Shifting Target

Research Article

735

Drug design against a shifting target: a structural basis for resistance to inhibitors in a variant of influenza virus neuraminidase Joseph N Varghese1*, Paul W Smith2, Steven L Sollis2, Tony J Blick1, Anjali Sahasrabudhe1, Jennifer L McKimm-Breschkin1 and Peter M Colman1 Background: Inhibitors of the influenza virus neuraminidase have been shown to be effective antiviral agents in humans. Several studies have reported the selection of novel influenza strains when the virus is cultured with neuraminidase inhibitors in vitro. These resistant viruses have mutations either in the neuraminidase or in the viral haemagglutinin. Inhibitors in which the glycerol sidechain at position 6 of 2-deoxy-2,3-dehydro-N-acetylneuraminic acid (Neu5Ac2en) has been replaced by carboxamide-linked hydrophobic substituents have recently been reported and shown to select neuraminidase variants. This study seeks to clarify the structural and functional consequences of replacing the glycerol sidechain of the inhibitor with other chemical constituents. Results: The neuraminidase variant Arg292→Lys is modified in one of three arginine residues that encircle the carboxylate group of the substrate. The structure of this variant in complex with the carboxamide inhibitor used for its selection, and with other Neu5Ac2en analogues, is reported here at high resolution. The structural consequences of the mutation correlate with altered inhibitory activity of the compounds compared with wild-type neuraminidase.

Addresses: 1Biomolecular Research Institute, 343 Royal Parade, Parkville, 3052 Australia and 2GlaxoWellcome Research and Development Limited, Stevenage, Herts, SG1 2NY, UK. *Corresponding author. E-mail: [email protected] Key words: antiviral, drug resistance, influenza, inhibitors, neuraminidase Received: 9 March 1998 Revisions requested: 25 March 1998 Revisions received: 6 April 1998 Accepted: 9 April 1998 Structure 15 June 1998, 6:735–746 http://biomednet.com/elecref/0969212600600735 © Current Biology Ltd ISSN 0969-2126

Conclusions: The Arg292→Lys variant of influenza neuraminidase affects the binding of substrate by modification of the interaction with the substrate carboxylate. This may be one of the structural correlates of the reduced enzyme activity of the variant. Inhibitors that have replacements for the glycerol at position 6 are further affected in the Arg292→Lys variant because of structural changes in the binding site that apparently raise the energy barrier for the conformational change in the enzyme required to accommodate such inhibitors. These results provide evidence that a general strategy for drug design when the target has a high mutation frequency is to design the inhibitor to be as closely related as possible to the natural ligands of the target.

Introduction Influenza is still a major disease of humans and some other animals. The pathogen, an orthomyxovirus, has a segmented negative-strand RNA genome that codes for two surface glycoproteins [1], one of which is the enzyme neuraminidase (NA) [2,3]. NA cleaves terminal sialic acid 1 (N-acetyl neuraminic acid; Figure 1) from glycoconjugates found on the surface of molecules on target cells in the upper respiratory tract of susceptible mammals [4,5]. These sialo-glycoconjugates are the receptors for influenza virus which binds to them via the viral haemagglutinin (HA) [6], the other surface glycoprotein on the virus particle. NA destroys the HA receptors, allowing the elution of progeny virus particles from infected cell surfaces [7] and preventing aggregation by HA of freshly synthesised viral glycoproteins

via sialylated carbohydrates. It is also thought that NA facilitates passage of virus through the protective mucin covering target cells by desialylation of the sialic acid rich mucin [8]. Following the determination of the X-ray structure of influenza virus NA in Type A [9,10] and Type B strains [11], there has been a renewed interest in using NA as a molecular target for anti-influenza drug design. Earlier attempts [12–15] to develop an NA inhibitor as an antiinfluenza drug, using a transition-state analogue Neu5Ac2en 2 (Figure 1), led to enzyme inhibition at micromolar levels. Drug design [3] based on the X-ray structure of influenza virus NA [16] and its complex with sialic acid and Neu5Ac2en [17] yielded two novel and potent NA inhibitors, 4-amino-Neu5Ac2en and 4-guanidino-Neu5Ac2en

736

Structure 1998, Vol 6 No 6

Figure 1 The chemical structures of influenza neuraminidase inhibitors: 1, sialic acid (Nacetylneuraminic acid (Neu5Ac)); 2, 2-deoxy2,3-dehydro-N-acetylneuraminic acid (Neu5Ac2en); 3, 4-amino-Neu5Ac2en; 4, Zanamivir, 4-guanidino-Neu5Ac2en; 5, 5-Nacetyl-4-guanidino-6-methyl(propyl) carboxamide-4,5-dihydro-2H-pyran-2carboxylic acid; 6, 5-N-acetyl-4-amino-6diethyl carboxamide-4,5-dihydro-2H-pyran-2carboxylic acid; and 7, GS4071, 4-N-acetyl-5amino-3-(1-ethylpropoxy)-1-cyclohexene-1carboxylic acid.

(3 and 4; Figure 1) [18,19], with Ki values in the nanomolar range. These inhibitors have been shown to be effective antiviral agents in cell culture against all type A and B strains tested [18,20–22]. In addition, one of these inhibitors 4-guanidino-Neu5Ac2en 4 (Zanamivir, GG167; Figure 1) has been shown to be effective in animal studies and clinical trials for the treatment of influenza virus infections in humans [23,24].

The broad spectrum efficacy of Zanamivir is believed to derive from one of its design features, that is, that it should (and does) interact only with residues in the active site of NA that are conserved in all currently known wild strains of influenza [16], dating back to the isolation of the virus in 1933 [25]. The discovery of effective NA inhibitors marks a turning point in the evolution of influenza virus because it is now possible to apply

Research Article Inhibitor-resistant variant of neuraminidase Varghese et al.

selection pressure on these conserved active-site residues, and some inhibitor- and drug-resistant mutants have emerged during multiple in vitro passaging of virus in the presence of drug [26–31]. An NA variant, E119G, that has an active-site mutation of Glu119→Gly, has reduced affinity to 4, and virus bearing this variant was able to replicate in vitro in the presence of 4 [28,29]. Alternatively, Zanamivir-resistant mutations were found to occur in HA around the sialic acid binding sites [27]. These mutations appear to lower the binding affinity of HA for sialic acid receptors on cells. Mutant viruses are able to elute from infected cells in the presence of 4 because of a lowered affinity of HA for sialic acid [27]. To date, however, no resistant virus has emerged in the course of normal clinical trials. Experience with the amantadine class of anti-influenza virus drugs has shown that viruses resistant to these drugs arise rapidly [32] by mutations in the haemagglutinin and in the ion channel protein M2. For the human immunodeficiency virus (HIV), drug resistance has been a severe problem, only recently addressed by using multidrug therapy [33–35]. A new class of NA inhibitors, exemplified by 5-N-acetyl4-guanidino-6-methyl(propyl)carboxamide-4,5-dihydro-2Hpyran-2-carboxylic acid (5; Figure 1) and 5-N-acetyl-4amino-6-diethyl carboxamide-4,5-dihydro-2H-pyran-2-carboxylic acid (6; Figure 1) has recently been reported, in which hydrophobic substituents have replaced the glycerol moiety at the 6-position [36–38]. A small conformational change in the active site of NA occurs to enable these inhibitors to be accommodated. Glu276 changes its position to form a salt link with Arg224, and thereby creates the necessary hydrophobic pocket for the carboxamide substituents [37,38]. A carbocyclic analogue of sialic acid 4-N-acetyl-5-amino-3-(1-ethylpropoxy)-1-cyclohexene-1-carboxylic acid (GS4071) [39] (7; Figure 1) which has a hydrophobic group attached to the 6-position

