All Restriction endonucleases were purchased from New England BioLabs (NEB, Beijing, China). Pfu DNA Polymerase was purchased from Thermo Fisher Scientific (San Jose, CA, United States). All other reagents and chemicals were analytically pure grade from Takara (Otsu, Japan).

All plasmids and primers used in this study were listed in Table 1 and Supplementary Table S1, respectively. The truncations and mutants of pBT-SmpB and pN-SN were from our previous work (Liu et al., 2016). The peptide aptamer library (pTRG-SN-peptides) was constructed and comprised of approximate 2 × 10⁷ clones which expressed the scaffold protein and the random exposed loop (Liu et al., 2016). In brief, the DNA fragment encoding SN was inserted into pTRG, and expressed as a fusion protein with α-subunit of RNA polymerase as scaffold protein, in which the constrained loop composed of the residues S₆₃L₆₄R₆₅K₆₆A₆₇ was replaced by 16 random amino acids. For construction of pET-28a-SmpB, the DNA fragment encoding SmpB was amplified from genomic DNA of A. veronii C4 using the primers F1/R1, at the end of which contained 5′-Nco I and 3′-Xho I restriction sites, and then digested and ligated into pET-28a to yield pET-28a-SmpB. For constructions of pET-28a-SN and pET-28a-PA-1, the DNA fragments were amplified using pTRG-SN or pTRG-PA-1 as templates and F2/R2 or F3/R3 as primers, respectively, followed by digestion and ligation with pET-28a. The plasmid pN-PA-1 was constructed using F4/R4 as primers according to our previous strategy (Liu et al., 2016).

Bacterial strains were listed in Table 2. A. veronii, A. veronii (pRE112), and A. veronii (pN-SN) were provided in our lab (Liu et al., 2016), and A. veronii (pN-PA-1) was constructed using the same method of A. veronii (pN-SN). Escherichia coli WM3064 was supplied as donor strain for genetic manipulation on pRE112 conjugative machinery in A. veronii. E. coli XL1-Blue MRF’ was applied to reproduce pBT and pTRG derivatives. E. coli XL1-Blue MR was used for bacterial two-hybrid system. E. coli BL21 (DE3) was used for the inducible expression of pET derivatives.

The plasmid pBT-SmpB was used as bait to screen specific peptide aptamers by Bacterial two-hybrid system. The peptide aptamer was selected and the interactive sites between SmpB and PA were identified as described previously (Liu et al., 2016).

Escherichia coli strain BL21 (DE3) was transformed with pET-28a-SmpB, pET-28a-SN and pET-28a-PA-1, respectively. The strains were grown until OD₆₀₀ of 0.4 in LB containing 50 μg/ml kanamycin, followed by supplementing with 0.1 mM isopropyl-β-d-thiogalactopyranoside (IPTG), and cultured overnight at 16°C. Cells were harvested, and resuspended in suspension buffer (10 mM PBS, pH 7.4) for sonication. The supernatant was collected and loaded onto nickel-iminodiacetic acid-agarose (Ni-IDA) column which was pre-balanced with equilibration buffer (Invitrogen, Frederick, MD, United States). Subsequently the column was washed with wash buffer (50 mM PBS, 10 mM imidazole, pH 7.4) until no further ultraviolet-absorbing values could be detected. Finally, the target protein was collected with elution buffer (50 mM PBS, 250 mM imidazole, pH 7.4), and estimated by SDS–PAGE. After the imidazole has been removed by dialysis, the concentration of protein was determined by BCA assay (Thermo Fisher Scientific, San Jose, CA, United States).

The wells of enzyme-linked immunosorbent assay (ELISA) plate were coated with 100 μl of SmpB (100 μg/ml) at 4°C overnight. Concurrently, 3% BSA was chosen as the control. The following day the wells were washed three times with 200 μl of TTBS (20 mM Tris–HCl, pH 8.0, 0.05% Tween-20, 150 mM NaCl), and blocked at 37°C for 1 h with 200 μl of 3% BSA in PBS. The aliquots of SN or PA-1 (1.6 μM) were incubated with the wells which were pre-coated with SmpB at 4°C overnight. After the wells were washed three times in TTBS, the polyclonal rabbit antibody against 12 residues of SN was added into the wells at 37°C for 2 h, and subsequently anti-rabbit immunoglobulin G (IgG) was appended for 1 h, followed by the addition of 100 μl of TMB substrate reagent for 30 min and 100 μl of TMB termination buffer for cancellation. The absorbance at 450 nm was measured with Microplate Readers (BioTek, Winooski, VT, United States). Assays were performed in triplicate and the dissociation constants K_d were analyzed with GraphPad Prism version 6.0 (GraphPad, CA, United States).

