key: cord-103015-3dxwbmd2
authors: Shengjuler, Djoshkun; Chan, Yan Mei; Sun, Simou; Moustafa, Ibrahim M.; Li, Zhen-Lu; Gohara, David W.; Buck, Matthias; Cremer, Paul S.; Boehr, David D.; Cameron, Craig E.
title: The RNA-binding site of poliovirus 3C protein doubles as a phosphoinositide-binding domain
date: 2017-08-04
journal: bioRxiv
DOI: 10.1101/172742
sha: 
doc_id: 103015
cord_uid: 3dxwbmd2

Some viruses use phosphatidylinositol phosphate (PIP) to mark membranes used for genome replication or virion assembly. PIP-binding motifs of cellular proteins do not exist in viral proteins. Molecular-docking simulations revealed a putative site of PIP binding to poliovirus (PV) 3C protein that was validated using NMR spectroscopy. The PIP-binding site was located on a highly dynamic α-helix that also functions in RNA binding. Broad PIP-binding activity was observed in solution using a fluorescence polarization assay or in the context of a lipid bilayer using an on-chip, fluorescence assay. All-atom molecular dynamics simulations of the 3C protein-membrane interface revealed PIP clustering and perhaps PIP-dependent conformations. PIP clustering was mediated by interaction with residues that interact with the RNA phosphodiester backbone. We conclude that 3C binding to membranes will be determined by PIP abundance. We suggest that the duality of function observed for 3C may extend to RNA-binding proteins of other viruses. Highlights A viral PIP-binding site identified, validated and characterized PIP-binding site overlaps the known RNA-binding site PIP-binding site clusters PIPs and perhaps regulates conformation and function Duality of PIP- and RNA-binding sites may extend to other viruses In Brief The absence of conventional PIP-binding domains in viral proteins suggests unique structural solutions to this problem. Shengjuler et al. show that a viral RNA-binding site can be repurposed for PIP binding. PIP clustering can be achieved. The nature of the PIP may regulate protein conformation.