737

via an ether link, has also been shown to inhibit NA and virus replication in vitro. Resistance to the carboxamide inhibitor 5 (Figure 1) has been investigated in type A influenza with the N9 subtype NA, and an active-site mutant NA, Arg292→Lys (R292K), has been isolated [30]. The specific activity of the R292K variant is only 20% of that of wild type, and virus bearing this mutant NA shows decreased sensitivity to all NA inhibitors [30]. The R292K mutant has also been reported to arise in an avian N2 background after passaging in the presence of Zanamivir [40]. Here we present structural studies of this N9 mutant viral NA complexed with the inhibitors 1–7. These studies suggest a structural basis for the differential resistance of the R292K variant to all of these inhibitors. They also provide evidence for a design principle for drugs directed at moving targets.

Results The three-dimensional structures of 1–7 complexed with wild-type N9 NA and the R292K variant of N9 NA have been determined by X-ray crystallography. For the complexes reported here, inhibitors were soaked into crystals of wild type or R292K NA at 18°C and flash frozen to –166°C in a cold nitrogen stream (except for the complex of 5 with wild type, which was collected at 18°C). Details of the experimental conditions and determination of the atomic structure of these complexes are given in the Materials and methods section. The R292K variant structure

The mutation of Arg292 to lysine results in a very local change of structure when compared to the wild-type enzyme (Figure 2). In the wild-type NA, Arg292, together with Arg118 and Arg371, engages the 2-carboxylate group of sialic acid in the active site and is partly responsible for distorting the pyranose ring from a chair to a boat

Figure 2 Stereo drawing of the local region around residue 292 in the active site of R292K (yellow) and wild-type (green) N9 NA. The dotted lines represent noncovalent interactions of residue 292 in the local environment (yellow for the R292K mutant, and green for the wild type). Two water molecules, W1 and W2, that appear in the mutant structure are also shown. These water molecules are positioned near the locations of the distal guanidinyl nitrogens of Arg292 in the wild-type enzyme. The mainchain atoms are represented by a tube.

E276

E276

N294

N294 E277

E277 K292

K292

W2

W2

W1

W1

Y406 R371

Y406 R371

Structure

738

Structure 1998, Vol 6 No 6

conformation [17]. The distal primary guanidinyl amide of Arg292 interacts with Glu277 and the tyrosyl oxygen of Tyr406 (Figure 2), and the other two guanidinyl nitrogens interact with Asn294. The guanidinyl group of Arg292 is parallel with the peptide plane of mainchain atoms of residues 348–349. Two water molecules in the active-site cavity form hydrogen-bond interactions to the primary amides of Arg292 before the entry of substrate into the active site. In the mutant structure, the amino group of Lys292 forms an ionic interaction with Glu276. In wild-type N9 NA this interaction is insignificant, as the guanidinyl nitrogens of Arg292 are some 4 Å distant. In the variant, the distal guanidinyl nitrogens in the wild-type Arg292 are replaced by two water molecules (W1 and W2 in Figure 2) about 1 Å from these nitrogen atoms. W1 connects the lysyl amino to Tyr406, Glu277 and Arg371 via hydrogen bonds, and W2 mediates the interaction between Asn294, the carbonyl oxygen of residue 348 and the amino group of Lys292. In this conformation the distance of the lysine nitrogen to the Oδ oxygen of Asn294 increases by about 0.8 Å, and the distance from the carboxylate oxygen of Glu277 increases by about 0.3 Å when compared with the guanidinyl nitrogens of the wild-type structure. Some of the minor changes in the conformation of the local region around Lys292 are a 0.6 Å movement of the hydroxyl of Tyr406 away from Lys292, a 0.3 Å movement of the mainchain oxygen of Gly348 closer to the nitrogen of Lys292, and a 0.4 Å movement of the carboxylate oxygen of Glu276 towards Lys292.

The loop around residue 246 has a slight movement (0.2 Å) away from Lys292, and the water molecules in the neighbourhood of Lys292 are perturbed by about 0.5 Å from their positions in the wild-type NA. Complexes of R292K with compounds 2–4

When complexed with the R292K N9 NA, Neu5Ac2en, 4-amino-Neu5Ac2en and Zanamivir (2–4) are positioned almost identically in the active-site pocket. Interactions with NA are similar to those in complexes with wild-type enzyme [41], especially for the C4 and C5 substituents, with some differences around the 2-carboxylate where a water molecule is now interposed between that group and Lys292. Small differences are also observed at the site of interaction of the 6-glycerol group with Glu276, all attributable to the Lys292–Glu276 salt link in the variant. In the wild type, the glycerol group binds deeper into the pocket, moving about 0.3 Å towards the floor in the direction of the Arg292 and Glu276 residues. In the mutant enzyme, the Lys292 amine nitrogen directly interacts with the O8 hydroxyl of the 6-glycerol group and a carboxylate oxygen of Glu276 (which moves 1.1 Å from the position in the wild-type enzyme, while keeping the other carboxylate oxygen approximately in the same position interacting with the O9 hydroxyl of the glycerol) preventing the glycerol from penetrating as deeply as in the wild type (see 3 in Figure 3a). The glycerol group of Neu5Ac2en in the mutant, however, binds about halfway between its position in the wild-type enzyme and its location in the complexes of 3 and 4 in the mutant enzyme (Figure 3b).

Figure 3

(a)

A246

A246

R37

R3 R224 H274

R224 H274 E276

Y E277

E276

Y E277

(b)

Structure

The binding of sialic acid and O4-substituted Neu5Ac2en analogues in the active sites of the wild-type and R292K mutant of N9 NA. (a) Stereo drawing of the 4-aminoNeu5Ac2en inhibitor complexed with the wildtype (yellow) and the R292K mutant of N9 NA (blue), superimposed on each other in the region of the inhibitor-binding pocket. Dotted green lines represent the nonbonded interactions of the 6-triol of 4-aminoNeu5Ac2en and the active-site residues in the mutant complex. (b) Stereo drawing of the superposition of sialic acid, Neu5Ac2en, 4amino- and 4-guanidino-Neu5Ac2en complexed in the wild-type N9 NA (yellow), and in the R292K mutant (red). The Neu5Ac2en–R292K N9 NA complex is shown in green; atoms are in standard colours. The classes of binding are evident. All of the compounds fail to penetrate the active site of the mutant as deeply as they do that of the wild type.

