordered structures (Latendorf et al., 2019).It is also suggested that directed co-aggregation of an amyloidogenic peptide and a target protein may lead to its antimicrobial effect (Galzitskaya, 2021).Additionally, there is growing evidence that oligomers of amyloidogenic peptides exhibit antimicrobial activity (Kagan et al., 2012); they have similarities in function, such as membrane remodelling processes including pore formation and fusion (Last & Miranker, 2013;Lee et al., 2020).Amphiphilic AMPs can self-assemble into super helical protofibrils, which form structural scaffolds for the ordered presentation of immune ligands such as DNA and dsRNA (Lee et al., 2020).
The interaction between positively charged AMPs and the negatively charged bacterial outer membrane is the fundamental molecular mechanism of function for most AMPs.They can then form pores in the membrane and cause cell lysis (Falanga & Galdiero, 2017).Apart from their membranedisrupting capabilities, antimicrobial peptides are now widely accepted as having the potential to influence microbial survival through the interruption of critical intracellular processes or interactions with intracellular targets (Le et al., 2017).However, finding specific intracellular targets of AMPs is more complex (Benfield & Henriques, 2020).The discovery of targets inside microbial cells and the mechanism by which AMPs can permeate the microbial cell without disrupting it, has gained much research interest.Some recent reports even showed the presence of intracellular targets of the AMPs (Cardoso et al., 2019;Le et al., 2017).Hence it is important to investigate the molecular targets of the AMPs showing antibacterial activity against A. baumannii, as it will help us to determine their working mechanism against multidrug-resistant bacteria.Therefore, the present study aims to identify the intrinsically disordered amyloid-forming antimicrobial peptides and their probable molecular targets effective against multidrug-resistant A. baumannii.

Antimicrobial AMPs against A. baumannii were selected
A literature survey was conducted to find the AMPs showing antibacterial activity against A. baumannii.We have enlisted 54 such antimicrobial peptides.The MIC of the selected AMPs varies from 1.6 mg/ml to 64 mg/ml against the A. baumannii (Supplementary table 1).

Seven selected AMPs have intrinsic disorder regions
One of the significant challenges of using AMPs is their structural stability.As intrinsically disordered peptides lack a defined three-dimensional structure and are highly flexible, the chances of their denaturation under physiological conditions are very low.Hence selected antimicrobial AMPs against A. baumannii were further investigated to check for intrinsic disorder using DisEMBL and GlobPlot2 web tools.
The seven antimicrobial peptides showed intrinsically disordered regions in their sequences (Supplementary Fig. 1).All seven antimicrobial peptides (Bactenecin, Cath BF, Citropin 1.1, DP7, NA-CATH, Tachyplesin and WAM-1) are disordered as per the Hot-loops definition of the DisEMBL server.To further confirm the prediction, GlobPlot2 results were also compared, and the antimicrobial peptides Cath BF, Citropin 1.1, NA-CATH, and WAM-1 show disorder probability in their sequences by Russell/Linding definition.The results showed that AMPs Bactenecin, Tachyplesin and DP7 are fully intrinsically disordered, while other AMPs have partially intrinsically disordered regions.

Seven selected AMPs have a propensity to form an amyloid structure
The amyloid structure is energetically more stable than any other protein structure (Ke et al., 2020); hence the amyloidforming capacity of selected AMPs was predicated using PASTA 2.0 and AGGRESCAN.The PASTA 2.0 results (Supplementary Figure 2 and Figure 3) showed that the seven antimicrobial peptides are showing amyloid-forming hotspot regions in their sequences.The Bactenecin AMPs have the highest number of amyloid fibril forming regions, that is, 20, followed by Tachyplesin (14 regions), DP7 (5 regions), WAM-1 (5 regions), Citropin 1.1 (3 regions), Cath BF (1 region), and NA-CATH (1 region) (Supplementary Figure 2).The lower the PASTA energy, the better probability of those residues forming amyloid fibrils.All our selected AMPs showed low energies between À 5 and À 8. Bactenecin had the lowest energy of À 8.7 PEU (1 PASTA Energy Unit (PEU) ¼1.192Kcal/mol).DP7 showed the highest percentage of disorder (75%) followed by Citropin 1.1 (56.25%),WAM-1 (33.33%),Cath BF (30.0%),Bactenecin (25.0%),Tachyplesin (11.76%) and NA-CATH (8.82%).The table also showed the percentage of the residues predicted to be in an a-helix, b-strand and coil state.AGGRESCAN output provided the aggregation-propensity values per amino acid (a3vSA) and the predicted number of hotspots.The sequence stretches with the highest projected aggregation propensity are highlighted in red and appear as peaks in the profile plots (Supplementary Fig. 3).In the profile plot, the green line represents the aggregation propensity for all the residues, the blue line represents the hotspot threshold, and the red line is the average of the aggregation propensity over a sliding window of certain residues based on total sequences (a4v).The results (Supplementary Fig. 3) showed that Bactenecin has an average aggregation propensity per amino acid of 0.5, and it has one probable hotspot.Cath BF also was observed to have one hotspot and a value of 0.1 for average aggregation propensity.For Citropin 1.1, residues 10 to 16 have been highlighted red, forming a peak in the plot and hence forming a hotspot.NA-CATH has two hotspot regions.In the case of DP7, residues 5-12 show a probability of forming one amyloid hotspot.Similarly, in Tachyplesin, one hotspot region can be observed between residues 1 and 14.For the AMP WAM-1, the average aggregation propensity per amino acid was À 0.054, and no hotpot region was observed through AGGRESCAN.

Possible molecular drug targets in Acinetobacter baumannii
Selected AMPs have antibacterial activity against A. baumannii, but their molecular target has not yet been investigated.

Modelling of antimicrobial peptides and molecular targets and its refinement
The three-dimensional structures of the antimicrobial peptides and probable molecular targets were unavailable.Hence, models of seven AMPs and seventeen molecular targets were created using I-TASSER (Iterative Threading ASSEmbly Refinement) web server and the Phyre2 web server, respectively.Figure 1 depicts the models of each selected AMP, and supplementary Figure 4 shows the structure of molecular targets.The models obtained were further refined using the GalaxyWEB Refine online tool.Among the five refined models of each protein, we have selected the one model with the lowest RMSD, MolProbity, and higher Ramachandran favoured region and GDT-HA.The refined models were then superimposed with the existing models to observe the changes (Supplementary Figure 5 and Supplementary Figure 6).

