In Vitro Assessment on Designing Novel Antibiofilms of Pseudomonas aeruginosa Using a Computational Approach

An anti-biofilm that can inhibit the matrix of biofilm formation is necessary to prevent recurrent and chronic Pseudomonas aeruginosa infection. This study aimed to design compounds with a new mechanism through competitive inhibitory activity against phosphomannomutase/phosphoglucomutase (PMM/PGM), using in vitro assessment and a computational (in silico) approach. The active site of PMM/PGM was assessed through molecular redocking using L-tartaric acid as the native ligand and other small molecules, such as glucaric acid, D-sorbitol, and ascorbic acid. The docking program set the small molecules to the active site, showing a stable complex formation. Analysis of structural similarity, bioavailability, absorption, distribution, metabolism, excretion, and toxicity properties proved the potential application of ligands as an anti-biofilm. In vitro assessment with crystal violet showed that the ligands could reach up to 95.87% inhibition at different concentrations. The nitrocellulose membrane and scanning electron microscopic visualization showed that the untreated P. aeruginosa biofilm was denser than the ligand-treated biofilm.


Introduction
Pseudomonas aeruginosa (P. aeruginosa) is an environmental, opportunistic bacterium associated with various diseases with various clinical spectrums. It primarily causes ventilator-associated pneumonia (VAP) with high mortality/morbidity rates [1][2][3][4]. VAPassociated biofilm is resistant to antibiotics, and it has a protective capability from host immunity, thereby inducing recurrent and chronic infection. Thus, it is difficult to treat [5,6]. Novel strategies have been developed to manage P. aeruginosa biofilms, particularly in Pseudomonas biofilm-associated pneumonia.
Biofilm is a virulence factor of P. aeruginosa, which promotes bacterial resistance, thereby making bacteria more difficult to eradicate. Biofilm is a community of microorganisms encased in a matrix of extracellular polymeric substances (EPSs). Exopolysaccharides are essential components of EPS. Bacterial biofilms are sessile cellular communities. Hence, their properties differ from their free form (planktonic). Moreover, they are highly resistant to immune response and thousand times more resistant to antibiotics [6][7][8]. Antibiotics can kill planktonic bacteria; however, sessile bacteria in the form of biofilm must be eradicated with an anti-biofilm to ensure the life cycle of biofilm formation, the physiological properties of the bacteria, and the accessibility of biofilm. Furthermore, at present, no specific drugs have been developed for biofilm eradication.
P. aeruginosa biofilm contains three main exopolysaccharides: alginate, polysaccharide synthesis locus, and pellicle formation. They play an essential role in maintaining the

Determination of the PMM/PGM Active Site via Redocking Analysis
The selected sphere cluster was taken at a radius of 10.0 Å from the TLA coordinates, and the cluster sphere could represent the active site receptor region ( Figure 1A). Redocking analysis showed good superposition with the root mean square deviation (RMSD) value less than 2.0 Å ( Figure 1B). In addition, the TLA coordinates obtained from redocking (pose) were similar to the TLA in the crystal form. Therefore, the RMSD value close to 0 Å indicated better pose coordinates.
Molecules 2022, 27, x FOR PEER REVIEW 3 of 17 application of novel therapeutic strategies (combined antibiotics and competitive inhibitor-based anti-biofilm) against P. aeruginosa-associated pneumonia.
EvdW and Eele indicated that electrostatic energy had the most significant contribution to binding affinity, and these two energy components in the native receptor-ligand interaction TLA conformation showed an excellent binding affinity to a grid score of −71.288 kcal/mol ( Figure 1C). In addition, ligand-residue interaction analysis at the receptor active site identified three amino acid residues (Lys114, Arg243, and His304) and one non-standard residue (Zn +2 ) as a cofactor of PMM/PGM ( Figure 1C). Moreover, as presented in Figure 1D, ligand-receptor interactions had the following five hydrogen bonds: E vdW and E ele indicated that electrostatic energy had the most significant contribution to binding affinity, and these two energy components in the native receptor-ligand interaction could be assessed via footprint analysis [17,19]. The analysis result was obtained by comparing each energy between crystal (reference) and redocked (pose) ligands ( Figure 2). could be assessed via footprint analysis [17,19]. The analysis result was obtained by comparing each energy between crystal (reference) and redocked (pose) ligands ( Figure 2).