Research Article Inhibitor-resistant variant of neuraminidase Varghese et al.

739

Figure 4 Stereo view superimposition of the carboxamide inhibitor 5 in the active site of the R292K mutant NA structure (blue) and the uncomplexed R292K mutant NA structure (yellow). Note the movement of the carboxylate of Glu276, from pointing into the active-site cavity to forming a salt bridge with Arg224 in the complex, and the hydrophobic pocket with which the 6-carboxamide group interacts. Hydrogen bonds are shown as green dotted lines.

R224

R224

A246

R152

H274

A246

R152

H274

E276

E276

N294

N294 K292 W2

D151

E277

W1

K292 W2

D151

E277

W1

R371

Y406

R118

R371

R118 Y406

Structure

Complexes of R292K with compounds 5 and 6

The inhibitor 5, which was used to select R292K, binds in a similar manner to the variant as do other carboxamides to wild type [37,38]. The 4-guanidino group of 5 interacts with Glu119, Glu227 and Asp151, together with mainchain carbonyl oxygens of residues 178 and 151. The nitrogen of the 5-N-acetyl group interacts with a water molecule, the carbonyl oxygen with Arg152, and the methyl group with Trp178. All of these interactions are isosteric with those of Zanamivir in complex with the R292K variant (see above). The carboxamide sidechain of 5 forms different interactions with the variant protein than does the glycerol of the natural substrate, however. The propyl group of 5 displaces three water molecules in the uncomplexed enzyme, and the methyl group forms a hydrophobic interaction with Glu276. A hydrophobic pocket is created by the formation of an internal salt bridge between Glu276 and Arg224 (Figure 4). This allows the methyl

and propyl groups of the carboxamide respectively to enter the cis and trans hydrophobic pockets bounded by the Glu276–Arg224 salt link, Ile222 and Ala246. In the uncomplexed enzyme, Lys292 forms an ionic interaction with a carboxylate oxygen of Glu276 (3.1 Å), the other carboxylate oxygen forming a hydrogen bond (2.8 Å) with a primary guanidinyl nitrogen of Arg224. This lysine–glutamate interaction is broken in the complex, whereas the carboxylate oxygen forms an ionic interaction with the secondary guanidinyl nitrogen of Arg224 and the Nε2 of His274. This involves a movement of 2.5 Å of this carboxylate oxygen on forming the complex. The other carboxylate oxygen of Glu276 retains its interaction with a primary guanidinyl nitrogen of Arg224. Compound 6 is some ten times less effective than 5 as an inhibitor of wild-type NA (Table 1). It nevertheless forms very similar complexes with both wild type and R292K to those observed with 5. The salt link between Glu276 and Arg224 is formed and the ethyl substituents enter the

Table 1 Resistance indices for inhibitor binding to the R292K variant versus wild type. Inhibitor*

1 2 3 4 5 6 7

Ki (mM) wild type/variant

Ki resistance

IC50 (mM) wild type/variant

IC50 resistance

55/1820 2.64/280 0.148/14 0.002/0.033 0.004/2.160 nd nd

33 106 95 16 540 nd nd

nd 20/410 0.075/2.5† 0.002/0.11† 0.02/5.26† 0.23/230 0.002/13

nd 20 33† 55† 263† 1000 6500

*Inhibitors are as defined in Figure 1. The index is defined as the ratio of the Ki (or IC50) for the variant to the Ki (or IC50) for the wild type. Ki and IC50 values from [30], except for the values marked with a dagger

which have been determined according to methods described in [30]. nd, not determined.

740

Structure 1998, Vol 6 No 6

newly created hydrophobic environment. The weaker binding of 6 compared with 5 may relate to energy penalties associated with accommodating the bulkier ethyl group in the cis pocket. In all complexes the plane of the carboxamide is normal to the plane of the pyranose, and this preferred orientation restricts the location of both the cis and trans substituents of the carboxamide. Complexes of R292K and wild-type NA with compound 7

The X-ray structure of a complex of GS4071 (7) and N9 NA has recently been determined at 2.8 Å resolution [39]. We have extended the resolution to 2 Å and note that 7, a carbocyclic sialic acid analogue, binds to the active site of wild-type NA in a similar way to the carboxamide inhibitor 6. The hydrophobic pentyl ether group of 7 binds in the hydrophobic pocket created by the rotation of Glu276 to form the salt link with Arg224. This result is different from the previous structure [39] in that the orientation of the carboxylate of Glu276 is reported to be slightly different, and the distances of the pentyl group from Glu276 are longer, probably a result of the lower resolution of that structure. In wild type, 6 and 7 display significantly different structures as a consequence of the planar carboxamide in 6 and its preferred orientation with respect to the ring. Thus one of the ethyl groups of 6 is more deeply buried than its counterpart in 7. Furthermore, the two ring structures are slightly displaced from each other (Figure 5). In the mutant enzyme, 7 fails to induce the formation of the internal salt link between Glu276 and Arg224. The hydrophobic pocket that receives the pentyl ether group in the wild type is thus not established in the mutant. This might be attributed to Glu276 being more tightly anchored to the mutant Lys292 than to the wild-type Arg292. As a consequence of the hydrophobic pocket not being formed, the pentyl ether group fails to enter the site

as deeply as it does in the wild type, but interactions with the carboxylate, the amino group and the acetamido group of the inhibitor with the enzyme are retained (Figure 6). The comparison with 6 bound to the mutant further illustrates the subtlety of the system, because in this complex the internal salt link is still formed, despite the interaction of Lys292 with Glu276. One is left to argue the importance of the carboxamide as a linking structure which, because of its preferred orientation normal to the pyranose ring, can ‘force’ the entry of one of the ethyl groups into its hydrophobic pocket. Differential resistance of inhibitors to R292K

Table 1 summarises the published data [30] on the inhibitory properties of the compounds in Figure 1 against wild-type and variant (R292K) NA activity. These data are either in the form of IC50 values or Ki values. We define a ‘resistance index’ as the ratio of the Ki (or IC50) for the variant to that for the wild type, and observe that resistance increases through the series of compounds 1–7, in accordance with their decreasing similarity to the transition-state analogue 2. Thus influenza viruses bearing the NA mutant R292K can be expected to have higher ‘resistance’ to 7 than to other inhibitors studied here. The complex of R292K N9 NA with compound 1

X-ray diffraction analysis of crystals, flash frozen to –166°C, of the R292K N9 NA with 20 mM Neu5Ac (1) soaked for 4 h at room temperature (18°C), revealed two sialic acid moieties bound to the NA. The first is in the conserved active site in a twisted boat conformation as reported previously [17], except that the interactions between the R292K enzyme and the carboxylate are as described for inhibitors 2–5; that is, a water molecule has been introduced between Lys292 and the carboxylate, compared with the direct interaction with Arg292 in the wild type.