Validation of models of AMPs and molecular targets
To investigate the structural quality of the refined models and to validate them, Ramachandran plots, G factor and Verify3D scores were obtained using the PSVS server.Through the Ramachandran plots, it was observed that >90% of the residues were located in the regions which were allowed for all the models.The ProSa web tool was used to further examine the models' quality and recognize the errors, if any.It also provided us with the Z-scores for each 3D model.Ramachandran plot analysis of the AMP models are shown in Supplementary Figure 7.All the selected antimicrobial peptides had >90% residues in allowed regions and none in the disallowed region, proving them to be good quality models.Also, the residues in most of the AMPs are either in the beta-sheet or alpha-helix forming region on the plots.This further proves that the selected AMPs have the probability of forming cross-b or cross-a amyloid structures.Similarly, the possible molecular target protein models were also analysed by the Ramachandran plot, and the results are shown in Supplementary Figure 8.All our target models had more than 90% of their residues in the allowed region.Hence, the Ramachandran map explains that these models are good quality models.
In addition to the Ramachandran plot, G factor analysis was also performed.In the given plots G-factor is calculated for the dihedral angles in the given AMP model.Most of the residues of our selected AMPs and molecular targets had a positive G-score, which proves that the 3D models had a good overall geometry (Supplementary Figure 9, Supplementary Figure 10).A few residues in the targets BfmR, LpxC, MurE, OmpW and Porin showed a G-factor score of below À 4. A low G-factor in all these cases suggests a feature corresponding to a low-probability confirmation when applied to a specific residue.Residues in the Ramachandran plot's forbidden regions will have a negative (or low) G-factor.The same is true for unfavourable chi1-chi2 and chi1 values.If a protein has many residues with low G-factors, it suggests the overall geometry isn't right.
Additionally, we have also analysed the model using Verify3D analysis.In supplementary Figure 11, the Verify3D scores for all the selected AMPs have been plotted.From the analysis of the results, it is clear that most of the residues in the AMPs, show a negative Verify3D score.These negative Verify3D scores indicate a folding error in the peptides.This corresponds with the fact that all the selected AMPs here lack a proper 3D structure and are intrinsically disordered.Supplementary Figure 12 shows the Verify 3D plots of the molecular targets.Even though some of the individual residues in the targets show a negative score in the plots, they all had an overall positive Verify3D score.
Z-Scores were calculated for the refined model using the Protein Structure Assessment (ProSA-web) server.The results are shown in Supplementary Figure 13.The various groups of structures solved by X-ray or NMR are displayed in distinctive colours, and a black dot represents the Z-score of our model.The obtained Z-scores have been listed in Supplementary Table 3.Most of our AMPs had a negative Z-score between 0 to À 2.17, except Citropin 1.1, which had a positive Z-score of 0.42.This positive Z-score might indicate the intrinsic disorder of Citropin 1.1.
All of our targets had a negative Z-score, thus proving to be good quality models (Supplementary Figure14).The top negative Z-scores observed were for the targets AdeB, MurE and GyrA, which were À 12.56, À 11.77 and À 11.61, respectively (Supplementary Table 4).One possible explanation for lower z-score values than usual might be the presence of hydrophobic binding sites on the target protein, which is not favoured by the ProSa web server.
The plot of residues scores analysis was also performed.Positive values generally refer to problematic or incorrect sections of the input structure.As we have taken the model of antimicrobial peptides with intrinsic disorders, so there are positive values in the plots (Supplementary Figure 15).A plot of energies of a single residue usually contains a lot of fluctuations and isn't very useful for evaluating models.As a result, the plot is smoothed by calculating the average energy for each 40-residue fragment.In the case of Cath BF, Citropin 1.1, NA-CATH, Tachyplesin and WAM-1, we can observe a lot of fluctuations.This is because peptides are very small in length, and as the average is calculated over 40 residues, the plot shows too many fluctuations.

Antimicrobial peptides were docked with probable molecular targets
Based on literature survey, selected AMPs are known for their antimicrobial activity against resistant bacteria, but their molecular targets are still unknown.Molecular interactions between AMPs and their targets play an essential role in the action of AMPs.Hence, we have performed molecular docking to determine the probable target of selected AMPs against A. baumannii.Each AMP was docked against all the molecular targets, and the docking scores were calculated for each complex using the HDOCK web tool.Figure 2 shows the docked complexes for selected AMPs with their probable targets.
In the case of the docking of Bactenecin with the targets, the top six complexes with the best docking scores are with OMP33-36 (À 251.97),AdeB, TopoIV, MurE, CARomp, and GyrA (Supplementary Table 5).For the docking of Cath BF with the targets, the top six complexes with the best docking scores are with 33-36 kDa OMP (À 278.22)

Binding energy and dissociation constant calculation to select the possible molecular targets of AMPs
Binding affinity helps us to find the probable target for the AMPs which can bind selectively and specifically to each other.Binding energy, dissociation constant, and docking score helps to select the suitable interacting molecular target of AMPs.Hence, we have calculated the binding affinity and dissociation constant of the best-docked pose (Supplementary table 12).Based on the selection criteria, that is, AMP-target complexes must have a dissociation constant, K d <10 À 7 M at 25 � C; binding energy, DG < À 10 kcal/ mol and docking energy score < À 200, we have selected the top two probable targets for each AMPs (Table 2).The interaction of AMPs with the selected molecular targets was further validated by MDS analysis.

Molecular dynamics simulation analysis confirms the molecular target of AMPs
Molecular docking can anticipate the optimal binding mode, but it treats proteins as stiff and rigid, preventing them from adjusting their conformation throughout the docking process.To overcome this limitation and to learn more about interactions, the optimal or best docking pose can be exposed to molecular dynamics simulation (MDS) investigations till 100 nanoseconds.MDS was performed for the top two targets for each AMP using the WebGro server.For simulating the complex, the force field GROMOS96 43a1 was used.The default settings on the server were chosen.The result of MDS was analysed for RMSD, RMSF, the radius of gyration and hydrogen bond formation to evaluate the compactness, folding properties and structural stability of the AMP-target complexes.Figure 3 shows MDS analysis (RMSD, RMSF, Rg, H-bonding) for the Bactenecin complexed with MurB.Detailed information on all these docked complexes is provided in supplementary data (supplementary figures 11 to 14).

