We performed sequence analysis by conducting BLASTp searches in NCBI's non-redundant protein sequence database. After removing non-clustered sequences, we optimized the codons. Subsequently, Alphafold2 was employed to model the structural conformations of these sequences.

Fig.1 The structure prediction diagram of cecropin B

Using the IGEM Registry, we selected CecB as our antimicrobial peptide sequence. Through NCBI's BLASTp database, we compared the protein sequences obtained from the IGEM Registry. Subsequently, we optimized its codon usage using the company's codon optimization software. The synthesized primer was ligated into a CecB fragment and inserted into the pET-28a vector for expression, which served as the bacterial host for subsequent cultivation.

Plasmid design:

Fig. 2 pET-28a-cecB

Fig. 3 pET-28a-yin-cecB

Fig. 4 pET-28a-pelB-yin-cecB

All the primers were put into the PCR system for amplification to synthesize fragments without template, and the correct fragments were recovered from the gel.

We extracted the plasmid from the synthesized gene and transformed it into E.coli BL21 (DE3). The transformed single colony was selected and cultured on the plate, and the bacterial colony PCR was used to verify whether it contained the recombinant plasmid.

Fig. 5 cecB colony PCR

Fig. 6 yin-cecB colony PCR

Fig. 7 pelB-yin-cecBcolony PCR

After culturing the colonies with confirmed results, the activated bacterial suspension was added to 50 mL LB medium for induction. Following the induction process, the cells were homogenized and the supernatant was purified through ultrafiltration. Total protein content, supernatant liquid, and sediment were analyzed using Western blot (WB) experiments (Fig.8,9) to determine their inhibition zones and Minimum Inhibitory Concentration (MIC)(Fig.10,11,12).

After induction, the supernatant and precipitate of cell pulverization were respectively subjected to SDS-PAGE

Fig. 8 CecB WB image

Fig. 9 PelB-Yin-CecB WB image

The leftmost lane is 10-180kDa Marker, lane 1 is ultrafiltration protein sample, lane 2 is total protein sample, and lane 3 is precipitation protein (using 5mL PBS pH=7.4 to resuspend the precipitation)

We found that the effect of direct lyophilization was better than that of purified ultrafiltration. Therefore, the crushed protein was added to the lysis solution with twice its volume to break the anion antioxidant peptide, and CecB protein was obtained after freeze-drying for MIC and inhibition effect determination.

Fig. 10 Results of the inhibition circle experiment

1：PelB-Yin-CecB；2: Yin-CecB-1；3：CecB；4：Yin-CecB-2

After determination, it was found that the target band could not be obtained in WB, and no obvious inhibitory effect of CecB was found in both inhibition zone and MIC results.

Fig. 11 Yin-CecB MIC experiment

Fig. 12 PelB-Pin-CecB MIC

After determination, it was found that the target band could not be obtained in WB, and no obvious inhibitory effect of CecB was found in both inhibition zone and MIC results.

Fermentation optimization

We investigated the fermentation temperature of the above PelB-Yin-CecB protein. The fermentation was carried out in LB medium, and 16℃ 25℃ 30℃ 37℃ of the fermentation broth was taken respectively for crushing and purification, protein concentration was determined, SDS-PAGE was used to verify the expression and optical density was determined.

Since the concentration of IPTG is also one of the important factors affecting the expression during induction, we induced the protein at a gradient of 0.25 mM, 0.5 mM and 1 mM. The fermentation broth was broken and purified to obtain the protein-like liquid for testing the optical density of the SDS-PAGE.

To explore the effect of culture medium on CecB, we also used TB culture medium and LB culture medium as control.

Fig. 13 Optical density image of temperature

Fig. 14 Optical density image of temperature

Fig. 15 Optical Density image of IPTG Concentration

Fig. 16 Optical Density image of IPTG Concentration

Fig. 17 Optical density image of media optimization

Fig. 18 Optical density image of media optimization

According to the optical density image, when the temperature is about 16℃, more bands can be obtained in the supernatant. The percentage image shows that the protein content in the supernatant is higher, so we choose 16℃ for induction of expression in the following.

According to the optical image of IPTG concentration test, when the OD₆₀₀ of bacterial solution reached 0.6, IPTG was added for induction to make the final concentration of 0.5 mM.

Through optimized medium cultivation, we observed that when pelB was added, the natural content of silkworm toxin in the precipitate became minimal. However, with TB medium, the toxin was fully detected in the supernatant. Notably, in plasmid without signal peptides, protein levels in the supernatant of TB medium also increased. Yet, when freeze-drying without ultrafiltration, no significant antibacterial effects were observed a situation that not only wastes time and financial resources but also contradicts our research expectations.

