Development of peptides targeting receptor binding site of the envelope glycoprotein to contain the West Nile virus infection

West Nile virus (WNV), re-emerging neurotropic flavivirus, can cross the blood–brain barrier (BBB) and cause fatal encephalitis and meningitis. Infection of the human brain microvascular endothelial cells (hBMECs), building blocks of the BBB, represents the pivotal step in neuroinvasion. Domain III (DIII) of the envelope (E) glycoprotein is a key receptor-binding domain, thus, it is an attractive target for anti-flavivirus strategies. Here, two combinatorial phage display peptide libraries, Ph.D.-C7C and Ph.D.-12, were panned against receptor-binding site (RBS) on DIII to isolate peptides that could block DIII. From series of pannings, nine peptides (seven 7-mer cyclic and two 12-mer linear) were selected and overexpressed in E. coli SHuffle T5. Presence of disulfide bond in 7-mer peptides was confirmed with thiol-reactive maleimide labeling. Except for linear peptide 19 (HYSWSWIAYSPG), all peptides proved to be DIII binders. Among all peptides, 4 cyclic peptides (CTKTDVHFC, CIHSSTRAC, CTYENHRTC, and CLAQSHPLC) showed significant blocking of the interaction between DIII and hBMECs, and ability to neutralize infection in cultured cells. None of these peptides showed toxic or hemolytic activity. Peptides identified in this study may serve as potential candidates for the development of novel antiviral therapeutics against WNV.


Results
Selection of DIII binding peptides from combinatorial phage libraries. A phage display was employed using two combinatorial Ph.D. libraries (New England Biolabs, USA) to isolate 7-mer cyclic or 12-mer linear peptides, which bind to the DIII of the E protein. Libraries were panned either against RBS ( 299 GTTYG-VCSK 307 -biotin, ~ 915 Da) 45 present on the DIII or against recombinant DIII (rDIII; Supplementary Fig. S1). Details of each round of panning are summarized in Table 1. Phages eluted from each round of panning were amplified and purified with PEG/NaCl precipitation. The interaction of the amplified phages from the last round of panning with rDIII was confirmed by semi-quantitative phage ELISA, wherein phages from both libraries showed binding affinity to rDIII (Ph.D.-C7C: A 450 = 1.026; Ph.D.-12: A 450 = 2.477; Fig. 1A). Please note that amplified phages used in ELISA contain a repertoire of all phages eluted in the last panning. Negative controls in ELISA did not show absorbance more than 0.37, which indicates the specific binding of the amplified phages to the rDIII (Fig. 1A).
Amplified phages were then serially diluted up to 10 -10 and seeded on LB plates supplemented with X-gal and IPTG along with E. coli ER2738 (soft agar overlay method) to obtain well-separated phage plaques. Thirty plaques for each library were randomly selected and fragment encoding 7-mer cyclic or 12-mer linear peptides were PCR amplified using vector-specific primers (Supplementary Fig. S2 and Supplementary Table S1) and sequenced. In the case of C7C-mer phage clones, sequences were grouped into 7 clusters ( Supplementary  Fig. S3A), in which the peptide CTKTDVHFC was the most enriched candidate (68% frequency), followed by CTNANHYFC (12% frequency). The peptides CIHSSTRAC, CMQTQRAHC, CTYENHRTC, CDPRHSKFC, and CLAQSHPLC were presented by 4% frequency. Regarding 12-mer linear phage clones, sequences were clustered into two groups representing SGVYKVAYDWQH (67% frequency) and HYSWSWIAYSPG (33% frequency) peptides ( Supplementary Fig. S3B).
