Cationic Geminoid Peptide Amphiphiles Inhibit DENV2 Protease, Furin, and Viral Replication

Dengue is an important arboviral infectious disease for which there is currently no specific cure. We report gemini-like (geminoid) alkylated amphiphilic peptides containing lysines in combination with glycines or alanines (C15H31C(O)-Lys-(Gly or Ala)nLys-NHC16H33, shorthand notation C16-KXnK-C16 with X = A or G, and n = 0–2). The representatives with 1 or 2 Ala inhibit dengue protease and human furin, two serine proteases involved in dengue virus infection that have peptides with cationic amino acids as their preferred substrates, with IC50 values in the lower µM range. The geminoid C16-KAK-C16 combined inhibition of DENV2 protease (IC50 2.3 µM) with efficacy against replication of wildtype DENV2 in LLC-MK2 cells (EC50 4.1 µM) and an absence of toxicity. We conclude that the lysine-based geminoids have activity against dengue virus infection, which is based on their inhibition of the proteases involved in viral replication and are therefore promising leads to further developing antiviral therapeutics, not limited to dengue.


Introduction
Dengue is responsible for close to 400 million infections worldwide per year, of which 25,000 are fatal [1,2]. Currently, there is no specific therapy available for dengue, and vaccine development has been proven difficult, exemplified by its yielding only limited immunity analogues with saturated alkyl chains such as palmitoyl-Lys-(Ala or Gly) n -Lys-hexadecyl, denoted as n-C 15 H 31 C(O)-K(X) n K-(NH)-n-C 16 H 33 with X = A or G (compounds 1-3, shorthand representation C 16 -KX n K-C 16 ), were designed for complexation of polynucleotides and their transfer across biological membranes [27], with the ultimate goal of transfection, gene therapy [29], and RNA inhibition (RNAi) [30]. For such applications, lipids must be cationic to interact with and compensate for the negative charge of the phosphates in the nucleotides, and Lys is preferred as the cationic amino acid over Arg because the positive charge of the latter is permanent, whereas that of the former is pH-dependent, i.e., it is involved in protonation equilibria (procationic), a factor which promotes endosomal escape of the polynucleotide upon uptake in the cell by endocytosis [31].
classical geminis [28] due to the asymmetry of the peptide, which has an acyl and an alkyl (amine) moiety appended to the N-and C-termini, respective representatives of this novel class of compounds, such as oleoyl-Ser-Pro-Ly (ol-SPKR-ol) and analogues with saturated alkyl chains such as palmitoyl-L Gly)n-Lys-hexadecyl, denoted as n-C15H31C(O)-K(X)nK-(NH)-n-C16H33 with X (compounds 1-3, shorthand representation C16-KXnK-C16), were des complexation of polynucleotides and their transfer across biological membrane the ultimate goal of transfection, gene therapy [29], and RNA inhibition (RNA such applications, lipids must be cationic to interact with and compensate for t charge of the phosphates in the nucleotides, and Lys is preferred as the cati acid over Arg because the positive charge of the latter is permanent, whereas former is pH-dependent, i.e., it is involved in protonation equilibria (procation which promotes endosomal escape of the polynucleotide upon uptake in endocytosis [31]. Because of the preference of DENV protease and furin for substrates w amino acids, we investigated whether geminoids with Lys (compounds 1-3) co the activity of these proteases and whether any selectivity could be detected their similar substrate preferences. Here, we show that geminoids of the C16-K and C16-KGnK-C16 (3) series, in particular with A and n = 1 (2a) or 2 (2b), a inhibitors of DENV2 protease and the host protease furin and explore their sele another clinically relevant protease, trypsin. The inhibitors are also shown t against DENV2 infection in a cellular context at non-toxic concentrations.