Interactions between viral proteins would then recruit the full complement of the genome-23 replication machinery. In 2010, Altan-Bonnet and colleagues discovered that a picornavirus 24 induced the synthesis of the phosphoinositide, phosphatidylinositol 4-phosphate (PI4P), and 25 showed that PI4P tracked with sites of genome replication (Hsu et al., 2010) . This observation 26 inspired the hypothesis that a cellular lipid rather than a cellular protein might be the lure used 27 to recruit viral proteins (Hsu et al., 2010) . 28 Ninety-four percent of the docking solutions in the major cluster exhibited a similar head 85 group orientation. Red spheres indicate the position of the phosphate at position four of the 86 inositol ring ( Figure 1A ). In contrast, the orientation of the fatty acyl chains varied for each 87 docking solution. Docking results indicated that K12, R13, and R84 of the major cluster interact 88 PI4P ( Figure 1A) . Similarly, K108 and R143 of the minor cluster are positioned to interact with 89 PI4P ( Figure 1A) . These residues are highly conserved in the 3C proteins of all prototype viruses 90 in the Enterovirus genus of the Picornaviridae family ( Figure 1B) . Even though the sequence 91 similarity between members of the genus ranges from 40-60%, the basic nature of K12, R13, R84, 92 and R143 were conserved, whereas that of K108 was not ( Figures 1B) . 93 Validation of PIP-binding sites by NMR 94 In order to test the validity of our docking observations, we titrated 15 N-labeled 3C protein 95 with soluble dibutyl-PI4P to observe potential NMR chemical shift perturbations (CSPs), which 96 would indicate chemical environment changes in the presence of PI4P (Figure 2) . Out of the three 97 basic residues of the major cluster that were predicted to interact with the PI4P, R13 showed the 98 largest CSP (Figure 2A ). There were also smaller CSPs associated with K12 and R84 (Figure 2A) . 99 However, we did not observe any CSPs for the minor cluster residues K108 and R143. 100 As seen in Figure 2B , the addition of PI4P resulted in a number of CSPs beyond that of 101 K12, R13, and R84. CSPs at least one standard deviation above the mean were considered to be 102 significantly different (CSP >0.019 ppm). Many of these residues were not in the immediate 103 vicinity of a docked PI4P molecule. For example, A9, M10, and N14 are located at the N-terminus 104 of 3C at a distance from PI4P but all showed significant CSPs ( Figure 2B ). These and other 105 perturbations could be caused by other structural changes conferred by PI4P binding. 106 Broad PIP-binding activity of 3C observed in solution 107 In order to assess PIP binding to 3C empirically and evaluate specificity of any observed 108 binding, we used a fluorescence polarization (FP)-based PIP-binding assay (Ceccarelli et al., 2007;  contrast, the bound form of the fluorescent ligand has a larger molecular volume, which reduces 113 the rotation such that the emitted light remains in the same plane as that used for excitation for 114 a longer period of time ( Figure 3A ). This approach was validated by using the PLC-δ1 PH domain 115 as a positive control, which showed PI(4,5)P2 binding specificity, as expected ( Figure S1 ) (Harlan 116 et al., 1994) . 117 We measured binding of phosphatidylinositol (PI) and mono-, bis-, or tris-phosphorylated 118 PIPs to 3C ( Figure 3B ). Binding of PI and all mono-phosphorylated PIPs was very weak. However, 119 binding of bis-and tris-phosphorylated PIPs was substantially stronger and without any apparent 120 order or preference ( Figure 3B ). Given the absence of an observed binding preference of 3C, we 121 determined the apparent dissociation constant (Kd,app) for PI(4,5)P2 binding to 3C. The value was 122 5 ± 1 µM ( Figure 3C ). 123 To assess the contribution of 3C residues implicated in PIP-binding to the actual binding 124 event monitored computationally (Figure 1A) , we changed K12, R13, or R84 to Leu. We used Leu 125 instead of Ala to limit any collateral damage that might be caused by creating a void in the protein 126 that could be filled by water. One expectation was that loss of any of these residues would 127 diminish PIP binding. However, we observed that the K12L derivative behaved as WT (Kd,app = 5 ± 128 1 µM) ( Figure 3D ). The R13L derivative reduced binding by ~4-fold relative to WT (Kd,app = 23 ± 14 129 µM) ( Figure 3E ). The R84L derivative actually increased PIP-binding affinity by ~10-fold (Kd,app = NMR experiment, dibutyl-PI4P was used instead. We monitored residues that were implicated in 142 PIP binding: K12, R13, and R84. It should be noted that the CSPs induced by addition of PI4P 143 (Figure 2A ) or RNA ( Figure 3I ) exhibited different trajectories. The addition of PI4P to the 3C-RNA 144 complex ( Figure 3I ) led to resonance positions similar to that of the 3C-PI4P complex (Figure 2A) , 145 suggesting that PI4P competed out RNA and that PI4P and RNA binding were mutually exclusive. 146 Broad PIP-binding activity of 3C observed on membranes 147 It is possible that the context in which a PIP is presented contributes to binding affinity 148 and/or specificity. To address this question, we employed a pH-modulation assay to monitor 149 protein binding to a planar supported lipid bilayer (SLB) in a microfluidic device (Jung et al., 2009; 150 Shengjuler et al., 2017) . The SLB is a stable model membrane system, and the composition of the 151 membrane can be easily tailored to probe interactions of specific lipids with proteins. The SLB 152 used here employed three lipids, the structures of which are illustrated in Figure 4A . A pH-153 sensitive dye, ortho-Sulforhodamine B-POPE (oSRB-POPE), is embedded into the SLB and used to 154 probe protein binding by detecting changes in the local electric field (Figures 4B). Upon binding 155 of 3C, which is positively charged at the experimental pH, the interfacial potential increases and 156 more hydroxide ions are recruited locally. As a result, the fluorescence signal decreases ( Figures  4B and S2) . 158 In these studies, we used 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine (POPC) 159 as the primary component of the SLB. We performed experiments in the presence of Mg 2+ 160 divalent cations to better mimic the cellular environment. 3C did not bind to pure POPC 161 membranes ( Figure 4C ). The addition of PI4P to the POPC membrane revealed the capacity of 3C 162 to bind to this mono-phosphorylated PIP ( Figure 4C ). The value for the Kd,app was 2.4 ± 0.4 µM 163 (Table 1 ). In addition to PI4P, 3C bound to POPC membranes containing PI3P and PI(4,5)P2, all 164 with essentially the same affinity (Table 1) . PIP binding to 3C in the context of a membrane also 165 required the phosphates because phosphatidylinositol (PI) binding was weak ( Table 1) . Worth 166 noting, PI binding did occur when compared to POPC alone; however, we were unable to produce 167 a complete binding isotherm because of the extremely high concentration of protein required. 168 The presence of PI will produce a net negative charge on the membrane, so the interaction with 169 the basic surface of 3C makes sense. To test the possibility that negative charge alone can 170 promote binding of 3C to the membrane, we performed an experiment in a membrane 171 containing 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine (POPS). The presence of POPS 172 will also produce a net negative charge on the membrane. The affinity of 3C for the POPS-173 containing membrane was on par with that observed for the PI-containing membrane ( Table 1) . (Figures 5A and 5B) . Although the same residues of 3C were used to interact with 183 both PIPs, the orientation of 3C differed slightly when bound to PI4P compared to that bound to 184 PI(4,5)P2 (Figures 5A and 5B) . For reference, residues of the protease active site have been 185 indicated by black spheres. The orientation of the active site in Figure 5A is clearly different than 186 observed in Figure 5B . This analysis expanded the number of residues interacting with PIP 187 headgroups to include residues in regions observed by NMR, for example D32, K156, and R176 188 ( Figures 5A and 5B) . Interestingly, the contribution of D32 to 3C-PIP interface was mediated by 189 two Na 1+ ions in the case of PI4P ( Figure 5A ) and one Na 1+ ion in the case of PI(4,5)P2 ( Figure 5B ). 190 A surface representation of 3C interacting with PIPs in the context of the membrane showed that 191 3C is propelled above the membrane by the headgroups (Figures 5C and 5D) . The N-terminus of 192 3C was an exception. Rotation of 3C to bring the N-terminus into view showed that the N-193 terminus has a more substantive interaction with the outer leaflet of the bilayer, albeit of a non-194 specific nature (Figures 5E and 5F ). The conformation and interaction of the N-terminus with 195 membrane was also influenced by the PIP present (Figures 5E and 5F ). 197 For the N-terminal residues of 3C to exhibit such conformational flexibility and diverse 198 interactions with membranes, the solution behavior would have to be more dynamic than 199 predicted by the crystal structure (Mosimann et al., 1997) . In order to evaluate the solution 200 beavior of 3C, we performed small-angle X-ray scattering (SAXS) experiments. Figure 6A shows 201 the scaled scattering profiles of 3C at concentrations ranging from 3.2 mg/mL to 13.0 mg/mL. The 202 overlap of the scaled curves at the low ends of scattering angles was consistent with the protein 203 remaining monomeric over this concentration range in agreement with the NMR and DLS studies 204 (Chan et al., 2016) . The scattering data at the low-angle end showed a linear correlation, 205 satisfying the Guinier approximation (qRg < 1.3), from which we determined a radius of gyration 206 (Rg) of 17.26 Å ( Figure 6A ). The Rg derived from the Guinier approximation was in good 207 agreement with that estimated from the pair-distance distribution function P(r) calculated by 208 GNOM (real-space Rg = 17.72 Å, reciprocal-space Rg = 17.76 ± 0.13 Å) ( Figure 6B ). From the 209 calculated P(r), the maximum particle dimension of the protein was estimated to be 65 Å ( Figure   210 6B). The asymmetric shape of the distribution function suggested the presence of a tail in the 211 solution structure of 3C. The values of Rg and Dmax from SAXS data also suggested that the 212 protein exists as a monomer in solution. Furthermore, the estimated molecular mass of 22.2 ± 213 0.9 kDa (derived from Porod volume) or 20.5 ± 1.1 kDa (derived from SAXS MoW calculation) was 214 in very good agreement with the calculated molecular mass of a monomer (19.6 kDa). 215 We compared the calculated scattering profiles of the crystal structure and the average 216 MD structure to the experimental data. As shown in Figure 6C , both structures agreed well with 217 the experimental data. However, the calculated curve from the average MD structure showed a 218 slightly better fitting to the SAXS data relative to the crystal structure (χ = 1.57 for crystal 219 structure vs. χ = 1.47 for MD structure). Next, we built the ab initio low-resolution SAXS model 220 using DAMMIN and DAMMIF programs; 10 independent models from each program were 221 generated and averaged. The models were highly similar; the average values of the normalized 222 spatial discrepancy (NSD), which represents the similarity among the models, were 0.563 ± 0.01 223 and 0.698 ± 0.064 for DAMMIN and DAMMIF models, respectively. Superimposing the average 224 MD structure onto the SAXS model showed a very good agreement between these two models 225 with an NSD of 0.66. The crystal structure showed less agreement with the SAXS model with an 226 NSD of 0.92. The reconstructed low-resolution SAXS model clearly revealed the presence of an 227 extended and highly dynamic N-terminal helix of 3C, the volume of which exceeded the volume 228 of a helix that is 14 residues in length ( Figure 6D ). Worth noting, the amphipathic nature of this 229 helix ( Figure S3 ) may also contribute to the alternative conformations observed on membranes 230 (Figures 5E and 5F ).

