2.1. Isolation, Identification, and Characterization of the Isolate
Isolation of thermophilic bacterial strains from Zara hot spring sample was performed at 60 °C and the isolate
Geobacillus sp. (designated as
Geobacillus sp. ZGt-1, GenBank accession no. KT02696) was identified based on 16S rRNA gene sequencing (
Table S1), which showed 99.9%–100.0% identity to
G. thermoleovorans and
G. kaustophilus, respectively. Since the assignment of
G. thermoleovorans and
G. kaustophilus into distinct species has been questioned previously [
12], we did not affiliate strain ZGt-1 to any of the species, and designated it as
Geobacillus sp. ZGt-1. Its cells appeared as single rods or in pairs after an overnight cultivation on R2A/Mueller Hinton agar. The sporulating cells of ZGt-1 showed one terminal endospore per cell. The colonies on Mueller Hinton (MH) agar were yellowish, rounded, raised, entire, and shiny, while they were creamy white, rounded, raised, entire, and opaque on R2A agar. Strain 10 was also isolated from the same hot spring and identified as
G. stearothermophilus (GenBank accession no. KU933578) (
Table S1).
2.2. Antibacterial Activity of Geobacillus sp. ZGt-1
The ability of
Geobacillus sp. ZGt-1 to inhibit the growth of the closely related
G. stearothermophilus strain 10 was confirmed using the agar-deferred spot method (
Figure 1). The antibacterial activity disappeared after treatment with proteinase K, indicating the proteinaceous nature of the secreted antibacterial agent. Testing of the antibacterial activity of ZGt-1 against mesophilic bacteria revealed inhibition zones in the case of
Bacillus subtilis and the pathogenic
Salmonella typhimurium CCUG 31969 (
Figure 1), but not with
Escherichia coli 1005,
Staphylococcus aureus NCTC 83254,
Staphylococcus epidermidis, and
Proteus vulgaris.
The inhibition zone of ZGt-1 developed against the Gram-positive
B. subtilis was more prominent than that against Gram-negative
S. typhimurium (
Figure 1), which may be ascribed to the different cell wall structures. In contrast to the Gram-negative bacteria, in which the cell wall peptidoglycan (PG) layer is protected by an outer membrane mainly composed of lipopolysaccharides (LPS), the PG in the Gram-positive bacteria is directly exposed to the external factors, including the cell wall lytic enzymes and antimicrobial peptides (AMPs) [
13,
14,
15]. Although AMPs, but not cell wall lytic enzymes, can interact with the negatively charged outer membrane, some Gram-negative bacteria can develop species-specific mechanisms to eliminate the effect of the AMPs under certain environmental conditions [
16,
17,
18]. Moreover, the variations in the structure of LPS, especially in Lipid A, among the different bacteria influence the AMP affinity and insertion into the outer membrane [
16,
17]. For example, there are differences in the polysaccharide chains and in Lipid A between
E. coli and
S. typhimurium; the latter has an additional fatty acid and different substituents of phosphate groups in Lipid A [
19].
The differences in the susceptibility of the various Gram-positive bacteria (
B. subtilis and
S. aureus) to the antimicrobial action of ZGt-1 could be ascribed to the differences in the degree of crosslinking between the stem peptides (short peptide chains of 4–5 amino acids) linked to
N-acetyl muramic acid residues in the PG [
15]. In case of
B. subtilis, 56%–63% of the stem peptides contribute to the cross-links, in contrast to
S. aureus, in which up to 90% of the stem peptides are involved in the cross-linkages in the cell wall [
15]. Moreover, the cross-links in
B. subtilis as well as in
G. stearothermophilus PG are direct (known as diaminopimelic acid (DAP)-direct), between
d-alanine in the stem peptide of one glycan strand and the
meso-diaminopimelic acid (
meso-DAP) in another [
20], while the cross-links in
S. aureus form a pentaglycine bridge between
l-lysine in the stem peptide of one glycan strand and
d-alanine in another [
21]. Even the charged polymers, teichoic- and teichuronic acids, which are covalently linked to the PG of Gram-positive bacteria, can influence the sensitivity of the cell wall to the lytic enzymes and AMPs, as the structure and amount of those polymers are strain- and species-specific [
14,
22,
23]. Under certain conditions,
B. subtilis produces atypical teichoic acid, which contains only the negatively charged phosphate groups that help in attracting the cationic AMPs to the cell surface [
14]. However, some bacteria, like
S. aureus, can alter the net surface negative charges of the bacterial surface or produce proteolytic enzymes capable of degrading AMPs as a resistance mechanism [
14].
