Modeling of the human rhinovirus C capsid suggests possible causes for antiviral drug resistance

Abstract

Human rhinoviruses of the RV-C species are recently discovered pathogens with greater clinical significance than isolates in the RV-A+B species. The RV-C cannot be propagated in typical culture systems; so much of the virology is necessarily derivative, relying on comparative genomics, relative to the better studied RV-A+B. We developed a bioinformatics-based structural model for a C15 isolate. The model showed the VP1–3 capsid proteins retain their fundamental cores relative to the RV-A+B, but conserved, internal RV-C residues affect the shape and charge of the VP1 hydrophobic pocket that confers antiviral drug susceptibility. When predictions of the model were tested in organ cultures or ALI systems with recombinant C15 virus, there was a resistance to capsid-binding drugs, including pleconaril, BTA-188, WIN56291, WIN52035 and WIN52084. Unique to all RV-C, the model predicts conserved amino acids within the pocket and capsid surface pore leading to the pocket may correlate with this activity.

1. Introduction

The human rhinoviruses (RV) are positive sense RNA viruses in the Enterovirus genus of the Picornaviridae family (Palmenberg et al., 2009). They are the most frequent causative agents of the “common cold” and responsible for millions of lost personnel hours in the workplace each year. The best studied isolates belong to the RV-A and RV-B species, where they are binned together if they share greater than 75% nucleotide identity (88% amino acid identity) in the VP1 region of their polyproteins. Each species further divides its isolates into multiple numbered genotypes. Originally, ~100 types from clinical panels archived by the American Type Culture Collection were indexed after assessment of antigenic crossreactivity or serotyping in rabbits. RV-A87 was subsequently reassigned to the Enterovirus D (EV-D68) after reevaluation of genetic, immunogenic and receptor properties (Savolainen et al., 2002). Common to the original RV-A (74 serotypes) and RV-B (25 serotypes) is the use of ICAM-1 or LDLR for cell attachment and entry (Vlasak et al., 2005). They are labile at low pH (<5), and grow predominantly in sinus and upper airway tissues (for reviews, see (Bochkov and Gern, 2012; Ashraf et al., 2013)).

Because of their medical and economic importance, considerable resources have been expended developing therapeutics against the RV-A+B. The ubiquitous nature of these viruses and the many serotypes, preclude the practical use of vaccines. Directed drugs that target protein elements in the RV replication cycle (e.g. rupintrivir), can be effective (Binford et al., 2007). But the preferred strategy is to target the virus before infection, usually by exploiting unique “pocket” features characteristic of all enterovirus virions. The RV capsids are icosahedral (pseudo T=3), composed of 60 copies each of four structural proteins, VP1, VP2, VP3 and VP4. The three largest proteins, VP1–3, assume similar 8-stranded, anti-parallel β-barrel motifs, despite being formed from very different sequences (Fig. 1). Protomer subunits containing mature copies of VP1–4 spontaneously self-assemble into pentamers with the VP1 proteins assuming symmetry around the 5-fold axes. When the pentamers coalesce into particles, encapsidating the genome RNA, the VP2–3 proteins alternate around the 3-fold and 2-fold axes (Fig. 1A). A deep groove within each protomer, formed where VP1–3 abut, creates a contiguous “canyon” circling each pentamer (Fig. 1B). The canyon topography is characteristic of all enteroviruses, and marks the thinnest portion of the capsid shell. The “north” (5-fold) and “south” (2-fold) walls of the canyon (Fig. 1C) are lined with residues that confer receptor recognition and type-specific immunogenicity (Arnold and Rossmann, 1990).

Fig. 1.

RV-C15 capsid model. (A) The C15 model with VP1 (blue), VP2 (green) and VP3 (red) proteins, around 5-fold, 2-fold and 3-fold axes of symmetry (Basta et al., this issue). The short VP4 protein (yellow) is internal. (B) A triangular crystallographic (PDB) subunit orients the VP1+VP2 from one biological subunit, with VP3 from the adjacent, counterclockwise unit. The hydrophobic drug binding pocket extends from a pore at the base of the canyon into the VP1 central core. (C) The VP1–3 proteins have similar 8-stranded β-barrels with extended connecting loops. The canyon is a depression in the surface topography, ringing the 5-fold. The north and south walls are landmark features. Pleconaril (gray spheres) is modeled into the VP1 pocket. Figure is taken from Hadfield et al. (1995).