Figure 2.
Footprint analysis of the primary amino acid residues responsible for the interaction between L-tartaric acid and phosphomannomutase/phosphoglucomutase. Footprint analysis of three amino acid residues proved an excellent binding affinity, which is supported by the EvdW + Eele value (kcal/mol) for each residue, which was less than zero (negative value). For example, EvdW and Eele of TLA are −13.776 and −57.512 kcal/mol, respectively.

Ligand Exploration by Using TLA as Template
Based on redocking analysis of the active site of PMM/PGM and the grid score of TLA, we identified three ligands i.e glucaric acid [GA], D-sorbitol [DS], and ascorbic acid [AA] via ligand exploration based on the established criteria. These criteria include nontoxic, easy to obtain and reported safety for use ( Figure 3).  Footprint analysis of three amino acid residues proved an excellent binding affinity, which is supported by the E vdW + E ele value (kcal/mol) for each residue, which was less than zero (negative value). For example, E vdW and E ele of TLA are −13.776 and −57.512 kcal/mol, respectively.

Ligand Exploration by Using TLA as Template
Based on redocking analysis of the active site of PMM/PGM and the grid score of TLA, we identified three ligands i.e glucaric acid [GA], D-sorbitol [DS], and ascorbic acid [AA] via ligand exploration based on the established criteria. These criteria include nontoxic, easy to obtain and reported safety for use ( Figure 3). could be assessed via footprint analysis [17,19]. The analysis result was obtained by comparing each energy between crystal (reference) and redocked (pose) ligands ( Figure 2). Footprint analysis of three amino acid residues proved an excellent binding affinity, which is supported by the EvdW + Eele value (kcal/mol) for each residue, which was less than zero (negative value). For example, EvdW and Eele of TLA are −13.776 and −57.512 kcal/mol, respectively.

Ligand Exploration by Using TLA as Template
Based on redocking analysis of the active site of PMM/PGM and the grid score of TLA, we identified three ligands i.e glucaric acid [GA], D-sorbitol [DS], and ascorbic acid [AA] via ligand exploration based on the established criteria. These criteria include nontoxic, easy to obtain and reported safety for use ( Figure 3).  Phosphomanomutase/phosphoglucomutase competitive inhibitors as potential ligand candidates.  The higher the negative grid score, the more vital the ligand interaction on the receptor's active site from a thermodynamic aspect. The ligands and receptor showed conformation fitness, interacting via several receptor residues on the receptor's active site, including GA-PMM/PGM (Arg11, Arg16, His105, Arg243, His304, His325, and Zn), DS-PMM/PGM (Arg11, Lys114, His304, and Zn), and AA-PMM/PGM (Arg16, His105, Lys114, Asp242, His325, and Zn). In particular, the hydrogen bond interaction with each ligand plays an important role in the target protein inhibition mechanism because the H-bond parameter is a strong category in ligand-receptor interactions [20,21]. The molecular docking analysis results showed potential H-bond interactions with amino acid residues to the receptor's active site ( Figure 4C  The grid score of each ligand was determined (TLA, −71.288; GA, −74.762; DS, −74.741; and AA, −32.495). This score is associated with the binding affinity (kcal/mol). The higher the negative grid score, the more vital the ligand interaction on the receptor's active site from a thermodynamic aspect. The ligands and receptor showed conformation fitness, interacting via several receptor residues on the receptor's active site, including GA-PMM/PGM (Arg11, Arg16, His105, Arg243, His304, His325, and Zn), DS-PMM/PGM (Arg11, Lys114, His304, and Zn), and AA-PMM/PGM (Arg16, His105, Lys114, Asp242, His325, and Zn). In particular, the hydrogen bond interaction with each ligand plays an important role in the target protein inhibition mechanism because the H-bond parameter is a strong category in ligand-receptor interactions [20,21]. The molecular docking analysis results showed potential H-bond interactions with amino acid residues to the receptor's active site ( Figure 4C   H-bond interactions were further visualized in the form of a hydrogen bond acceptor (HBA) and hydrogen bond donor (HBD). As shown in Figure 5, the visualization was supported by classifying HBA and HBD areas on the receptor's active site. Meanwhile, analysis of the hydrophobic area on the active site showed that each inhibitor was in a hydrophilic area ( Figure 6). In addition, this result was indicated by a low hydrophobicity value (<0, blue area) for each ligand [22]. H-bond interactions were further visualized in the form of a hydrogen bond acceptor (HBA) and hydrogen bond donor (HBD). As shown in Figure 5, the visualization was supported by classifying HBA and HBD areas on the receptor's active site. Meanwhile, analysis of the hydrophobic area on the active site showed that each inhibitor was in a hydrophilic area ( Figure 6). In addition, this result was indicated by a low hydrophobicity value (<0, blue area) for each ligand [22].   H-bond interactions were further visualized in the form of a hydrogen bond acceptor (HBA) and hydrogen bond donor (HBD). As shown in Figure 5, the visualization was supported by classifying HBA and HBD areas on the receptor's active site. Meanwhile, analysis of the hydrophobic area on the active site showed that each inhibitor was in a hydrophilic area ( Figure 6). In addition, this result was indicated by a low hydrophobicity value (<0, blue area) for each ligand [22].   Hydrophobic pocket areas for molecular docking of phosphomannomutase/phosphoglucomutase by several ligands. Brown and blue indicate the hydrophobic and hydrophilic pocket areas, respectively.

