Analysis of the antimicrobial mechanism of porcine beta defensin 2 against E. coli by electron microscopy and differentially expressed genes

Porcine beta defensin 2 (pBD2) is a cationic antimicrobial peptide with broad spectrum antibacterial activity, which makes it a potential alternative to antibiotics to prevent and cure diseases of pigs. However, development of pBD2 as an effective antibiotic agent requires molecular understanding of its functional mechanism against pathogens. In this study, we investigated the functional mechanism of pBD2 antibacterial activity. Escherichia coli was incubated with different pBD2 concentrations for different times. Electron microscopy was used to analyze the locations of pBD2 and its induced morphological changes in E. coli. Gene expression analysis was also performed to further understand the molecular changes of E. coli in response to pBD2 incubation. The results demonstrated that E. coli membranes were broken, holed, and wrinkled after treatment with pBD2, and pBD2 was located on the cell membranes and manly in the cytoplasm. Furthermore, 38 differentially expressed genes (DEGs) were detected, successfully sequenced and confirmed by quantitative real-time PCR (qRT-PCR). Most of the known functional DEGs were associated with DNA transcription and translation and located in the cytoplasm. Collectively, the results suggest that pBD2 could have multiple modes of action and the main mechanism for killing E. coli might be influence on DNA transcription and translation by targeting intracellular molecules after membrane damage, although transport and metabolism proteins were also affected.

Scanning electron microscopy. The E. coli cells were incubated for 1 h and 4 h with different concentrations of pBD2 (0, 37.5, 75 and 150 μg/mL). Morphological changes were analyzed and are shown in Fig. 2. Bacterial cells from the control group were intact and smooth without noticeable damage ( Fig. 2A,E). In contrast, E. coli cells treated with different concentrations of pBD2 were damaged, as evident by outflow of cell contents (Fig. 2C), deep craters and burst cells (Fig. 2B,F), debris of cells and cell death (Fig. 2D,F-H), and so on. Morphological damage occurring in response to different doses of pBD2 was similar. The damage was more serious and noticeable at 4 h than that at 1 h. Furthermore, the numbers of cells were reduced with increased pBD2 concentrations. The results demonstrated similar morphological changes irrespective of pBD2 concentration, whereas more pronounced effects were observed with increased incubation time.
Cellular localization of pBD2 peptides. Immuno-fluorescence analysis was used to determine pBD2 localization in bacterial cells (Fig. 3). The results showed higher concentration of pBD2 on bacterial membranes at both ends of E. coli, and some of the pBD2 peptide was localized inside the cells, suggesting penetration of pBD2 into bacterial cells. Moreover, immuno-gold TEM analysis was used to confirm the precise localization of the peptide in bacterial cells (Fig. 4). E. coli cells analyzed after culturing with 150 μg/mL pBD2 for 4 h demonstrated collapsed cell membranes and cytoplasmic localization of pBD2, as evidence by the presence of gold particles mainly in the cytoplasm (Fig. 4b,d). Our results showed several other striking structural alterations in E. coli treated with pBD2 including indistinct inner membranes in many regions, loss of the double membrane structure and cell contents from the damaged membranes, particularly from the cross-section (Fig. 4c,d). The results from both approaches demonstrated that pBD2 was possibly located on the membrane and mainly in the cytoplasm of E. coli.  Identification of DEGs by ACP-based RT-PCR. Spectrophotometric analysis of RNAs extracted from treated and untreated E. coli showed that the OD 260 /OD 280 ratios were between 1.8 and 2.2 and that concentrations were above 1 μg/μL, suggesting the high purity and optimum quantity of RNA suitable for further downstream analysis. Using 20 arbitrary primers with ACP-based RT-PCR analysis, the results showed a variety of differential gene expression among control and pBD2 treatment groups. As the patterns of all the PCR gels were similar at different pBD2 concentrations and times for each arbitrary primer, except for a few DEGs that were different in different treatments, a subset of the results is shown in Fig. 5. The DEGs were cloned into a T-Vector and sequenced. From the 20 tested arbitrary primers, more than 160 sequences were obtained, and more than 90 sequences were successfully sequenced.