MDS analysis of Bactenecin complexed with MurB
The docking score observed for the Bactenecin-MurB complex was À 215.67, binding energy was À 11 kcal mol À 1 , and the dissociation constant was 8.8E À 09 M.This indicates a strong binding, and forming a stable complex.The RMSD value of this complex increased up to 0.5 nm for the first 25 ns of the simulation, then stabilized at around 0.5 nm for the rest of the simulation.The RMS fluctuation decreased from around 1 nm at the N-terminal to around 0.2 nm towards the C-terminal.
The lower RMSD and fluctuation indicate a stable complex formation.We can see that the radius of gyration decreases at the beginning of the simulation, but it is relatively invariant as the simulation proceeds.This indicates that the protein remains stable.Peak hydrogen bonds formed were observed to be around 270, also indicating a stable complex.Hence, the docking studies, binding energy and dissociation constant calculations, and molecular dynamics simulation results explain the interaction of AMPs Bactenecin with UDP-N-acetylenolpyruvoylglucosamine reductase (MurB).

MDS analysis of Cath BF complexed with MurB
The docking score observed for the Cath BF-MurB complex was À 256.50, and binding energy was À 11.6 kcal mol À 1 .The disassociation constant was 3.0E À 09 M.This indicates a strong binding, and forming a stable complex.The RMSD value of the complex remained stable at around 0.25 nm except some spikes observed at around 50 and 80 ns.The RMS fluctuation decreased from around 2 nm at the N-terminal to around 0.2 nm towards the C-terminal.The lower RMSD and fluctuation indicate a stable complex formation.The radius of gyration was observed to be stable at around 2.05 nm, with some spiking observed at 50 ns and towards the end of the simulation.Peak hydrogen bonds formed were observed to be around 295 which further indicates a stable complex.Hence, the docking studies, binding energy and dissociation constant calculations, and molecular dynamics simulation results explain the interaction of Cath BF with UDP-N-acetylenolpyruvoylglucosamine reductase (MurB).

MDS analysis of Citropin 1.1 complexed with Omp33-36
The docking score observed for the Citropin 1.1 -3336Omp complex was À 205.83, and the binding energy was À 11.8 kcal mol À 1 .The disassociation constant was 2.2E À 09 M.This indicates a strong binding, and forming a stable complex between AMP and target.The RMSD of the complex was observed to be constant at below 0.5 nm, with a single spike observed after 50 ns.The RMS fluctuation decreased from around 0.6 nm at the N-terminal to below 0.2 nm towards the C-terminal.The lower RMSD and fluctuation indicate a stable complex formation.The radius of gyration was observed to be stable at just above 1.8 nm throughout the simulation, with a single spike just after 50 ns.Peak hydrogen bonds formed were observed to be around 230. Overall this indicates that a stable complex was formed.Hence, the docking studies, binding energy and dissociation constant calculations, and molecular dynamics simulation results explain the interaction of Citropin 1.1 with 33-36 kDa outer membrane protein (Omp33-36).

MDS analysis of DP7 complexed with Porin SubF
The docking score observed for the DP7-PorinSubF complex was À 255.21, and binding energy was À 11.9 kcal mol À 1 .The disassociation constant was 2.0E À 09 M.This indicates a strong binding with the formation of a stable complex.The RMSD of the complex was observed to be constant at below 0.5 nm, with a few spikes observed at 25, 50 and near 100 ns.
The RMS fluctuation decreased from around 1.5 nm at the Nterminal to around 0.5 nm towards the C-terminal.The lower RMSD and fluctuation indicate a stable complex formation.The radius of gyration was observed to be stable between 2 and 2.1 nm throughout the simulation, with spikes at 25, 50, 80 and 100 ns.Peak hydrogen bonds formed were observed to be around 300. Overall this indicates the formation of a stable complex.Hence, docking studies, binding energy and dissociation constant calculations, and molecular dynamics simulation results explain the interaction of DP7 with the Porin Subfamily Protein (PorinSubF).

MDS analysis of NA-CATH complexed with MurB
The docking score observed for the NA CATH-MurB complex was À 254.99, binding energy was À 10.0 kcal mol À 1 .The disassociation constant was 4.9E À 08 M.This indicates a strong binding, and forming a stable complex.The RMSD value of the complex remained stable at around 0.5 nm except some spikes observed at around 35 and 80 ns.The RMS fluctuation decreased from around 2.7 nm at the N-terminal to around 0.5 nm towards the C-terminal.The lower RMSD and fluctuation indicate a stable complex formation.The radius of gyration was observed to be stable at around 2.1 nm, with some spiking observed at 35 ns and towards the end of the simulation at 80 ns.Peak hydrogen bonds formed were observed to be around 280 which further indicates a stable complex.Hence, the docking studies, binding energy and dissociation constant calculations, and molecular dynamics simulation results explain the interaction of NA-CATH with UDP-N-acetylenolpyruvoylglucosamine reductase (MurB).

MDS analysis of Tachyplesin complexed with OMP33-36
The docking score observed for the Tachyplesin-3336Omp complex was À 271.00, binding energy was À 11.3 kcal mol À 1 .The disassociation constant was 5.0E À 09 M.This indicates a strong binding forming a stable complex.The RMSD of the complex was observed to be constant at below 0.5 nm, with spikes observed after 50 ns.The RMS fluctuation decreased from around 2.3 nm at the N-terminal to around 0.25 nm towards the C-terminal.The lower RMSD and fluctuation indicate a stable complex formation.The radius of gyration was observed to be stable at just below 2.0 nm throughout the simulation, with spikes observed after 50 ns.The peak hydrogen bonds formed were observed to be around 250. Overall this indicates that a stable complex was formed.Hence, the docking studies, binding energy and dissociation constant calculations, and molecular dynamics simulation results explain the interaction of Tachyplesin with 33-36 kDa outer membrane protein (OMP33-36).