Prediction of Drug-Likeness, Bioavailability, Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET)
Data on the prediction of drug-likeness of each ligand molecule with a competitive inhibitory activity against PMM/PGM based on docking analysis are shown in Table 1. All assayed molecules had good permeability as a potential anti-biofilm. However, one violation occurred in the topological polar surface area (TPSA) and HBD parameters for GA and DS inhibitors, respectively. Furthermore, pharmacokinetic analysis predicted that each ligand molecule had good permeability as a potential anti-biofilm because of the presence of only one violation. The violations occurred in TPSA and HBD for GA and DS, respectively.  Figure 7 shows the prediction of oral bioavailability, which can provide theoretical information about the physicochemical properties of each potential anti-biofilm. The parameters measured included lipophilicity (−0.7 < X logP3 < 5.0), size (150 D < Mw < 500 D), polarity (20 Å 2 < TPSA < 130 Å 2 ), insolubility (0 < ESOL < 6), saturation (0.25 < Csp3 < 1), and flexibility (0 < rot bonds < 9) [23,24]. Furthermore, based on the prediction results, GA had one violation compared with TLA, DS, and AA (had no violation; Tables 1 and 2). Overall, the bioavailability score shows that each inhibitor has good pharmacokinetic properties with a value of 0.55-0.56.

Prediction of Drug-Likeness, Bioavailability, Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET)
Data on the prediction of drug-likeness of each ligand molecule with a competitive inhibitory activity against PMM/PGM based on docking analysis are shown in Table 1. All assayed molecules had good permeability as a potential anti-biofilm. However, one violation occurred in the topological polar surface area (TPSA) and HBD parameters for GA and DS inhibitors, respectively. Furthermore, pharmacokinetic analysis predicted that each ligand molecule had good permeability as a potential anti-biofilm because of the presence of only one violation. The violations occurred in TPSA and HBD for GA and DS, respectively.  Figure 7 shows the prediction of oral bioavailability, which can provide theoretical information about the physicochemical properties of each potential anti-biofilm. The parameters measured included lipophilicity (−0.7 < X logP3 < 5.0), size (150 D < Mw < 500 D), polarity (20 Å 2 < TPSA < 130 Å 2 ), insolubility (0 < ESOL < 6), saturation (0.25 < Csp3 < 1), and flexibility (0 < rot bonds < 9) [23,24]. Furthermore, based on the prediction results, GA had one violation compared with TLA, DS, and AA (had no violation; Tables 1 and 2).
Overall, the bioavailability score shows that each inhibitor has good pharmacokinetic properties with a value of 0.55-0.56.    Table 3 shows the results of predictive analysis of ADMET. The prediction of absorption indicated that the anti-biofilm candidate had a low absorption ability on average, given its poor Caco-2 permeability and negative intestinal absorption-Human (HIA). However, AA had a positive HIA category > 30%, which indicated that AA is absorbed in the small intestine of humans. Distribution parameter analysis of each anti-biofilm candidate could predict blood-brain barrier (BBB) permeability (category: negative BB) with a log BB of less than −1.0. Therefore, all anti-biofilm candidates were not distributed, and they did not affect the central nervous system. Table 3. Absorption, distribution, metabolism, excretion, and toxicity (ADMET) prediction properties using the pkCSM server. Meanwhile, the predicting metabolic parameters showed that the antibiofilm candidates did not inhibit cytochrome isoenzymes (CYP), including CYP1A2, CYP2C19, CYP2C9, CYP2D6, and CYP3A4. The absence of these isoenzymes indicated that each antibiofilm candidate was a promising drug candidate because it did not interfere with the CYP activity. Therefore, the antibiofilm candidate will not induce unwanted side effects on metabolic processes. In addition, the prediction of excretion showed that antibiofilm candidates were included in renal organic cation transporter-2 (OCT2) prediction, a subcategory of the non-substrate.