Inhibition of DENV2 Protease and Furin by Geminoids Studied with MCA Subs
The IC50 values of the geminoid peptide amphiphiles 1-3 ( Figure 1, with n-C15H31, R2 = n-C16H33) for DENV2 protease and furin with MCA substrates a the left part of Table 1 (see Figures S1 and S2 in the Supporting Information fo representations). The geminoids with Ala (2) were better inhibitors than thos (3), and C16-KAK-C16 was found to be the better inhibitor for DENV2 protease to C16-KA2K-C16 (IC50 values of 0.66 resp. 0.80 µM), while for furin it was the around (IC50 values of 3.57 resp. 2.14 µM). Because of the preference of DENV protease and furin for substrates with cationic amino acids, we investigated whether geminoids with Lys (compounds 1-3) could inhibit the activity of these proteases and whether any selectivity could be detected in spite of their similar substrate preferences. Here, we show that geminoids of the C 16 -KA n K-C 16 (2) and C 16 -KG n K-C 16 (3) series, in particular with A and n = 1 (2a) or 2 (2b), are effective inhibitors of DENV2 protease and the host protease furin and explore their selectivity with another clinically relevant protease, trypsin. The inhibitors are also shown to be active against DENV2 infection in a cellular context at non-toxic concentrations.

Inhibition of DENV2 Protease and Furin by Geminoids Studied with MCA Substrates
The IC 50 values of the geminoid peptide amphiphiles 1-3 ( Figure 1, with y = 16, R 1 = n-C 15 H 31 , R 2 = n-C 16 H 33 ) for DENV2 protease and furin with MCA substrates are given in the left part of Table 1 (see Figures S1 and S2 in the Supporting Information for graphical representations). The geminoids with Ala (2) were better inhibitors than those with Gly (3), and C 16 -KAK-C 16 was found to be the better inhibitor for DENV2 protease compared to C 16 -KA 2 K-C 16 (IC 50 values of 0.66 resp. 0.80 µM), while for furin it was the other way around (IC 50 values of 3.57 resp. 2.14 µM).  Figure S4.
The IC 50 values determined for inhibition of trypsin for a number of selected geminoids were more than an order of magnitude higher than those for DENV2 protease (Table 1, Figure S3). The serine proteases that are highly susceptible to inhibition by cationic geminoids have a preference for substrates that contain cationic amino acids [23,24] and are active on proteins that are located in the membrane of the endoplasmic reticulum [6].

Effect of Lipid Aggregation on the Inhibition
The inhibition of furin by the C 16 -KG n K-C 16 (3) compounds with Ac-RVRR-MCA as the substrate in competitive inhibition experiments had a non-linear dependence on inhibitor concentration (see Figure S4 for the example of 3b). We observed the following three phases: (i) a decrease in activity by 30-40% in the inhibitor concentration range of 0-4 µM; (ii) a plateau in the region of 4-12 µM; (iii) a steep decrease to full inhibition above 12 µM. We determined the critical micelle or aggregate concentration (CMC) of a number of effective inhibitors by studying the fluorescence of pyrene as a probe (see Figure S5) [32,33]. Because a possible explanation for the multi-phase behaviour would be that the CMC of the geminoid corresponds to the transition between phases (ii) and (iii) and that the last phase represents a very efficient inhibition by inhibitor aggregates, which would imply that the 1st and 2nd phases represent the maximum degree of inhibition attainable with non-aggregated monomer. The CMC values found were, however, all in the order of 10-100 µM (Table 1), which is typical for geminis [28]. They decreased with the length of the spacer, in line with what is observed for gemini surfactants with alkyl spacers with more than 4-6 methylene groups [34], but did not appear to be correlated to the type of amino acid (Ala or Gly) in the spacer. The CMC values of the most effective inhibitors are well above the IC 50 values for both DENV2 protease and furin for these compounds (Table 1). Although the CMC values are determined in pure water and could be affected by solutes in the various assay buffers, we conclude that micelle formation or aggregation probably does not play a major role in the inhibition assays.

Inhibition of DENV2 Protease by Geminoids Studied with Tyr(3-NO 2 ) Substrate
To further explore the dependence of DENV2 protease inhibition on the choice of substrate, a selected group of geminoids was investigated with the aforementioned Tyr(3-NO 2 ) substrate [22,[35][36][37]. The IC 50 values determined in this assay (Table 1, Figure 2) showed that the geminoids were effective inhibitors in this assay as well, but contrary to the results with Z-RR-MCA (Table 1), the superiority of the geminoids with Ala residues in the spacer (2) over those with Gly (3), in particular, 3a (C 16 -KGK-C 16 ), was less pronounced with this substrate; moreover, the order appeared to be reversed, as 2b (C 16 -KA 2 K-C 16 ) was a better inhibitor than 2a (C 16 -KAK-C 16 ).