MDS analysis of WAM-1 complexed with MurE
The docking score observed for the WAM-1-MurE complex was À 258.29, binding energy was À 12.3 kcal mol À 1 .The disassociation constant was 1.0E À 09 M.This indicates a strong binding forming a stable complex.The RMSD of the complex was observed to be constant at around 0.25 nm, with a few spikes observed around 100 ns.The RMS fluctuation decreased from around 2.0 nm at the N-terminal to around 0.5 nm towards the C-terminal.The lower RMSD and fluctuation indicate a stable complex formation.The radius of gyration was observed to be stable between 2.5 and 2.75 nm throughout the simulation, with spikes around 100 ns.Peak hydrogen bonds formed were observed to be around 450.
Overall this indicates the formation of a stable complex.
All the MDS results suggest that MurB is the most preferred target for the AMPs investigated, followed by OMP33-36, MurE and Porin.They have a good docking score, binding affinity and dissociation constant, as well as validated through the MDS analysis.The RMSD analysis of some complexes also suggests some fluctuation in the complex that needs to be further analysed.Experimental validation needs to be done with the purified targets and pure AMPs before considering their targets.

In silico investigation of the amyloid-forming capability of selected AMPs
To further validate the amyloid formation capability of selected AMPs, the amyloid polymers (dimers, trimers, tetramers and pentamers) of each antimicrobial peptide were created.We calculated their docking score through the HDOCK server.The binding energy and dissociation constant of each complex was also calculated using the PRODIGY server (Table 3).
The docking score for the polymers of Bactenecin showed an increasing trend from dimers to pentamers except the trimer, which also has favourable docking.Although the binding affinities for the polymers were observed to be low, still they were favourable.Bactenecin trimers had the highest binding affinity of À 4.0 kcal/mol and a very low dissociation constant.In the case of Cath BF, the highest docking score was observed for its tetramers.The binding affinity for the oligomers was observed to be favourable.Similar results were observed for the AMPs NA-CATH, DP-7 and WAM-1.All of them had favourable binding affinities for the formation of oligomers.Citropin 1.1 did not show favourable binding affinities.The docking score for the oligomers was also quite low.Even though this antimicrobial peptide showed the probability of forming amyloid, the results here are contrary.Tachyplesin oligomers showed the best docking scores among all the AMPs.The binding affinity for the trimer formation was À 8.1 kcal/mol, and the dissociation constant was 1.2E À 06 M. The probable reason for such low energies for the formation of oligomers from different AMPs may be the short size of the selected peptides.All these results show that selected AMPs have some oligomeric capability that supports the selection of these peptides.The biochemistry of oligomerization needs to be investigated by experimental analysis.

In silico investigation of the interaction of oligomers with probable targets
We further validated the probable targets by analysing the interaction (docking, binding energy and dissociation constant) between selected probable targets and different oligomers of selected AMPs using the HDOCK server and PRODIGY server (Table 4).In the case of the interaction between Bactenecin oligomers with its potential target MurB, the docking score was observed to increase from dimers to hexamers.The binding affinity and dissociation constant were favourable for all the oligomer-MurB complexes.
Similarly, for Cath BF, the docking scores increased from dimers to heptamers when docked with its probable target, MurB.The best docking score was observed for the Cath BF Heptamer-MurB complex, with a score of À 287.29.Citropin 1.1 oligomers showed a varied trend in docking scores with OMP33-36.The highest docking score was À 282.58 for the dimer-OMP33-36 complex, followed by À 276.63 for the hexamer-OMP33-36 complex.The binding affinity was also very favourable for all the oligomer-MurB complexes, along with the dissociation constant.For DP7, the trimer-Porin complex had the highest docking score of À 330.53.The binding affinities and dissociation constant values were constant for all the oligomer-Porin complexes, with the best energies observed for the dimer-porin complex.In the case of NA-CATH, the best docking score of À 254.41 was observed for its heptamer-MurB complex, followed by À 255.73 for the tetramer-MurB complex.The binding affinity was also very favourable for all the oligomer-MurB complexes, along with the dissociation constant.Tachyplesin oligomers showed an increasing trend in docking scores, with the best score observed for its heptamers.The binding affinity and dissociation constant were also favourable.The best docking score for WAM-1 was observed for its dimer-MurE complex, which was À 275.23.The binding affinities and dissociation constant were also favourable.The best binding affinity was observed for the trimer-MurE complex with a value of À 9.8 kcal/mol and a dissociation constant of 6.2E À 08 M. In a comparative analysis of all the AMPs; it was also observed that the oligomeric state bactenecin (pentamer and hexamer) and WAM-1 (Trimer) showed very good interaction with its selected molecular targets MurB and MurE respectively.All these results show that the oligomeric forms of the selected AMPs interact well with the investigated probable targets and hence support the experiment's hypothesis.Furthermore, combining the results of amyloid forming capabilities of the AMPs and the interaction of these oligomers with probable targets, we can conclude that Bactenecin and Tachyplesin III showed the best interaction with the targets MurB and 33-36 kDa OMP, respectively.