Analysis of the DEGs. The sequences were analyzed by BLAST analysis using the NCBI and ENSEMBL databases. The results showed that 38 transcripts were differentially expressed in response to pBD2 treatment and that 27 of them corresponded to 28 functional genes (one sequence covers two genes); 7 fragments were between two genes; and 4 were unknown functional protein sequences (hypothetical proteins). The results are listed in Table 1. Twenty-eight known DEGs are classified, abbreviated and listed in Table 2. Among the 28 known functional DEGs, 6 genes were related to ion transport or protein trans-membrane transport, all of which are membrane proteins; 13 genes were related to DNA transcription and translation (including two fragments coding tRNA), all of which are in the cytoplasm; and 9 genes were related to metabolism, where among them, 1 gene is a membrane protein, 3 genes are in the cytoplasm, and 5 genes have been predicted to be in localized in the cytoplasm according to their functions. These results suggested that among the 28 known functional genes, 7 genes regulate the function of cell membranes, and 21 genes play an important role in the cytoplasm. Among the 6 ion or protein transport genes, kdpB, FhuC and pstS are related to ion transport, secY and metI are related to protein translocation channels and methionine import, and msbA is involved in lipid A export and for biogenesis of the outer membrane. Among the 9 metabolic genes, sucC, otsB and frmA are related to carbon metabolism, and frmB and eutB are related to amino acid metabolism. eptC is related to LPS core biosynthesis. udhA and ubiH are related to redox balance; udhA has an impact on the redox balance of cells, which could convert NADPH to NADH, through which it is involved in the respiratory chain for energy generation. ubiH has oxidoreductase activity, acting on paired donors, with incorporation or reduction of molecular oxygen. Thirteen genes are related to DNA transcription and translation, and they are all located in the cytoplasm. 10SaRNA, CrsB, cb1, lac1, sixA and rsmH are all related to regulation of DNA transcription. 10SaRNA (also known as SsrA RNA or tmRNA) has properties of tRNA and mRNA; it binds to SsrA-binding protein (the gene is adjacent to sequence 5) and rescues stalled ribosomes, known as trans-translation. CrsB is a non-coding RNA molecule that antagonizes the effects of CsrA in vivo. CsrA modulates glycogen synthesis and catabolism, among other process, cb1 regulates two operons consisting of ABC transporters that are part of the cys regulon and is an ssuEADCB and tauABCD operon transcriptional activator. 23sRNA, rplN, lysS, hisS and rpoB, and tRNA are involved in DNA translation or protein synthesis. There are 14 genes adjacent to the 7 differently expressed fragments; among them, there are 2 hypothetical proteins and 12 known functional genes (also listed in Table 2). They are also related to transport (TonB, sat and syd, in the membrane), DNA transcription and translation (lysR and smpB, in the cytoplasm), and metabolism (phD, metF, katG, csiD, gk2, citD and citC). Among the 4 unknown genes, hypothetical protein 3 (HP3, 417 bp, 198 aa) is known to be associated with bacterial flagellum assembly, and HP2 (408 bp, 135 aa) is linked to the flavinator of succinate dehydrogenase family proteins.