The amino acid sequences of SmpB and PA-1 were aligned online using PROMALS3D, and optimal templates of SmpB (PDB code 1k8hA) and SN (PDB code 1jokA) were selected and applied to predict 3D-structures of SmpB and PA-1 by the iterative threading assembly refinement (I-TASSER) webserver (Yang et al., 2015). The most stable structures were projected for SmpB and PA-1 docking via the High Ambiguity Driven biomolecular DOCKing (HADDOCK) webserver¹ (van Zundert et al., 2015). All the protein structures and docking complexes were visualized using the software PyMol Version 1.7.0.0.

The plasmid pN-PA-1 was transformed into E. coli WM3064 and then transferred into A. veronii C4 by conjugation as described before (Liu et al., 2016). The growth curves of wild type A. veronii C4, A. veronii C4 (pRE112), A. veronii C4 (pN-SN) and A. veronii C4 (pN-PA-1) were measured with a UV-spectrophotometer (Mapada UV-1800, Shanghai, China) at regular intervals. The LB culture media were supplemented individually as follows: 2.5 mM CaCl2, 25 mM MgCl₂ and 0.0–5.0% NaCl (Liu et al., 2016).

The wild type and engineered A. veronii C4 were grown to stationary phase in LB supplemented with either 50 μg/ml ampicillin or 50 μg/ml ampicillin and 25 μg/ml chloramphenicol simultaneously at 30°C. The total amount of RNA were extracted for relative expression analysis of genes, including smpB, three type secretion dependent effector (aexU), outer membrane protein (ompA), histidine kinases (bvgS), aerobactin (aer), serine protease (ahp gene), outer membrane channel (tolC), hemolysin (trh), low calcium response V (lcrV), RNA-binding protein (hfq), flagella basal body protein (fliL), universal stress protein A (uspA), and pilus assembly protein (flpL) (Liu et al., 2016). The primers of qRT-PCR were listed in Supplementary Table S2. The threshold cycle (Ct) values of targets were normalized utilizing 16S rRNA as internal standard. And the relative expression quantity was calculated using the equation 2^-ΔΔC_t, where ΔΔCt = (Ct target - Ct 16S rRNA)_Treatment - (Ct target - Ct 16S rRNA)_control (Liu et al., 2016).

All animal experiments were approved by the Committee of the Ethics on Animal Care and Experiments at Hainan University, and all the animal experiments were carried out in accordance with the approved guidelines.

All zebrafish which were provided at the age of 4 months, average weight of ∼0.3 g and length of ∼3 cm, were fed with the basal diet and allowed to acclimate for at least 7 days before use. To evaluate 50% lethal dose (LD₅₀), 10 zebrafish were intraperitoneally injected with 0.01 ml bacterial suspensions of wild type or A. veronii (pN-PA-1) in triplicate, and monitored at 25°C for 7 days, in comparison to the negative control with saline only. In the meantime, survival condition of zebrafish was recorded daily, and eventually LD₅₀ was calculated by the method of Reed and Muench (1938). In brief, logLD50 = αlogβ+γ, where α = (The mortality higher than 50%-50%)/(The mortality higher than 50%-The mortality lower than 50%), β = dilution rate, in current experiment b = 10^-1, γ = the log of minimum dilution rate, when the mortality higher than 50%.