Figure 5

R152

A246

R152

A246

R224

R224

H274

H274 N294

N294 D15

E276

E277 R371

D151

E276

R371 E277

R118

R118

E Structure

Y406

E Y406

Stereo view overlay of the wild-type complexes with 6 (blue) and 7 (yellow). Note that the orientation of the carboxamide in 6 results in one of its ethyl groups penetrating the pocket more deeply than its counterpart in 7. Dotted red lines represent hydrogen bonds; green dotted lines represent noncovalent interactions.

Research Article Inhibitor-resistant variant of neuraminidase Varghese et al.

741

Figure 6 Stereo drawing of inhibitor 7 complexed with the wild-type (yellow) and the R292K mutant (blue) of tern N9 NA. The structures were superimposed on each other in the region of the inhibitor-binding pocket. Interactions of 7 with the wild-type active-site residues are shown as green dotted lines, and interactions of Glu276 in the R292K mutant are shown as red dotted lines.

R152

A246

R152

A246

R224

R224

H274

H274

N294

N294 D15

E276

E277 R371

D151

E276

R371 E277

R118

R118

E Structure

The second sialic acid binding site is observed in a chair conformation similar to that found in the sialic acid binding sites of HA [42]. The location of this second site is the same as that observed in wild type [43] soaked in 1 for 4 h at 4°C. This site is unoccupied in the room temperature soak in the wild-type enzyme. The location of this site had been predicted from the properties of antibody-selected variants [44] in wild-type A/tern/Australia/G70C/75 N9 NA [45], and in A/FPV/Rostock/34 N1 NA [46]. The structure of the wild-type NA–Neu5Ac complex at the second site of NA for the 4°C soak is similar to the structure of the R292K N9 NA–Neu5Ac complex in the 18°C soak (root mean square deviation [rmsd] of 0.12 Å over residues 367 to 373, 400 to 403 and 432 in the neighbourhood of the second site). However, four additional water molecules were found interacting with 1 in the second site in the wild type, and there was tighter electron density of the Neu5Ac moiety as a result of the higher resolution of the wild-type N9 NA–Neu5Ac complex. This observation is not central to the current findings on drug-resistant variants, but does again illustrate that subtle effects accompany mutations. The Arg292→Lys substitution is 14 Å away from the second binding site and yet it modulates its affinity for sialic acid such that room temperature (18°C) binding is now observed, whereas in the wild type low temperature (4°C) is required to capture sialic acid.

Discussion The active site of influenza virus NA has been conserved in all currently known field strains of the virus [10,16], including the recently described avian influenza in humans [47]. Evidently, it is not necessary to alter the active-site residues to produce variants that can escape antibody binding. It has been proposed [2] that antibodies are unable to exert selection pressure on the active site because an antibody footprint is typically larger than the

Y406

E Y406

active site, and mutations occurring outside the active site are sufficient to abolish binding [48–51]. It has also been suggested [41] that the highly conserved active site is a result of the virus optimising its ability to spread in the environment of the upper respiratory tract. This involves overcoming immobilisation in the protective mucin layer, and facilitation of the elution of progeny from infected cells. This would require a balance of the rate of desialylation by NA and the rate of attachment by HA to sialylated glycoconjugates, implying that alterations in the enzyme characteristics of NA could be critical to the viability of the virus; hence the highly conserved nature of the enzyme active site. The selection in vitro by a NA inhibitor of virus with HA mutations and decreased receptor-binding affinities of HA [27] is further indication of the interplay between HA binding and NA activity. The NA inhibitors Zanamivir 4 [18] and 5 [36–38] do apply selection pressure in vitro on the enzyme activesite residues. The mutant Glu119→Gly (E119G) was selected after serial passage of the virus in the presence of Zanamivir. This variant confers resistance by affecting interactions of the 4-guanidinium group of Zanamivir with Glu119 [29] while at the same time retaining the interactions with the natural substrate sialic acid. E119G has the catalytic activity of wild type but its stability is compromised [52]. The R292K mutant has arisen after passage of virus in 5 [30]. It has also been selected by 4 in a different genetic background (H4N2) [40]. This indicates that the R292K mutation successfully reduces the binding of both 4 and 5 in the mutant enzymes, implying that the mutation of Arg292→Lys alters the binding not only of the 6-carboxamide substituent but also of other elements of 5 and 4, respectively. In fact, compared with wild-type NA, the R292K variant exhibited reduced binding to all of the inhibitors studied here (see Table 1). For the purposes of

742

Structure 1998, Vol 6 No 6

this discussion, the inhibitors can be classified into one of three groups depending on the nature of the substituent at position C6: the first class have a glycerol sidechain at position C6 (1, 2, 3 and 4); the second a carboxamide derivative (5 and 6); and the third an ether-linked derivative (7). The consequences of the mutation at Arg292 for inhibitor binding appear to fall into two categories. Firstly, and common to all inhibitor groups, is the effect of the substitution on interactions with the inhibitor carboxylate. The hydrogen bonding of Arg292 in the wild type to the inhibitor carboxylate is replaced for all inhibitors by hydrogen bonding of Lys292 via a bound water molecule in the variant. This water molecule (W2) is also a structural feature of the unliganded variant. This suggests that the common effect for all inhibitors is that the energy of the interaction between the inhibitor carboxylate and the protein is smaller in the variant than in the wild type. This may, at least in part, also explain the decrease in the specific enzyme activity to 20% of wild-type values [30]. Secondly, but differently for each of the three groups, is the effect of altering the interactions in the pocket which accommodates the sidechain at C6. For the first inhibitor group (1–4), interactions between Lys292 and the C8-hydroxyl prevent the inhibitors from settling as deeply into the binding site as they do in the wild type. Estimating the energetic effect of this is difficult because of the balancing effect of somewhat higher solvent exposure of the inhibitor with the enthalpic gain of an additional hydrogen bond (to Lys292) of the C8-hydroxyl. The net effect of the mutation on the glycerol-containing inhibitors is to reduce binding by one to two orders of magnitude (see Table 1). For the second inhibitor group (5,6), the conformation and position of the carboxamide is identical for wild type and variant, but in the case of the variant, the salt link between Lys292 and Glu276 must be broken in order to create the hydrophobic binding pocket for the carboxamide substituents. Glu276 undergoes a similar conformational change in wild type, but its movement is not constrained as it is in the variant because Arg292 does not interact directly with Glu276. The penalty associated with breaking the Lys292–Glu276 interaction and loss of hydrogen-bond interactions by Glu276 and Lys292 with the glycerol compared with only van der Waals interaction with the carboxamide group, may contribute to an additional loss of binding of 5 and 6 to the variant when compared with the group 1 inhibitors; that is, two to three orders of magnitude compared with one to two orders of magnitude (see Table 1). Finally, the C6 substituent of 7 does not bind in the same position, or with the same conformation, to wild type and variant. In the wild type, the conformational change of Glu276 occurs unimpeded but, in the variant, the interaction between Glu276 and Lys292 is not disrupted and the binding pocket for the pentyl ether group is not created. This

