Intersubunit Interactions Modulate pH-Induced Activation of Membrane Fusion by the Junin Virus Envelope Glycoprotein GPC

The mature arenavirus envelope glycoprotein GPC is a tripartitecomplex comprising a stable signal peptide (SSP) in additionto the receptor-binding (G1) and transmembrane fusion (G2) subunits.We have shown previously that SSP is a key element in GPC-mediatedmembrane fusion, and that GPC sensitivity to acidic pH is modulatedin part through the lysine residue at position 33 in the ectodomainloop of SSP (J. York and J. H. Nunberg, J. Virol. 80:7775-7780,2006). A glutamine substitution at this position stabilizesthe native GPC complex and thereby prevents the induction ofpH-dependent membrane fusion. In efforts to identify the intersubunitinteractions of K33, we performed alanine-scanning mutagenesisat charged residues in the membrane-proximal ectodomain of G2and determined the ability of these mutations to rescue thefusion deficiency in K33Q GPC. Four second-site mutations thatspecifically complement K33Q were identified (D400A, E410A,R414A, and K417A). Moreover, complementation was also observedat three hydrophobic positions in the membrane-spanning domainof G2 (F427, W428, and F438). Interestingly, all of the complementingmutations restored wild-type pH sensitivity to the K33Q mutant,while none themselves affected the pH of membrane fusion. Ourstudies demonstrate a specific interaction between SSP and G2that is involved in priming the native GPC complex for pH-inducedmembrane fusion. Importantly, this pH-dependent interactionhas been shown to be vulnerable to small-molecule compoundsthat stabilize the native complex and prevent the activationof membrane fusion. A detailed mechanistic understanding ofthe control of GPC-mediated membrane fusion will be importantin guiding the development of effective therapeutics againstarenaviral hemorrhagic fever.

INTRODUCTION

Top

INTRODUCTION

Arenaviruses are endemic in rodent populations worldwide (37),and several species of the virus can be transmitted to humansto cause severe hemorrhagic fevers (30, 33). These include Lassafever virus (LASV) in western Africa and Junín virus(JUNV) in the New World. The arenaviruses are enveloped, negative-strandRNA viruses whose bipartite genomes encode the ambisense expressionof the four viral proteins (7, 10). The envelope glycoproteinGPC promotes virus entry into the host cell and thereby providesa viable target for therapeutic intervention (3, 27, 28, 44).GPC is synthesized as a precursor glycoprotein that is proteolyticallymatured by the cellular SKI-1/S1P protease (2, 26, 29) to generatethe receptor-binding (G1) and transmembrane fusion (G2) subunits,which remain noncovalently associated in the mature complex.Arenavirus entry is initiated by G1 binding to a cell surfacereceptor. The pathogenic New World arenaviruses utilize transferrinreceptor-1 for entry (34, 35), whereas Old World viruses bind ${alpha}$ -dystroglycan (8, 18, 38). The virion subsequently is endocytosed(4), and fusion of the viral and cellular membranes is initiatedby acidification in the maturing endosome (9, 12, 13). The detailsof the acid-induced activation of GPC-mediated membrane fusionare unknown.

The arenavirus GPC is unique among viral envelope glycoproteinsin that the mature complex includes a cleaved and stable signalpeptide (SSP) as a third subunit (6, 15, 48) (Fig. 1). The 58-amino-acidSSP spans the membrane twice, with both termini in the cytosol(1), and likely is retained in the complex by the formationof an intersubunit zinc finger structure with the cytoplasmicdomain of G2 (46). Importantly, we have shown that the shortectodomain loop of SSP plays a central role in the responseof the GPC complex to acidic pH and, thereby, in pH-dependentmembrane fusion (44, 47). Despite the unusual involvement ofSSP in pH-triggered activation, the subsequent mechanics ofGPC-mediated membrane fusion are likely similar to those ofother class I viral fusion proteins, such as influenza virushemagglutinin (HA) and the retrovirus envelope glycoprotein(Env) (17, 20, 43). A generally accepted model for membranefusion by these proteins (reviewed in references 14, 22, and42) posits that the native envelope glycoprotein exists in ametastable prefusion state that is established on proteolyticmaturation. Destabilization of the native complex by acidicpH in HA (or receptor binding by Env) initiates a sequence ofconformational changes that ultimately lead to formation ofthe now-classical six-helix bundle structure in the ectodomainof the transmembrane fusion subunit. In the process of formingthis highly stable postfusion structure, the virus and cellmembranes are brought into apposition and membrane fusion ensues.HA is the best-studied pH-dependent class I fusion protein,but even though the prefusion and postfusion structures areknown in atomic detail, the molecular basis for the pH-inducedactivation of the complex remains elusive (14, 22, 42).