Effect of Geminoids on DENV2 Replicon Activity in HeLa Cells
To investigate the effect of geminoids on viral replication, we used HeLa cells containing a DENV2 replicon. Instead of the structural proteins of DENV2, the replicon encodes a luciferase reporter that can be used as a readout for DENV2 protease dependent virus replication [38]. Cell viability was assessed on the same cells using a colourimetric assay based on 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2H-tetrazolium (MTS) reduction, and by light microscopy. The most effective compounds from Table 1 were tested at concentrations ranging from 0.3 to 10 µM, well below their CMC values. 2a (C16-KAK-C16) proved most effective in this system, with a 39% reduction of luciferase activity at 3 µM and 62% inhibition at 10 µM ( Figure 3, top panel). We cannot exclude that this reduction at the highest concentration tested is partly due to cytotoxicity (Figure 3, bottom panel). Other compounds, 2b (C16-KA2K-C16) and 3a (C16-KGK-C16) (MTS assay >80%, relative to DMSO control) with a similar toxicity profile, were less effective (47% and 38% inhibition at 10 µM, respectively). 3b (C16-KG2K-C16) induced considerable cytotoxicity, and we thus could not establish DENV2 inhibition by this compound in this assay. With the exception of 3b (C16-KG2K-C16), these geminoids have a concentration window for which luciferase activity is reduced with cell viability values of >80%. Inhibition of DENV replicon follows the trends in IC50 found in the studies on the inhibition of the DENV2 protease construct with the MCA substrate ( Table 1). The viability of unmodified HeLa cells after treatment with these compounds was also tested in a separate experiment using the Celltiter Blue Viability Assay (Promega, see Supporting Information) in the concentration range of 0.8-50 µM. No CC50 could be calculated at this concentration range ( Figure S6), and only 3b (C16-KG2K-C16) showed slight toxicities at the highest concentration. The geminoids appear to be less toxic to unmodified HeLa cells than to the replicon-containing cells.  Table 1 for averaged IC 50 values from curve fits.

Effect of Geminoids on DENV2 Replicon Activity in HeLa Cells
To investigate the effect of geminoids on viral replication, we used HeLa cells containing a DENV2 replicon. Instead of the structural proteins of DENV2, the replicon encodes a luciferase reporter that can be used as a readout for DENV2 protease dependent virus replication [38]. Cell viability was assessed on the same cells using a colourimetric assay based on 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2H-tetrazolium (MTS) reduction, and by light microscopy. The most effective compounds from Table 1 were tested at concentrations ranging from 0.3 to 10 µM, well below their CMC values. 2a (C 16 -KAK-C 16 ) proved most effective in this system, with a 39% reduction of luciferase activity at 3 µM and 62% inhibition at 10 µM ( Figure 3, top panel). We cannot exclude that this reduction at the highest concentration tested is partly due to cytotoxicity ( Figure 3, bottom panel). Other compounds, 2b (C 16 -KA 2 K-C 16 ) and 3a (C 16 -KGK-C 16 ) (MTS assay >80%, relative to DMSO control) with a similar toxicity profile, were less effective (47% and 38% inhibition at 10 µM, respectively). 3b (C 16 -KG 2 K-C 16 ) induced considerable cytotoxicity, and we thus could not establish DENV2 inhibition by this compound in this assay. With the exception of 3b (C 16 -KG 2 K-C 16 ), these geminoids have a concentration window for which luciferase activity is reduced with cell viability values of >80%. Inhibition of DENV replicon follows the trends in IC 50 found in the studies on the inhibition of the DENV2 protease construct with the MCA substrate ( Table 1). The viability of unmodified HeLa cells after treatment with these compounds was also tested in a separate experiment using the Celltiter Blue Viability Assay (Promega, see Supporting Information) in the concentration range of 0.8-50 µM. No CC 50 could be calculated at this concentration range ( Figure S6), and only 3b (C 16 -KG 2 K-C 16 ) showed slight toxicities at the highest concentration. The geminoids appear to be less toxic to unmodified HeLa cells than to the replicon-containing cells.