Quantitative real-time PCR confirmation for selected genes. qRT-PCR was used to confirm the DEGs of E. coli in response to different concentrations of pBD2 treatments. The relative expression level of each gene was normalized by GAPDH gene expression. The results of qRT-PCR are consistent with the results of ACP-based RT-PCR and are shown in Fig. 6. The expression analysis of transport genes is shown in Fig. 6A. The genes secY and metI were upregulated 2-to 16.2-fold, respectively, in all the treatment groups compared to that of the control (P < 0.01) at 1 h and 4 h post treatment with the exception of secY gene expression at 4 h, which had no change with the maximum dose of pBD2 treatment. kdpB expression was higher at 1 h (P < 0.01) in response to only the highest concentration (150 μg/mL) of the pBD2 treatment but was consistently upregulated at 4 h in all the treatment groups. Expression of msbA was reduced in E. coli cells treated with 37.5 μg/mL of pBD2 compared to that of the untreated control (P < 0.01). The expression of Fhu differed at 1 h and 4 h. As shown in Fig. 6B, the levels of metabolism-related genes frmA and ubiH were increased, while sucC and eutB transcripts were basically decreased in response to pBD2 treatment. otsB was downregulated at 1 h and upregulated at 4 h (P < 0.05 and P < 0.01, respectively). eptC was upregulated in response to the 37.5 μg/mL pBD2 treatment (P < 0.01), and frmB and Cat genes were altered at different pBD2 concentrations and at different times. Expression analysis of genes involved in DNA transcription and translation are shown in Fig. 6C. Expression of 23srRNA, rplN, lysS, lac1, CrsB and rpoB was basically 1-to 8-fold upregulated (P < 0.05 or 0.01) in pBD2 treated-cells compared to that of the control. cb1 expression was lower at lower pBD2 concentrations (P < 0.05) and higher at high concentrations (P < 0.01). sixA was downregulated at 1 h and upregulated at low pBD2 concentration at 4 h (P < 0.01). In total, the proteins for DNA transcription and translation were influenced. Four hypothetical proteins are shown in   6D. HP3 was downregulated (P < 0.01). Interestingly, the expression level of HP3 was decreased gradually at 1 h and increased at 4 h with increased pBD2 concentrations. HP1 was upregulated at 150 μg/mL pBD2 treatment (P < 0.01). HP2 and HP4 varied at different times and concentrations. The longer sequences between two genes were selected and the expression levels were detected, including the two genes shown in Fig. 6E. Sequences 4 and 6 were basically upregulated (except for seq6 at 150 μg/mL for 1 h); sequence 5 was downregulated at 1 h and upregulated at 4 h (P < 0.01). Sequence 3 varied at different times and concentrations and was significantly different from the control (P < 0.01). The changes in two genes (HP5 and syd) near sequence 4 were downregulated (P < 0.01); moreover, the trends of their changes were consistent. The trends of transcript change in gk2 and lysR other than seq5 were also consistent, as was the case for HP6 and phD other than seq1, and katG and metF other than seq2. In total, the genes nearby besides differentially expressed fragments were changed themselves.
In summary, the results of qRT-PCR confirmed the results of ACP-based RT-PCR.

Discussion
pBD2 is one of the defensins secreted by pigs and shows high antibacterial activity against bacteria, making it a good candidate for antibiotics 15,16 . Veldhuizen (2008) reported that synthetic pBD2 showed high antibacterial activity against pathogenic intestinal bacteria and could inhibit E. coli, S. typhimurium, and P. aeruginosa and so on 16 . Our previous reports showed that pBD2 had high antibacterial activity against gram-negative E. coli and gram-positive S. aureus, including multi-resistant E. coli isolated from fowl 15 . The bacterial concentration for tests was usually approximately 10 4 -10 6 cfu/mL; in this test, the concentration was raised to 1 × 10 9 cfu/mL because a large number of bacteria were required in the later experiments. The results showed that expressed pBD2 could inhibit the growth of 10 9 cfu/mL bacteria. In addition, antibacterial activity was increased with pBD2 concentration and decreased with exposure time (P < 0.01), which was consistent withprevious results for 10 5 -10 6 cfu/mL bacteria in our lab 15 . After co-culture for 1 h and 4 h with different pBD2 concentrations, morphological damage occurring in response to different doses of pBD2 was similar. E. coli was seriously damaged and showed pronounced changes including leakage of contents and debris from cell death (Fig. 2), and the damage was more serious with increased pBD2 concentrations and exposure times. Some cells death appeared to be because of the cell contents leaking (Fig. 2C), some cells appeared to be burst with deep craters (Fig. 2B,F) and some cells deposited very small pieces of cell debris, appearing that the cells were dissolved (Fig. 2D,G). These results may imply that there is more than one mechanism causing cell death. Moreover, some white particles or blebs appeared on the cell surface (Fig. 2B,D), and it was difficult to distinguish whether they were either debris from burst cells or blebs because of cell damage from pBD2. In addition, pBD2 may have killed the E. coli one by one because some cells appeared normal, and other cells were dead in the same field of view. Some morphological changes of E.coli in this experiment are consistent with those of E. coli effected by human epididymis 2 (HE2) protein isoforms, HE2α/HE2β1/ HE2β2 32 . However, our results differed from those for antimicrobial peptides gramicidin S (GS) and PGLa acting on E. coli in hypotonic medium on silicon platelets 33 . Our results showed morphological changes were more complex, implying that the actions of pBD2 might be complex, which requires further research.