When the 99 historical RV-A+B types were tested for sensitivity against a panel of antiviral capsid-binding therapeutics they were found to subdivide, roughly along species lines, into two experimental groups (Andries et al., 1990). The structures of 28 virus-drug complexes have been determined to atomic resolution (Suppl. Table S1). The Group-1 viruses (predominantly RV-B) have long, narrow pockets interior to their VP1 proteins, which accommodate matching long chain hydrophobic drugs like WIN52084 (W84). The Group-2 viruses (most RV-A) have shorter, wider VP1 hydrophobic pockets, and therefore accept an alternate cohort of drugs, like R61837 (JEN). Each determined drug-virus structure shows a pore-like opening connecting each VP1 pocket to the deepest portions of the canyon, providing an entry port for the relevant drug (Fig. 1B). Native RV-A+B in the absence of drugs, have “pocket factors”, commonly modeled as sphingosine, in same interior VP1 locations. The intrinsic occupancy of these factors contributes to capsid stability (Oliveira et al., 1993). Displacement of the native factors with efficacious drugs prevents a required VP1 transition during the uncoating process, making such drugs inhibitory to infection. Compounds that exploit this mechanism and bind both Group pockets have shown clinical promise on a range of RV. For example, during Phase 3 evaluations, pleconaril (WIN63843) shortened the duration of the common cold and reduced the window of virus shedding. Despite this efficacy, the FDA disapproved pleconaril for over-the-counter use, citing concerns about interference with the metabolic pathways of other drugs (Hayden et al., 2003). A newer analog, BTA-188, is reportedly more effective in reducing RV infections for many laboratory and clinical strains (Barnard et al., 2004).

While a “cure” for the common cold might have seemed in the offing from this work, the discovery in 2006 of a completely new RV species, put optimism on hold. The RV-C are clearly rhinoviruses, but unlike RV-A+B, they are not readily propagated in typical cell culture systems (Bochkov et al., 2011). The currently recognized 51 types of RV-C were detected and characterized by direct sequencing from patient effluents. As with the RV-A+B, each RV-C type includes those isolates whose VP1 sequences exceed 87% pairwise identity at the nucleotide level (Simmonds et al., 2010; McIntyre et al., 2013). The RV-C have special clinical relevance since it is now recognized these strains are associated with up to half of infections in young children (Bochkov et al., 2011). They grow readily in both the lower and upper airways and tolerate higher growth temperatures in culture (Ashraf et al., 2013). Moreover, the RV-C use cell receptors that are not common to the RV-A+B (Bochkov et al., 2011). Unfortunately, these receptors are apparently lost, whenever primary tissue snippets are transitioned to undifferentiated monolayers. RV-C can be grown in mucosal organ cultures, but this technique requires the availability of primary human donor samples (Bochkov et al., 2011). Parallel work with differentiated sinus or bronchial epithelial cells at air–liquid interface (ALI) is promising (Ashraf et al., 2013; Hao et al., 2012), but neither technique has yet produced enough virus for extensive biological studies. Instead, RV-C information relies heavily on comparative sequence analysis to maximize data from limited experimental samples.

Current computational tools include a deep alignment of full-genome datasets for >350 RV-A+B+C isolates. It was nucleated by superimposition of determined capsid structures for multiple RV-A+B, then extended to the RV-C with Markov-based profile fits (Palmenberg et al., 2009). Combined with the RV-A+B structures, the alignment helped to develop a high resolution 3D model for C15 (Basta et al., this issue), an isolate which has been cloned into cDNA, and tested for biological activity in mucosal and ALI cultures (Ashraf et al., 2013; Bochkov et al., 2011). The new model displays an altered surface topography for all RV-C, accounting for differential receptor use and diverse immunogenicity (Basta et al., this issue). It also shows the essential Cα backbone of the RV-C VP1 β-core that marks the capsid drug-binding pocket is superimposable on the RV-A+B, regardless of whether the model was tweaked by Group 1 (e.g. B14) or Group-2 (e.g. A16) structures. While this was reassuring in terms of model veracity, the outcome was puzzling in terms of biology, because in preliminary studies (below), recombinant C15 virus was known to behave somewhat differently from Group-1 or Group-2 RV.