2.3. Production of the Antibacterial Substance(s) Using Immobilized Cells in Sequential Batch Mode with Cell Recycling
Although
Geobacillus sp. ZGt-1 grown on solid MH agar medium exhibited good antibacterial activity against
G. stearothermophilus, no antibacterial activity was observed on analysis of the culture supernatant when the isolate was grown overnight in a liquid medium (reaching an OD
620 of 1.5). Similar observations have been reported previously with other strains [
4,
24]. Hence, it seems that the cell growth and production of antimicrobial agents is facilitated on the solid-state medium. To confirm this and to enable cultivation in larger scale for the production of antimicrobial molecule(s), the ZGt-1 cells were immobilized by entrapment in agar beads that were suspended in a liquid culture medium. This approach resulted in the appearance of antibacterial activity in the medium. The immobilized cells could be recovered and recycled for sequential batch cultivations in fresh medium with increasing antibacterial activity over consecutive cycles up to the 14th batch (
Figure 2a–d). The bacterial growth during the cultivations was noted by the appearance of free cells in the medium. Entrapment inside the gel beads provides a protective environment for the cells that are more active at producing certain metabolites than the free cells and show increased tolerance to inhibitory compounds, which might otherwise limit the cell growth [
25,
26]. After the 14th cycle, however, the antimicrobial activity started to decrease and then disappeared almost completely during the 25th batch (
Figure 2d,e). This decrease may be due to the nutritional limitations for the larger number of immobilized cells present and also the competition by the free cells that are not efficient in production of the antimicrobial metabolites [
26], or due to difficulties with exporting the antimicrobial compound to the extracellular environment.
2.4. Antimicrobial Proteins Produced by Strain ZGt-1
Protease treatment gave us an indication of the antimicrobial activity to be associated to protein(s). Therefore, we proceeded with isolation of the proteins from the cell-free supernatant collected from the sequential batches with immobilized ZGt-1 cells by precipitation with ammonium sulphate. The protein precipitate obtained with 60% salt saturation was dialyzed against distilled water, and its antibacterial activity was confirmed using the spot-on-lawn method (
Figure S1). The activity was found to be stable after heating at 70 °C for 45 min but was lost on heating to 80 °C for 10 min (
Figure S2). The proteins in the sample were resolved on SDS-PAGE and the gel was subjected to the antibacterial activity test (
Scheme 1). An area corresponding to 15–20 kDa molecular mass displayed the inhibition zone with
G. stearothermophilus (
Figure 3 and
Figure S3).
Interestingly, the protein fraction was seen to be active even after being exposed to SDS under reducing conditions (with dithiothreitol; DTT), which was confirmed during at least eight separate SDS-PAGE-antibacterial activity assays performed using two different batches of the protein molecular weight standards. Higher abundance of SDS-resistant proteins has been reported in thermophilic microorganisms than that in the mesophilic ones [
10]. Production of such proteins seems to be a defense strategy developed by prokaryotes, especially thermophiles, to protect some of their proteins against aggregation and premature degradation in order to save energy for thriving in extreme environments [
10]. Moreover, the insensitivity of the antimicrobial activity to the reducing agent is possibly due to the lack of cysteine residues and hence the disulfide bridges, as is the case for several antimicrobial peptides [
16].
2.5. Targeted Proteomics Analysis of the Antibacterial Protein-Containing Gel Samples
The gel area corresponding to 15–20 kDa, which displayed the inhibition zone against
G. stearothermophilus strain 10, was analyzed for its protein content by liquid chromatography tandem mass spectrometry (LC-MS/MS). At the same time, the genome of
Geobacillus sp. ZGt-1 was sequenced and annotated (GenBank accession no. LDPD00000000.1) [
27], and used to construct a ZGt-1 strain-specific protein database. The collected MS/MS data were searched against the locally created ZGt-1 database to identify the proteins present in the excised gel samples. To be considered true protein identification, the following criteria had to be fulfilled: significant peptide sequences greater than or equal to 5, a total protein score greater than or equal to 200, and an individual peptide score greater than or equal to 20. Many proteins larger than 20 kDa were identified (
Table S2), which could possibly be attributed to proteolysis. The high resolving power and sensitivity of the mass spectrometric analysis carried out in this study enables the detection of peptides of truncated or degraded proteins even if present in a small amount. Some proteolysis is indeed to be expected due to the action of bacterial proteases during the sequential batch cultivations and subsequent processing and storage of the extract at 4 °C until the time of analysis. Besides proteolysis, SDS-resistance could be another factor underlying the identification of proteins larger than 20 kDa. SDS-resistant proteins preserve their compact structure and display an apparent molecular weight in the gel that is smaller than their theoretical molecular weights [
28,
29].