Inhibition of DENV2 Replication in LLC-MK2
To assess the antiviral activity, we studied the inhibitors with wildtype DENV2 (DENV2 NGC) in LLC-MK2 (rhesus monkey epithelial kidney) cells with an immunochemical assay, which reports the percentage of infected cells. In this assay (Table 1, Figure 4), Ala-containing geminoids 2a and 2b (C 16 -KA n K-C 16 with n = 1 and 2) were much more effective than the Gly-containing 3a (C 16 -KGK-C 16 ), while 3b (C 16 -KG 2 K-C 16 ) was not active. Toxicity was monitored microscopically. Slight toxicity was observed for 2b (C 16 -KA 2 K-C 16 ), 3a (C 16 -KGK-C 16 , and 3b (C 16 -KG 2 K-C 16 ), but none for 2a (C 16 -KAK-C 16 ). 2a (C 16 -KAK-C 16 ) is, therefore, the most promising compound, even though its IC 50 and EC 50 for respectively DENV2 protease inhibition with the Tyr(3-NO 2 ) substrate and DENV2 replication are slightly less favourable than those of 2b (C 16 -KA 2 K-C 16 ).  compound 3b). The assay, which reports the percentage of infected cells by an immunochemical approach, was performed in duplicate (series-1 and series-2) at 8 concentrations (three-fold dilutions), see Table 1 for averaged EC50 values with standard deviation from curve fits; n. a., not active.

Discussion
We have found that the geminoids (gemini-like peptide amphiphiles) with two lysines separated by one or two Ala residues (2a-b) are strong inhibitors, with some IC50 values below micromolar, of serine proteases involved in the maturation of DENV capsids, DENV2 protease and furin, with resp. Z-RR-MCA and Ac-RVRR-MCA as the substrates (Table 1). Further studies showed that geminoids of this type inhibit DENV2 replication and viral infection in cultured cells. It is of interest to consider the effect of the amino acids between the linkers; the substitution of hydrogen for a methyl group going from Gly to Ala makes the head group more hydrophobic but also introduces more conformational rigidity, which is reflected in the preferences of the amino acids to be found in certain secondary protein structure elements, where Gly is mostly found in βturns and Ala in α-helices. The higher polarity of the headgroup in the Gly-containing geminoids might result in less effective cell penetration, which might explain the relatively poor performance of 3a in the cell infection assay, whereas the rigidity of the Ala-containing geminoids 2 probably favours efficient recognition by the substratebinding site of DENV2 protease. It should be noted, however, that in the in vitro assays of DENV2 protease, the FRET substrates gave similar results for the geminoid inhibitors but different relative efficacies (Table 1). With Z-RR-MCA, 2a (C16-KAK-C16) was a better inhibitor than 2b (C16-KA2K-C16), and the geminoids with Gly 3a and 3b were relatively poor, whereas with the Tyr(3-NO2) substrate, 2b (C16-KA2K-C16) was the best inhibitor, and 3a (C16-KGK-C16) was more effective than 2a (C16-KAK-C16). This difference may be related to the interaction of the DENV2 protease NS2B and NS3 domains around the active site. NS3 alone is active on relatively small substrates such as Z-RR-MCA, whereas association with NS2B is required for the recognition of larger peptide substrates [8]. The DENV2 protease used in this study is a construct [6,23] in which the NS2B and NS3   compound 3b). The assay, which reports the percentage of infected cells by an immunochemical approach, was performed in duplicate (series-1 and series-2) at 8 concentrations (three-fold dilutions), see Table 1 for averaged EC 50 values with standard deviation from curve fits; n. a., not active.