The lack of primary organ cultures and inconsistencies among donor samples make tests against the RV-C difficult. We reported that WIN56291 (W91), a short chain (Group-2) compound, was efficacious against C15 when tested in sinus organ cultures (Bochkov et al., 2011), but it was not clear whether those results would be common to the more reliable ALI systems (Ashraf et al., 2013). Another study reported that the usually-more-potent pleconaril was only weakly effective against C15 when tested in ALI cultures (Hao et al., 2012), although it readily inhibited A16 controls. Therefore, it is indeterminate whether the RV-C are indeed susceptible to those capsid-binding drugs categorized by existing Groups, or if they represent a new category with a different pocket efficacy. To gain more insight, the C15 capsid model was re-probed against all known RV PDB datasets, looking for structural inconsistencies and/or conserved sequences that might explain observed drug reactivity. Parallel experiments in mucosal organ culture and ALI cultures re-tested the efficacy of W91, as well as W84, WIN52035 (W35), pleconaril and BTA-188, added or preloaded into C15 and A16 virions. The model, sequences and mutagenesis experiments suggest there are multiple residues, including those at the entrance to the VP1 pore, that co-vary among all RV-C and contribute to a novel pocket environment. These “Group-3” pockets could explain why all RV-C are probably refractive to strategies with the current antiviral drugs.

2. Results and discussion

2.1. C15 drug susceptibility

Pleconaril and BTA-188 are effective against many RV-A+B types. One study reported that BTA-188 inhibited 75% (of 56) tested strains, including A16 with an MIC₅₀ of ~8 nM and B14 with MIC₅₀ of ~68 nM (Barnard et al., 2004). Pleconaril was reported effective on 93% (of 101) RV-A+B, with an EC₅₀ of 0.59 μM for A16, and 0.16 μM for B14 (Ledford et al., 2004). Drug testing for the RV-C is more difficult. Studies have described recombinant C15 as resistant to pleconaril (Hao et al., 2012), but susceptible to a related compound, W91 (Bochkov et al., 2011). Neither report did extensive testing because the RV-C can only be titered by qRT-PCR, and the yield in organ cultures (Bochkov et al., 2011) or ALI (Ashraf et al., 2013; Hao et al., 2012) is relatively low. For organ cultures particularly, variability among donors creates significant reproducibility issues (Bochkov et al., 2011) as well as a paucity of viable samples for proper controls. In new attempts with this technique, W84 and W35 were tried in sinus organ culture against A1, A16 and C15. Both drugs seemed partially effective against A1 (n=1), and the yield of A16 also showed some reduction with W84. But neither reduced the replication of C15 (Suppl. Fig. S1A).

The inherent inconsistencies with organ cultures warranted a change to human bronchial (HBE) or sinus (HSE) epithelial ALI cultures derived from primary tissue snippets. While again, cell expansion is limited, with successful cultures, the method has less variability and more samples can be tested simultaneously. Preexperiments with WisL cells and A16 determined the lowest concentration of drugs that prevented cytopathatic effect (CPE) and virus replication without overt toxicity (e.g. rounding or lifting from the plate). For pleconaril and BTA-188, those were 10 μg/ml and 2 μg/ml, respectively (Suppl. Fig. S1B). When A16 and C15 were incubated with these drugs, then titered in HSE-ALI cultures (Fig. 2A), the controls showed typical virus replication as evidenced by the 1–2 Log increase in genome signal (RNA Copies) between 4 h (gray bars) and 24 h (blue/red bars) relative to a 0-time binding control (white bar). Pleconaril or BTA-188 significantly reduced A16 amplification (blue bars), but neither drug had the same effect on C15 (red bars). The same result was obtained with a subsequent, independent ALI culture (Fig. 2B). Whether the drugs were present or not, C15 bound equivalently to the cultured cells, established productive infections and amplified normal progeny numbers. In parallel experiments, C15 was also resistant to W91 at a concentration (2 μg/ml) that reduced A16 growth by 1.3 Log (Fig. 2C). The previous partial susceptibility to this compound in organ culture could not be repeated in ALI culture. It is possible the reported phenomenon was unique to virus growth in that particular donor organ sample. In short, throughout multiple attempts, none of the 5 tested capsid-binding drugs had efficacy against C15.

Fig. 2.

Drug tests. (A) Recombinant A16 and C15 were titered for growth in HSE-ALI cultures in the presence of pleconaril or BTA-188. Colored bars (blue, red) are from samples harvested after 24 h PI. Gray bars were harvested after 4 h PI. White bars are 0-time binding controls. (B) Similar to A, this series used an independent HSE-ALI culture. (C) Similar to A, virus samples incubated with W91 were titrated for growth in HSE-ALI cultures. (D) Recombinant A16 and C15 transcripts were transfected into HeLa cells with (or without) BTA-188. After 24 h, virus was harvested and titered by qRT-PCR. (E) Output virus from D was infected into HBE-ALI culture as in A, except no additional drug was added during this phase. In all panels, qRT-PCR values (RNA copies) were determined in duplicate. Experimental repeats are indicated (n). Bars indicate average error from n exps.