The membranes forming the genome-replication organelle of enteroviruses are enriched 233 in PI4P. It is known that the RNA-dependent RNA polymerase domain, 3D, of enteroviruses binds 234 to PI4P, although it is not clear where on the protein this binding occurs (Hsu et al., 2010) . 235 Because the 3CD protein accumulates to a much higher level in PV-infected cells than 3D and 3CD 236 has been implicated in formation and/or function of the genome-replication organelle (Oh et al., 237 2009), we used the available structural information for these proteins (Marcotte et al., 2007; 238 Mosimann et al., 1997; Thompson and Peersen, 2004) to screen for PI4P-binding sites by using a 239 molecular-docking algorithm (Goodsell et al., 1996; Morris et al., 1998) . The observation of a 240 putative PI4P-binding site on 3C was unexpected but exciting as it was now very possible that 241 trafficking of PV proteins to the site of genome replication might be facilitated by use of new 242 structural classes of PI4P-binding domains (Figure 1 ). Our focus on 3C was motivated by the 243 arsenal of structural and computational resources available to study PV 3C that would facilitate 244 identification and characterization of the PI4P-binding site (Amero et al., 2008; Marcotte et al., 245 2007; Mosimann et al., 1997; Moustafa et al., 2015) , including a near-complete backbone NMR 246 resonance assignment of the 2D (Amero et al., 2008) . 247 Using a solution-and bilayer-based assay for PIP binding, we observed PIP binding by 3C 248 (Figures 3 and 4) . Titration of PI4P into a solution containing 3C caused CSPs that were consistent 249 with the major PI4P-binding site observed computationally (Figure 2A) . However, the CSPs also 250 included other residues (Figure 2B) , in particular residues known to contribute to the RNA-251 binding activity of 3C (Amero et al., 2008) . PIP-binding was indeed competitive with RNA binding 252 (Figures 3H and 3I) . For a small molecule like PI4P to cause so many CSPs or to compete with a 253 substantially larger RNA, either multiple molecules of PI4P bind to the protein or a single 254 molecule of PI4P induces a large change in the conformation and/or dynamics of the protein. 255 Analysis of NMR spectra were inconsistent with large-scale changes in conformation or dynamics 256 as most of the resonance positions and intensities were not affected by PI4P binding (Figure 2) . 257 MD simulations were consistent with multiple PIPs binding to PV 3C along the shallow cleft used 258 for RNA binding (Figure 5) . While the negative charge of the phospholipid headgroup contributed 259 to PIP binding, the display of the charge on the inositol ring contributed to high-affinity binding 260 ( Table 1) . In this regard, neither the location of the phosphate on the inositol ring nor the number 261 of phosphates on the inositol ring mattered for 3C binding ( Table 1) , although the nature of the 262 PIP influenced the conformation of 3C bound to membrane observed computationally ( Figure 5) . 263 MD simulations indicated that one contributor to the PIP-dependent conformation is the 264 flexibility of the N-terminus (compare Figures 5E and 5F ). Such conformational flexibility of the 265 3C N-terminus was also observed in solution using SAXS (Figure 6 ).