Phage display may generate molecules (e.g. peptides, antibodies, etc.) with a high non-specific binding affinity to the solid phase used in panning, substances presented in buffers, or contaminants. Therefore, consecutive panning steps were performed using different solid phases (various plastic surfaces), washing buffers, and different www.nature.com/scientificreports/ elution strategies to avoid non-specific binders (Table 1). Further, amino acid sequences of peptides were analyzed using the TUPScan tool (SAROTUP server), wherein none of the peptides was classified as a target-unrelated peptide (TUP). Similarly, in a search performed for peptides characterized by other researchers (Peptide search tool in UniProtKB database) no match was found with any of the deposited sequences (Table 2). Phage clones bearing 7-mer and 12-mer peptides were amplified and purified to test the binding ability of individual clones to the rDIII in semi-quantitative phage ELISA. All representative phage clones bearing 7-mer cyclic peptides showed significantly higher binding affinity to rDIII compared to the negative controls, wherein rDIII was omitted (Fig. 1B). Clones CC2 (phage clone carrying 7-mer cyclic peptide -CC), CC4, and CC6 were the most promising binders (in all cases A 450 > 1.3), while CC15 showed the least affinity to rDIII (A 450 = 0.518). In the case of the clones carrying 12-mer linear peptides, LC3 (phage clone carrying 12-mer linear peptide -LC) showed a significant binding (A 450 = 1.213) to rDIII (Fig. 1B). Although LC19 showed a strong interaction with rDIII (A 450 = 3.051), high absorbance in the negative control (A 450 = 2.985) indicated non-specific binding to the solid-phase (Fig. 1B).
Production of DIII-blocking peptides in E. coli expression system. The 7-mer and 12-mer peptides from the phage clones used in phage ELISA, mentioned above, were produced in the E. coli expression system. All peptides were N-terminal tagged, and the sequence of the overexpression cassette was as follows: 6xHis tag -BamHI restriction site (GS) -28 aa tag (devoid of cysteine and methionine residues) -Factor Xa cleavage site (IEGR) -SalI restriction site (VD) -GGGGS linker -enterokinase cleavage site (DDDDK) -[alanine (A)] -7-mer or 12-mer peptide sequence (7-mer sequences were flanked with cysteine at both ends) -GGGS linker -stop codon ( Supplementary Fig. S4). Please note that, all peptides had the C-terminal GGGS linker to mimic their presentation on the phage pIII protein. DNA encoding peptide sequence was ligated into the expression vector ( Supplementary Fig. S5) and electroporated to E. coli Shuffle. Correct insertion of DNA in the transformants confirmed with PCR and the purity of all tagged peptides confirmed with LDS-PAGE are presented in Fig. 2A and B. Molecular masses of the representative cyclic and linear peptides assessed by MALDI-TOF MS are presented in Fig. 2C and D. The molecular masses of all tagged peptides judged with MALDI-TOF MS are presented in Supplementary Fig. S6A. Peptides were cleaved off the N-terminal tag using enterokinase and the tag was removed with nickel affinity chromatography. Cleaved ACX 7 CGGGS or X 12 GGGS peptides were used for the assessment of rDIII binding ability and rDIII blocking potential. Precise molecular masses of peptides (without the tag) judged with MALDI-TOF MS are presented in Supplementary Fig. S6B. The observed molecular masses of all peptides were matched with theoretical masses predicted by Geneious Pro v9.1.8 software.
ACX 7 CGGGS peptides are in cyclic form. The peptides of Ph.D.-C7C phage library are presented on the phage pIII protein in cyclic form due to the formation of the disulfide bond between two cysteine residues. Thus, cyclization is essential to maintain the functionality of the peptides. To achieve the formation of the disulfide bonds, protein constructs were overexpressed in E. coli Shuffle, which are having enhanced capacity to correctly fold proteins with disulfide bonds in the oxidized cytoplasm. The presence of disulfide bond was verified with thiol-reactive maleimide labeling. In brief, free thiols in the protein constructs were blocked, the disulfide bond in the protein was cleaved with a reducing agent, and the revealed thiol groups were labeled with IRDye Maleimide. Non-reduced (oxidized) protein samples were used as a control. The reduced form of the protein was readily conjugated with maleimide IRDye, while the oxidized form of the same protein failed to get labeled (Fig. 2E). These results confirm the presence of the disulfide bonds in ACX 7 CGGGS peptides, which indicates that peptides are in cyclic form.