Discussion
Antimicrobial resistance is on the rise as a result of improper antibiotic use.As a result, more powerful antibiotics must be created, yet their subsequent overuse creates a positive feedback loop that encourages the spread of AMR (Au et al., 2021).Because of antibiotic resistance's emergence and spread, A. baumannii is becoming one of the most challenging nosocomial infection pathogen.The World Health Organization listed A. baumannii in the priority or critical pathogen list, highlighting the substantial threat it poses to the public health (Asokan et al., 2019).A. baumannii has developed resistance to almost all antibiotics available, and multidrug resistance is well established.The emergence of antibiotic resistance has reduced the number of effective therapeutic drugs for A. baumannii infection (Xie et al., 2018).Now, many alternative therapies are under development to either prevent or treat antimicrobial resistance.These include phage therapy, lysins, antibodies, antimicrobial peptides, etc. (O'Neill, 2016).Phage therapy has several limitations.Due to the great specificity of phages for bacteria, it is necessary to keep enough variety of phages on hand to locate one, which is likely to help treat an infection.In addition, to isolating, characterising, and purifying phages, producers must also stabilise them to prevent them from mutating.Cross-contamination is another issue for producers, as phages might interact with bacteria that aren't targeted (Brives & Pourraz, 2020).Similarly, vaccines against antimicrobial-resistant pathogens are being developed, although vaccines for most antimicrobial-resistant diseases are currently unavailable.Developing vaccines take a long time (Micoli et al., 2021).
Because of their promise as a potential solution to the impending health tragedy of antibiotic resistance, many research groups are examining AMPs to determine how they exert their antimicrobial actions.Antimicrobial peptides are protein compounds found in all animals' innate immune systems.They contain broad-spectrum antibacterial activities and aid in preventing infection in the body.Furthermore, they do not induce resistance since their interactions with bacterial components do not involve particular protein binding sites i.e., they have not been observed to show specificity to any particular protein target (Wimley & Hristova, 2011).As the bacterial membrane is usually the primary target of most AMPs, to develop resistance against the AMPs, the bacteria need to change their membrane composition, which can prove to be harmful to them.AMPs have been preferred over antibiotics due to AMPs' diverse ways of action against bacteria as compared to the antibiotics' usage of a fixed set of targets; hence, it is possible to explain the absence or gradual development of resistance to microorganisms (Nicolas, 2009;Yu et al., 2018).Because AMPs are broken down into amino acids, they are thought to be less toxic compared to other treatments, which may produce potentially hazardous metabolites (Rima et al., 2021).AMPs are easier to synthesise and do not affect microbiota as compared to antibiotics (Rima et al., 2021).
Proteins can exist in several forms-native state, unfolded state, partially folded state, intrinsically disordered or in aggregates.Intrinsically disordered proteins (IDP) are unable to fold into a proper three-dimensional structure and can form amyloids.The complete protein can be disordered, or only specific regions can be disordered.One of the disadvantages of AMPs is that they are very unstable.Because AMPs primarily exploit electrostatic interactions with bacterial membranes to adopt ordered structures, ionic strength in the medication solution might alter structural stability.The cationic characteristics of the salt can be neutralised, preventing the initial electrostatic contact of AMPs with the bacterial surface (Kang et al., 2014).This might result in denaturation and activity loss of the AMP.For this reason, it is important to choose AMPs that are intrinsically disordered.If the peptides already lack an ordered 3D structure, their activity would not be affected by factors affecting their structure.Intrinsically disordered proteins play an important role in many cellular signalling and regulation.Disordered sequences can bind to multiple targets (Wright & Dyson, 2015).It is interesting to note that an IDP folds when it has to function, and this folding is influenced by its surrounding environments and interaction with other molecules like proteins or nucleic acids.A pre-requisite for an IDP to aggregate is by taking up a partially folded structure (Uversky et al., 2008).So even if the AMPs considered here are intrinsically disordered, to function as an antimicrobial and to attach to a target protein, it has to partially fold into a molten globular or pre-molten globular state.Positively charged intrinsically disordered AMPs show antimicrobial activity against methicillin-resistant S. aureus, P. aeruginosa, and antifungal activity too (Latendorf et al., 2019).
Similarly, the amyloid structure is based on connections among main polypeptide chains, unlike many protein assemblies that can generate intermolecular contacts only when each polypeptide chain folds into a higher-order structure.As a result, the creation of amyloid fibrils does not require folding each constituent protein or peptide into a specific tertiary structure (Chatani et al., 2021).Hence, the intrinsically disordered AMPs capable of forming amyloids will be far more stable than regular AMPs.In a study, exploring the potential of aggregating peptides as antibiotics, it was observed that aggregating peptides specifically targeted bacteria within mammalian cells.The aggregating peptides proved to be toxic to the intracellular bacterial cells, but the mammalian host cell remained fully healthy (Bednarska et al., 2016).Therefore even though the AMPs that we screened included peptides like Cercopin P1 with a MIC of 1.6 mg/ml, CA-MEL, Pexiganan, NK-2, NK-27 with a MIC of 2 mg/ml each (Supplementary Table 1), and many other AMPs with lesser MICs compared to the selected AMPs, we did not select them because they lack the intrinsic disorder and amyloidforming properties, making them comparatively less stable.There are reports of how some of the AMPs enter the bacterium (Nicolas, 2009;Park et al., 1998) hence, they might have some intracellular targets, which are not identified yet for most of them.Hence, we have selected 17 probable targets on A. baumanni, including targets inside the cell and on the membrane.
Ab initio modelling of AMPs, and homologous modelling of selected target proteins were performed, and the results were validated.It is interesting to note that some of the AMPs have partial secondary structure but bactenecin, Tachyplesin, and DP7 do not have any partial secondary structure (see Figure 1).Some of the natural intrinsically disordered proteins (e.g., LL-37, melittin, etc.) are known to take up a-helix or b-sheet structures (Argudo & Giner-Casares, 2021).The modelled AMPs were then docked against all the targets, and the docking score, binding energy and dissociation constant for each complex were calculated.The top two probable targets for each AMP selected based on the best scores were further submitted for molecular dynamics simulation.Based on each complex's RMSD and RMS fluctuation values, we finally decided the most probable target for each of our selected AMPs.MurB is a potential target for the AMPs Bactenecin, Cath BF and NA-CATH.The OMP33-36 can be one of the targets for the antimicrobial activity of the AMPs Citropin 1.1 and Tachyplesin.MurE and Porin SubF are the most probable targets for WAM-1 and DP7, respectively (Figure 4).As the molecular targets selected are significant for peptidoglycan synthesis, nutrient synthesis, and other essential bacterial activities, blocking these targets by using AMPs may justify their action.Further, by looking at the amino acid sequence of AMPs (supplementary table 1), it was observed that all the seven peptides have R or K residues which provide a positive charge to AMPs at physiological pH.They contribute to the electrostatic interaction and hydrogen bond interactions.Due to the more favourable interactions of the guanidinium moiety of arginine with membranes, lysine-to-arginine substitution further increases the antibacterial activity of several natural AMPs (Luong et al., 2018).It has also been seen that Trp residues function as Arg-rich AMPs' natural aromatic activators through ionpair interactions, which improve peptide-membrane interactions (Huan et al., 2020).The presence of Trp was observed in Tachyplesin and DP7, which may somehow affect their activity.The ROS-dependent mechanism is also used by many antimicrobials (Tiwari et al., 2020;Verma et al., 2022), but there is not much research related to the ability of these specific AMPs to react with ROS.But there is evidence of other AMPs in the literature that stimulate ROS generation, like the AMPs Mo-CBP3-PepI, Mo-CBP3-PepII, and Mo-CBP3-PepIII (Oliveira et al., 2019).Some AMPs are also capable of inducing apoptosis by reactive oxygen species (Jaeyong et al., 2012).
The protein MurB or UDP-N-acetylenolpyruvoylglucosamine reductase was observed to be the most probable target for three of the selected AMPs, namely Bactenecin, Cath BF and NA-CATH.UDP-N-acetylenolpyruvoylglucosamine reductase is an oxoreductase belonging to the Mur family which contains around eight enzymes.This enzyme is located in the cytoplasm of A. baumanii and has a vital role in the peptidoglycan biosynthetic pathway.In the last step of the peptidoglycan biosynthesis, MurB acts as a catalyst and, in presence of NADPH, catalyses UDP-N-acetylmuramic acid (UDP-MurNAc) formation from enolpyruvyl-UDP-N-acetylglucosamine.Flavin adenine dinucleotide (FAD) acts as a cofactor in this reaction (Amera et al., 2020).Mutant strains of S. aureus demonstrated decreased peptidoglycan synthesis in a study, which is consistent with MurB protein having UDP-N-acetylenolpyruvylglucosamine (UDPGlcNAcEP) reductase activity, which is critical in peptidoglycan biosynthesis.Furthermore, mutant cellular extracts displayed reduced specific UDPGlcNAcEP reductase activity compared to wild-type extracts (Matsuo et al., 2003).Hence, mutating the MurB gene in a bacteria inhibits its growth.So, MurB being the most probable intracellular target for the AMPs-Bactenecin, Cath BF, and NA-CATH means that these three AMPs might work by intracellularly binding to MurB and inhibiting cell wall synthesis in A. baumannii.
33-36 kDa outer membrane protein (3336OMP) is an aquaporin located on the outer membrane of A. baumannii, which plays a role in imipenem and carbapenem resistance in the bacteria (Clark, 1996).It acts as a channel or passage for water and other nutrients into the cell and the removal of toxins in the cell.It has a b-barrel structure (Rumbo et al., 2014).Studies have observed that reduced synthesis or expression of the 33-36 kDa OMP was also linked to resistance to amikacin, imipenem, tobramycin, ceftazidime and ciprofloxacin among A. baumannii isolates (Clark, 1996).This protein was the most probable target of the AMPs Citropin 1.1 and Tachyplesin.This indicates that these AMPs work by binding to a membrane target on A. baumannii and probably cause lysis of the bacterial cell by forming pores on the outer membrane.
Porin Subfamily Protein (PorinSubF) is located in A. baumannii's outer membrane.Porins are proteins that can generate channels allowing molecules to pass through bacteria's lipid bilayer membranes with poor hydrophilic solute permeability.The first stage in an antimicrobial's travel to its target site is translocation over the porin channel.Certain chemical compounds known as polyamines have been observed to block the porins in bacteria and thus reduce the cell membrane's permeability (Iyer & Delcour, 1997).This protein was identified as most probable target of DP7 AMP, hence DP7 might attach to a surface protein on A. baumannii and cause the death of the bacteria by disrupting the membrane.
MurE or UDP-N-acetylmuramoyl-L-alanyl-D-glutamate-2,6diaminopimelate ligase is a Mur family enzyme that aids in the formation of cell walls.This protein is encoded by the MurE gene and is present in the cytoplasm.During the synthesis of peptidoglycan in the bacterial cell wall, it acts as a catalyst to add meso-diaminopimelic acid to the precursor of nucleotide, UDP-N-acetylmuramoyl-L-alanyl-D-glutamate (UMAG) (Amera et al., 2019).Researchers are trying to develop inhibitors of Mur enzymes implicated in peptidoglycan formation to combat several bacterial illnesses.MurC, MurD, MurE, and MurF inhibitors include phosphonates, phosphinates, and sulphonamides.A study showed that an antibacterial from Hypericum acmosepalum could inhibit MurE, which assisted in stopping Mycobacterium tuberculosis growth (Osman et al., 2012).MurE was the probable target for the antimicrobial peptide WAM-1 in our current study.So, WAM-1 might have MurE, as a likely intracellular target, and thus it might work against A. baumannii by inhibiting its cell wall biosynthesis process.
Based on the interaction study of the AMP oligomers with the targets, the best interaction was observed for Tachyplesin III with 3336 OMP and for Bactenecin with MurB.Tachyplesin III is a cationic AMP which shows antimicrobial activity against A. baumannii (Neshani et al., 2020).This AMP was first isolated from horseshoe crab (Muta et al., 1990).Tachyplesins typically work to kill bacteria in a number of ways, the most notable of which are: making bacterial membranes more permeable; possibly attaching to DNA, and entering the microbial cell and affecting the synthesis of unsaturated fatty acids by binding and inhibiting the 3ketoacyl carrier protein reductase hence potentially disrupting membrane integrity (Neshani et al., 2020).If Tachyplesin binds to 33-36 kDa OMP as per our results, the mechanism it targets might be making the bacterial membranes more permeable, which can further disrupt membrane integrity and assist in forming pores.Bactenecin is a bovine derived antimicrobial peptide and is also cationic in nature (Romeo et al., 1988).Studies on various bacteria have demonstrated that Bactenecin's primary antibacterial activity is attributable to membrane damage caused by depolarization and permeabilization.Additionally, it is assumed that Bactenecin targets intracellular targets via penetrating the membrane (Lee et al., 2009).Binding of Bactenecin to MurB can lead to inhibition of the cell wall biosynthesis pathway.Further experimental studies have to be performed to validate these in silico results.
There are several studies going on currently regarding AMPs and their mode of action.The fungal defensin, plectasin works by attaching directly to the bacterial cell-wall precursor Lipid II.The synthesis of cell wall is the route targeted by plectasin, according to various genetic and biochemical techniques.Lipid II was identified as the unique cellular target in vitro cell-wall production experiments (Schneider et al., 2010).Cathelicidin-BF is a cationic, amphiphilic, and a-helical AMP with high antibacterial activity against MDR clinical pathogens.However, due to its in vivo protease sensitivity, it cannot be utilised systemically.Through studies, it was observed that Cathelicidin-BF might have multiple intracellular targets (Liu et al., 2017).As the selected AMPs have already been tested for their antibacterial activity, the present study will add more information to their antibacterial role and future use as therapeutics against A. baumannii.