The PIP-binding site structure, corresponding specificity and extreme capacity to cluster 267 PIPs differs substantially from known cellular PIP-binding domains. Most cellular PIP-binding 268 domains contain a solvent-accessible, basic cavity to which a PIP binds (Lemmon, 2008) . In certain 269 cases, spatial restriction within the cavity enables stereospecificity to the interaction with a PIP (Chukkapalli et al., 2010; Johnson et al., 2016; Saad et al., 2006) . The role of viral matrix 301 proteins is to bridge the genome or some genome-containing ribonucleoprotein complex to the 302 membrane used to envelope the virus as it buds from the cell. Both of these viral proteins appear 303 to bind specifically to PI(4,5)P2. PIP-binding signals to the protein that it has reached the 304 destination for assembly. In the case of MA, PIP binding triggers the exposure and insertion of 305 its myristoylated N-terminus into the plasma membrane (Saad et al., 2006) . In the case of VP40, 306 PIP binding promotes oligomerization (Johnson et al., 2016) . Both HIV MA and Ebola VP40 307 assemble into oligomers, which cluster PIPs, but not to the same extreme as predicted for 3C 308 (Johnson et al., 2016; Saad et al., 2006) . Both of these matrix proteins bind RNA. In the case of Non-structural protein 5A (NS5A) from hepatitis C virus binds PI(4,5)P2 and to lesser 312 extent PI(3,4)P2 through its N-terminal amphipathic α-helix (Cho et al., 2015) . The motif used has 313 been referred to as the basic amino acid PIP2 pincer (BAAPP) domain (Cho et al., 2015) . This motif 314 is nothing more than two basic amino acids flanking a series of hydrophobic residues displayed 315 on a helix, presumably that penetrates into the bilayer (Cho et al., 2015) . The BAAPP motif can 316 be found in amphipathic α-helices of cellular and viral proteins; however, the function is not 317 known. Indeed, such a motif might exist in the N-terminus of 3C ( Figure S3 ). In the case of 3C, 318 we propose that this helix just augments membrane binding without any specificity (Figures 5E 319 and 5F). How the BAAPP domain confers specificity is not known. It is worth noting that NS5A is 320 an RNA-binding protein. The organization described for 3C, amphipathic α-helix followed by an 321 RNA-binding domain, therefore applies to NS5A. It is compelling to speculate that this RNA-322 binding domain might also interact with PIPs. Our computational studies suggested that 3C interacted with a single PIP mediated by 340 K12 and R84. (Figure 1A) . Although R13 was nearby, a role for this residue in PIP binding was not 341 suggested ( Figure 1A) . Removal of the positive charge from K12 had no impact on PIP binding, 342 removal from R84 increased PIP-binding affinity, but removal from R13 weakened PIP-binding 343 affinity ( Figures 3C-3F) . A role for R13 and R84 in PIP binding was supported by NMR experiments, 344 but the evidence for K12 was weak ( Figure 2B) . So, while the molecular-docking simulation 345 pointed us in the right direction, the molecular details of the putative interaction were not Deliberate interrogation of this hypothesis is warranted. 364 In conclusion, in an effort to begin to identify proteins and mechanisms that PV uses to 365 recognize PIPs, we have discovered that nature has been creative and adapted a viral RNA-366 binding surface to perform this additional task. The outcome is a new structural class of PIP-367 binding proteins with the capacity for multivalent binding to any mono-, bis-or tris-phosphorylated PIP, but with a unique conformational state determined by the nature of the 369 bound PIP. It is likely that a similar mechanism exists in noroviruses and coronaviruses, suggesting 370 a role for PIPs in the biology of these viruses even though empirical support of this possibility is 371 needed. CSPs at least one standard deviation above the average were considered to be substantial (Δtotal = 0.019 ppm). These results are in agreement with the docking analyses, which show PI4P-binding towards the N-terminus of 3C. 