Screening of the binding ability of peptides to rDIII. This screening was necessary to show that 7-mer cyclic peptides (CP) and 12-mer linear peptides (LP) produced by us possess DIII binding potential, which is crucial for blocking the interaction between DIII and hBMECs. The screening was also important to show that Table 2. Bioinformatics analysis of identified peptides obtain from phage display. TUPScan tool (http://i. uestc. edu. cn/ sarot up3/ cgi-bin/ TUPSc an. pl): a tool of the SAROTUP web server for the prediction of TUPs based on the match with any known TUP motif. Peptide search tool (https:// www. unipr ot. org/ pepti desea rch/): a tool for the scan of peptides that match with sequences deposited in the UniProtKB database. CC -phage clone carrying 7-mer cyclic peptide; LC -phage clone carrying 12-mer linear peptide. www.nature.com/scientificreports/ the binding ability of peptides is retained in the absence of pIII phage protein. The screening was performed with ELISA, wherein the peptides were coated into microtiter wells through their C-terminus. For all 7-mer cyclic peptides, the absorbance > 1.9 acknowledge that these peptides represent strong DIII binders (Fig. 3). From two 12-mer linear peptides, only LP3 showed a strong binding to rDIII (A 450 = 2.564), whereas LP19 did not retain its binding ability to rDIII (A 450 = 0.082, Fig. 3). Therefore, LP19 was excluded from further assays.
Peptides are able to block the binding of rDIII to hBMECs (ELISA). The 7-mer cyclic and 12-mer linear peptides that showed strong binding ability to rDIII (Fig. 3) were assessed for their potential to inhibit the interaction between rDIII and proteins of hBMECs. Before the blocking of rDIII-hBMECs interaction, the minimum concentration at which rDIII is able to interact with hBMECs proteins was determined. Absorbance higher than 1.0 was set as a threshold to select the minimal concentration of rDIII. It was observed that a minimum of 2 µg (0.15 nM) of rDIII was sufficient to detect the interaction ( Supplementary Fig. S7). Thus, rDIII at this concentration (0.15 nM) was pre-incubated with a tenfold molar excess (1.  Fig. 4A). Absorbance in the positive control (without pre-blocking of rDIII) was A 450 = 1.064. CP7 showed 37% inhibition, however, the blocking of the interaction was not significantly different (P = 0.0101) from the positive control. Therefore, CP7 was not included in further assays. When rDIII was pre-incubated with a linear peptide LP3, 39% inhibition (A 450 = 0.647) in its interaction with hBMECs was observed (Fig. 4A).     Supplementary Fig. S9). Thus, in the blocking assay, 2.5 µg (0.19 nM) of rDIII were pre-incubated with a tenfold molar excess (1.9 nM) of each peptide prior to incubation with hBMECs strips. The inhibition of the interaction was confirmed in the case of CP2, CP4, CP12, and C16 (Fig. 4B). CP6, CP15 and LP3 did not show clear inhibition of the interaction (Fig. 4B). Therefore, these peptides were excluded from the further assays. Positive control (unblocked rDIII) included in the Western blot showed a signal at ~ 15 kDa, while in the negative control (rDIII excluded from assay), no signal was observed (Fig. 4B).
Peptides are able to block the binding of rDIII to hBMECs (immunocytochemistry). Based on ELISA and Western Blot results, CP2, CP4, CP12, and CP16 were selected for further assessment by immunocytochemistry. To block the adhesion of rDIII to the culture of hBMECs, the minimum concentration of rDIII needed to detect visible interaction with the endothelial cells was determined. Five micrograms (0.39 nM) of rDIII were sufficient to visualize adhesion of rDIII on endothelial cells culture in immunocytochemistry (data not shown). rDIII was pre-blocked with a tenfold molar excess (3.9 nM) of each peptide before incubation with the cells. All four peptides were able to inhibit the adhesion of rDIII on the endothelial cells, which ensures the blocking potential of selected anti-WNV peptides (Fig. 5A).