The location of pBD2 in E. coli was detected by IFM. The results indicated that pBD2 was distributed on the membrane, specially focused at both ends of E. coli, and that some was found inside of the bacteria. The locations of pBD2 in bacteria cells were further confirmed by immune-gold TEM, which demonstrated that pBD2 was a cell-penetrating peptide. Antimicrobial peptide magainin 2, chicken cathelicidin-2, and SpHyastatin among others have been reported to pass through the cell membrane and are located in the cytoplasm of bacteria [34][35][36][37][38] 25 . These results indicate that AMPs have different mechanisms, even when they have similar structures. Our results indicated that pBD2 was a cell-penetrating peptide and could enter the E. coli cells, perhaps targeting intracellular molecules and playing important roles in the cytoplasm. In addition, E. coli undergoes fission, starting from one end of the bacteria, which may be the most easily invaded location by defensins, and this may be the reason why pBD2 mainly focused on the ends of the E. coli. These results demonstrated that pBD2 could disrupt the membrane and enter the cytoplasmic membrane.
It is well known that the first step for antimicrobial peptides to kill bacteria is interacting with the negatively charged lipids on the membrane surface. We had predicted the structure of mature pBD2, composed of 37 amino acids, based on the structure of hBD1. The peptide has a typical beta defensin shape with one alpha helix (His2-Lys8) and three beta sheets (Thr11-Asn13, Phe20-Tyr28 and Lys31-Arg37). The structure of recombinant pBD2 with a His-tag was predicted to have another alpha helix in its N-terminal 15 . The three disulfide bonds make the peptide flexible and impart a stabilizing force. These stable structures are more likely to disrupt surface tension and create membrane defects that ultimately break down the bilayer integrity of bacteria 30 . pBD2 may interact with the outer membrane and then may either aggregate on the membrane or, once a critical concentration is reached, lead to the disruption of the membrane, followed by pBD2 interacting with the inner membrane and entering into the cells.
In this study, RNAs of E. coli were extracted according to the reported method with some modifications 41 , and ACP-based RT-PCR without adding poly(A), DEGs were found. The results proved that the prokaryotic RNA contained polyadenylic acid. To obtain polyadenylated RNA, the RNAs were not isolated by conventional procedures involving deproteinization by phenol extraction, which appeared to cause specific losses of poly(A)-RNA 41 . ACP-based RT-PCR technology is an easy differential display technology (detected by agarose  gel electrophoresis). The core of the technology is an annealing control primer, which binds highly specifically to the target template and can provide greater annealing specificity during PCR 42 . The system is used to identify DEGs without generating false positives. Therefore, this technique has been widely used to identify DEGs under different treatment conditions and for the development of organisms [43][44][45] . By ACP-based RT-PCR, the DEGs were detected and were very samilar in response to different pBD2 concentration treatments and exposure times with each annealing control primer, which indicated ACP-based RT-PCR was credible and repeatable. By this technique, 38 DEGs were obtained, and 28 known genes from these were analyzed. Most changed genes were related to DNA transcription and translation, and mainly were in the cytoplasm. These results suggest that the pBD2 influenced genes mainly in the cytoplasm and with DNA transcription and translation functions. A great deal of researches has indicated that AMPs act with intracellular molecules, such as DNA or ribosomes, which can subsequently result in influence on DNA or RNA synthesis, such as PR-39, indolicidin and its analogs, buforin 2 and its anologues BF2-A/C and so on 21,[23][24][25]46 . Most of the DEGs of E. coli were in the cytoplasm and related to DNA transcription and translation, and the results implied that pBD2 kills E. coli by influencing DNA or RNA through the cell membrane, which is consistent with the location of pBD2. Among the 28 known genes besides the genes in the cytoplasm, 7 known genes on the membrane were influenced, and they are mainly integral components of the membrane and related to transport of potassium ions, iron ions, phosphate ions, and so on. In addition, the two genes next to seq6 were involved in transport. These results suggest that transport proteins in the membrane also changed, and there have been some reports that AMPs killed bacteria based on the ion mechanism [47][48][49] . The genes for metabolism were changed, along with those involved in carbon (energy) metabolism, and amino acid and redox balance, which are all involved in the actions of AMPs killing bacteria 29 .