The strains of A. veronii C4 (pN-PA-1) were grown in LB at 30°C, and harvested by centrifugation and re-suspension. The immunizations were consistently proceeded with three independent repeats, of which 20 zebrafish were injected with 1/10 LD₅₀ A. veronii C4 (pN-PA-1) in the total amount of 1.62 × 10⁶ CFU/g, in contrast to negative control saline at 25°C. IgM antibody levels were determined in zebrafish sampled from 1 to 28 days by following the instructions of Fish IgM ELISA Kit (Mlbio, Shanghai, China). In brief, each sample including individual zebrafish was cut, weighed and frozen in liquid nitrogen and stored at -80°C for subsequent use. After the tissues were homogenized according to the proportion of 0.1 g per 1 mL PBS buffer (pH 7.4), the supernatant was collected. In the meanwhile, standards of purified IgM were diluted with TTBS buffer (containing 3% BSA) from 16 to 1 μg/mL using multiple proportion dilution method. Subsequently 50 μl of each standard or sample was loaded to the 96 micro-well plate pre-coated with an antibody specific for IgM for 30 min at 37°C. Each well was washed five times with TTBS, and incubated with 50 μl of diluted detection antibody for 30 min at 37°C. Having been washed five times, the plates were appended to 50 μl both of Chromogen Solution A and B for 15 min at 37°C. Eventually the absorbance at 450 nm was measured using microplate reader.

Having been vaccinated with 1/10 LD₅₀ A. veronii (pN-PA-1) for 14 days, the challenge was conducted with 100 LD₅₀ A. veronii C4 in the total amount of 4.98 × 10⁷ CFU/g. Mortality was examined, and dead zebrafish were removed in subsequent 7 days. The relative percent survival (RPS) was determined according to the following formula. RPS=[1-(% mortality of immunized fish/% mortality of control fish)]× 100 (Byon et al., 2005).

Statistical data were analyzed using the statistical Package for the Social Science (SPSS) version 20.0 (SPSS, Chicago, IL, United States) and GraphPad Prism version 6.0 (GraphPad, San Diego, CA, United States), and presented as mean values of three independent experiments with standard deviation (SD) using one-way analysis of variance (ANOVA). P-values less than 0.05 or 0.01 were considered as significant or extremely significant.

Bacterial two-hybrid selection was used to identify peptide aptamers that bound specifically to SmpB protein in vivo (Figure 1A). Three clones that might interact with SmpB were isolated from 2 × 10² transformants and designated as PA-1, PA-2, and PA-3, respectively (Figure 1B). The sequencing results of the plasmids conferring the expressions of peptide aptamers revealed that PA-1 was the best candidate with correct open reading frame, while a stop codon existed in its encoded region of PA-2 and a frameshift mutation occurred in PA-3 (Figure 1B). The interaction between PA-1 and SmpB was confirmed again by Bacterial two-hybrid system (B2H) (Supplementary Figure S1).

The overexpressed SmpB, scaffold protein SN and PA-1 were purified on Ni-IDA column and verified using SDS–PAGE (Figure 2A). The ELISA was performed to evaluate binding affinity of SmpB with PA-1, SN or BSA (Figure 2B). PA-1 showed stronger interaction with SmpB than those controls of SN and BSA. The binding curve of PA-1 interacting with SmpB was plotted, and equilibrium dissociation constant (K_d) was calculated by employing a model of one site binding-saturation analysis (Martinez-Archundia et al., 2012). The results showed that the binding of PA-1 to SmpB was stronger with K_d of 0.691 μM, in contrast to the binding of SN to SmpB with K_d of 1.380 μM (Figure 2C).

Bacterial two-hybrid system was performed to study the interaction of SmpB and PA-1. Expectedly, SmpB and PA-1 had no self-activations and toxicities. Although scaffold protein SN interacted with SmpB, PA-1 displayed much stronger interplay with SmpB (Figure 3A). In order to further explore the region of SmpB to interact with PA-1, SmpB truncations including pBT-SmpB ΔN34, pBT-SmpB ΔN34C30 and pBT-SmpB ΔC30 were co-transformed with pTRG-PA-1, respectively. The results implied that N- and C-terminal residues of SmpB were required for its interaction with PA-1 (Figure 3B). Subsequently the conservative sites of N- and C-terminal SmpB were aligned from different pathogenic bacteria using WebLogo 3 (Figure 3C), and a series of pBT-SmpB mutants were constructed (Liu et al., 2015). When pTRG-PA-1 was co-transformed with pBT-SmpB (T₁₄I₁₅), pBT-SmpB (F₂₆I₂₇) and pBT-SmpB (D₁₃₈K₁₃₉R₁₄₀), respectively, the cells showed growth defects on selective medium (Figure 3D), indicating the residues T₁₄I₁₅, F₂₆I₂₇ and D₁₃₈K₁₃₉R₁₄₀ of SmpB were essential for the interaction with PA-1.