Peptides are able to block infection in cultured cells. The ability to block the infection was assessed
by the neutralization test (Fig. 5B, Supplementary datasets 1  Cytotoxic and hemolytic activity of peptides. Four DIII blocking peptides (CP2, CP4, CP12, and CP16) were tested by XTT for cytotoxic effect induced in eukaryotic cells. First, the cell viability was standardized in untreated (100% viability) and Triton treated cells (< 2.8% viability; Fig. 6A). The toxicity of each peptide was determined at different concentrations (1 μM, 2 μM, 4 μM, and 6 μM). None of the peptides was toxic for eukaryotic cells even at 6 μM concentration (Fig. 6A). Among all peptides, the maximum decrease in viability was observed in the case of CP16 (6 μM, 10.6% decrease), however, the reduction was not significant to designate CP16 as a toxic peptide (P = 0.1158). Further assessment of toxicity was performed by evaluating the hemolytic activity of peptides on sheep red blood cells (RBCs). Hemolysis > 10% was set as a threshold to designate the peptide with hemolytic activity. . None of the DIII WNV blocking peptides (CP2, CP4, CP12, and CP16) showed hemolytic activity as the maximum release of the hemoglobin was < 0.15% (Fig. 6B).

Discussion
The neuroinfections caused by viruses represent a severe threat worldwide. Till to date, antiviral therapeutics are available only against few viruses (reviewed in [60][61][62][63] ). In the case of WNV infection, no specific antiviral therapies are available and the treatment is restricted to supportive care. Different stages in the viral life cycle can be targeted to develop antiviral molecules, however, inhibition of the initial steps, the binding and entry into the host cells, is an attractive strategy 64,65 . In case of flaviviruses, E glycoprotein is the only protein that mediates the binding of viral particles to the host cells. It is composed of 3 ectodomains (DI -DIII) 26,32,66 , while the DIII is a receptor-binding functional determinant for the viral entry 31,36,42,43,[66][67][68] . This makes DIII a good candidate to stop the viral entry and a favorable target for the development of anti-flavivirus therapy. The binding of DIII to endothelial cells of the brain vasculature was also confirmed in our previous study 45 . hBMECs were selected as target cells in this study since they represent a primary barrier to viral dissemination into the CNS. It was demonstrated that hBMECs are sensitive to WNV and can facilitate entry of the cell-free virus into the CNS without compromising the BBB integrity, suggesting a key role of the endothelial layer in the virus spread via the hematogenous route 9,69 . Therefore, it is supposed that the adhesion of WNV to hBMECs is one of the crucial steps in flavivirus neuroinvasion.
We believe that the entry-inhibiting strategy could be an effective way to limit WNV neuroinvasion. The viral entry-inhibition strategy is adopted against various viruses (reviewed in [70][71][72] ). One possible advantage of this entry-inhibiting approach is that it becomes difficult for the virus to develop resistance to entry-inhibiting agents. The resistance to various antiviral agents, mainly to those based on the viral-enzyme inhibition strategy, has been reported earlier [21][22][23][24] . Compared with inhibitors targeting viral enzymes, entry inhibitors can directly bind to virions in the blood without the need for penetration into cells. Furthermore, since the entry-inhibiting compounds block the very first step of the viral life cycle wherein the host cell membrane permeability is not necessary, they have the potential to be less toxic. Several anti-WNV monoclonal antibodies were developed previously as entry-blocking agents [73][74][75][76] , however, their cross-reactivity was the major limiting factor. Therefore, more specific and efficacious therapeutic tools against WNV are highly required. The high specificity can be achieved by selecting the molecules specifically against RBSs. Thus, the RBS 299 GTTYGVCSK 307 located at the www.nature.com/scientificreports/ N-terminus of DIII 45 was targeted to develop plausible entry-blocking peptides in this study by panning two different combinatorial phage display libraries. Although the antiviral peptides targeting E protein of various flaviviruses were isolated previously 28,59,77,78 , anti-WNV peptides 28 were generated by using full-length E protein. E glycoprotein is a large molecule (53-60 kDa) 79 with several surface-exposed regions not taking part in the receptor-binding. Panning of combinatorial libraries against this target may lead to the selection of peptides that can strongly bind to the E protein, however, fail to occupy RBSs. Antiviral peptides can be obtained employing various approaches such as natural sources (e.g. from plants), computational methods (e.g. molecular docking), or biological source (e.g. display techniques). From all available display techniques, phage display represents a widely used methodology for the selection of peptides with selective affinity for a specific target. The success of a selection (a panning) in display technologies relies on the complexity of the library. The greater is the diversity of clones, the more likely the library contains sequences that will be a target-specific binder. Combinatorial phage display libraries used in this study consist of 10 9 -10 10 random variants of peptide sequences displayed on pIII phage protein in 5 copies represent valuable tools for the selection of peptides. Both combinatorial phage libraries (C7C-mer and 12-mer) are constructed using the same M13 bacteriophage backbone, however, the 12-mer library can be propagated more efficiently than the C7C-mer library 80 . On the other hand, studies have reported that high-affinity ligands are readily selected from the loop-constrained cyclic peptide libraries, like C7C-mer, than from linear libraries [81][82][83] . Thus, both combinatorial libraries were used in this study to obtain effective DIII blocking peptides.