2.2. Capsid-drugs fit C15 models

The activities of enterovirus capsid-binding drugs have been reviewed extensively (e.g. Shih et al., (2004; Rossmann, 1994). When such drugs intercalate into the VP1 hydrophobic pocket, they can deform the floor of the canyon to an extent that receptor binding and attachment is inhibited (e.g. B14 and pleconaril) (Pevear et al., 1989). Alternatively, they may enter the pocket with minimum canyon deformation, but prevent required VP1 transitions for entry and uncoating (e.g. A1 and W56) (Rossmann, 1994). Assuming biostability is not an issue, drug failure comes down to one of three mechanisms. The selected drug might be a poor steric fit for a given pocket, and rejected. The drug might fit the pocket but fail to prevent required VP1 changes. Or, the drug might be excluded from the pocket altogether, either because it cannot enter, or cannot displace putative resident pocket factors.

By definition, none of the determined RV-A+B structures exhibit steric clashes between the VP1 proteins and their ligands. These would have been resolved at coordinate deposition. Steric considerations that might explain C15 drug resistance were explored with bioinformatics and the new capsid models. Since output coordinates can adjust slightly, dependent on precise template selection, four C15 VP1 models were created by I-TASSER (Basta et al., this issue), using independent A16 and B14 determined structures that did or did not include bound plenonaril. An additional model was computed on the full I-TASSER database, containing 31 RV files and 29 EV files. Each model was asked to extract from these structures, the subset of 10 best ligand fits for that C15 VP1 conformation. Collectively they identified 12 synthetic compounds covering both Group-1 and Group-2 capsid-binding drugs, and three natural compounds, myristate, succinate and lauric acid, which were co-crystalized as natural ligands (Suppl. Table S2). For each C15 dataset, the identified ligands were superimposed into chimeric PDB files then assessed in MacPyMOL for steric clashes. A negative “C15 Fit” was defined as any single VP1 atom lying closer than 0.35 Å to a ligand atom. W35 and W91 were included for completeness although they were not directly selected by I-TASSER. The BTA-188 structure (in RV-A2) was a personal communication (Biota). As summarized in Table 1, many of these drugs (or natural ligands) were excellent fits for the C15 pocket. Pleconaril and BTA-188 in particular, usually matched the cavity with ease (e.g. Fig. 3B). Indeed, all 5 C15 models selected pleconaril as a preferred ligand, sometimes choosing multiple related structures (e.g. 1c8m and 1nd3) from the database.

Table 1.

Summary of C15 Drug/Model Conflicts.

Drug/ liganda	Returned C15 models without steric clashesb	Returned C15 models with steric clashesc	Conflicting C15 VP1 residuesd
BTA-188	1	0	None
Pleconaril	8	2	Phe132(1), Met204(1)
W91	1	0	None
W54	1	0	None
W84	0	6	Phe96(6), Asn202(5), Met204(6)
S57	1	1	Phe132(1)
W01	2	0	None
W03	3	0	None
W35	1	0	None
W71	2	0	None
J77	1	0	None
W56	1	4	Phe132(3), Met204(4)
W8R	0	2	Phe96(2), Met180(2), Met204(2)
SD8	1	0	None
JEN	1	4	Met116(1), Tyr178(3), Met204(3), Leu207(2)
Myristate	5	0	None
Lauric  Acid	3	0	None
Succinate	2	0	None

^aLigands selected by I-TASSER for any C15 model (Suppl. Table S1).

^bNumber of models returned by I-TASSER with no steric clashes (Suppl. Table S1, “C15 Fit” is “Yes”).

^cNumber of models returned by I-TASSER with 1 or more clashes.

^dC15 VP1 residues with clashes (observations).

Fig. 3.

Key RV residues and pleconaril. (A) Pleconaril (spheres) modeled in the drug pocket of C15 highlights the distribution of any potential residue that could cause a steric clash with any modeled drug as per Table 1 (purple). Location of Phe₁₃₂ and Thr₁₇₂, the Ledford et al. (2005) residues that confer resistance to B14 are highlighted in orange (all panels). (B) C15 model, (C) A16 (1ncr) and (D) B15 (1ncq) with pleconaril are shown as cutaways through the drug pocket. The 5-fold is to the right. The 2-fold, canyon and entry pore (not always visible here) are at the upper left of the pockets.