Based on the antibacterial activity being limited to a molecular weight of 15–20 kDa, we decided to screen for proteins within the range of 10–30 kDa to take into account the “gel shifting” phenomenon, as noted above for the SDS-resistant proteins [
28,
29,
30].
Table 1 lists the identified proteins, most of which are not known to display antibacterial activity. Only three uncharacterized/hypothetical proteins within 10–30 kDa were identified in two of the three excised gel samples, with 5–8 significant peptide sequences and a score of 292–517 (
Table S3). We combined two approaches in a non-mutually exclusive way to predict the antibacterial activity of the uncharacterized proteins: calculation of the physiochemical properties of the proteins based on their amino acid sequences and comparing the inferred properties to those of known antimicrobial peptides/proteins, and
in silico prediction of their antibacterial activity by employing the web-based antimicrobial peptides/proteins prediction algorithms, in an approach similar to that reported earlier [
31,
32] with some modifications.
2.6. Prediction of Antibacterial Potency Based on Physicochemical Properties
Table 2 summarizes the physicochemical properties depicted from analysis of the amino acid sequences of the three uncharacterized proteins. The proteins are in the molecular weight range of approximately 14–19 kDa, and a size range of 129–173 amino acid residues, which is in the range (10 to 100 s) for AMPs [
33].
The antibacterial activity of many AMPs is brought about by lysis of the cells or membrane permeabilization due to electrostatic and/or hydrophobic interaction with the membrane of the target cell [
34]. This interaction requires the AMP to have a net positive charge at neutral pH to help in initiating the interaction with the anionic surface of the bacterial cell, and to be rich in hydrophobic amino acid residues for insertion into the hydrophobic core of the bacterial cell membrane to destabilize the lipid bilayer and eventually kill the bacteria [
32,
34,
35,
36]. A minimum net charge of +2 is one of the unique features of cationic α-helical AMPs [
37]. The three proteins have isoelectric points (pI) in the range of 7.72–8.80. Two of the proteins (ID 6_35, and 23_543) have a net charge of +2 at neutral pH, while all have a hydrophobic ratio of 36%–40% (
Table 2).
The AMPs typically adopt an amphipathic conformation, i.e., the hydrophobic and polar residues are located on opposite sides. Linear amphipathic α-helical AMPs represent an important class of AMPs [
34,
38]. Analyzing the amino acid sequences of the three uncharacterized proteins using “The Antimicrobial Peptide Database (APD3)” showed that the three proteins have the potential to form amphipathic α-helices, where a certain number of hydrophobic amino acid residues group on the same side of the α-helix (
Table 2), they are free of disulfide bridges, and have a linear amphipathic α-helical conformation. This is in agreement with the insensitivity of the antimicrobial effect to the reducing environment, as observed above (
Section 2.4).
Grand average of hydropathy (GRAVY) index, a measure of the peptide/protein solubility, is within the range of −0.257 to −0.044 for the uncharacterized proteins (
Table 2), in accordance with the predominant range of −1 to 0 for AMPs listed in the BIOPEP database as well as the predicted AMPs from milk proteins [
32]. The Boman index values, reflecting the potential of a peptide/protein to interact with other proteins [
38], for 6_35, 23_543, and 4_4 are 1.12, 1.57, and 1.19, respectively (
Table 2), which also fit within the range of 1–2 for AMPs in the BIOPEP database [
32].
In vitro stability of the AMPs reflects their bioavailability over a period of time [
32]. The proteins 6_35, 23_543, and 4_4 have instability index values of 14.04, 17.61, and 36.7, respectively (
Table 2), and are classified as stable in vitro, according to a criterion of instability index below 40 to be good for stability [
39]. The tolerance to heat and SDS shown above could be a reflection of the stability of the proteins.
The Alpha index is a measure of the relative volume of the aliphatic amino acids side chains (alanine, leucine, isoleucine, and valine) [
32,
40], and has a positive correlation with thermostability [
40]. The proteins 6_35, 23_543, and 4_4 scored 80.08, 110.33, and 96.76 on the aliphatic index, respectively (
Table 2), again within the predominant range of 40–120 for AMPs in the BIOPEP database [
32].
The potential of the AMPs to form aggregates at their site of interaction with the bacterial cell membranes is a necessary step for their mechanism of action [
35]. AGGRESCAN software [
41] predicted that there are three, six, and six putative aggregation “hot spots” within the sequences of the proteins, 6_35, 23_543, and 4_4, respectively (
Table 2). The Na
4vSS values were 5.3, −6.6, and −2.4 for 6_35, 23_543, and 4_4, respectively (
Table 2). Previously reported AMPs have Na
4vSS values within the range of −40 to 60 [
36,
42].