The pre-trained large language model is utilized for task-oriented fine-tuning. The fine-tuning dataset consists of known antimicrobial peptide sequences and their functional annotations, aiming to enhance the model's recognition capabilities for key physicochemical characteristics (charge distribution, amphiphilicity, secondary structure propensity). Charge-enhancing modifications: By replacing positively charged residues such as Lys/Arg, interactions with negatively charged targets are enhanced;

Structural optimization: improve the folding and groove compatibility by replacing hydrophobic or small volume residues.

All candidate sequences are standardized and organized into a table file for subsequent calculation and verification.

The model generated multiple single-point mutation sequences, categorized as charge-enhancing and structure-optimizing types. Charge-enhancing mutations predominantly involved positive residue substitutions, while structure-optimizing mutations mainly featured small-volume hydrophobic residue replacements. The overall sequence retained its original backbone structure, with mutation sites distributed along the peptide chain's outer side to facilitate interfacial interactions. The binding free energy of these mutants to bacterial receptor proteins was also calculated.

Table 1. Results combined with free energy

System	ΔVDW	ΔEEL	ΔG_solv	ΔH	-TΔS	End G binding
WT + 4RHB	-128.3	-955.1	998.6	-84.8	71.3	-13.5
CecB-E10K + 4RHB	-127.5	-1102.3	1126.3	-103.5	85.0	-18.5
CecB-N15K + 4RHB	-125.9	-1180.7	1205.1	-101.5	80.5	-21.0
CecB-V29W + 4RHB	-129.8	-950.4	996.5	-83.7	71.5	-12.2
WT + 1DPE	-95.6	-680.2	715.3	-60.5	50.1	-10.4
CecB-E10K + 1DPE	-93.8	-695.1	729.4	-59.5	50.0	-9.5
CecB-N15K + 1DPE	-90.1	-705.8	742.5	-53.4	45.0	-8.4
CecB-V29W + 1DPE	-145.2	-675.5	720.8	-99.9	82.3	-17.6

After induction, the bacterial cells were relatively clear, and there was almost no precipitation after fragmentation. Moreover, it was verified by SDS-PAGE that there was no obvious band. Subsequent corresponding expression results showed no obvious effect. Although we still constructed the double-point mutation plasmid in the later stage, no relevant band or effect was found.

Based on preliminary experiments, Cecropin B didn't provide substantial benefits at this stage of our investigation. However, leveraging the rapid advancement of artificial intelligence, this study adopted a deep learning process to mine AMP (numbered 7, 4, 2) with brand-new sequence features from an extreme environment bacillus sequence library, and selected the optimal structural tandem configuration through molecular dynamics (MD) simulation. As a result, several monomers with potential performance, namely NJT-Lyy-7, NJT-Lyy-4, and NJT-Lyy-2, were obtained. For convenience, they will be referred to as "7", "4", and "2" in the following context.

We first integrated positive samples (AMPs) from multiple authoritative databases (such as APD3, DADP, DBAASP), and selected negative samples (non-AMPs) from the UniProtKB/Swiss-Prot database to construct a balanced dataset containing over 10,000 sequences. We adopted a bidirectional long short-term memory network (Bi-LSTM) with an attention mechanism as the core model architecture, and trained an integrated model through a five-fold cross-validation strategy to ensure its prediction accuracy and generalization ability.

After training, the integrated model was used to conduct a high-throughput scan on a genomic sequence library from extremophilic Bacillus.

All permutations of the three sites (A(3,3) = 6) were carried out to obtain six configurations: "742, 724, 472, 427, 274, 247"; for each configuration, the initial three-dimensional structure model and rapid energy minimization were conducted, and then the subsequent simulation screening was carried out. (Fig.19)

Fig.19 RMSD curves from MD simulations of the six tandem configurations

Then, using pET-28a(+) as the vector, a tandem expression cassette containing the pelB signal peptide sequence and codon-optimized antimicrobial peptides (AMP2, AMP4, and AMP7) was constructed, with the three peptides connected by a flexible linker (GGGGS). This recombinant plasmid was named pET-28a-pelB-742, or pET-28a-pelB-NJT-LYY-742.

Fig. 20 Map of the pET-28a-pelB-742 plasmid

The correct colonies will be inoculated into two 50 mL LB media bottles. After reaching an OD₆₀₀ of 0.6, they will be subjected to ultrasonic disruption. The protein from one of the bottles will be directly freeze-dried, while the other bottle will undergo purification and ultrafiltration before conducting an SDS-PAGE experiment. After the freeze-dried samples are melted, the inhibition zone test and MIC determination will be carried out.