“Fit” was assessed negatively if even a single atom overlapped, but in fact, every determined compound in the RV database was identified at least once by an I-TASSER model, as a potential ligand for C15. Of the 41 observed clashes, most involved a limited subset of residues (Fig. 3A, Table 1). Only 1 of these (pleconaril and Met₂₀₄) originated from a ligand resolved as part of an RV-A structure (1nd3). Invariably, the conflicting atoms were contributed by a B14-bound drug, a suboptimal template for C15 modeling (Basta et al., this issue). All conflict locations were peripheral to the core of the C15 β-barrel, and scattered throughout its length (Fig. 3A). W84, W56, W8R, were selected several times among the highest ranking ligands, even though they showed the most frequent (2–3) steric clashes. Perhaps not surprising, these same (Group-1) drugs have the longest chain lengths (Suppl. Table S1). They can be made to fit perfectly into C15, if allowed only slightly less rigid superimposition parameters (not shown). Indeed, any of the observed conflicts could probably be resolved easily by minor bond rotations (if permitted). Basically, the full modeling exercise was unable to identify any single, overt steric impediment that might prevent the well-studied cohort of capsid drugs from intercalating into the C15 VP1 pocket. In terms of physical constraint alone, the C15 model(s) do not immediately predict resistance, either to Group-1 or Group-2 drugs.

2.3. Drug pocket environment

The pocket residues important for A16 and B14 drug binding, especially for pleconaril, are well described (Zhang et al., 2004; Ledford et al., 2005). An inclusive contact list within the VP1 β-core was generated with Endscript (Gouet and Courcelle, 2002), by evaluating the 28 RV-A+B structures co-crystalized with drugs and returning every residue within 4 Å of any drug, in any virus. The 43 C15-drug models were scored in parallel. The tabulation, summarized by WebLogo (Fig. 4), shows species variation and frequency according to the deep RV alignments, for every putatively-involved amino acid. The C15 positions identified by PyMOL as possible steric clashes are designated in purple. Below each position, a stacked graph records the number of structures (RV-A, blue, N_max=7; RV-B, red N_max=22) or models (RV-C, light green, N_max=43), placing that residue within 4 Å of any drug. Defined, pleconaril-reactive residues (Zhang et al., 2004) are boldfaced. The RV isolates with Group-1 and Group-2 drug reactivity correlate roughly with the B and A species, respectively (see Section Materials and methods).

Fig. 4.

Drug pocket residues. WebLogo depictions (Crooks et al., 2004) for the RV-A+B+C were tabulated for all VP1 residues measured within 4 Å of any determined or modeled ligand as defined by Endscript (Section Materials and methods). The sequence set was the refined genome alignment. The graphs show the number of determined structures (RV-B in blue, RV-A in red) and RV-C models (green) that placed that residue near a ligand. Position numbers are those for native A16, B14 and C15 VP1 sequences. Published interactions with pleconaril (Zhang et al., 2004) are shown in bold. Residues identified in Table 1 with potential steric clashes are in purple. The 2 pleconaril-resistant locations, identified by Ledford et al. (2005) are highlighted in orange. The category value assigned to each position is described in Section Results and Discussion.

How do these pockets differ? The 36 profiled sites divide into obvious categories. The first (Category-1) includes about half the residues, marking the positions least frequently identified by the 4 Å proximity criterion. If only 1–10 models or structures returned these sites (N_max=72), they are probably not key drug resistance determinants, even though all of them must certainly contribute to the overall RV-C pocket environment. The remainder had proximity thresholds that were returned by at least 11 models or structures. These include, among others, all site analogs previously associated with pleconaril interactions (bold numbers). This cohort itself subdivides into those where the residues are conserved among all RV-A+B+C (Category-2), or where conservation is between the RV-C and either the RV-A or RV-B (Category-3), or where the dominant RV-C sequence(s) is unique (Category-4). The RV common sites (Category-2) include Ile₉₄, Phe₁₁₄, Met₁₃₁, Ser₁₅₅, Phe₁₆₇, Tyr₁₇₈, Asn₂₀₂ and Met₂₀₄. Category-3 includes Gln₉₇, Val/Ile₁₁₈, Phe/Tyr₁₃₂, Pro₁₅₄, Val₁₅₆, Met₁₈₀, and Leu₂₀₇. Collectively, Categories 2+3 contain 5 of the 7 residues with incidents of (putative) steric clashes (Tyr₁₇₈, Met₁₈₀, Asn₂₀₂, Met₂₀₄, and Leu₂₀₇). Again, these modeled clashes are unlikely to be critical determinants of RV-C drug resistance, because the same residues are frequently displayed equivalently in the pockets of many susceptible RV-A+B. The same could be said for virtually all of the Category-2+3 residues, in that these sequences, shared freely among at least 2 species, do not interfere with the defined Group-1 or Group-2 drugs.