Parameters
Analysis of toxicity showed remarkable results, where all antibiofilm candidates showed non-toxicity and met the following criteria: non-AMES toxic, non-hepatotoxicity, and non-skin sensitization [25]. Thus, overall prediction with the pkCSM server showed that all ligands had promising ADMET, which could be considered as drug candidates.
2.5. In Vitro Assessment of Ligands against P. aeruginosa Biofilm 2.5.1. Microtiter Plate Assay The activity of each ligand as an anti-biofilm was evaluated at several concentration ranges using crystal violet assay. Based on the data presented in Table 4, one of the ligands could reach up to 95.87% inhibition at different concentrations. The minimal biofilm inhibitory concentration (MBIC) of TLA had >50% anti-biofilm activity at 2 × 10 3 µg/mL. By contrast, GA or calcium D-saccharate tetrahydrate had >50% inhibitory biofilm formation activity at 2.5 × 10 2 µg/mL. DS had a >50% inhibitory effect on biofilm formation at 4 × 10 5 µg/mL. AA had a >50% inhibitory activity against P. aeruginosa biofilm formation at 1.25 × 10 4 µg/mL, and its inhibitory activity reached 91.5% (Figure 8). and non-skin sensitization [25]. Thus, overall prediction with the pkCSM server showed that all ligands had promising ADMET, which could be considered as drug candidates.

Microtiter Plate Assay
The activity of each ligand as an anti-biofilm was evaluated at several concentration ranges using crystal violet assay. Based on the data presented in Table 4, one of the ligands could reach up to 95.87% inhibition at different concentrations. The minimal biofilm inhibitory concentration (MBIC) of TLA had >50% anti-biofilm activity at 2 × 10 3 µ g/mL. By contrast, GA or calcium D-saccharate tetrahydrate had >50% inhibitory biofilm formation activity at 2.5 × 10 2 µ g/mL. DS had a >50% inhibitory effect on biofilm formation at 4 × 10 5 µ g/mL. AA had a >50% inhibitory activity against P. aeruginosa biofilm formation at 1.25 × 10 4 µ g/mL, and its inhibitory activity reached 91.5% (Figure 8).   Figure 9 shows the macroscopic morphology of the P. aeruginosa biofilm treated with each ligand upon the nitrocellulose membrane. Moreover, each ligand-treated biofilm was smaller and thinner than the biofilm produced in the control experiments without ligand treatment.  Figure 9 shows the macroscopic morphology of the P. aeruginosa biofilm treated with each ligand upon the nitrocellulose membrane. Moreover, each ligand-treated biofilm was smaller and thinner than the biofilm produced in the control experiments without ligand treatment.     Figure 9 shows the macroscopic morphology of the P. aeruginosa biofilm treated with each ligand upon the nitrocellulose membrane. Moreover, each ligand-treated biofilm was smaller and thinner than the biofilm produced in the control experiments without ligand treatment.     Figure 10a shows that the untreated P. aeruginosa biofilm comprised bacterial cells with some exopolysaccharides (EPSs). The bacterial cell walls appeared intact and closed together, attached, and covered by an EPS matrix. Figure 10b-e shows the ability of ligands to reduce biofilm formation. The bacterial cells were grown on coverslips for 8 h in the attachment stage, continually treated with the addition of ligands in Brain Heart Infusion Broth (BHIB) media, and incubated for 24 h. Rodrigues et al., (2017) showed the characteristics of polymicrobial biofilms associated with pneumonia, with bacterial-fungal polymicrobial biofilms colonizing the endotracheal surface. P. aeruginosa, which is commonly associated with pneumonia, interacts and interplays with other microbes in polymicrobial biofilms, thereby enhancing pathogenesis and affecting antimicrobial therapy. Therefore, the high mortality/morbidity rate associated with pneumonia and the increased antibiotic resistance worldwide have prompted studies on novel therapeutic strategies to combat polymicrobial infection in pneumonia. Rodrigues et al., (2017) proposed combinational antimicrobial therapy using one antibiotic (polymyxin B, PolyB) and anti-fungal (amphotericin B, AmB) agent against the polymicrobial biofilms of P. aeruginosa and Candida albicans [2]. Meanwhile, we developed an antimicrobial therapy with an anti-biofilm to eradicate the polymicrobial biofilms of P. aeruginosa and C. albicans. In addition, we evaluated the performance of several anti-biofilms of C. albicans that enhance the anti-fungal action of fluconazole [26].