Chemical structures of the supported lipid bilayer (SLB) components; pH-sensitive ortho-Sulforhodamine B-1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (oSRB-POPE); L-αphosphatidylinositol 4-phosphate (PI4P), and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC). "R" represents the fatty acyl chains. (B) Schematic diagram illustrating the principles of the SLB-binding experiment. In the absence of 3C, the fluorescent probe is in its "on state" (left).

Upon binding to the membrane, the interfacial potential is increased, causing the fluorescent probe to switch to its "off state" (right). The pH response curve of the fluorescent probe in a bilayer containing 92 mol% POPC, 7.5 mol% PI4P, and 0.5 mol% oSRB-POPE is shown. (C) 3C binding to PI4P-containing SLBs. Change in fluorescence intensity was observed as a function of 3C concentration, which was normalized to a reference channel. Shown is a hyperbolic fit of the data set. The apparent dissociation constant for 3C-PI4P interaction is 2.36 ± 0.38 µM. 3C was unable to bind to pure POPC membranes. All experiments were conducted at 20 mM HEPES at pH 7.0, 100 mM NaCl, and 5 mM magnesium acetate. Error bars represent the SEM (n = 3). to the experimental SAXS data (grey). Both structures fit well; however, the average MD structure showed slightly better fitting compared to the crystal structure as indicated by its lower χ-value.

The reconstructed SAXS filtered model calculated by DAMMIN is shown as an orange transparent surface with the crystal structure superimposed. The SAXS model clearly shows an extended N-terminus compared to the crystal structure. The calculated scattering profile of the SAXS model (orange) fitted to the experimental data (grey) is also shown. 1.4 ± 0.1 4.7 ± 0.5 0.3 ± 0.1 PI(4,5)P2

1.1 ± 0.1 2.2 ± 0.5 0.3 ± 0.1 a Each membrane also contained 0.5 mol% oSRB-POPE. b PI and POPS binding as a function of 3C concentration was in the linear range. Assuming an amplitude similar to that of PI4P binding, yields the indicated values for Kd,app, which represent a lower limit. 

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Craig E. Cameron (cec9@psu.edu).

Cell Lines Rosetta™(DE3) Competent Cells (Millipore Sigma) were used for PLC-δ1 protein purification and grown in NZCYM media, pH 7.6, by IPTG induction.

Poliovirus proteins were expressed and purified from E. coli BL21(DE3)pCG1 (Shen et al., 2008) in NZCYM media, pH 7.6, by IPTG induction.

Materials.

BODIPY-FL-labeled and unlabeled water soluble phosphoinositides were from Echelon Biosciences, Inc. Construction of expression plasmids.

The 3C construct used in this study, pET26Ub-PV-3C-C147A-C153S-CHis6, was previously described (Amero et al., 2008) . The QuikChange site-directed mutagenesis kit (Stratagene) was used to introduce mutations into the 3C coding sequence. The oligonucleotides employed were as follows: The docking study of PI4P to 3C and 3D proteins was carried out using AutoDock 4.2 suite of programs (Goodsell et al., 1996; Morris et al., 1998) . The crystal structure of the 3C protein (PDB 1L1N) (Mosimann et al., 1997) was obtained from the Protein Data Bank. Only chain A of the two identical monomers in the crystal structure was chosen for the docking. The protein was prepared for the docking runs as follows:

the structural water molecules and any non-protein atoms were deleted from the crystal structure; explicit hydrogen atoms were added to the protein and the structure was subjected to quick minimizations (500 steps) using AMBER force field ff99SB in Chimera (Hornak et al., 2006; Pettersen et al., 2004) . Next, the AutoDockTools (ADT) was used to complete the preparation of the target protein by merging the non-polar hydrogen atoms, adding Kollman charges, and creating the PDBQT files for the docking runs. The structure of the PI4P ligand was prepared by modifying the structure of PI(4,5)P2, extracted from the NMR complex structure of HIV-1 matrix protein bound to PI(4,5)P2 (PDB 2H3Z), in Chimera (Saad et al., 2006) . The ligand was subjected to quick energy minimizations (100 steps) in Chimera after adding hydrogen atoms and Gasteiger-Marsili atomic partial charges; ADT was then used to prepare the ligand PDBQT file, the flexible ligand has 20 active torsions.

For the target protein, the affinity grid field was generated using the auxiliary program AutoGrid. Small angle X-ray scattering (SAXS) analysis.

A non-His-tagged version of 3C protein was expressed and purified. Protein samples were prepared at concentrations of 3.2, 6.5, and 13.0 mg/mL in 40 mM Tris-HCl buffer at pH 7.4 containing 200 mM NaCl, 10% glycerol, 1 mM EDTA and 2 mM DTT. Synchrotron SAXS data were collected on the F2-line station at MacCHESS at 293 K using dual Pilatus 100K-S detector and a wavelength of 1.224 Å. The data were collected using exposure times of 5 minutes in ten 30-second frames and covered a momentum transfer range (q-range) of 0.01 < q < 0.8 Å -1 . The program RAW (Nielsen et al., 2009) was used for data reduction and background subtraction. The radius of gyration (Rg) and forward scattering I(0) were calculated using Guinier approximation. The GNOM program (Svergun, 1992) was used to calculate the pair-distance distribution function P(r) from which the maximum particle dimension (Dmax) and Rg were estimated.