Residue Phe/Tyr₁₃₂ is an exception, however. In 2005, as part of a pleconaril study, Ledford et al. (2005) noted that not all RV-A+B were susceptible to this drug. B4, B5, B42, B84, B93, B97 and B99 are resistant. When they compared the relative sequences of 25 key B14 VP1 residues, it was noticed that susceptible viruses invariably displayed Tyr₁₅₂ and Val₁₉₁ in their drug pockets, while the resistant viruses had Phe and Leu at the same positions (orange). Between B14 (susceptible), and B5 or B42 (resistant), these were the only sequence changes in the pocket (or the pore). Recombinant exchange of these residues in a B14 context, reversed pleconaril susceptibility. The study concluded that Leu₁₉₁ was the “major driver” of reduced susceptibility, but the effect was most profound if both sites were changed simultaneously.

Among the RV-C, none, including C15, has the B14 susceptible profile for these 2 residues. The B14 Tyr₁₅₂ is equivalent to C15 Phe₁₃₂. The B14 Val₁₉₁ is equivalent to C15 Thr₁₇₂, a Category-4 position, marking it as unique to the RV-C. Unfortunately, the Ledford study never defined why these particular residues conferred apparent resistance. The A16 analog, Tyr₁₄₄, undergoes the largest displacement in the pocket when pleconaril binds this virus (Zhang et al., 2004). The Leu₁₈₄ (in A16) or Val₁₉₁ (in B14) localize nearby to the same Ring B segment of the drug (Zhang et al., 2004) (Fig. 3C and D). Not only do the RV-C lack a susceptibility profile at these sites, they have 4 additional drug-proximal residues in Category-4 that again, mark them as unique. The C15 Phe₉₆, Met/Leu₁₁₆, Ile₁₃₀ and Ile₁₆₉ positions are rarely (or never) shared with the other species. The RV-C conserve these selections at 93%, 57/43%, 93% and 100%, respectively. In the various C15 models, the pocket floor (Fig. 3A and B), displaying Phe₉₆ and Met₁₁₆ caused occasional steric clashes with the longest (W56, W8R, W84) or fattest (JEN) drugs (Table 1), although as mentioned above, only a few of the drugs and a few of the models recorded such conflicts.

2.4. Drug pocket pores

For capsid drugs to be effective, they must traverse a narrow pore at the base of the canyon into the VP1 core. The opening is a clear, resolved feature in all B14 structures (Fig. 5B). Sited in the very deepest portion of the canyon, the pore is immediately adjacent to the COOH end of VP3 (in red, Fig. 5A–C). Several participating B14 residues, including Asn₁₀₅, Asn₂₁₉ and His₂₂₀ have been described with natural mutations conferring drug resistance or drug selectivity (Hadfield et al., 1995; Kolatkar et al., 1999). The observed mutations did not prohibit drug entry, but instead, partially compensated for drug-induced VP1 changes by providing tighter receptor interactions in the overlapping ICAM-1 footprint (Hadfield et al., 1995). The pore contributions are not as well studied for A16, but again, every resolved structure shows a distinct opening in the same location (Fig. 5A). From the proper angle, the tails of VP1-embedded drugs, pleconaril in these illustrations (dark blue) can be glimpsed through the A16 and B14 holes. To better describe these pores, all canyon residues within 7–9 Å of these tails, were compiled (MacPyMOL), then queried within the sequence alignment for analogs among the species. In the RV-C, 5 of the 9 residues showed Category-2 or Category-3 conservation, including the 3 mutant-defined B14 locations (C15) Asn₉₅, Asn₂₀₂, Asp₂₀₃, as well as Ile₉₄ and Ser₉₉ (Fig. 4D). The other residues, Glu₉₃, Lys₂₄₅, Tyr₂₄₆ and Ser₂₄₇ were Category-4, and showed higher diversity within and between species. The RV-C do not use the same ICAM-1 footprint as the RV-A+B (Bochkov et al., 2011), so some diversity was expected here. What was not expected, was the visible absence of any discernible opening that even resembled a pore in the C15 model. In fact, none of the models, no matter which PDB templates refined them, showed an opening into the VP1 core at this locale (e.g. Fig. 4C). The change in mass, blocking the opening was contributed by several residues, but primarily by Tyr₂₄₆ (purple), conserved as Tyr or Phe in 88% of the RV-C. Every RV-B encodes Gly here (B14 267, 100%), while the RV-A have His (A16 260, 68%). Sterically, C15 Tyr₂₄₆ is reasonably confined to this orientation as are its surrounding neighbors. If this model’s predictions were true for real virions, drug entry might be slowed or even precluded by the tighter passageway. Conceivably though, an appropriately effective drug (pleconaril is modeled here) might find an alternate route elsewhere in the canyon into the VP1 core, or even persevere here by wiggling through tenaciously, despite the altered sequences.