manifests itself as a further compromise to binding (see Table 1), as 7 is unable to cause the necessary conformational changes to accommodate the pentyl ether group. There is a qualitative correlation between the effect of the Arg292→lysine substitution on the structure of complexes with inhibitors and on the binding properties of the inhibitors. The largest structural consequences occur with the inhibitors whose binding is most severely compromised, that is 7 and then 6 and 5. Smaller structural alterations are associated with inhibitors that retain the glycerol sidechain, and these compounds are only marginally impaired as inhibitors as a consequence of the Arg292 replacement by lysine. Even more interesting is the qualitative correlation evident in Table 1 between the degree of resistance of the variant over wild type and the degree of structural dissimilarity of the inhibitor to the transition-state analogue 2. Structural similarity is a complex issue and can be measured in many ways. Here we note without quantification that glycerolcontaining inhibitors are less compromised in their binding to R292K than are nonglycerol-containing inhibitors. Insofar as Lys292 interacts directly with the glycerol moiety, or with C6 substituents of inhibitors, there is some structural logic to this observation as well as some implication for drug design against rapidly mutating targets. Microorganisms develop resistance to drugs in a variety of ways, only the simplest of which we consider here, and that is by the selection of point mutations. Even in such cases, unexpected complexities arise such as the appearance of a mutation in a protein that is not the drug target, for example the HA variants selected with the NA inhibitors [27]. Thus, our discussion of drug resistance, and how it might be addressed at the level of drug design, is limited to resistance arising through point mutations in the target. In this restricted context, successful drug-resistant variants are characterised by a high resistance index (as defined in Table 1) for the drug, with retention of the essential functional attributes of the wild type. The R292K variant has retained only 20% of wild-type enzyme activity, but its high resistance to 7, 6 and 5 suggests that this mutant will be more successful against these inhibitors than against those that retain a glycerol moiety at C6. We have also examined the resistance index of the inhibitors in Figure 1 for the NA variant E119G, which was selected with 4, and find that only compounds 4 and 5 are significantly resistant (approximately 200-fold) for this variant. Clearly, resistance is a function of both the inhibitor and the variant. The fact that only inhibitors with a guanidinium substitution at C4 are resistant to a variant at position 119 that interacts directly with the

Research Article Inhibitor-resistant variant of neuraminidase Varghese et al.

guanidinium is consistent with the principle that inhibitors that are more closely related to the natural substrate (or transition-state analogue) will be less resistant than others. With respect to the E119G variant, alterations to the glycerol pocket have little effect on resistance. The introduction of new drugs to combat infectious microorganisms often results in the emergence of drug-resistant variants, and in some cases the point mutations that give rise to the resistant phenotype are well documented but poorly understood. For influenza virus and the drug amantidine, there is no known correlation between the drug and any natural ligand for the sites to which it binds. Variants arise rapidly, both in vivo and in vitro [32]. Inhibitors of the HIV reverse transcriptase are of two types, nucleoside and non-nucleoside analogues, which bind respectively in the dNTP-binding site and remotely to that site [53]. Analysis of inhibitor-resistant mutations is complicated by incomplete understanding of the interactions with natural substrates. HIV protease inhibitors also lead to drug resistance with quite complex results, affecting either the binding of the inhibitor or the inherent catalytic rate of the enzyme [54]. These consequences can be compensatory. Comparison of the drugs for similarity with substrate is complicated by the promiscuity of the cleavage-site specificity [55,56] and also the dynamics of the system, which requires large flap movements of the protease to allow substrate entry to the active site.

743

antiviral drug therapy. The RNA genome of influenza virus encodes two surface glycoproteins: neuraminidase (NA) and haemagglutinin (HA). NA cleaves terminal sialic acid from glycoconjugates found on the surface of target cells. These sialo-glycoconjugates are the receptors for influenza virus which binds to them via the viral HA. Several anti-influenza drugs, for example Zanamavir, have been designed to inhibit viral NA. Two drug-resistant mutations (Glu→Gly and Arg→Lys) in the conserved active site of NA, have been observed after multiple passages of the virus in vitro in the presence of these inhibitors. These mutations have resulted in lower thermostability, in the case of the E119G mutation, and altered NA activity in the R292K mutation. This raises the question as to whether drug-resistant mutations can arise that leave the virus viability unchanged, and why only a few mutations have occurred to date. As no field strain is known to have mutated in the active-site pocket for the past 65 years, it could be inferred that selection pressure is operating to balance the rate of desialylation of the viral environment compared with the rate of attachment of the HA to sialic acid containing receptors. This balance is altered by NA inhibitors, and consequent selection pressure on the virus leads to the emergence of resistant strains in vitro.

In conclusion, both the structural and binding effects of the Arg292→Lys mutation are most pronounced on inhibitors that least resemble the natural ligand. This suggests that inhibitor design, at least for targets that have high mutation frequency, should aim to retain as many features as practicable of the natural ligand. The principle behind such a design strategy is simple — variants in the natural ligand binding site must retain some binding to that ligand to preserve the protein function.

Viable variants, selected under drug pressure, will generally retain wild-type function. This requires retention of binding interactions with the functional ligands at the same time as loss of binding interactions with the drug: such an outcome will be more likely if the drug and the natural ligand bind to the target in chemically different ways. The molecular interplay between the target site, the natural ligand, and the inhibitor or drug, is reminiscent of interactions between macromolecules in which two different proteins can form complexes with a common target site on a third protein [51]. The determinants of binding between macromolecules, and between macromolecule and small ligand, are chemically degenerate. The data and argument presented here suggest that an optimum strategy for drug design in the face of high mutation frequency of the target is a minimal strategy that preserves as many structural features of the natural ligand as practicable. It is not sufficient to target evolutionarily conserved residues in the ligand-binding pocket. This work provides evidence that one should aim to target drugs via chemical interactions that resemble, as closely as possible, those used by the natural ligand. Such a strategy minimises the possibility of mutations, selectively abrogating interactions with the drug while retaining interactions with natural ligands that are required for protein function.

Biological implications

Materials and methods

Understanding the ability of a virus to become resistant to drugs that interfere with its biological activity is of great importance to the development of sustainable

The R292K N9 NA mutant of the NWS/G70C virus was isolated by serial passage of virus in the presence of 5 [30] in MDCK cells. The N9 NA from influenza virus A/NWS/G70C and the R292K N9 NA mutant virus were purified as described previously [57] and crystallised by

None of the complicating features found in these other systems, where there is nevertheless a wealth of drugresistance data, is present in the system we describe here. Thus, although the design principle of minimal departure from the natural ligand is, in retrospect, self-evident, we believe these data provide the first experimental demonstration of its validity. Design features that will conflict with this principle include the desirability of creating ligands with higher affinity for the target than the natural ligand and in some cases the requirement for selectivity of a drug for a microbial target over the homologous protein in the host.