The ab initio low-resolution models were reconstructed using DAMMIN (Svergun, 1999) and DAMMIF (Franke and Svergun, 2009) programs using data in the range 0.031 < q < 0.40 Å -1 ; ten independent models generated from each program were averaged using DAMAVER (Volkov and Svergun, 2003) . FOXS Expression and purification of PV 3C.

3C protein was expressed in BL21(DE3)pCG1 competent cells as previously described (Shen et al., 2008) .

Cells were harvested by centrifugation at 5,400 x g at 4°C for 10 min and washed with a buffer containing 10 mM Tris and 1 mM EDTA at pH 8.0 and then re-centrifuged. The cell pellet was resuspended in Buffer A [20 mM HEPES, 10% glycerol, 5 mM imidazole, 5 mM β-mercaptoethanol (BME), 1 mM EDTA, 500 mM NaCl, 1.4 μg/mL pepstatin A, 1.0 μg/mL leupeptin, at pH 7.5] at a 5 mL Buffer A per 1-gram cell pellet ratio. Isotopic labeling and sample preparation for NMR.

Small unilamellar vesicle (SUV) preparation.

Lipids were mixed at the desired mole ratio in chloroform in a glass scintillation vial. The chloroform was removed by continuously purging the vial with N2 gas. Desiccation was performed under vacuum for 2-3 hours to remove any residual organic solvent. The dried lipid films were hydrated with 20 mM HEPES, 100

mM NaCl, at pH 7.0 followed by sonication in a water bath at room temperature to obtain 0.5 mg/mL lipid suspensions. These suspensions were then subjected to 10 freeze−thaw cycles with liquid N2 and 40°C

water bath, and 10 extrusion cycles through a 100 nm track-etched polycarbonate membranes Glass cleaning procedure.

The glass substrates used for supporting fluid lipid bilayers were first boiled in a 7-fold diluted 7X cleaning solution (MP Biomedicals) and water for one hour. This was followed by rinsing the glasses with copious amounts of purified water before drying thoroughly with nitrogen gas. The coverslips were then annealed for five hours at 550°C before being stored until use. Photomask design.

The microfluidic device was designed with a drafting software (AutoCAD v.2016). This device consists of eight channels with independent inlets and outlets. Each channel has a dimension of 1 cm x 50 µm (length x width) and each channel is separated from each other by a 25 µm gap. The design with black background and clear features was printed at a 20,000 dpi resolution on a transparent mask (5 x 7 in) by CAD/Art Services.

Silicon mold fabrication.

The mold containing the microfluidic patterns was fabricated in the Nanofabrication Laboratory at Penn State in University Park, PA. A 4-inch silicon wafer was dehydrated (1 min at 95°C). The SU8-50 (MicroChem Corp.), negative tone photoresist, was poured onto the center of the wafer via static dispensing method, spun (5 sec at 500 rpm, 35 sec at 3,000 rpm), pre-baked (5 min at 65°C), soft-baked (15 min at 95°C), exposed to UV light with the MA/BA6 mask aligner for 1 min (4 x 15 sec/exposure) at a power density of 8.0 mW/cm 2 to produce a positive relief of photoresist on the wafer, and then postbaked (1 min at 65°C and 4 min at 95°C) to selectively cross-link the UV-exposed portions of the film. The wafer was developed in an SU8 developer for 6 min (without agitation), rinsed with isopropyl alcohol (IPA), and dried with N2 gas. The wafer with photoresist pattern was hard-baked for 30 sec at 65°C, 30 sec at 95°C, and 1 min at 150°C.

In the simulations, a membrane consisting of 244 POPC and 20 PI4P lipid molecules, and a membrane composed of 244 POPC and 20 PI(4,5)P2 were prepared. The model membranes were created by the CHARMM-GUI (Wu et al., 2014) and equilibrated for 50 ns. The protein was placed above the membrane with the center of mass of the protein about 5 nm away from the center of membrane in the Z direction.

The closest distance between the protein and the lipid molecules is around 0.5 nm. The starting orientation of the protein towards the membrane is based on the docking prediction, with few vital PIPinteracting residues such as R13 and R84 facing the membrane. The protein and membrane were solvated in a box of TIP3P water. Sodium and chloride ions were added at a near-physiological ion concentration of 150 mM.

The CHARMM36 force field was used for the simulation atoms (Huang and MacKerell, 2013; Klauda et al., 2010) . A time step of 2 fs was employed. The van der Waals interactions was cut at 1.2 nm and the electrostatic interactions was treated with Particle-Mesh Ewald (PME) method. The SHAKE algorithm was used for length constraint on bonds involving hydrogen. The simulations were carried out under conditions of constant temperature at 310 K and constant pressure at 1 bar. The simulation system was subjected to energy minimization for 2000 steps, followed by 1 ns constraint simulation with a harmonic potential applied on protein atoms. The production simulations of 400 ns in length were then performed.

The initial 30 ns of all simulations was simulated using the NAMD/2.10 package (Phillips et al., 2005) . Then it was moved to the Anton supercomputer that is optimized for MD simulation for another 370 ns simulation (Shaw et al., 2009) . The trajectories of the last 300 ns were used for analysis.