Fig. 5.

Drug pocket pores. (A) A16 (1aym), (B) B14 (4rhv), and (C) C15 (model) space-filling surfaces show VP1 residues at the base of the canyon, guarding the entrance to the drug-binding pockets. The COOH tail of VP3 (red) and pleconaril (dark blue) orient these figures clockwise, 90° relative to Fig. 1, with the 5-fold to the right, and 2-fold to the left. (D) WebLogo depictions (Crooks et al., 2004) for the RV-A+B+C were tabulated for these residues. The putatively obstructive C15 Tyr₂₄₆ and its analogs are highlighted in purple.

2.5. C15 grown with drug

During sample preparation for crystallography, it is common to diffuse capsid-binding drugs into preformed virus arrays or to cocrystalize the materials. Should this not achieve sufficient occupancy for diffraction, the virus can be grown in the presence of the drug and then crystallized (Zhang et al., 2004). If capsid drugs were prevented from entering C15 when mixed with virus, it seemed reasonable that the impediment(s) might be resolved through co-culture conditions. Recombinant transcripts for A16 and C15 were transfected into HeLa cells in the presence or absence of BTA-188 then incubated for 24 h. After cell lysis and treatment with RNAse, qRT-PCR showed both preparations had produced stable virions. The progeny outputs (n=6) were similar whether or not the drug was present (Fig. 2D). But when the derived lysates were tested for infectivity to ALI cultures, only the pre-treated A16 samples were unable to replicate (Fig. 2E). Exposure to BT-188, even during the packaging of C15 progeny, did not inhibit their subsequent growth potential.

2.6. Predictions for RV-C drug reactivity

All of the Category-4 analogs in pockets and pores of the RV-A+B, as well as C15 Phe₁₃₂, are defined, important players in drug selection or binding, especially for pleconaril. The collective sequence changes, notably those involving polar residues (e.g. C15 Met₁₁₈ vs. B14 Ser₁₂₈, or C15 Thr₁₇₂ vs. B14 Val₁₉₁), and size changes (e.g. C14 Phe₉₆ for B14 Leu₁₀₆, or C15 Ile₁₆₉ for B14 Val₁₈₈, C15 Tyr₂₅₆ for B14 Gly₂₆₇) must surely change the underlying character of the RV-C VP1, let alone its structural flexibility. Short of an experimental C15 structure with an active drug, it is impossible to anticipate the relative contributions of these residues to the broader question of drug resistance. The observed Category-2+3 residues are not wholly common or predictive of Group-1 or Group-2 isolates either. Instead, the sequence modeling, residue comparisons, and obviously the drug tests themselves, suggest the RV-C should be generalized with novel “Group-3” reactivity. Without labeled drug tracers, which are beyond the scope of this study, it cannot even be determined whether BTA-188 (and presumably pleconaril) actually intercalates into the C15 pockets. Possibly these drugs are indeed bound, but their inclusion is simply irrelevant to this virus life cycle. The RV-C do not use ICAM-1 or LDLR receptors which probe deeply to the canyon floor and are therefore sensitive to VP1 deformations (Rossmann, 1994). Alternatively, the Group-3 pockets may just be more strongly dependent on the presence of the native pocket factor(s) than the RV-A+B. If the RVC preferentially incorporated and retained these factors, the essential, capsid-stabilizing properties would be relinquished only reluctantly and infrequently, to any drug replacement, even during protomer assembly. A putatively narrower or absent pore and the lack of reactivity even when C15 was grown in the presence of BTA-188 are consistent with this interpretation.

3. Materials and methods

3.1. Sequences and alignment

Refined alignments (337 seqs) of complete RV genome sequences are based on foundation superimposition of determined protein structures as described (Palmenberg et al., 2009; Basta et al., this issue). A translated polyprotein alignment with species, type and accession numbers are available in fasta and meg formats from http://virology.wisc.edu/acp/aligns/. The set includes RV-A (77 types, 203 seqs), RV-B (25 types, 69 seqs), RV-C (30 types, 65 seqs), and EV outgroups (4 species, 10 seqs). Current RV nomenclature designates the species letter (A, B or C), and type number (e.g. A16). Strain designations are unique to each accession number. Group-1 and Group-2 drug specificities are as defined (Andries et al., 1990). Briefly, Group-1 includes all RV-B plus A8, A13, A32, A43, A45, A54, and A95. Group-2 includes all remaining RV-A except A100–103, which like the RV-C were not tested in the original study. The amino acid numbering system is for C15 protein VP1 (GU219984), unless otherwise specified. Analogous residues for A16 (L24917) and B14 (L05355) in the same alignment column are referred to with their native (ungapped) sequence designations.