The results of qRT-PCR confirmed the results of ACP-based RT-PCR. The expression of membrane transporters was basically increased in E. coli. Expression of kdpB changed, and suggesting that potassium ion balance was disrupted. Yang et al. have reported that human beta defensin 2 (hBD2) is able to inhibit potassium channels 47 , and perhaps pBD2 has this activity, leading to higher kdpB levels. Ferric siderophore receptors provide an advantage for the cells to scavenge iron but can also be exploited by antibacterial compounds 48 . In E. coli, antibacterial peptide MccJ25 can bind outer membrane ferric siderophore receptor FhuA and enters the target bacteria and kills them through a TonB-dependent process 48,49 . FhuC, one of the ferric siderophore receptors, and the TonB-dependent siderophore receptor family protein (TonB) next to sequence 6 were upregulated at 4 h, which implied that pBD2 killed bacteria by influencing these processes. Protein CsiD combined with Fe 2+ , which had lower expression, implied that perhaps soluble ferrous iron (Fe 2+ ) was in short supply, was oxidized to insoluble ferric iron (Fe 3+ ) and was associated with the high expression of Fhu 50 . Meanwhile, SecY and metI were also upregulated, which suggests that translocation of the proteins and amino acids changed. Expression of msbA, involved in the transport of lipid A across the inner membrane and for biogenesis of the outer membrane, was lower than that of the control, suggesting that the biogenesis of membranes was influenced, potentially inhibiting the growth of E. coli. The genes for metabolism changed in different ways: expression of frmA, which influences energy metabolism, changed more than 100-fold compared to that of the control at 1 h pBD2 treatment, which implied that the energy station may be deeply affected, including the changes in sucC and otsB. Antimicrobial peptide LL-37 was reported to killing the bacteria mainly by target energy metabolism 29 . eptC is related to LPS core biosynthesis and had high expression at low pBD2 concentration and lower expression at high concentration, implying that bacteria may survive at low pBD2 concentrations. ubiH has oxidoreductase activity, and its high expression implied that the redox balance was disrupted.
The changed genes for DNA transcription/translation were mainly affected in response to pBD2. rplN, lysS and rpoB were upregulated, which enhance DNA transcription. Moreover, upregulated CrsB also leads to enhanced DNA transcription. CrsB antagonizes the effects of CsrA in vivo, and CsrA can facilitate glgC (glycogen biosynthesis) mRNA decay in vivo and inhibit in vitro glg gene expression posttranscriptionally 51 . The functions of 10sa RNA and smpB (SsrA-binding protein, the gene next to sequence 3, Fig. 6F) were to rescue the stalled ribosomes 52,53 . Their high expression at low pBD2 concentrations at 1 h implied that bacteria may be able to rescue the stalled ribosomes caused by pBD2 exposure, while lower expression at higher pBD2 concentration implied that the ribosomes might be inactivated. Roy et al. reported that proline-rich peptides, such as Bac7 1-35 , oncocins, and two oncocin derivatives, blocked the peptide exit tunnel of the ribosome and interfered with the initiation step of translation, which led to inactivated protein synthesis 54,55 . Florin et al. reported that Api137, a derivative of the insect-produced antimicrobial peptide apidaecin, bound the Escherichia coli ribosome and resulted in a global shutdown of translation termination 56 . sixA was apparently downregulated at 1 h, which enhanced DNA transcription, while its lower expression (at 37.5 μg/mL at 4 h) inhibited related gene expression. cb1 regulates two operons consisting of ABC transporters, and its high expression at higher concentration enhanced gene expression, while its lower expression at lower pBD2 concentration inhibited expression. Our results appear contradictory; it is difficult to conclude that pBD2 enhances or reduces gene expression, perhaps indicating that genes related to DNA transcription and translation had dynamic changes in response to pBD2 concentrations and exposure times, which requires further study.