In the present study, among 9 unique panned peptides (7 cyclic and 2 linear), 8 peptides showed specific binding to rDIII (Fig. 3). Among those 8 peptides only four cyclic peptides (CP2 -CTKTDVHFC, CP4 -CIHSSTRAC, CP12 -CTYENHRTC, CP16 -CLAQSHPLC) were able to block the interaction between rDIII and hBMECs judged by ELISA, Western blot and immunocytochemistry ( Fig. 4 and Fig. 5A). Those four peptides also inhibited the infection in cultured cells assessed by neutralization assay (Fig. 5B). None of the linear peptides showed potential to inhibit the interaction. The fact that other phage display-selected peptides (CP6 -CMQTQRAHC, CP7 -CTNANHYFC, CP15 -CDPRHSKFC, LP3 -SGVYKVAYDWQH) with good DIII binding activity (in all A 450 > 1.5) did not block the interaction between rDIII to hBMECs was not surprising (Fig. 4). It may be possible that those peptides did not occupy the RBSs sufficiently to be able to hinder rDIIIreceptor interaction.
From overall experimental approach in the present study, 4 cyclic peptides and none of the linear counterpart were able to block the interaction between DIII and hBMECs. Historically, structurally constrained peptidebased drugs are more successful as therapeutics than linear peptides [84][85][86] , and it is tempting to speculate that our finding is in line with this fact. Cyclic peptides have better binding affinity toward the target molecules because of the conformation 87 , which allow them to occupy the larger surface area of RBSs necessary to hinder the ligandreceptor interaction effectively 88,89 . Moreover, cyclic peptides are resistant to proteolysis, demonstrate effective crossing of the membranes, and are less toxic 90,91 . To rule out the possible toxicity of the 4 cyclic peptides generated in the present study, the XTT and hemolytic test were used. The therapeutic antiviral peptide should be specific for virus-related processes with no or fewer effects on cell metabolism and should not demonstrate acute or permanent toxicity against the host. It is generally accepted that dehydrogenase activity more than 70% compared to untreated control (ISO 10993-5: Biological evaluation of medical devices-Part 5: Tests for in vitro cytotoxicity) and hemolysis < 10% 92 define peptide as a safer candidate for translation of the biomolecules to therapy. In our study, all 4 cyclic peptides showed no signs of toxicity to hBMECs (Fig. 6A), and hemolytic activity (Fig. 6B) even at the concentration of 6 µM. This concentration is approximately 1500 times higher than the concentration required for blocking of rDIII-hBMECs interaction evaluated by immunocytochemistry, and approximately 2469 times higher than the concentration of CP2 required to neutralize VLP (EC 50 > 7290, 2.43 nM) in neutralization assay. In case of other three CPs, the concentrations required to exhibit EC 50 were ranged between 10-17 nM only. Overall results indicate that CPs can neutralize the infection in vitro at nanomolar concentration with no toxic and hemolytic effect even at several 100-fold higher concentration.