Discussion
The exploration of several P. aeruginosa anti-biofilms based on the competitive inhibitors of the PMM/PGM enzyme, which is responsible for extracellular matrix biosynthesis, was assessed. In the present study, substances with a structure similar to that of the PMM/PGM active site were identified, which served as ligands for this enzyme and a macromolecule. The ligands included the stereoisomer of the native substrate of PMM/PGM, mannose, and glucose and the ligand of S108A PMM/PGM (TLA). The results showed that three substances (AA, GA, and DS) matched the criteria. Based on the structure similarity between these ligands and TLA (as the natural substrate of PMM/PGM), the ligands should occupy the active site of PMM/PGM. Thus, these ligands may induce activity as PMM/PGM competitive inhibitors, which can be predicted via in silico analysis, thereby indicating its activity through in vitro assay.
Theoretical studies on the use of substrate analogs as PMM/PGM inhibitors based on their structural similarity (mimic) to native ligands and their commercial availability can provide an efficient approach to drug design. In addition, the mechanism of PMM/PGM competitive inhibition was analyzed on the basis of the interaction of the inhibitor with residues on the receptor's active site at a molecular level. Redocking determined the active site of PMM/PGM based on its interaction with TLA as the native ligand. Moreover, sphere cluster selection using Dock6 can facilitate easier identification of native ligand coordinates at the receptor active site [27,28].
Docking of the PMM/PGM ligand was conducted at the initial coordinates of the redocking results ( Figure 4A). In addition, ligands with a binding affinity of <0 (kcal/mol) have good interactions with receptors or targeted proteins [19,20,27]. The higher the negative value, the more robust the interaction with the inhibitor. Redocking had good feasibility in determining the active site of the receptor; thus, it could be used for docking other ligands. Molecular docking analysis of the PMM/PGM ligands showed that each ligand had an excellent conformation with a negative grid score.
The performance and characteristic prediction of competitive inhibitors were investigated via in silico pharmacokinetic studies. Furthermore, the pharmacokinetics of inhibitors were evaluated via several parameters, including drug-likeness, bioavailability, and AD-MET. These parameters can provide information on biological activities based on their structure. For example, drug-likeness analysis observes drug permeability by considering Lipinski's and Veber's rules. A drug candidate has good permeability if it does not deviate by more than one Lipinski's and Veber's rule [29].
Previous pharmacokinetic studies confirmed that each inhibitor meets the criteria of oral drugs, showing that the development of an excellent oral drug had ≤2 deviations [29]. Therefore, this information will be considered in selecting potential inhibitors based on oral bioavailability. Consequently, in silico pharmacokinetics study resulted in potential antibiofilms of P. aeruginosa. Then, this prediction has been proven via quantitative and qualitative in vitro assays.
The first in vitro assay using the crystal violet method confirmed that the four ligand compounds could inhibit P. aeruginosa biofilms. Using spectrophotometric analysis and the crystal violet method, biofilm formation was determined on the basis of optical density and [30,31]. In vitro experiments showed the potential application of these compounds as anti-biofilm agents.
Tartaric acid derived from ascorbic acid as the native ligand represents aldaric acid, which can be found in different plants. Tartaric acid and its salts (K, Na, Na/K, and calcium tartrate) are widely used in the food industry as flavoring agents, food additives, emulsifiers, antioxidant synergists, and sequestrates, which are safe for humans [32,33]. Crystal violet assay shows that TLA at 4 × 10 3 µg/mL inhibits biofilm formation by almost 100%. This result indicates the application potential of TLA as an anti-biofilm candidate against P. aeruginosa biofilm. Previous studies reveal that tartaric acid and its derivates have antimicrobial activity against several bacteria [34,35], but no published data have shown that tartaric acid can be an anti-biofilm candidate.