The noncoding sequences including the nearby genes were also changed. In addition, those genes are related to transport, DNA transcription and translation, and metabolism. TonB, sat and syd were related to transport. TonB was related to siderophore transport, similar to FhuC, and their expression levels were basically consistent. Protein Syd interacts with the SecY protein in vivo as a regulatory factor that negatively controls the translocase function of SecY, and their expression levels was inversely related (Fig. 6). The lysR family protein also regulated transcription and had high expression, which implied that DNA transcription was influenced. phD, metF, katG, csiD, gk2, citD and citC are related to metabolism; phD is related to the metabolism of phospholipids, and its expression level was lower at lower pBD2 concentrations and higher at higher concentrations, which indicates that pBD2 may influence membranes at different concentrations. citD and citC are related to citrate lyase; citD had high expression and is related to glucose metabolism, providing energy, same to gk2. Although there is not a clear connection between noncoding sequences and their nearby genes, the nearby genes were clearly changed, and the noncoding sequences may have regulated the expression of their nearby genes. The detected seq5 and seq6 are longer than 200 bp and may be long non-coding RNA, and their function requires further study. In short, from differential gene analysis, pBD2 affects the transport of cell membranes, DNA transcription/translation and metabolic activities. Specifically, the main effect is on DNA transcription and translation.
From the analysis above, pBD2 has multiple actions. It disrupted cell membranes, influenced the ion and energy balance, and affected DNA transcription and translation. Overall, the main action of pBD2 is the influence on DNA transcription and translation by targeting the intracellular molecules after membranes are damaged.

Conclusion
After incubation with pBD2, the contents of E. coli leaked. pBD2 was located in the cell membrane and cytoplasm based on IFM and predominantly in the cytoplasm according to immuno-gold TEM. DEGs of E. coli were detected by ACP-based RT-PCR, and they were related to transport and transmembrane transport proteins in the membrane and DNA transcription and translation and metabolic proteins in the cytoplasm. The known functional DEGs were mainly related to DNA transcription and translation. The results of qRT-PCR confirmed those of ACP-based RT-PCR; the transmembrane transport proteins were mainly upregulated; and the genes related to DNA transcription and translation were influenced. The main mechanism for pBD2 killing E. coli is the influence on DNA transcription and translation by entry into the cytoplasm after membranes are damaged, although other proteins are also affected. The results explained the mechanism of pBD2 killing E. coli, which will be the basis for further study and for the application of this AMP in pig disease.

Materials and Methods
Bacterial strains and growth conditions. E. coli ATCC 25922 used in the present study was purchased from the Beijing Ordinary Microbiology Strain Store Center, Beijing, China. The E. coli were cultured in Luria broth (LB) medium (1% peptone, 0.3% beef extract, and 0.5% NaCl) overnight and then inoculated into fresh LB medium for 2-4 h with constant shaking at 220 rpm.
Preparation of pBD2. The engineered strain E. coli BL-pET-pBD2 was cultured and induced according to the procedures described previously 15 . In brief, E. coli BL-pET-pBD2 was induced by 1 mM IPTG for 4 h and was lysed by sonication for 30 min. The recombinant pBD2 was purified by a Ni-NTA agarose column (His-tag affinity column) and dialyzed against 10 mM phosphate buffer or water and concentrated with polyethylene glycol 8000. The concentration of protein was determined by the BCA method.