Discussion
We have found that the geminoids (gemini-like peptide amphiphiles) with two lysines separated by one or two Ala residues (2a-b) are strong inhibitors, with some IC 50 values below micromolar, of serine proteases involved in the maturation of DENV capsids, DENV2 protease and furin, with resp. Z-RR-MCA and Ac-RVRR-MCA as the substrates (Table 1). Further studies showed that geminoids of this type inhibit DENV2 replication and viral infection in cultured cells. It is of interest to consider the effect of the amino acids between the linkers; the substitution of hydrogen for a methyl group going from Gly to Ala makes the head group more hydrophobic but also introduces more conformational rigidity, which is reflected in the preferences of the amino acids to be found in certain secondary protein structure elements, where Gly is mostly found in β-turns and Ala in α-helices. The higher polarity of the headgroup in the Gly-containing geminoids might result in less effective cell penetration, which might explain the relatively poor performance of 3a in the cell infection assay, whereas the rigidity of the Ala-containing geminoids 2 probably favours efficient recognition by the substrate-binding site of DENV2 protease. It should be noted, however, that in the in vitro assays of DENV2 protease, the FRET substrates gave similar results for the geminoid inhibitors but different relative efficacies (Table 1). With Z-RR-MCA, 2a (C 16 -KAK-C 16 ) was a better inhibitor than 2b (C 16 -KA 2 K-C 16 ), and the geminoids with Gly 3a and 3b were relatively poor, whereas with the Tyr(3-NO 2 ) substrate, 2b (C 16 -KA 2 K-C 16 ) was the best inhibitor, and 3a (C 16 -KGK-C 16 ) was more effective than 2a (C 16 -KAK-C 16 ). This difference may be related to the interaction of the DENV2 protease NS2B and NS3 domains around the active site. NS3 alone is active on relatively small substrates such as Z-RR-MCA, whereas association with NS2B is required for the recognition of larger peptide substrates [8]. The DENV2 protease used in this study is a construct [6,23] in which the NS2B and NS3 fragments are connected by a flexible [39] GGGGSGGGG linker. The DENV2 protease assay with the MCA substrate was carried out in the presence of glycerol, required to stabilise the enzyme in an aqueous solution [40], whereas the buffer for the assay with the Tyr(3-NO 2 ) substrate contained ethylene glycol and the non-ionic detergent Brij ® 58 (polyoxyethylene (20) cetyl ether). Importantly, DENV2 protease inhibition by geminoids persists under the latter conditions, which have been designed to suppress the inhibition by non-selective inhibitors [36]. The hydrophobic character of the geminoids promotes the formation of nanoparticles in an aqueous environment, but their CMCs are in the high micromolar range, i.e., considerably higher than the IC 50 for protease inhibition in vitro. In the attempts to determine the IC 50 for the inhibition of furin by the Gly geminoids 3, we observed that the dependence of residual protease activity on inhibitor concentration in vitro showed a disproportional decrease above a concentration of approx. 12 µM ( Figure S4). This is quite close to the CMC, at 30 µM for 3b (C 16 -KG 2 K-C 16 ), which is the lowest value found for the selected compounds ( Table 1). The enzyme assays with their variety of solutes, buffer salts, glycerol, ethylene glycol, or detergent, are not designed for micelle forming inhibitors. In cells, however, the geminoids are more likely associated with the lipid phase of the membranes than with micelles. No evidence of non-linearity was observed in the concentration range for the experiments in cellular models (Figures 3 and 4), and inhibition of both furin and DENV2 protease was evident at concentrations considerably below the CMC (Table 1. The CMC for 3b (C 16 -KG 2 K-C 16 , 30 µM) is too high to correspond to either of the transitions in Figure S4 at 1 and 12 µM. Thus, despite the apparent hydrophobic attraction between the molecules of the amphiphilic inhibitors in water, the transitions in Figure S4 cannot be explained by their aggregation alone. In addition to the recognition of the peptide sequence in the enzyme's active site, hydrophobic interactions between amphiphilic inhibitor and enzyme probably play a role. The interaction of NS3 and NS2B, which is required for full catalytic activity, has recently been identified as a target for allosteric inhibition of DENV2 and Zika proteases [41][42][43]. Although the interactions between the fragments in the open, inactive, ligand-free (DENV2) [39] and closed, active, ligand-bound (DENV3) [44] enzyme conformations are mainly electrostatic, the common structural feature of the first inhibitors that are recognised as allosteric [41][42][43] is that they contain multiple apolar aromatic groups. It is therefore very likely that the apolar alkyl tails of the geminoid inhibitors play a similar role. The relatively good performance of compound 3a in the DENV2 protease assay with the larger (i.e., the Tyr(3-NO 2 )) substrate may be explained by the aforementioned expected higher flexibility of this Gly-containing geminoid, allowing it to interact with both the substrate and allosteric sites.
Because of their two alkyl tails, geminoids are likely to interact with biological membranes, allowing efficient access of relatively large polar peptide substrates to the endoplasmic reticulum, where they can be presented to a viral or host enzyme at the membrane surface. We suggest that this is a likely explanation for the efficiency of this novel class of amphiphilic inhibitors of viral maturation. In an earlier study on peptide inhibitors, the positive effect of N-acylation on the inhibition of furin in cells was ascribed to the improved access to the intact cell and linked to the affinity of furin for membranes [45]. The amphiphilic nature of this class of inhibitors could have multiple advantages for their application as drugs and for their possible translocation into the cell. It is likely that single molecules or nanoparticle aggregates of the amphiphilic cationic peptides can be taken up by the cell by endocytosis, analogous to what has been proposed for lipoplexes with cationic gemini surfactants and geminoids in transfection [30]. The application of additional functional elements such as selected oligosaccharides and peptides would allow receptormediated targeting and cellular trafficking [46,47]. The formation of mixed nanoparticles such as those with PEGylated lipids would allow stabilized and targeted delivery from the blood [48].
In the cellular context of the DENV2 replicon assay (Figure 3) in HeLa cells, most of the geminoid compounds that showed activity in the in vitro protease inhibition assays inhibited viral replication. 2a (C 16 -KAK-C 16 ) proved most effective at low toxicity. Furthermore, the compounds significantly reduced wild-type dengue virus's replication in LLC-MMK2 cells ( Figure 4); the geminoids with Ala 2 were much more effective in these experiments than those with Gly 3 (Table 1). Both the DENV2 protease and host proteases, including furin, are involved in the maturation of the viral polyprotein [6] and are inhibited by C 16 -KA n K-C 16 geminoids 2a and 2b. Thus, we cannot exclude that both serine proteases are targeted in the inhibition of the virus replication. Importantly, however, this is achieved without the adverse effects expected upon complete furin inhibition. This is consistent with the recent finding that furin inhibitors inhibit the replication of the hepatitis B virus [49] and highly pathogenic avian influenza virus [50] without apparent toxicity.
The discovery that geminoid molecules, originally designed for polynucleotide delivery, are active protease inhibitors that suppress viral replication in a variety of cells is a starting point for the design of the next generation of geminoids with peptide sequences optimised for the interaction with the active sites of the target proteases, and, if considered necessary, for selectivity of inhibition of various viral proteases over host proteases such as furin. For this approach, advantage can be taken of the available X-ray crystallographic structures of the DENV2 protease construct [39,51] and furin [52,53].