744

Structure 1998, Vol 6 No 6

Table 2 Data collection conditions and statistics for X-ray diffraction from R292K (mt) and from mutant and wild-type (wt) N9 NA complexes with inhibitor. Crystal/ inhibitor* mt mt/1 mt/2 mt/3 mt/4 mt/5 wt/5* mt/6 wt/6 mt/7 wt/7

Number of observations

Unique data

Resolution (Å)

Cell edge (Å)

Rsym

Rmerge

Soaking conditions

156,483 163,196 250,274 187,582 227,793 133,428 111,968 330,993 106,146 170,194 154,336

51,358 35,098 48,031 37,126 29,925 36,084 28,605 45,702 32,767 33,424 37,216

1.7 1.8 1.6 2.0 2.0 1.9 2.0 1.8 2.0 1.8 1.8

181.37 180.73 180.79 180.91 180.85 181.41 182.80 180.89 180.86 180.98 180.95

0.074 0.099 0.079 0.077 0.084 0.075 0.060 0.074 0.121 0.076 0.088

0.076 0.100 0.089 0.089 0.099 0.077 0.087 0.073 0.125 0.081 0.092

None 20 mM 4 h 20 mM 4 h 20 mM 18 h 20 mM 3 h 10 mM 18 h 10 mM 18 h 10 mM 1.5 h 10 mM 1 h 20 mM 1 h 10 mM 1 h

*Inhibitors 1–7 are as defined in Figure 1; mt, mutant, wt, wild type. All soaks were carried out at 18°C, and data were collected from crystals

flash frozen to –166°C. The data set was collected at the Photon Factory (Tsukuba, Japan) at 18°C.

established procedures [45]. The crystals were transferred to 20% glycerol while maintaining the concentration of the phosphate buffer prior to freezing in a cold stream of nitrogen gas at –166°C. X-ray diffraction data were collected on a Rigaku R-axis II Image Plate X-ray detector mounted on a MAC Science SRA M18XH1 rotating anode X-ray generator, operating at 47kV and 60mA with focussing mirrors. One dataset, compound 5 complexed with wild type, was collected at room temperature on beamline 6A2 at the Photon Factory. The soaking conditions of the inhibitor–NA complexes are given on Table 2, together with the X-ray data collection statistics.

active site removed, as a starting atomic model, using the crystallographic refinement program X-PLOR [60]. Water molecules in the active site and elsewhere were added by examining the difference Fourier peaks of observed and calculated structure factors of the atomic model that were greater than 5σ. The final refinement statistics are given in Table 3.

The structure of the R292K mutant NA was determined by examination of the difference Fourier of the X-ray data from crystals of R292K mutant and wild-type N9 using phases obtained from the refined N9 structure [29,58,59]. This resulted in the largest negative peak (–10.3σ) at the guanidinium group of Arg292 and the largest positive peak (+8.3σ) near Arg292 corresponding to the Nζ of a lysine substituted for Arg292. This was consistent with a mutation of R292K. Apart from some water molecules near the lysine mutation, no other large differences were noted, indicating that this mutation was the only one in the N9 NA head of the mutant NWS/G70C virus. The structure was then refined to 2 Å resolution using the wildtype structure modified by the lysine substitution at position 292, and the water molecules in the

The structure of the inhibitor–R292K mutant NA complexes were determined by examining the difference Fourier of structure factors from the inhibitor complexes and the uncomplexed R292K mutant NA crystals, using the phases of the uncomplexed refined structure with active-site water molecules removed. The inhibitors appeared as a large positive feature in the difference Fourier (> 5σ) which clearly showed the location of all the atoms in the inhibitor, and active-site water molecules. Atomic models were built from these difference Fourier features, and refined by X-PLOR. Additional waters and/or modifications to sidechain orientations were carried out in the atomic models as in the refinement of the uncomplexed structure described above, and the structures refined to 2 Å resolution or higher. Refinement statistics of these inhibitor complexes are given on Table 3. In all the above X-PLOR refinements all charges in charged amino acids were set to zero, as well as charged groups on the inhibitors. The stereochemical restraints used were of Engh and Huber [61], and all the structures retained good stereogeometry (see Table 2).

Table 3 Structure refinement statistics for R292K (mt) and wild-type (wt) N9 NA and inhibitor complexes. Structure*

mt mt/1 mt/2 mt/3 mt/4 mt/5 wt/5 mt/6 wt/6 mt/7 wt/7

Rms† deviations from ideal

Resolution

Selected

Final

(Å)

reflections

R factor

Bonds (Å)

6–1.7 6–2.0 6–1.6 6–2.0 6–2.0 6–1.9 6–2.0 6–1.8 6–2.0 6–1.8 6–2.0

47,115 28,729 46,535 30,018 25,513 34,738 25,550 39,147 27,445 31,985 30,431

0.171 0.166 0.173 0.165 0.149 0.167 0.184 0.183 0.177 0.172 0.172

0.013 0.013 0.014 0.013 0.013 0.014 0.015 0.013 0.014 0.013 0.014

*Inhibitors 1–7 are as defined in Figure 1; mt, mutant, wt, wild type. †Rms, root mean square.

Angles (°) 1.82 1.91 1.81 1.87 1.86 1.89 1.96 1.93 2.00 1.93 1.97

Research Article Inhibitor-resistant variant of neuraminidase Varghese et al.

Figure 7

1.0 0.9 0.8

Completeness

R or C

0.7 0.6 0.5 0.4 0.3

R factor

0.2

0.25 0.20 0.15 0.10

0.1 0

1/d 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 Structure

Completeness of X-ray data and the final R factors of the refined atomic models as a function of resolution 1/d (where d is the resolution in Å) of the R292K mutant NA and complexes with inhibitors (solid lines). The traces are colour-coded: native, black; 1, Neu5Ac grey; 2, Neu5Ac2en blue; 3, 4-amino-Neu5Ac2en orange; 4, Zanamivir red; 5, green; 6, purple; and 7, yellow. Inhibitor–wild-type complexes are represented as dashed lines. The blue dashed lines represent the Luzzati error contours from 0.1 to 0.3 Å in steps of 0.05 Å.

The completeness of the data and the final R factor as a function of resolution is given in Figure 7.

Accession numbers Coordinates for the 11 structures have been deposited with the Brookhaven Protein Data Bank (accession numbers 2QWA, 2QWB, 2QWC, 2QWD, 2QWE, 2QWF, 2QWG, 2QWH, 2QWI, 2QWJ, 2QWK).

Acknowledgements We thank Bert van Donkelaar and Pat Pilling for technical assistance, N Sakabe for assistance at the Photon Factory, Tsukuba, Japan, the Australian National Beam Line Program for travel support and Biota Holdings Ltd and Glaxo-Wellcome for support.