Kill kinetics. The kill kinetics were performed by the turbidimetric method as reported previously 15,57 .
Optical density of the bacterial cells was measured at 600 nm, and the absorption value of OD 600 was adjusted to approximately 1 (the final density was approximately 10 9 cfu/mL). E. coli were incubated with pBD2 (final concentrations were 0, 37.5, 75, and 150 μg/ml) in polypropylene 96-well microtiter plates for different time durations (0 h, 1 h, 4 h, and 8 h). The optical density was measured at 630 nm (reference 405 nm) by a microplate reader (Stat Fax 2100, Awareness Technology Inc. USA). The OD value at 0 h was adjusted to 0, the increased OD values were calculated, and all assays were carried out in triplicate. All values were expressed as the mean ± standard deviation (SD).
Bacterial treatments. E. coli at 10 9 cfu/mL was suspended in different concentrations of purified recombinant pBD2 (0, 37.5, 75, and 150 μg/ml) at 37 °C for different lengths of time (1 h and 4 h), and bacteria were harvested after centrifugation at 3000 rpm for 5 min. After washing three times with 10 mM PBS, bacterial cells were used for RNA extraction, SEM, IFM and immuno-gold TEM.
Scanning electron microscopy. SEM analysis was performed as described previously 33  Immuno-fluorescence analysis. Immuno-fluorescence analysis was performed according to methods published with some modifications 38 . E. coli cells were spread on anti-off slides (Wuhan Boster Biological Engineering Co., Ltd, China) and air-dried. After fixing with 4% paraformaldehyde for 2-4 h at 4 °C, the samples were treated with 0.5% Triton-100 in 10 mM PBS buffer for 15 min. Samples were then washed with 10 mM PBS buffer for 3 × 5 min and were incubated with 10% normal goat serum in PBS for 30 min at room temperate in a wet box. After drying with absorbent paper, bacterial samples were incubated with 50 µl polyclonal pBD2 (prepared by us from rabbit 58 , diluted to 1:500) at 4 °C overnight. Afterwards, cells were incubated with fluorescently (FITC) labeled secondary antibodies for 1 h at 37 °C in the box, washed with PBS for 3 × 5 min and then stained by DAPI solution for 5 min in the dark. After staining, samples were washed 3 times with PBS and blocking agent was added containing anti-fluorescence quencher and sealed with cover slips. Controls for immune-fluorescence were performed by incubating samples without primary antibodies. Cells were observed by microscopy (DM6000B, Leica, Germany).
Immuno-gold transmission electron microscopy. Immuno-gold TEM was used to further determine the localization of the peptide as reported previously 36 . The cells were embedded in 2% low-melting point agarose, trimmed into small pieces (1 mm 3 ), and fixed in fixative buffer for 2 h on ice. The samples were first dehydrated in 30% methanol for 15 min at 0 °C and then dehydrated with gradient concentrations of methanol at −20 °C. After infiltration in the mixtures of 100% methanol and Lowicryl K4M (the ratios were 1:1, 1:2, and 1:3) for 60 min and in pure Lowicryl K4M at −20 °C overnight. Samples were then embedded in gelatin capsules, filled with resin and capped tightly, polymerized under UV light (360 nm) and placed 20 to 30 cm away from the UV source for 48 h at −35 °C, after which they are placed for 12 h at room temperature.
The materials were cut into 0.1-μm thick sections by a CM1900 Frozen Slicer (Leica, Germany). Sections were mounted on 100 mesh copper Formvar-carbon coated grids and incubated with 10 mM PBS (pH 7.4) containing 0.05% Triton X-100, 0.05% Tween and 1% bovine serum albumin for 5 min at 20 °C. Further, sections were incubated with pBD2 antibody (1:200) for 2 h at RT and exposed to protein-A gold (1:50) for 1 h at RT, and specimens were then stained with 3% uranium acetate for 4 min and lead citrate for 1 min. For all electron microscopy, a JEM-1400 electron microscope was used. Cells incubated with PBS without primary antibodies were used as a negative control.