Walaszek (1997) reported that GA has anticancer and antioxidant properties. However, its anti-biofilm and antibacterial activities have not been evaluated. Calcium D-glucarate is a supplement for cancer prevention, liver detoxification, and hormone regulation. Moreover, it is a calcium salt of D-GA, a non-toxic compound found in several fruits and vegetables, particularly grapefruit, apples, oranges, and broccoli [36,37]. The in vitro anti-biofilm test of GA was constrained by the presence of calcium compounded with the active ingredient (GA). Hence, its solubility in the aquadest solvent was low. However, calcium-bound GA could be assayed at a maximum concentration of 5 × 10 2 µg/mL. GA in this assay has shown an inhibitory activity of >50% when the concentration is below 5 × 10 2 µg/mL.
Sorbitol, a polyol group, inhibited biofilm formation in S. mutant bacteria [38]. The polyol group has an osmotic effect that causes osmolarity to remain constant at each concentration tested without any significant effect on absorbance. Abbas et al., (2012) showed that sorbitol could inhibit P. aeruginosa biofilm formation at 4 × 10 5 µg/mL. Sorbitol is sugar alcohol with a sweet taste which is slowly digested in humans. This compound can be produced by reducing glucose, which converts the aldehyde group (-CHO) into primary alcohol or the hydroxyl group (-C(OH)H 2 ). Sorbitol is widely found in fruits, and it is used in industries as an additive to food, cosmetics, and medicine [39].
The inhibitory ability of AA describes the mechanism of PMM/PGM inhibition via substrate analogs. Several studies have shown the capability of AA to interfere with the physiology of microorganisms. For example, Novak and Fratamico (2004) reported the capability of AA to inhibit quorum sensing in C. perfringens. Meanwhile, Serry et al., (2008) showed that AA served as an efflux pump inhibitor. Finally, Abbas et al., (2012) showed that sodium ascorbate inhibited planktonic growth in P. aeruginosa at 5-20 mg/mL.
The second in vitro assay used the colony biofilm method. This method is suitable for assessing the strength of an anti-biofilm or resistance to antibiotics. The nitrocellulose membrane is a good surface for growing P. aeruginosa biofilms [40]. Biofilm preparation using this method can promote the supply of new nutrients by simply transferring membrane-grown cells to fresh agar plates. Therefore, one can easily change the carbon source or the type of anti-biofilm treatment without washing the cells. Figure 9 shows that the macroscopic appearance of each ligand was different from that of the control. The untreated P. aeruginosa biofilm was denser than that of the ligand-treated biofilm. Based on this qualitative experiment, each test ligand reduced the formation of P. aeruginosa biofilms. More detailed evaluations of microscopic structures were conducted by scanning electron microscopy (SEM).
The third in vitro assay was performed using an electron microscope. Figure 10 shows that biofilm formation by P. aeruginosa on glass coverslips was monitored via SEM. This section showed that the cell morphology has changed with ligand treatment. The preparation of SEM comprises several steps that can prevent changes in the morphological structure, although the cell undergoes death [41]. SEM analysis revealed that untreated P. aeruginosa biofilms (control) comprise bacterial cells and an EPS matrix. P. aeruginosa biofilms grown in vitro are formed through several stages. First, planktonic bacteria attach to abiotic or biotic surfaces through flagella and adhesion. Second, an irreversible attachment will occur with the help of the SadB protein.
Third, microcolonies are formed in the complex layer of biomolecules and the EPS matrix. Fourth, the biofilm matures. Fifth, bacteria are released from the biofilm to form a new biofilm [10].
All tested ligands were added after the first stage when bacterial cells were attached to the coverslips' surface. Based on the stages of biofilm formation [10], the EPS matrix will begin to form in the second stage. Therefore, ligands were added to inhibit this process through their function as a substrate analog of PMM/PGM. The TLA with a concentration of 4 × 10 3 µg/mL showed sparse cells without EPS, which is similar to the addition of GA (5 × 10 2 µg/mL), DS (4 × 10 5 µg/mL), and AA (5 × 10 4 µg/mL). The appearance of the bacterial cell wall in each treatment varied. The TLA and GA-treated cells and the cell walls of some bacteria were not intact and lysed. Meanwhile, the DS-treated cells were affected by osmotic pressure, thereby causing the cells to become turgid. Meanwhile, the cell walls remained intact. However, EPS was not found in AA-treated cells.