Conclusion and future prospects
With the rise in resistance against traditional antibiotics in microorganisms, AMPs are emerging as promising therapeutic drugs because of their capacity to cause delayed resistance.Due to a synergistic action, when combined with conventional antibiotics, AMPs have been shown to inhibit drug-resistant bacteria previously resistant to the antibiotics.Overall, the current data imply that AMPs are intriguing molecules that could be effective A. baumannii inhibitors and provide useful information for further experimental studies.We investigated the probable targets for seven such AMPs.MurB is a potential target for the AMPs Bactenecin, Cath BF and NA-CATH.The OMP33-36 can be one of the targets for the antimicrobial activity of the AMPs Citropin 1.1 and Tachyplesin.MurE and Porin SubF are the most probable targets for WAM-1 and DP7, respectively.As the molecular targets selected are significant for peptidoglycan synthesis, nutrient synthesis, and other essential bacterial activities, blocking these targets by using AMPs may justify their action.Studying further interaction of the AMP oligomers suggested MurB and 3336 OMP to be the most probable target of Bactenecin and Tachyplesin.
In this study, we could only consider seventeen molecular drug targets of A. baumannii to find the probable targets of our selected AMPs.In future, more targets can be considered, which may prove to be better targets for the AMPs.The present study mainly focuses on in silico study, so further experimental validation needs to be done before confirming whether the probable targets for each AMP are actually acting as its target.Using AMPs in conjunction with standard antibiotics may give a strategy for overcoming resistance caused by permeability constraints or drug efflux, as well as extending the usable life of older and novel compounds.This vast field of study offers numerous avenues for additional investigation.
Along with their advantages, these AMPs come with some limitations too.They have a very short half-life in the serum.Once they enter the cell, they are susceptible to easy degradation by proteases.Some peptides may also be inactive at the cell's physiological salt concentrations.They may also be unstable in the cell.Many of them are cytotoxic to the hosts, including humans.There have not been many studies regarding the targeted delivery of these AMPs into the cell.It's also challenging to deliver these AMPs to the infection's target site (Sarkar et al., 2021).
Antimicrobial peptides have long been investigated as a potential alternative to antibiotics.However, there have been a few notable successes.Our inability to characterise their mode of action, which is valuable for future design and engineering applications, might be a critical factor in developing new antimicrobial peptide medicines.Because of their adverse pharmacokinetics, possible immunogenicity, and metabolic instability, peptides are used as therapeutics infrequently.Their ability to engage in extremely precise interactions with receptor proteins, on the other hand, makes them extremely useful tools in drug development.Chemical changes in the peptide backbone and amino acid side chains can be used to create small molecular weight peptidomimetics (compounds that functionally and structurally imitate peptides and have drug-like effects).