SCientifiC RePoRtS | (2018) 8:14711 | DOI:10.1038/s41598-018-32822-3 RNA extraction of E. coli. The total RNA of E. coli was extracted according to the procedure reported previously with some modifications 41 . Briefly, 1 mL of cells at 10 9 cfu/mL were washed with water and digested with lysozyme for 40 min in 15 mM Tris-HCl (pH 8.0), then protoplasts were collected and were lysed by 0.5% SDS in 80 mM Tris-HCl buffer (pH 7.5). Then, the proteins were removed by saturated sodium chloride solution. The RNA was precipitated and washed with 70% ethanol. After drying at room temperature, RNA pellets were then dissolved in RNase free water. The extracted total RNA was digested with DNase for 30 min and subjected to an ethanol-based purification before use in the downstream applications. The concentration and purity of RNA was determined by UV absorption at OD260 and OD280 by spectrophotometer (2800 UV/VIS, Shanghai, China). The quality of RNA was determined by 1% agarose gel electrophoresis.

Identification of DEGs by ACP-based RT-PCR.
Genes differentially expressed in bacteria in response to pBD2 treatment were identified using the GeneFishing TM DEG Kit (Seegene, Seoul, South Korea) following the manufacturer's instructions using 20 different arbitrary primers in combination with an oligo-dT anchor primer. Total RNAs extracted from the bacteria were converted into first-strand cDNAs using reverse transcription-PCR (RT-PCR) in 20 μl reaction containing 1 μl RNA, 4 μl of 5 × PrimeScriptII buffer, 1 μl of dNTP mixture (each 10 mM), 1 μl of 10 μM dT-ACP1, 0.5 μl of RNase inhibitor (40 U/μl), and 1 μl of PrimeScriptII RTase(200U/μL) at 42 °C for 1.5 h. First-strand cDNAs were stored at −20 °C until use for PCR. DEGs were screened with the ACP-PCR based method using primers that anneal specifically to the template in 20 μl of reaction containing 1 μl of cDNA, 1 μl of dT-ACP2 (10 μM), 1 μl of 10 μM arbitrary ACP, and 13 μl of premix. The PCR protocol for second-strand cDNA synthesis included one cycle at 94 °C for 5 min, followed by 50 °C for 3 min, and 72 °C for 1 min. The second-stage PCR amplification was then performed using 40 cycles of 94 °C for 40 sec, followed by 65 °C for 40 sec, 72 °C for 40 sec, and a 10 min final extension at 72 °C. The amplified PCR products were separated and analyzed on 2% agarose gels stained with ethidium bromide. The ACP-based RT-PCR was repeated three times. The differentially amplified PCR fragments were extracted from the gel using a DNA purification kit (TIANGEN, China) and directly cloned into a pMD18-T vector and transformed into Escherichia coli TG1 cell. The cloned fragments were subjected to DNA sequencing to characterize the identity of the DEGs.
Analysis of the DEGs. All sequences were subjected to BLAST analysis using the bacterial genome database in GenBank and the Ensembl (http://bacteria.ensembl.org) genome browser to characterize the functional annotations.
Quantitative real-time PCR. Reverse transcription was performed as described above. Differently expressed genes identified in ACP-RT-PCR were further confirmed by a quantitative real-time PCR (qRT-PCR) method performed as follows: 40 cycles of denaturation at 95 °C for 15 sec, annealing at 60 °C for 40 sec, and extension at 72 °C for 15 sec, according to the manufacturer's protocol. Assays were performed with SYBR Green Dye (TaKaRa, Dalian, China) using a Bio-Rad CFX96 real-time PCR platform. All the samples were analyzed in triplicate and fold-changes of gene expression were calculated by 2 −△△CT methods with GAPDH as a reference gene.
Statistical analysis. An analysis of variance (ANOVA) was performed with SPSS 17.0 software. LSD or Dunnett's method was used to compare treatment means. Statistical significance was defined as P < 0.05.