Statistical analysis and nonlinear regression was provided by GraphPad Prism v.6. Error bars represent the SEM.

the N-terminus of poliovirus 3C. Two clusters were identified by docking dibutyl-PI4P (red spheres) on 3C surface (grey ribbon). The major cluster (94% of the trials) encompasses residues K12, R13, and R84; the minor cluster (6% of the trials) encompasses residues K108 and R143. Total of 100 docking runs were performed. The crystal structure of 3C (PDB 1L1N) was obtained from the Protein Data Bank (Mosimann et al., 1997) . The structure of the PI4P ligand was prepared by modifying the structure of phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2), extracted from the NMR complex structure of HIV-1 matrix protein bound to PI(4,5)P2 (Saad et al., 2006) . Phosphates on the inositol head group are depicted as red spheres; basic residues in the major and minor clusters are depicted as sticks. (B) Multiple-sequence alignment of enteroviral 3C proteins shows that the majority of the residues predicted by docking (K12, R13, R84, and R143) and/or the basic charge are conserved; K108 is not conserved (colored in blue). Residues implicated in phosphoinositide-binding ( ); conserved residues (*); PV, poliovirus; CV, coxsackievirus; EV, enterovirus; BEV, bovine enterovirus; PEV, porcine enterovirus; SEV, simian enterovirus; SV, simian virus. N heteronuclear single quantum correlation (HSQC) spectra of free 3C (black) and 3C in a complex with dibutyl-PI4P (red). (Right) Close up view of the chemical shift perturbations (CSPs) for K12, R13, and R84, which are implicated in PI4P-binding by docking. The NMR experiments were conducted at 25°C in a buffer containing 10 mM HEPES at pH 8.0, 50 mM NaCl, and 10% D2O. The final concentration of 3C and PI4P were 0.2 mM and 2.0 mM, respectively. (B) A bar graph showing CSPs induced throughout 3C by PI4P binding. The total chemical shift change (Δtotal) depends on the chemical shift changes in the 1 H (ΔH) and 15 N (ΔN) dimensions according to Δtotal = (ΔH 2 + 0.2 ΔN 2 ) 0.5 . Residues with significant CSPs (above the red line) are indicated. CSPs at least one standard deviation above the average were considered to be substantial (Δtotal = 0.019 ppm). These results are in agreement with the docking analyses, which show PI4P-binding towards the N-terminus of 3C.

The principles of the fluorescence polarization (FP)-based PIP-binding assay. Unbound form of the fluorescent ligand (PIP-BODIPY-FL) yields low polarization value due to its small size and rapid rotation. In contrast, the bound form yields a higher polarization value due to the overall size of the complex and its slow rotation. All FP-based PIP-binding experiments were conducted in a buffer containing 20 mM HEPES at pH 7.5 and 10 mM NaCl. 3C-PI(4,5)P2 samples were incubated for 30 s inside a chamber at 25 °C prior to measuring the millipolarization (mP). (B) PIP-binding specificity of 3C. 3C binds to both bis-and tris-phosphorylated PIP species. Binding was tested at a fixed 3C concentration (5 µM) using 0.4 nM of each PIP-probe as indicated. Error bars represent the SEM (n = 3). (C) WT 3C and 3C variants: (D) K12L, (E) R13L, and (F) R84L bind to phosphatidylinositol 4,5-bisphosphate (PI(4,5)P2) with varying affinities. Shown is a hyperbolic fit for each data set from which each apparent dissociation constant (Kd,app) was obtained. Values are provided in Table 1 . (G) The principles of the fluorescence polarization (FP)-based competition assay. 3C-RNA (fluorescently-labeled 11-mer) complex is pre-formed, which yields Chemical structures of the supported lipid bilayer (SLB) components; pH-sensitive ortho-Sulforhodamine B-1palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (oSRB-POPE); L-α-phosphatidylinositol 4phosphate (PI4P), and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC). "R" represents the fatty acyl chains. (B) Schematic diagram illustrating the principles of the SLB-binding experiment. In the absence of 3C, the fluorescent probe is in its "on state" (left). Upon binding to the membrane, the interfacial potential is increased, causing the fluorescent probe to switch to its "off state" (right). The pH response curve of the fluorescent probe in a bilayer containing 92 mol% POPC, 7.5 mol% PI4P, and 0.5 mol% oSRB-POPE is shown. (C) 3C binding to PI4P-containing SLBs. Change in fluorescence intensity was observed as a function of 3C concentration, which was normalized to a reference channel. Shown is a hyperbolic fit of the data set. The apparent dissociation constant for 3C-PI4P interaction is 2.36 ± 0.38 µM. 3C was unable to bind to pure POPC membranes. All experiments were conducted at 20 mM HEPES at pH 7.0, 100 mM NaCl, and 5 mM magnesium acetate. Error bars represent the SEM (n = 3).