The band size of pET-28a-pelB-NJT-LYY-742 is approximately 14.8 kDa. (Fig.21)

Fig.21 pelB-742 SDS-PAGE image

n the far right is the 10-180 kDa protein marker. Lane 1: Total protein of PelB-742; Lane 2: PelB-742 supernatant; Lane 3: PelB-742 precipitate; Lane 4: Sample after ultrafiltration of PelB-742

Fig. 22 pelB-742 Antimicrobial zone

Fig. 23 pelB-742 MIC Test

The experimental results showed that the antibacterial effect of pET-28a-pelB-742 was significantly improved, which proved that our series was effective and effective. The results also showed that when the concentration of antibacterial peptide was close to 3.13 µg/mL, it still had antibacterial effect on the indicator bacteria.

Molecular partners are a class of proteins that can remove, degrade and label misfolded proteins to maintain protein homeostasis. They can promote the correct folding and soluble expression of proteins, effectively reduce the formation of inclusions, but do not affect the activity of proteins.

Four kinds of molecular chaperone plasmids, pGro7, pKJE7, pG-Tf2 and pTf16 were co-expressed with pET-28a-NJT-LYY-742 plasmid to promote the correct folding of foreign protein in prokaryotic cells and increase the solubility of foreign protein, so as to obtain a large amount of soluble protein and further explore the biological function of the protein.

Four molecular chaperone plasmids were co-transfected into Escherichia coli BL21(DE3) host cells with the pET-28a-pelB-NJT-LYY-742 plasmid. Primary screening was conducted on Kan and Cm antibiotic-resistant plates. Single colonies PCR-positive were selected for induction and protein extraction purification. After induction, cell homogenized supernatants and precipitates underwent SDS-PAGE (Fig.24) for analysis.

Fig. 24 SDS-PAGE co-expressed by chaperone and pET-28a-pelB-NJT-LYY-742

M:maker(10-180kDa); Lane 1:PG-Tf12/pelB-NJT-LYY-742-1 supernatant; Lane 2: PGro7/pelB-NJT-LYY-742-1 supernatant; Lane 3:PGro7/pelB-NJT-LYY-742-1 precipitate; Lane 4: pTf16/pelB-NJT-LYY-742-1 supernatant; Lane 5:pTf16/pelB-NJT-LYY-742-1 precipitate; Lane 6:PKJE7/pelB-NJT-LYY-742-1 supernatant; Lane 7:PKJE7/pelB-NJT-LYY-742-1 precipitate; Lane 8:pTf16/pelB-NJT-LYY-742-2 supernatant; Lane 9:pTf16/pelB-NJT-LYY-742-2 precipitate; Lane 10:PKJE7/pelB-NJT-LYY-742-2 supernatant; Lane 11:PKJE7/pelB-NJT-LYY-742-2 precipitate; Lane 12:PGro7/pelB-NJT-LYY-742-2 supernatant; Lane 13:PGro7/pelB-NJT-LYY-742-2 precipitate; Lane 14:PG-Tf12/pelB-NJT-LYY-742-2 supernatant

We conducted SDS-PAGE experiments (Fig.24) using the supernatant and precipitate. Through experimental verification, we found that molecular chaperones have a good effect on inclusion bodies. The bands in the supernatant were significantly higher than those without the addition of molecular chaperones. Moreover, we believe that the co-expression of PKJE7 and pelB-NJT-LYY-742 has the best effect.

Antimicrobial formulations were selected from three different basic hydrogel formulations. The following three hydrogels were prepared and tested:

Sodium alginate/gelatin hydrogel (SOP 1)

PVA/agarose hydrogel (SOP 2)

The hydrogel of chitosan/sodium alginate (SOP 3, 1:1 ratio) was made into circular gel sheet with a diameter of 6 mm for testing. It was found that only SOP 3 had effective effect. Therefore, the composition ratio of chitosan/sodium alginate hydrogel was optimized to enhance its antibacterial effect.

Based on SOP 3, three different mass ratios of chitosan were prepared: sodium alginate hydrogel: 1:2, 1:1(Control), 2:1

In accordance with SOP 3 process, only the raw material quality was adjusted to obtain the above three proportions of samples.

Repeated inhibition circle experiment.

Fig 25. Inhibitory circle test

P：2:1 T: 1:1 N: 1:2

1:2 ratio: the diameter of the antibacterial circle is about 8 mm.

1:1 ratio: the diameter of the antibacterial circle is about 12 mm.

2:1 ratio: the maximum diameter of the antibacterial circle is about 18 mm.

Fig. 26. The MIC of Hydrogels

Increasing the proportion of chitosan can significantly enhance the antibacterial effect. The effect of 2:1 can reach 6.25 mg/mL. The optimal formula is 2:1 ratio of chitosan and sodium alginate.