References 1. Lamb, R.A. (1989). Genes and proteins of the influenza virus. In The Influenza Viruses. (Krug, R.M., ed.), pp. 1–87, Plenum, New York. 2. Colman, P.M. (1994). Influenza virus neuraminidase: structure, antibodies, and inhibitors. Protein Sci. 3, 1687-1696. 3. Varghese, J.N. (1997). The design of anti-influenza virus drugs from the X-ray molecular structure of influenza virus neuraminidase. In Structure-Based Drug Design: Diseases, Targets, Techniques and Developments, Vol 1 (Veerapandian P., ed.), pp. 459-486, Marcel Dekker, New York. 4. Klenk, E., Faillard, H. & Lempfrid, H. (1955). Uber die enzymatische Wirkung von Influenzavirus. Z. Physiol. Chem. 301, 235-246. 5. Gottschalk, A. (1957). Neuraminidase: the specific enzyme of influenza virus and Vibrio cholerae. Biochim. Biophys. Acta 23, 645-646. 6. Ward, C.W. (1981). Structure of influenza virus hemagglutinin. Curr. Top. Microbiol. Immunol. 94/95, 1-74. 7. Palese, P., Tobita, K., Ueda, M. & Compans, R.W. (1974). Characterisation of temperature sensitive influenza virus mutants defective in neuraminidase. Virology 61, 397-410.

745

8. Burnett F.M. (1947). Mucins and mucoids in relation to influenza virus action. IV. Inhibition by purified mucoid of infection and haemagglutinin with the virus strain WSE. Aust. J. Exp. Biol. Med. Sci. 26, 381-387. 9. Varghese, J.N., Laver, W.G. & Colman, P.M. (1983). Structure of the influenza virus glycoprotein antigen neuraminidase at 2.9 Å resolution. Nature 303, 35-40.22. 10. Colman, P.M., Varghese, J.N. & Laver, W.G. (1983). Structure of the catalytic and antigenic sites in influenza virus neuraminidase. Nature 303, 41-44. 11. Burmeister, W.P., Ruigrok, R.W.H. & Cusack, S. (1992). The 2.2 Å resolution crystal structure of influenza B neuraminidase and its complex to sialic acid. EMBO J. 11, 49-56. 12. Meindl, P. & Tuppy, H. (1969). 2-Deoxy-2,3-dehydrosialic acids. I. Synthesis and properties of 2-deoxy-2,3-dehydro-N-acetylneuraminic acids and their methyl esters. Monatsh. Chem. 100, 1295-1306. 13. Meindl, P., Bodo, G., Palese, P., Schulman, J. & Tuppy, H. (1974). Inhibition of neuraminidase activity by derivatives of 2-deoxy-2,3dehydro-N-acetylneuraminic acid. Virology 58, 457-463. 14. Palese, P., Schulman, J.L., Bodo, G. & Meindl, P. (1974). Inhibition of influenza and parainfluenza virus replication in tissue culture by 2deoxy-2,3-dehydro-N-trifluroacetylneuraminic acid (FANA). Virology 59, 490-498. 15. Palese, P. & Schulman, J.L. (1977). Inhibitors of viral neuraminidase as potential antiviral drugs. In Chemoprophylaxis and Virus Infections of the Upper Respiratory Tract, Vol 1. pp. 189-205, CRC Press, Boca Raton, Florida. 16. Varghese, J.N. & Colman, P.M. (1991). Three-dimensional structure of the neuraminidase of influenza virus A/Tokyo/3/67 at 2.2 Å resolution. J. Mol. Biol. 221, 473-486. 17. Varghese, J.N., McKimm-Breschkin, J.L., Caldwell, J.B., Kortt, A.A. & Colman, P.M. (1992). The structure of the complex between influenza virus neuraminidase and sialic acid, the viral receptor. Proteins 14, 327-332. 18. von Itzstein, M., et al., & Penn, C.R. (1993). Rational design of potent sialidase-based inhibitors of influenza virus replication. Nature 363, 418-423. 19. von Itzstein, M., Wu, W.-Y. & Jin, B. (1994). The synthesis of 2,3didehydro-2,4-dideoxy-4-guanidinyl-N-acetylneuraminic acid: a potent influenza virus inhibitor. Carbohydr. Res. 259, 301-305. 20. Woods, J.M., et al., & Penn, C.R. (1993). 4-guanidino-2-4-dideoxy-2,3dehydro-N-acetylneuraminic acid is a highly effective inhibitor both of the sialidase (neuraminidase) and growth of a wide range of influenza A and B viruses in vitro. Antimicrob. Ag. Chemother. 37, 1473-1479. 21. Ryan, D.M., Ticehurst, J., Dempsey, M.H. & Penn, C.R. (1994). Inhibition of influenza virus replication in mice by GG167 (4-guanidino2,4-dideoxy-2,3-dehydro-N-acetylneuraminic acid) is consistent with extracellular activity of viral neuraminidase. Antimicrob. Ag. Chemother. 38, 2270-2275. 22. Thomas, G.P., Forsyth, M., Penn, C.R. & McCauley, J.W. (1994). Inhibition of growth of influenza virus in vitro by 4-guanidino-2,4-dideoxy-2,3dehydro-N-acetylneuraminic acid. Antiviral Res. 24, 351-356. 23. Hayden, F.G., Rollins, B.S. & Madren, L.K. (1994). Anti-influenza activity of the neuraminidase inhibitor 4-guanidino-Neu5Ac2en in cell culture and in human respiratory epithelium. Antiviral Res. 25, 123-131. 24. Hayden, F.G., et al., & Wightman, K. (1997). Efficacy and safety of the neuraminidase inhibitor Zanamivir in the treatment of influenza virus infections. New Engl. J. Med. 337, 874-880. 25. Smith, W., Andrews, C.H. & Laidlaw, P.P. (1933). A virus obtained from influenza patients. Lancet 2, 66-68. 26. McKimm-Breschkin, J.L., Marshall, D. & Penn, C.R. (1994). Phenotypic changes observed in influenza viruses passaged in 4-amino or 4guanidino-Neu5Ac2en in vitro. Abstracts of the 9th International Conference of Negative Strand Viruses, Estoril, Portugal, p. 260. 27. McKimm-Breschkin, J.L., et al., & Penn. C.R. (1996). Generation and characterisation of variants of the NWS/G70C influenza virus after in vitro passage in 4-amino-Neu5Ac2en and 4-guanidino-Neu5Ac2en. Antimicrob. Ag. Chemother. 40, 40-46. 28. Gubareva, L.V., Bethell, R., Hart, G.J., Penn, C.R. & Webster, R.G. (1995). Characterization of mutants of influenza virus selected with 4guanidino-Neu5Ac2en. J. Virol. 70, 1818-1827. 29. Blick, T.J., et al., & McKimm-Breschkin, J.L. (1995). Generation and characterization of an influenza virus variant with decreased sensitivity to the neuraminidase specific inhibitor 4-guanidino-Neu5Ac2en. Virology 214, 475-484. 30. McKimm-Breschkin, J.L., et al., & Varghese, J.N. (1998). Mutations in a conserved residue in the influenza virus neuraminidase active site decreases sensitivity to Neu5Ac2en derivatives. J. Virol. 72, 2456-2462.