2.7. In Silico Prediction of the Antibacterial Potency
The bioinformatics tools used for predicting the antibacterial potency of the uncharacterized proteins based on the amino acid sequences were the sequence-based prediction tools available on “The Collection of Anti-Microbial Peptides (CAMP
R3)” [
43], the antimicrobial peptide calculator and predictor tool available on the “APD3” [
44], and the “AMPA” application [
45].
The CAMP
R3 prediction tool uses machine learning algorithms to predict the antimicrobial activity of peptides/proteins including the novel ones. CAMP
R3 uses four different prediction models (Support Vector Machines (SVM), Random Forests (RF), Artificial Neural Network (ANN), and Discriminant Analysis (DA)). SVM, RF, and DA predict the antimicrobial activity and state the AMP probability, while ANN makes a qualitative description of the peptide/protein as either an AMP (for “antimicrobial peptide/protein”) or NAMP (for “non-antimicrobial peptide/protein”). The accuracy of the prediction results for the models is within the range of 87%–93% [
32,
46].
APD3 predicts the potential of an amino acid sequence (up to 200 residues) to have an antimicrobial activity by analyzing each amino acid residue and comparing the physicochemical properties of the sequence with those of the natural AMPs already deposited in the APD3 database [
32,
44,
47]. AMPA uses a sliding window algorithm that calculates an “antimicrobial index” for individual amino acid residues and estimates the tendency of the amino acid to be found within an AMP sequence. By doing so, AMPA identifies the antimicrobial domain within the protein/peptide and hence predicts the overall antimicrobial activity [
45]. The algorithms indicated that the three uncharacterized proteins have antimicrobial potential to different extents (
Table 2).
Combining the prediction results inferred from the sequence-derived physicochemical properties with the results of the six prediction algorithms used (the four algorithms of CAMP
R3, the APD3 algorithm, and the AMPA algorithm), we can conclude that protein 23_543 is the most likely antibacterial protein candidate (
Table 2). In the case of the protein 6_35, not all six prediction algorithms confirmed its potential as an antimicrobial protein. Protein 4_4 does not fulfill all the physicochemical properties of AMPs because its pI is 7.72 and its positive net charge is +1 [
37]. Moreover, not all six prediction algorithms confirmed the potential of protein 4_4 as an antimicrobial protein. However, the probability of the antimicrobial potential of these two proteins cannot be ignored since most of the sequence-derived properties match with the features known about reported AMPs, and four out of the six algorithms predicted them to display antimicrobial activity (
Table 2).
2.8. Identified Protein Sequences Matching Parts of Antimicrobial Enzymes
As described above (
Section 2.5), many proteins larger than 20 kDa were identified by LC-MS/MS analysis. Among these are two enzymes (26_23 and 2_3 in
Table S2) already reported as antibacterial;
N-acetylmuramoyl-
l-alanine amidase (referred to as amidase; EC.3.5.1.28), and serine-type
d-alanyl-
d-alanine carboxypeptidase (referred to as
dd-carboxypeptidase; EC: 3.4.16.4). The predicted amino acid sequences of amidase and
dd-carboxypeptidase in strain ZGt-1 give a theoretical molecular weight of approximately 87 and 47 kDa, respectively, after cleavage of the enzyme’s signal peptide. These enzymes are known to catalyze the lysis of bacterial cells by hydrolyzing the covalent bonds in the peptidoglycan layer of the bacterial cell wall [
13,
48]. Amidase cleaves the amide bond between the glycan and the peptide chain [
13], while
dd-carboxypeptidase cleaves the terminal
d-alanyl-
d-alanine bond in the stem peptides of the peptidoglycan layer, resulting in the removal of the terminal
d-alanine [
49]. The lytic enzymes of bacterial origin are known to have a role in bacterial cell growth and division [
13,
48], and also act as antimicrobials by attacking the cell wall of competing bacteria [
13].
Matching of the MS/MS-detected significant peptide sequences of the ZGt-1 amidase (
Table S3) to the enzyme domains that we previously identified using InterProScan analysis [
50] of the predicted amino acid sequence, showed alignment with segments in the catalytic domain at N-terminus (data not shown). Similar analysis of ZGt-1
dd-carboxypeptidase domains showed a match between the MS/MS-detected significant peptide sequences of the enzyme (
Table S3) with N-terminus segments (the catalytic domain), C-terminus segments (supposed to act as the enzyme’s binding domain as indicated by the InterProScan tool), and parts of the region in between the two domains (data not shown). However, whether the detected partial sequences of the two enzymes would be responsible for the antibacterial activity and the clearance zone in the 15–20 kDa region is not clear.