Synthesis
The preparation of alkylated peptides of the geminoid type ( Figure 1) with C16-tails has been described elsewhere [26,27]. Details of the preparation and characterization of the new series of geminoids 2 (C 16 -KA n K-C 16 , 1 < n < 4) are given below (with NMR data, including 1 H and 13 C NMR spectra for 2a and 2b, Figures S7-S10, in the Supporting Information); the preparations of 1 and 3 (C 16 -KG n K-C 16 , 0 < n < 4, first mentioned in [26]) are given along with their characterization (including 1 H and 13 C NMR spectra for 3a and 3b, Figures S11-S14) in the Supporting Information. After washing with diethyl ether and drying the products were cleaved from the resins with 5% H 2 O in TFA for 2-3 h. The products with n = 1 and 2 were dissolved in methanol and purified using preparative reverse-phase HPLC; for n = 3 and 4 this was not possible due to solubility problems. The mobile phase started as water (0.01% TFA) and went in 15 min to 100% acetonitrile (0.01% TFA), which was retained for 5 min. The fractions with product were collected and dried in vacuo.

Critical Micelle Concentration (CMC)
Pyrene was used as a probe to study the changes in its fluorescence, in particular the ratio (I 1 /I 3 ) of the intensities I 1 and I 3 at between 373 and 383 nm, respectively [32,33]. An abrupt change in this ratio with increasing surfactant concentration points to an increase in hydrophobicity of the environment of the probe corresponding to the formation of aggregates. See Supporting Information for details.

Enzyme Expression, Purification, and Assay
DENV2 protease and human furin were expressed and purified as previously described in refs. [23,54]. See Supporting Information for details.

Furin Assay with MCA Substrate
Furin was dissolved at 0.76 nM concentration in 1 mL MES buffer (10 mM), 1 mM CaCl 2 , pH 7.0 at 36.5 • C. The substrate Ac-RVRR-MCA [25] was added at a concentration of 2.35 µM (10 times the K m ), and the inhibitor was added in increasing concentrations (as increasing volumes of 1.0, 5.0, 10.0, 20.0, and 40.0 µL) from a stock solution of 2 mg in 1 mL DMSO. The residual activity was measured as fluorescence at 460 nm following excitation at 380 nm in a Hitachi F2500 spectrofluorimeter, and plots were fitted using the Grafit ® software (Erithracus Software, Horley, Surrey, UK).