746

Structure 1998, Vol 6 No 6

31. Staschke K.A., et al., & Laver, W.G. (1995). Molecular basis for resistance of influenza viruses to 4-guanidino-Neu5Ac2en. Virology 214, 642-646. 32. Hayden, F.G. (1996). Amantidine and rimantidine — clinical aspects. In Antiviral Drug Resistance. (Richmond, D.D., ed.), pp. 69-77, John Wiley, Chichester, UK. 33. Kimberlin D.W., et al., & Whitley R.J. (1995). Antiviral resistance in clinical practice. Antiviral Res. 26, 423-438. 34. Hayden F.G. (1993). Update on antiviral agents and viral drug resistance. In Principles and Practice of Infectious Disease, Vol 2. (Mandell G.L., Douglas R.G. & Bennett J.E., eds), pp. 3-15, Churchill Livingston, New York. 35. Madren L.K., Shipman C. & Hayden F.G. (1995). In vitro inhibitory effects of combinations of anti-influenza agents. Antiviral Chem. Chemother. 6, 109-113. 36. Sollis, S.L., Smith, P.W., Howes, P.D., Cherry, P.C. & Bethell, R.C. (1996). Novel inhibitors of influenza sialidase related to GG167. Synthesis of 4-amino and 4-guanidino-4H-pyran-2-carboxylic acid-6propylamides; selective inhibitors of influenza A virus sialidase. Bioorg. Med. Chem. Lett. 6, 1805-1808. 37. Smith, P.W., et al., & Colman, P. (1996). Novel inhibitors of influenza sialidases related to GG167. Structure-activity, crystallographic and molecular dynamics studies with 4H-pyran-2-carboxylic acid 6carboxamides. Bioorg. Med. Chem. Lett. 6, 2931-2936. 38. Taylor, N.R., et al., & Varghese, J. (1998). Dihydropyran carboxamides related to GG167. A new series of inhibitors of influenza virus sialidases. Part 2: A crystallographic and molecular modeling study of complexes of 4-amino-4H-pyran-6-carboxamides and sialidase from influenza types A and B. Bioorg. J. Med. Chem. 41, 798-807. 39. Kim C.U., et al., & Stevens R.C. (1997). Influenza neuraminidase inhibitors possessing a novel hydrophobic interaction in the enzyme active site: design, synthesis, and structural analysis of carbocyclic sialic acid analogues with potent anti-influenza activity. J. Am. Chem. Soc. 119, 681-690. 40. Gubareva, L.V., Robinson M.J., Bethell, R. & Webster, R.G., (1997). Catalytic framework mutations in the neuraminidase active site of influenza viruses that are resistant to 4-guanidino-Neu5Ac2en. J. Virol. 71, 3385-3390. 41. Varghese, J.N., Epa, V.C. & Colman, P.M. (1995). Three-dimensional structure of 4-guanidino-Neu5Ac2en and influenza virus neuraminidase. Protein Sci. 4, 1081-1087. 42. Weiss W., Brown, J.H., Cusack, S., Paulson, J.C., Skehel, J.J. & Wiley, D.C. (1988). Structure of the influenza virus haemagglutinin complexed with its receptor sialic acid. Nature 333, 426-431. 43. Varghese, J.N., Colman, P.M., van Donkelaar, A., Blick, T.J., Sahasrabudhe, A. & McKimm-Breschkin, J.L. (1997). Structural evidence for a second sialic acid binding site in avian influenza virus neuraminidases. Proc. Natl Acad. Sci. USA 94, 11808-11812. 44. Webster, R.G., et al., & Laver, W.G. (1987). Antigenic structure and variation in an influenza N9 neuraminidase. J. Virol. 61, 2910-2916. 45. Laver, W.G., Colman, P.M., Webster, R.G., Hinshaw, V.S. & Air, G.M. (1984). Influenza virus neuraminidase with hemagglutinin activity. Virology 137, 314-323. 46. Hausmann, J., Kretzschmar, E., Garten, W. & Klenk, H-.D. (1995). N1 neuraminidase of influenza virus A/FPV/Rostock/34 has haemadsorbing activity. J. Gen. Virol. 76, 1719-1728. 47. Subbarao, K., et al., & Cox, N. (1998). Characterisation of an avian influenza A (H5N1) virus isolated from a child with fatal respiratory illness. Science 279, 393-396. 48. Colman, P.M., et al., & Webster, R.G. (1987). Three-dimensional structure of a complex of antibody with influenza virus neuraminidase. Nature 326, 358-363. 49. Colman, P.M., et al., & Air, G.M. (1989). 3-D structures of influenza virus neuraminidase-antibody complexes. Proc. R. Soc. Lond. B 323, 511-518. 50. Malby, R.L., et al., & Colman, P.M. (1994). The structure of a complex between the NC10 antibody and influenza virus neuraminidase and comparison with the overlapping binding site of the NC41 antibody. Structure 2, 733-746. 51. Colman, P.M. (1997). Virus versus antibody. Structure 5, 591-593. 52. McKimm-Breschkin, J.L., McDonald, M., Blick, T.J. & Colman, P.M. (1996). Mutation in the influenza virus neuraminidase gene resulting in decreased sensitivity to the neuraminidase inhibitor 4-guanidinoNeu5Ac2en leads to instability of the enzyme. Virology 225, 240-242. 53. Tantillo, C., et al., & Arnold, E.A. (1994). Locations of anti-AIDS drug binding sites and resistance mutations in the three-dimensional structure of HIV-1 reverse transcriptase. J. Mol. Biol. 243, 369-387.

54. Schock, H.B., Garsky, V.M. & Kuo, L.C. (1996). Mutational anatomy of an HIV-1 protease variant conferring cross-resistance to protease inhibitors in clinical trials. J. Biol. Chem. 50, 31957-31963. 55. Pettit, S.C., Simsic, J., Loeb, D.D., Everitt, L., Hutchison, C.A. & Swanstrom, R. (1991). Analysis of retroviral protease cleavage sites reveals two types of cleavage sites and the structural requirements of the P1 amino acid. J. Biol. Chem. 266, 14539-14547. 56. Griffiths, J.T., et al., & Kay, J. (1992). Different requirements for productive interaction between the active site of HIV-1 proteinase and substrates containing -hydrophobic*hydrophobic- or -aromatic*procleavage sites. Biochemistry 31, 5193-5200. 57. McKimm-Breschkin, J.L., Caldwell, J.B., Guthrie, R.E. & Korrtt, A.A., (1991). A new method for the purification of influenza virus A virus. J. Virol. Methods 32, 121-124. 58. Baker, A.T., Varghese, J.N., Laver, W.G., Air, G.M. & Colman, P.M. (1987). The three-dimensional structure of neuraminidase of subtype N9 from an avian influenza virus. Proteins 1, 111-117. 59. Tulip, W.G., et al., & Colman PM. (1992). Refined atomic structures of N9 subtype influenza virus neuraminidase and escape mutants. J. Mol. Biol. 221, 487-497. 60. Brünger, A.T. (1992). X-PLOR, Version 3.3: a System for X-ray Crystallography and NMR. Yale University Press, New Haven, CT. 61. Engh, R.A. & Huber, R. (1991). Accurate bond and angle parameters for X-ray protein-structure refinement. Acta Cryst. A 47, 392-400.