Methods
Identification of probable molecular targets against intrinsically disordered amyloid-forming antimicrobial peptides was initiated by selecting antimicrobial peptides showing antimicrobial activity against A. baumannii through a literature survey.Similarly, molecular targets present in A. baumannii were also screened.For this, we performed a systemic search on PubMed and Google Scholar search engines.All the articles in the search results and their references were screened to identify the required AMPs and molecular targets.

Prediction of intrinsic disorder and amyloid-forming properties of AMPs
The antimicrobial peptides showing activity against A. baumannii were then subjected to web servers DisEMBL (Linding et al., 2003) and GlobPlot2 (Linding et al., 2003).These two servers predicted which antimicrobial peptides were intrinsically disordered or had a disordered region in their sequence.DisEMBL is a programme that predicts disordered or unstructured areas in protein sequences.The prediction is done on the basis of several parameters that are already set in the tool, like the hot loops prediction (Linding et al., 2003).This tool uses three criteria to predict the disordered residues in any peptide sequence.(a) Loops/coils-protein disorders are mostly found in loops, so in this criteria, they assigned all the residues that form an alpha helix, 3 10 helix or beta strand to be ordered.The rest of the states were termed as loops/coils.(b) Hot loops-These are the same as above.C-alpha temperature determines which loops have a high degree of mobility and are considered disordered.(c) Missing coordinates (Remark À 465 PDB entries)-These are non-assigned electron densities frequently indicative of intrinsic disorder (Linding et al., 2003).GlobPlot is an online tool that lets users see how a protein's predisposition for order/globularity and disorder changes over time.It successfully recognises inter-domain segments, including linear motifs and ostensibly organised sections that lack a recognised domain.
GlobPlot has been successfully used to improve the expression, purification, and crystallisation of constructs, discover intrinsically disordered proteins (IDPs/IUPs) and forecast regions abundant in linear motifs (Linding et al., 2003).
The antimicrobial peptides showing intrinsic disorder were then further screened using web tools PASTA 2.0 (Walsh et al., 2014) and AGGRESCAN (Conchillo-Sol� e et al., 2007) to select those which could form amyloids.Prediction of amyloid structural aggregation (PASTA 2.0) is a web tool based on sequence-based parameters.For numerous input sequences, it predicts the most aggregation-prone sections and the related b-strand inter-molecular pairing.Furthermore, it accurately predicts intrinsic protein disorder and secondary structure, allowing for the evaluation of complementary sequence features.We can input our desired peptide sequence in FASTA format in the tool.As output, it shows a percentage of a-helix, coils, b-strand, intrinsic disorder and the best aggregation pairing energy (Walsh et al., 2014).Another software AGGRESCAN was also used to predict the amylogenic property of AMPs.AGGRESCAN is an online programme which predicts aggregation-prone portions in protein sequences, analyses the influence of mutations on protein aggregation propensities, and compares the aggregation properties of other proteins or sets of proteins.We can input our desired peptide sequence in the server in FASTA format, giving us the output showing the probable amyloid forming hotspot regions in the sequence.

Modelling of AMPs and molecular targets protein
The three-dimensional structures of the selected intrinsically disordered amyloid forming antimicrobial peptides and molecular targets were generated using the web-tools I-TASSER (Yang & Zhang, 2015) and Phyre2 (Kelley et al., 2015), respectively.The structures were visualised using Chimera (Pettersen et al., 2004).Using amino acid sequence, I-TASSER generated three-dimensional atomic models using a series of threading alignments and iterative structure assembly simulations (Roy et al., 2010).Phyre2 uses powerful distant homology recognition techniques to create 3D models, identify ligand binding sites, and analyse amino-acid changes in a protein sequence (Kelley et al., 2015).