In conclusion, the blocking peptides identified in this study (CTKTDVHFC, CIHSSTRAC, CTYENHRTC, CLAQSHPLC) could present the basic building block for the development of new entry-inhibiting antivirals against WNV neuroinvasion since specific therapy is still not available. The experimental approach presented here may, potentially, serve as a template to design mid-size antiviral therapeutics against other emerging viruses.  Table 1. In general, synthetic peptide (RBS) or rDIII were coated in 96-well plates overnight at 4 °C (as per manufacturer's instructions) and non-specific binding sites were blocked with blocking buffer for 1 h at room temperature. After three washings with washing buffer, 2 × 10 11 phages were allowed to interact with immobilized target for 1 h at room temperature. The wells were then washed 15 times with the same washing buffer and phages were eluted as presented in Table 1. After each elution, number of phages present in eluate were calculated with the plate titration method and the rest of the eluate was subjected for amplification of phages in E. coli ER2738 (New England Biolabs) for 4.5 h at 37 °C. Amplified phages were precipitated from cleared E. coli culture with PEG/NaCl precipitation as per the manufacturer's instructions (New England Biolabs). After the last round of panning, phages in eluate were amplified, precipitated, and their binding ability to rDIII was judged by phage ELISA (details are described in Supplementary Method S3). Amplified phages were then plated on LB agar plates containing 1 mM IPTG (Fermentas, Slovakia) and 1 mM X-gal (Sigma, USA) and 30 well-separated phage plaques were randomly picked and propagated as per the manufacturer's instructions (New England Biolabs). Individual phage clones were subjected to DNA sequencing and phage ELISA.

Methods
Isolation of phage DNA, sequencing, and in silico peptide sequences analysis. Genomic DNA from each phage clone was isolated as per the manufacturer's instructions (New England Biolabs). Sequence encoding phage-displayed peptide was amplified with PCR (vector-specific primers are presented in Supplementary Table S1) and sequenced on ABI PRISM 3100-Advanced Genetic Analyzer (Thermo Fisher Scientific, Slovakia). The amino acid sequences of peptides (presented on phage clones) were deduced with the help of Geneious Pro v9.1.8 software (Biomatters, USA). The bioinformatics analysis of identified peptide sequences was performed to exclude any non-specific binder (TUPs) with the help of the SAROTUP server (http://i. uestc. edu. cn/ sarot up3/ index. html) 93 . The UniProtKB database (Peptide search tool, https:// www. unipr ot. org/ pepti desea rch/) was used to search possible sequence match in the public repository.
Assessment of binding ability of individual phage clone with rDIII. Amino acid sequences of the peptides (presented on pIII phage protein) deduced above were aligned and grouped as per sequence homology. A representative phage clone from each group was amplified in E. coli ER2738 and purified with PEG/NaCl precipitation as described above. Each phage clone was then analyzed for its binding ability to rDIII using phage ELISA (Supplementary Method S4). Production of 7-mer cyclic and 12-mer linear peptides. The 7-mer cyclic and 12-mer linear peptides from the selected phage clones were produced in the E. coli SHuffle expression system (New England Biolabs). All peptides were flanked with GGGS sequence at C-terminus (e.g. X 12 GGGS), while 7-mer cyclic peptides contain additional alanine residue at N-terminus (ACX 7 CGGGS) as suggested by the manufacturer of Ph.D. libraries (Supplementary Figure S4). Nucleotide sequences of amplified fragments used for the digestion and ligation into the pQE-30-UA-mCherry-STOP expression vector (Supplementary Figure S5) are summarized in Supplementary Table S2. The protein constructs were overexpressed in the E. coli SHuffle T5 strain. Detail steps of cloning, transformation of E. coli, selection of clones, protein overexpression, purification, and quality control are presented in Supplementary Method S5, S6, S7, and S8. The concentration of recombinant protein constructs was measured with NanoDrop One (Thermo Fisher Scientific). Aliquots of purified protein constructs were vacuum-dried and stored at room temperature until further use.