Using HADDOCK docking program, the docking simulation was explored to verify the possible structural arrangements of PA-1 and SmpB complex, which were compatible with the previously identified residues from bacterial two-hybrid system. The HADDOCK between SmpB and PA-1 grouped the total of 146 structures into 9 clusters, which represented 80.3% of water-refined models, and the best model was selected when the lowest Z-score was -2.6. The docking result displayed that the residues T₁₄I₁₅, F₂₆I₂₇ and E₁₃₈K₁₃₉R₁₄₀ interacted with the variable regions of PA-1 (Figure 4).

The plasmids pN-PA-1 and pN-SN were constructed and introduced into A. veronii C4 for evaluating the function of PA-1 according to our previous methods (Liu et al., 2016). When treated at different NaCl concentrations (0.0, 0.3, 0.5, 1.0, 2.0, 3.0, 4.0, and 5.0%), the engineered strain A. veronii C4 (pN-PA-1) expressing PA-1 showed severely impaired growth compared to wild type appearing best, and both of A. veronii C4 (pRE112) and A. veronii (pN-SN) manifesting intermediary growth. Under 4.0 and 5.0% NaCl concentrations, all the strains were not able to grow anymore (Figure 5A). At 2.5 mM CaCl₂, the similar results were exhibited as demonstrated in 0–3.0% NaCl treatments (Figure 5B). At 25 mM MgCl₂, the growth velocity was ranked as follows, wild type, A. veronii C4 (pRE112), A. veronii C4 (pN-SN) and A. veronii C4 (pN-PA-1), of which A. veronii C4 (pN-PA-1) did not grow (Figure 5C),while the growth differences were not evident in negative control LB medium (Supplementary Figure S2).

After A. veronii C4 derivatives were grown to stationary phases, the total amount of RNA was extracted for relative expression analysis of virulence genes. The Quantitative Real-Time PCR (qRT-PCR) assays showed the transcriptional levels of ompA, aer, ahp, lcrV, fliL and uspA in A. veronii (pN-PA-1) were extremely significantly downregulated compared to wild-type A. veronii C4, and also showed significant differences with A. veronii (pN-SN) (Figure 6A). Besides, the levels of transcriptional downregulation of aexU, bvgS, hfq and flpL in A. veronii C4 (pN-PA-1) only had extremely significant differences compared to that of wild type (Figure 6B), while those of tolC, trh and smpB were identical among these strains (Figures 6C,D). The transcriptional level of arfA in A. veronii C4 (pN-PA-1) was downregulated significantly compared to those of wild type and A. veronii C4 (pN-SN) (Figure 6D), indicating not to compensate for the deficiency of SmpB function. Taken together, PA-1 interacted with SmpB and inhibited its function, thereby reducing the virulence gene expressions in A. veronii C4 (pN-PA-1).

To assess the medium lethal doses (LD₅₀ value) of A. veronii C4 derivatives, the adult zebrafish were applied as the animal model. The survival numbers of zebrafish were recorded after wild type and A. veronii C4 (pN-PA-1) were injected with a series of appropriate doses. As a result, the LD₅₀ value of wild type was 4.98 × 10⁵ CFU/g (Figure 7A), and the engineered strain A. veronii C4 (pN-PA-1) was 1.62 × 10⁷ CFU/g (Figure 7B), which was 33-fold higher than wild type. Next we investigated whether A. veronii C4 (pN-PA-1) could efficiently protect zebrafish against wild type attack. To explore whether A. veronii C4 (pN-PA-1) effectively stimulated fish immune response, the ELISA was performed to analyze the changes of immunoglobulin M (IgM) antibody level in zebrafish. The total tissues of zebrafish were collected for measuring IgM levels at different time points after vaccination. The result showed that the antibodies against IgM were significantly higher in the immune groups than the controls from 14 to 28 days (P < 0.01) (Figure 7C). Based on the varying patterns of IgM levels, a large amount of zebrafish were vaccinated with A. veronii C4 (pN-PA-1) for 14 days, and challenged with wild type A. veronii C4. The percent survival was recorded in the following 7 days, the vaccinated zebrafish were well protected with RPS of 65%, in contrast to 100% mortality of the control group which was injected with saline (Figure 7D).