Materials and Bacteria
The ligands used are commercially available, i.  10895) and was assayed to identify its biofilm formation capability. This isolate was stored at −80 • C in Tryptone Soya Broth (TSB) CMO129 (Oxoid, Basingstoke, UK) supplemented with glycerol 20% (Merck, Darmstadt, Germany). Before each assay, P. aeruginosa frozen stocks were freshly cultured to brain heart infusion agar and then incubated aerobically at 37 • C for 18-24 h.

Ligands as PMM/PGM Inhibitors
The PMM/PGM S108A mutant was used as the target protein to design the P. aeruginosa anti-biofilm, which was retrieved as a crystal structure (1K2Y) from the Protein Data Bank (PDB) (rscb.org/pdb) accessed on 23 August 2019. TLA was a PMM/PGM native ligand extracted from the 1K2Y crystal structure, which was used as a reference coordinate in determining the active site of PMM/PGM [16]. In addition, TLA, GA, DS, and AA had a structure similar to the substrate, and they were commercially available in the Sigma database. Ligand selection begins with exploring compounds in https: //pubchem.ncbi.nlm.nih.gov/ (accessed on 4 September 2019) that have similar structure to TLA, includes information about related compounds, chemical vendors, drug and medication information, food additives and ingredients, pharmacology and biochemistry, use and manufacturing, safety and hazards and toxicity. Ligands structure similar to TLA are hypothesized to have the same, or better, activity. The inhibitory activity of selected ligands against PMM/PGM were then screened and evaluated through molecular docking analysis and considered for commercial availability, so that they can proceed to in vitro experiment.

Receptors and Ligands
PMM/PGM served as the receptor during docking. The receptor was prepared by extracting amino acid residues and Zn +2 atoms from the molecular structure of PMM/PGM using Chimera version 1.13. The ligand geometry was optimized using Semi-empirical Quantum Method-Parametric Method 3 (SQM-PM3). Further optimization was performed with HyperChem 8.0 to calculate the electrostatic potential charge of each atom. Analysis was continually performed to determine the ligand and receptor parameters such as bonded, non-bonded, and charged using the AMBER FF14SB force field tool [42] and Austin Model 1-Bond Charge Correction [43] contained in Chimera version 1.13.