Removal of the tag and recovery of peptides. The N-terminal tag (6xHis tag -28 aa tag -GGGGS linker) was cleaved with His tagged bovine enterokinase as per the manufacturer´s instructions (GenScript Biotech, USA). Please note that tagged peptides were designed with the enterokinase cleavage site (highly specific DDDDK sequence). In short, 50 µg of tagged peptides were incubated with 5 µL of 10X cleavage/capture enterokinase buffer (200 mM Tris-HCl, 500 mM NaCl, 20 mM CaCl2; pH 7.4) and 2 µL of enterokinase (diluted 1:5 [v/v] in dilution/storage enterokinase buffer; 20 mM Tris-HCl, 200 mM NaCl, 2 mM CaCl2, 50% glycerol; pH 7.4) at 22 °C for 16 h. The mixture was incubated with His Mag Sepharose Ni beads (Thermo Fisher Scientific) for 2 h at room temperature with gentle agitation. During the incubation, the tag (6xHis tag -28 aa tag -GGGGS linker) and the enterokinase (please note that the enzyme is his tagged) were bound to the beads, while the unbound part containing ACX 7 CGGGS or X 12 GGGS peptide remained in solution. The solution containing peptides was collected in a new tube. The purity of the peptide was checked by MALDI-TOF MS (as described in Supplementary Method S8).
Assessment of the disulfide bonds in tagged peptides. The presence of disulfide bonds in tagged 7-mer cyclic peptides was judged with thiol-reactive IRDye 800CW Maleimide (LI-COR Biosciences, USA) labeling after the reduction of tagged peptides with tris(2-carboxyethyl)phosphine (TCEP, Sigma). Detail steps are presented in the Supplementary Method S9. After labeling reaction, tagged peptides were separated with LDS-PAGE with minor modifications (Supplementary Method S10) and signals were captured on the Odyssey CLx imaging system (LI-COR Biosciences).
Assessment of the binding of peptides to rDIII. The 7-mer cyclic and 12-mer linear peptides were covalently bound in CovaLink NH microwell plates and their binding to rDIII was corroborated by ELISA. Details are presented in Supplementary Method S11. www.nature.com/scientificreports/ Blocking of the interaction between rDIII and proteins of hBMECs. First, concentration-dependent ELISA and Western blot were performed to get the minimum concentration of rDIII required to show its interaction with hBMECs proteins. Details are presented in Supplementary Method S12 and Supplementary Method S13. Subsequently, the ability of 7-mer and 12-mer peptides to block the interaction between rDIII and proteins of hBMECs was tested. A tenfold molar excess of peptides was incubated with rDIII for 1 h at room temperature. Such pre-incubated rDIII was subsequently used in the assays. Details of both methods are described in Supplementary Method S12 and Supplementary Method S13.
Blocking of the adhesion of rDIII on cultured hBMECs. Cells were cultured as described in Supplementary Method S1. First, concentration-dependent immunocytochemistry was performed to reveal the minimum concentration of rDIII required to detect its adhesion to the endothelial cells. Details are presented in Supplementary Method S14. After that, the ability of peptides to block the adhesion of rDIII to the culture of hBMECs was judged. A tenfold molar excess of each peptide was incubated with rDIII for 1 h at room temperature. Pre-incubated rDIII was subsequently used in the assay. Details are described in the Supplementary Method S14.
Peptide cytotoxicity assay. The toxicity of peptides was determined by the XTT test (AppliChem, USA) as per the manufacturer's instructions. Details are presented in Supplementary Method S15.
Assessment of the hemolytic activity. Hemolysis assay was performed according to the method of Bender et al. 94 . Details are presented in Supplementary Method S16.

Assessment of the ability of peptide to block infection in cultured cells. The ability to block
infection in cultured cells was assessed using virus like particles (VLP) that possess C, M, and E proteins, and carry luciferase reporter gene. Details of the neutralization test are presented in Supplementary Method S17. In short, the peptides were serially diluted, which then mixed with VLP (400-500 TCID 50 /ml) and incubated for 90 min. Cultured cells were infected with the peptide-VLP mixture and incubated for 48 h. The amount of VLP entering the target cells was calculated by detecting the expression of luciferase, which then used to measure the neutralizing ability of the peptides, expressed in half maximal effective concentration (EC 50 ).

Data availability
Data generated in this study are available from the corresponding author on reasonable request.