3.2. Molecular modeling

An I-TASSER model, with output PDB file, was predicted for the capsid of C15. It is primarily based on the structure of A16 (1aym) as has been described (Basta et al., this issue). Refined, protomer and virion coordinates (hrvc) are available from VIPERdb (http://viperdb.scripps.edu/). Subunit illustrations were rendered in MacPyMOL The PyMOL Molecular Graphics System (2008). UCSF Chimera (Pettersen et al., 2004) created full capsid structures and pentameric assemblies from protomer files. The PDB entries for enteroviruses (CB3, PV3) and RV-A+B with antiviral drugs or identified pocket factors resolved within their VP1 pockets, are summarized in Suppl. Table 1. Drug names, abbreviations, references and published MIC₅₀ are indicated. Each listed set of C15 coordinates was generated by I-TASSER then submitted online (http://endscript.ibcp.fr/) to Endscript (Gouet and Courcelle, 2002). The Crystallography and NMR System (CNS) module within this program identified lists of residues within 4 Å of any bound compounds for the RV-A+B (Fig. 5), including (53×) at least one iteration of every crystallized drug or pocket factor ligand. Additional evaluations extracted list(s) of C15 residues within 4 Å of any (putative) ligands, should the chimeric model(s) be representative of analogous interactions (see Suppl. Table S2).

3.3. Viruses

Recombinant A16 was produced in WisL cells (Lee et al., 1995). When required, titration was by plaque assay on HeLa cells. C15 virus was prepared by reverse genetics, transfecting full-length RNA transcripts into HeLa or WisL cells, then purifying and concentrating the progeny as described (Bochkov et al., 2011).

3.4. Drugs

WIN56291, WIN52035 and WIN52084 were provided by Dr. Wai-Ming Lee. Pleconaril and BTA-188 were provided by BIOTA (Australia). Working concentrations were prepared fresh for each experiment.

3.5. Drug standardization

Pleconaril, BTA-188 and W91 were evaluated for minimum inhibitory concentrations and toxicity using A16 in WisL cells grown in Medium A (Eagle’s Minimum essential medium with 1% non-essential amino acids, 5% fetal calf serum and pen/strep) in 12-well plates. The drugs were diluted to 1, 2, 5 or 10 μg/ml, and then incubated with virus (2×10⁷ RNA copies, 15 min, room temperature). Aliquots (200 μl) were added to cell-containing wells. After 15 min at room temperature, then 45 min at 34 °C, the cells were washed (3×) with PBS then Medium A, containing the respective concentrations of drug. Incubation continued at 34 °C for 24 h. Control wells without virus assessed potential effects of DMSO or the drugs alone. At 24 h PI, CPE was recorded, if evident. The cells were frozen and thawed three times to release particles. They were treated with RNase A (Qiagen) then assayed by qRT-PCR for RNA content as described (Bochkov et al., 2011).

3.6. Drug testing

Capsid-binding drug tests for C15 in human organ cultures have been described (Bochkov et al., 2011). For ALI-based tests, cultures of differentiated human sinus epithelium (HSE) or human bronchial epithelium (HBE) were grown in 12-well plates for 30 days in ALI medium (1:1 BEGM (Lonza) and DMEM (Mediatech) with additives, then screened for C15 growth before use (Ashraf et al., 2013). Successful cultures were gently washed (3×) with PBS to remove mucus. Virus samples (2×10^7–8 RNA copies of A16 or C15) were incubated with pleconaril (10 μg/ml, 26.24 μM), BTA-188 (2 μg/ml, 5.43 μM) or W91 (2 μg/ml, 5.63 μM) or DMSO (controls) in ALI medium (15 min at room temperature) then inoculated into culture wells at the apical ALI surface. 15 min later (room temperature) the plates were transferred to 34 °C for 3 h 45 min. For samples harvested at this time, the cells were washed with PBS (3×), then lysed in RLT buffer (350 μl, Qiagen). The remaining cultures were fed basally with ALI medium (1 ml with drug) and incubation continued (to 24 h at 34 °C). At harvest, the medium was aspirated from the outer wells and cells were lysed by RLT buffer. RNA extraction and qRT-PCR were as described (Bochkov et al., 2011). Virus titer is expressed as RNA copies (log₁₀) averaged from duplicate qRT-PCR values and n=2–6. Average error is indicated for each sample type.