Refining of the modelled AMPs and target model protein
The GalaxyWEB Refine online tool (Ko et al., 2012) was used to refine the 3D models obtained.For each model submitted, five refined models were obtained, which were further analysed based on their RMSD, MolProbity, Rama favoured region and GDT-HA.The GalaxyWEB server uses templatebased modelling to predict protein structure from sequence and ab initio modelling to refine loop or terminal regions.After side-chain packaging changes, using short molecular dynamics simulations, GalaxyRefine remodels all side-chain conformations and continuously relaxes the structure.

Validation of the refined models of AMPs and target model protein
The refined models of the targets and the antimicrobial peptides were further analysed using PSVS (Bhattacharya et al., 2007) and ProSa Web server (Wiederstein & Sippl, 2007).The Ramachandran plots, G-factor plots, Verify 3D plots, Z-score and energy plots were used to validate each model.PSVS incorporates results from various structure-validation tools like RPF, PROCHECK, Verify3D, MolProbity, the PDB validation software and Prosa II (Bhattacharya et al., 2007).We used the PSVS server to validate the AMP and target structures.By submitting the PDB files of the proteins, we obtained their Ramachandran plots, three-dimensional profile plots, and the analysis of the G-factor for all dihedral angles.Ramachandran plot tells about the amino acid residues forming the secondary structure, whether it's an alpha helix, beta sheets, loops, or in the disallowed region.There should be over 90% most favoured regions in a good quality model.The G-factor tells us how unusual any given stereochemical property may be in a given model.The stereochemical property considered here is the dihedral angles in the peptide residues.Verify 3D tells us about the sequence assignment concerning the template structure.It also tells about whether the alpha helix, beta sheets, loops or polar contacts has been assigned properly or not.
To further validate and identify the quality scores of the 3D structures, the ProSa-web server was used.ProSA (Protein Structure Analysis) is a tool commonly used to check 3D models of protein structures for mistakes or errors.It displays scores and energy plots which help analyse the errors in the structures (Wiederstein & Sippl, 2007).The Z-score reveals how good a model is overall.Its value is plotted alongside the Z-scores of all experimentally obtained protein chains in the given PDB.Different colours distinguish groupings of structures from multiple sources (X-ray, NMR) in the Z-score plots.It can be used to see if the Z-score of the input structure is within the range of native proteins of similar size.Graphing energies as a function of amino acid sequence location, the plot of residue scores demonstrates local model quality.

Molecular docking of AMPs with the target model protein
Each selected antimicrobial peptide was docked with all the seventeen molecular targets using the web server H-DOCK (Yan et al., 2020), and docking scores for each complex were listed.To conduct protein-protein docking, the H-DOCK server employs a hybrid technique combining template-based modelling and ab initio template-free docking.Using input data, the service predicts the interaction of receptor and ligand molecules using a hybrid method of template-based and template-free docking (Yan et al., 2020).

Binding energy estimation and calculation of dissociation constant
PRODIGY server (Xue et al., 2016) estimated the binding energy and dissociation constant for each AMP with the target protein in the docked complex as per published methods (Tiwari et al., 2020(Tiwari et al., , 2022)).The PRODIGY (PROtein binDIng enerGY) web server is used to predict the binding strength of a protein-protein complex using its 3D structure.The server uses intermolecular interactions and attributes collected from non-interface surfaces.As an input, we need to provide the structure of the protein complex, and it will give us the predicted value of the dissociation constant and binding energy as an output (Xue et al., 2016).

Molecular dynamics simulation of AMPs and target proteins
The molecular dynamic simulation was performed for the top two docked complexes showing the best docking scores, binding energy and dissociation constant as per the published method (Tiwari et al., 2022).The WebGro server was used for MDS analysis using GROMOS96 43a1 force field with default settings (Christen et al., 2005).WebGro uses the GROMACS simulation programme.We have to submit the protein complex file, and the server performs the simulation and trajectory analysis.As an output, we get the RMSD analysis, RMSF analysis, Hydrogen bonds, radius of gyration, solvent-accessible surface analysis, etc. (Abraham et al., 2015).
We used 100 ns of simulation time for our simulation, and the rest of the parameters were kept as default.

Investigating oligomeric propensity of selected AMPs
The oligomer states of selected AMPs were formed using protein-protein molecular docking with the help of H-DOCK server, and docking scores were listed.Further, the oligomers' binding energy and dissociation constants were also estimated using the PRODIGY server.

Interaction of AMP oligomers with probable targets
To confirm the interaction between AMP oligomers and probable targets, oligomers for each AMP were docked to its potential target using the H-DOCK server.Binding energy and dissociation constants were also calculated to analyse the interaction between oligomers and their probable targets using the PRODIGY server.

Figure 2 .
Figure 2. Docked structures of all the selected antimicrobial peptides with their probable targets generated using H-DOCK server (a) Bactenecin-MurB complex, where bactenecin is represented in pink colour and MurB in cyan.(b) Cath BF-MurB complex, where CathBF is represented in yellow colour and MurB in purple.(c) DP7-Porin complex, where DP7 is represented in green colour and porin in gold.(d) Citropin 1.1-33-36kDa OMP complex, where Citropin1.1 is represented in red colour and 33-36 kDa OMP in cyan.(e) NA-CATH-MurB complex, where NA-CATH is represented in cyan colour and MurB in pink.(f) Tachyplesin-33-36 kDa OMP complex, where tachyplesin is represented in red colour and 33-36 kDa OMP in yellow.(g) WAM1-MurE complex, where WAM1 is represented in yellow colour and MurE in green.

Figure 4 .
Figure 4. Probable molecular targets of intrinsically disordered amyloidogenic AMPs in Acinetobacter baumannii based on the present in silico study.

Table 9 .
In the case of the docking of Tachyplesin with the targets, the top four complexes with the best docking scores are with CARomp (À 284.69),AdeB,MurE and 33-36 kDa Omp.Their docking scores are highlighted in Supplementary Table 10.In the case of the docking of WAM-1 with the targets, the top three complexes with the best docking scores are AdeB (À 286.54), 33-36kDaOmp and Porin.Their docking scores are highlighted in Supplementary Table11.

Table 1 .
List of probable molecular drug targets in Acinetobacter baumannii.

Table 2 .
Docking score, binding affinity and dissociation constant of selected AMPs with their molecular targets.

Table 3 .
Docking score, binding energy and dissociation constant values for different oligomeric state of AMPs.

Table 4 .
Docking score, binding energy and dissociation constant values of the complex of the AMP oligomers and their probable targets.