Molecular Docking
Molecular docking was performed in two stages, namely, redocking and docking of other test ligands, using the Dock6 package. During redocking, a cluster sphere with a radius of 10.0 Å was first selected to determine the native TLA ligand coordinates on the receptor active site. Next, a grid box with a grid spacing of 0.1 Å was created on the selected cluster sphere around the receptor active site. The native ligand RMSD value represented successful redocking (≤2.0 Å) [44]. Next, binding affinity and footprint were determined using the grid score function with a rigid conformation. The scoring function could clearly describe the conformation (pose) between the ligand and the receptor [27]. In addition, the grid score function could predict several energy contributions, including van der Waals energy (E vdW ) and electrostatic energy (E ele ).
Next, the microtiter plate was incubated for 24 h at 37 • C. This assay used meropenem (Minimum Inhibitory Concentration/MIC: 0.5 µg/mL) as the positive control and bacterial inoculum as the negative control. The treatment was replicated three times.
After incubation, the microtiter plate was washed with distilled water three times and then dried at room temperature for 5 min. Next, 125 µL of 1% crystal violet solution was added to each well and then incubated at room temperature for 15 min. The microtiter plate was washed three times under running water. Next, 200 µL of ethanol was added to each well [30,46]. Optical density was measured at a wavelength of 595 nm. Then, the MBIC was calculated in accordance with the following formula: % inhibition = (OD growth control − OD sample ) OD growth control × 100 4.6.2. Colony Biofilm Assay P. aeruginosa was cultured on 13 mm diameter nitrocellulose semipermeable membrane (Whatman, Buckinghamshire, UK) and placed on an agar plate at 36 • C. The preparation of the ligand test was similar to the previous in vitro method. In total, 20 µL of bacterial suspension that reacted with the optimal concentration of each ligand was dripped onto a nitrocellulose semipermeable membrane layer on BHI agar and incubated for 24 h. The macroscopic morphology of the biofilm was observed on the membrane surface.

Surface Morphological Investigation
P. aeruginosa biofilms were formed on a sterile 12 mm diameter round glass coverslip (Electron Microscopy Sciences, Hatfield, PA, USA) in a 24-well polystyrene microtiter plate. The preparation of biofilm formation and ligand treatment was similar to the previous procedures in the in vitro biofilm matrix extracellular inhibition assay. The coverslips were submerged in 400 µL of bacteria and ligand suspension and incubated overnight at 37 • C. A biofilm grown without ligand treatment was used as the control. SEM preparation was performed in accordance with the method of Hess et al. and Kazmierczak [41,47] with modifications. First, the suspensions were discarded and rinsed two times with 1% sterile Phosphate Buffered Saline (PBS) (G-Biosciences, St. Louis, MO, USA). Then, the 12 mm of coverslips were fixated with 200 µL of 2.5% (b/v) glutaraldehyde in 0.15 M cacodylate buffer (Electron Microscopy Sciences, Hatfield, PA, USA) overnight at room temperature. Furthermore, after the fixative agents were discarded, 200 µL of methanol (Merck, Branchburg, NJ, USA) was added for 30 min during dehydration. Finally, the coverslip was removed from wells, dried overnight, and coated with 99.9% Au using a JEOL JEC-3000FC auto fine coater for 120 s 3,4 Pascal. Then, a double-sided carbon tape was attached for examination via SEM (JEOL JSM-6510LA).

Conclusions
The in vitro experiments in this study proved that TLA, GA, DS, and AA had excellent activity against the biofilm of P. aeruginosa. These potential anti-biofilm agents can reach almost 100% inhibition values at different concentrations. TLA, GA, DS, and AA had a biofilm inhibition rate of >50% at 2 × 10 3 , 2.5 × 10 2 , 4 × 10 5 , and 1.25 × 10 4 µg/mL, respectively. Nitrocellulose membrane assay showed that controls and ligand-exposed biofilm had different appearances. SEM images revealed that the control biofilm could be covered in the EPS matrix. However, these images were not visible in treated biofilms. Molecular docking analysis showed that each ligand had an excellent binding affinity to the receptor, as supported by several hydrogen bonds (i.e., TLA [5 H-bonds act as HBA], GA [6 H-bonds act as HBA and 1 H-bond act as HBD], DS [3 H-bonds act as HBA], and AA [1 H-bond act as HBA and 1 H-bond act as HBD]). Furthermore, each inhibitor can be a promising drug candidate based on its pharmacokinetic properties. Analysis of drug-likeness and bioavailability showed good results, where all anti-biofilm candidates showed non-toxicity and met the following criteria: non-AMES toxic, non-hepatotoxicity, and non-skin sensitization. Data Availability Statement: All data that support the results of this study are available from the corresponding author upon reasonable request.