Figure 5. 3C interacts with multiple PI4P and PI(4,5)P2 ligands. All-atom molecular dynamics simulations of (A) 3C binding to a PI4P-containing membrane or (B) to a PI(4,5)P2-containing membrane. Snapshots of selected simulations show that 3C interacts with five, clustered PI4P or PI(4,5)P2 molecules. PI4P and PI(4,5)P2 are shown as dark grey sticks with phosphates colored orange (phosphorus) and red (oxygen). 3C is shown as a light grey ribbon; residue side chains are colored as follows: light blue (positive side chains) and red (negative side chains) sticks. Also shown are: sodium ions as magenta spheres; the amino-and carboxy-termini as blue and red spheres, respectively; 3C protease active-site residues as black spheres. Specific interactions are as follows, with number in parentheses referring to headgroup in panel: (1) R13 and R84; (2) D32, mediated by sodium ions; (3) D32 for PI4P in panel A mediated by sodium, or R176 for PI(4,5)P2 in panel B; (4) K156 and R176; (5) α-amino group of G1 for PI4P in panel A, or K12 and R13 for PI(4,5)P2 in panel B. (C) and (D) Surface representations of 3C (light grey) with PI4P-containing membrane in panel C or PI(4,5)P2-containing membrane in panel D (blue, red and dark grey sticks); phosphatidylcholine (light grey lines). Charged interactions between 3C and the respective ligands are shown as blue (positive) or red (negative) spheres to correspondingly colored sticks. (E) and (F) same as panels (C) and (D) rotated 90 degrees counter-clockwise about the vertical access of the page to reveal the amino terminus (blue) interacting with the membrane (light grey lines). PI4P-containing membrane was composed of 244 phosphatidylcholine (PC) and 20 PI4P lipids; PI(4,5)P2containing membrane was composed of 244 PC and 20 PI(4,5)P2 lipids. All lipids were equally distributed on each monolayer of the membrane. The simulation was conducted at 310K for 400 ns; the last 300 ns of the trajectories were used for analysis. Figure 6 . The N-terminus of 3C is dynamic as revealed by SAXS. (A) Scattering profiles of 3C obtained at three different protein concentrations (3.2 mg/mL, green; 6.5 mg/mL, red; 13.0 mg/mL, grey). Scaling factors were applied for the low-concentration data. The inset shows the linear fitting of the Guinier plot from which radius of gyration (Rg) was determined. (B) The transformed scattering profile calculated by GNOM from the pair-distance-distribution function, P(r), that is shown in the inset. The maximum particle dimension and the Rg obtained from P(r) are indicated. The estimated Rg from GNOM is in good agreement with that obtained from Guinier fitting. (C) The calculated scattering profiles of the crystal structure of poliovirus 3C monomer (cyan) (Mosimann et al., 1997) and the average MD structure (black) (Moustafa et al., 2015) fitted to the experimental SAXS data (grey). Both structures fit well; however, the average MD structure showed slightly better fitting compared to the crystal structure as indicated by its lower χ-value. (D) The reconstructed SAXS filtered model calculated by DAMMIN is shown as an orange transparent surface with the crystal structure superimposed. The SAXS model clearly shows an extended N-terminus compared to the crystal structure. The calculated scattering profile of the SAXS model (orange) fitted to the experimental data (grey) is also shown.

Purification was performed as described above. Purified 3C fractions were concentrated using Vivaspin-20 centrifugal concentrators (Sartorius Stedium Biotech) and buffer exchange was performed using Zeba spin desalting columns (7,000 MWCO) (Thermo Scientific)

N HSQC NMR spectra were recorded at 25°C on a 600 MHz Bruker Avance III spectrometer equipped with a 5mm

NMRPipe and NMRView software were used to process and analyze the NMR spectra

ΔωHN is the change in chemical shift in the amide proton dimension and ΔωN is the change in chemical shift in the nitrogen dimension. Chemical shift perturbations at least one standard deviation above average were considered to be substantial

Fluorescence polarization-based phosphoinositide binding assay

To test PIP-binding specificity, 5 µM of 3C or its mutants were added into a solution containing 0.4 nM of BODIPY-FL-labeled PIPs in a binding buffer [20 mM HEPES and 10 mM NaCl, at pH 7.5] at a 100 μL final reaction volume. Due to tight binding, PLCδ1 PH domain PIP-specificity was assessed at 100 mM NaCl and in the presence of 34 nM protein. For competition experiments

P-4504) were titrated into a solution containing a pre-formed 3C-RNA complex. 1 μM of 3C and 0.4 nM of 3'-fluorescein (FL) labeled 11-mer RNA (5'-AGU UCA AGA GC-3'-FL), corresponding to the PV oriI sequence, were added to the binding buffer

)P2) competes out the bound RNA, the millipolarization (mP) value decreases. (H) FP-based competition experiments. The unlabeled PI(4,5)P2 efficiently competes out the bound RNA, suggesting that these interactions are mutually exclusive. The competition experiment with phosphatidylinositol (PI) competitor is used to show that binding to 3C is required for RNA displacement. Binding reactions contained 1 µM 3C-WT, 0.4 nM 3'-fluorescein (FL)-labeled RNA

Changes in the direction of K12, R13, and R84 resonances (green arrow) in two-dimensional space suggest that RNA and PIP interactions are mutually exclusive. The NMR-based experiments were conducted at 25°C in a solution containing 10 mM HEPES at pH 8.0, 50 mM NaCl, and 10% D2O. The final concentration of 3C, RNA, and