DENV2 Protease Assay
The inhibition reported here was studied on an NS2B-NS3 construct derived from dengue serotype 2 (CF40-GGGGSGGGG-NS3) called DENV2 protease in this study. - With MCA substrate: The assay was carried out and analysed as described above for furin, but with DENV2 protease at 20 nM concentration in 50 mM Tris.HCl, pH 9.0, 20% glycerol, 37 • C, and with 20 µM Z-RR-MCA as the substrate; -With Tyr(3-NO 2 ) substrate: The applied assay protocol was described by [38]. IC 50 values were determined in CDD Vault [55] using the Levenberg-Marquardt algorithm for fitting a Hill equation to dose-response data [56,57].

Replicon Assay and Viability Test
The replicon assay was carried out as described earlier [38] using HeLa cells that contain a stably replicating DENV2 replicon expressing a luciferase reporter gene. The amphiphilic inhibitors were added as concentrated solutions in DMSO; the same amount of DMSO was used as the blank experiment, with the viral inhibitor ribavirin as a positive control. Luciferase activity and cell viability were assessed as described previously [38]. 4.6. DENV2 IPOX Cytoprotection Assay in LLC-MMK2 Cells 4.6.1. Cell Preparation LLC-MK2 (Monkey Rhesus Kidney cells; CCL-7.1) were passaged in assay medium (EMEM (Lonza Cat No: BESP069F) supplemented with 10% heat-inactivated FCS (Lonza), 2% Pen/strep (Gibco), 2% L-Glutamine (Gibco), 2% Hepes (Lonza), and 1% sodium bicarbonate (Lonza)) prior to use in the antiviral assay. Cells were seeded in 96-well plates (10 5 cells/well) in assay medium to be exposed 16-24 h later to compounds and viruses. The plates were incubated at 37 • C/5% CO 2 overnight to allow for cell adherence.

Compound Preparation
Compounds were solubilized in DMSO and evaluated using two-fold serial dilutions (8-points dose-response curves starting at a concentration of 50 µM) in duplicate for the antiviral assays. Compounds were diluted in assay medium at 1× test concentrations. Ribavirin (Sigma Aldrich, Amsterdam, The Netherlands) was evaluated as a positive control compound in the antiviral assays.

Virus Preparation and Cellular Infection
DENV2 New Guinea strain was grown in AP-61 insect cells (in-house cell bank) in complete Leibovitz medium containing 1% pen/strep (Gibco), 1% L-glutamin (Gibco), 0.5% Hepes (Lonza), 0.5% sodium bicarbonate (Lonza), and 10% tryptose phosphate for the production of stock virus pools. On the day of cellular infection, an aliquot of virus was removed from the freezer (−80 • C) and allowed to thaw in water in a biological safety cabinet. Virus was diluted into assay medium (10 4 TCID 50 ), and 100 µL of this was added to each well, resulting in a TCID 50 of 100. Cells were incubated for 2 h at 37 • C/5% CO 2 and washed 3 times with blank assay medium. Directly after washing, 100 µL of the compound dilutions were added to each well.

Plate Format
Each plate contained cell control wells (cells only), virus control wells (cells plus virus), duplicate drug toxicity wells per compound (cells plus drug only), as well as duplicate experimental wells (drug plus cells plus virus).

Immunoperoxidase Staining and Toxicity Determination
Virus-infected cells were visualised using a DENV2 immunoperoxidase staining protocol. Two days after infection, cells were inactivated with ethanol 70% for 30 min and washed with PBS. Fixed plates were incubated with PBS containing 0.05% H 2 O 2 for 20 min at 37 • C and washed again 3 times with PBS. Plates were incubated for 1 h with 50 µL monoclonal anti-DENV-2 NS1 antibody (Millipore; diluted 1:500 in EMEM). Samples were washed once with PBS containing 0.05% Tween20 and twice with PBS only. Secondary polyclonal goat anti-mouse IgG HRP (Dako; diluted 1:2000) was added 50 µL per well and incubated for 1 h at 37 • C in the dark. Following 3 washing steps with PBS, 100 µL AEC (3-amino-9-ethylcarbazole) substrate buffer (containing 0.03% H 2 O 2 , 3% DMF) was added to each well and incubated for 30 min at room temperature in the dark. Bidest water was added after removal of the substrate solution, and all virus-positive cells per well (marked by brown/red staining) were counted under a microscope. Visual scoring of toxicity per well was performed in parallel.