View larger version (24K):
[in this window]
[in a new window]

FIG. 1. Arenavirus GPC complex. The JUNV GPC open reading frame is illustrated at the top. Amino acids are numbered from the initiating methionine, and the SSP, G1, and G2 subunits are indicated. Membrane-spanning regions in SSP (h ${phi}$ 1 and h ${phi}$ 2; 1) and in G2 (TM) are shaded dark gray, and the N- and C-terminal heptad repeat regions in G2 (43) are in light gray. The amino acid sequence of the G2 membrane-proximal ectodomain and transmembrane domain (shaded) of JUNV is detailed below; positions studied in this report are indicated by dots. At the lower right is a diagram of the proposed subunit organization in the tripartite GPC complex. Thickened lines represent membrane-spanning domains in SSP and G2 and the heptad repeat regions in G2. The cytosolic N terminus of SSP is myristoylated (thin line) (48), and an intersubunit zinc finger (ball) is thought to link the C terminus of SSP with the cytoplasmic domain of G2 (46). Lysine 33 (K33) in SSP is marked. The drawing is representational and not to scale. In the lower left, we show the alignment of the arenavirus membrane-proximal G2 ectodomain regions. Accession numbers are the following: JUNV, D10072; LASV-Nigeria (LASV-N), P17332; Tacaribe virus (TCRV), NP_694849; Pichinde virus (PICV), AAB58484; Machupo virus (MACV), AAX99337; Sabiá virus (SABV), AAC55091; LASV-Josiah (LASV-J), AAG41802; Mopeia virus (MOPV), AAV54108; and lymphocytic choriomeningitis virus-Armstrong (LCMV-A), NP_694851. Charged residues are highlighted in blue (basic) or red (acidic).

Our previous studies of JUNV GPC identified the unusual roleof SSP in modulating the response of GPC to acidic pH. We demonstratedthat mutations that reduce the positive charge at lysine 33(K33) in the ectodomain loop of SSP correspondingly stabilizeGPC against pH-induced activation (47). Based on this finding,we sought to identify the pH-responsive elements in GPC withwhich K33 interacts. In the absence of structural information,we directed our investigation to the membrane-proximal ectodomainand transmembrane regions of G2, which may lie near the membrane-proximalK33 residue in SSP. By using the method of genetic complementation,we now have identified second-site mutations in G2 that rescuethe fusion deficiency arising from the substitution of K33 withglutamine. Our findings provide evidence that SSP and G2 interactin controlling the pH-induced activation of GPC-mediated membranefusion.

MATERIALS AND METHODS

Top

MATERIALS AND METHODS

Molecular reagents. GPC from the pathogenic JUNV strain MC2 (21) was expressed inVero cells by the Lipofectamine 2000 (Invitrogen)-mediated cotransfectionof two plasmids: CD4sp-GPC (in which SSP is replaced by theconventional signal peptide of CD4) and SSP-term (in which astop codon is introduced following the C-terminal SSP aminoacid T58) (47). SSP and the G1-G2 precursor associate in transand reconstitute the native GPC complex (15). This experimentalsystem obviates concerns regarding the potential effects ofmutation on the signal peptidase cleavage of full-length GPC(45). Mutations in GPC were introduced by QuikChange mutagenesis(Stratagene), and all constructs were verified by DNA sequencing.

Cell-cell fusion assay. A vaccinia virus-based β-galactosidase fusion reporterassay (31) was used to characterize the ability of GPC to mediatepH-dependent cell-cell fusion (47, 48). Briefly, Vero cellsexpressing GPC and infected by vTF-7 (which expresses bacteriophageT7 polymerase) (19) were maintained in medium containing 10µM cytosine arabinoside (araC) to limit vaccinia virusreplication. Target cells for syncytium formation were infectedwith a recombinant vaccinia virus bearing the β-galactosidasegene under the control of a T7 promoter (vCB21R-lacZ) (31) andsubsequently were seeded into 96-well microculture dishes (40,000cells/well) in the presence of 100 µg rifampin/ml to minimizevaccinia assembly and cytopathic effect (24). Prior to use inthe cell-cell fusion assay, target cells (40,000 cells/well)were added to the microcultures, allowed to settle for 30 min,and subjected to a brief low-speed centrifugation ( $~$ 25 x g) toensure cell-cell contact. These cocultures were continued for3 h in medium containing araC and rifampin. Membrane fusionwas initiated by incubation in medium that had been adjustedto low pH using HEPES and piperazine-N,N'-bis(2-ethanesulfonicacid) buffers. After 10 min at 37°C, the cells were restoredto neutral medium and cultured for 5 h to allow for the expressionof the β-galactosidase fusion reporter gene, which thenwas quantitated using the chemiluminescent GalactoLite Plussubstrate (Applied Biosystems) and a Tropix TR717 microplateluminometer.

RESULTS AND DISCUSSION

Top

RESULTS AND DISCUSSION

The pH-induced activation of GPC-mediated membrane fusion iscontrolled in part by the lysine at position 33 in the shortectodomain loop of SSP. Amino acid substitutions that reducepositive polarity at K33 systematically lower the pH necessaryto activate membrane fusion (47). The replacement of lysinewith glutamine (K33Q) reduces membrane fusion activity to 6%of the wild-type level (see below) while decreasing the pH ofmaximal fusion from the wild-type optimum of pH 5.0 to pH 4.5.The apparent stabilization of the mutant complex against acidicpH does not appreciably affect its intrinsic fusion activity,as K33Q GPC retains (at pH 4.5) as much as 40 to 50% of thewild-type GPC activity (at pH 5.0). The replacement of K33 withalanine not only abrogates membrane fusion at any relevant pHbut also stabilizes the complex against spontaneous dissociationof the G1 and G2 subunits (44). In order to identify pH-sensitivedeterminants that might interact with K33, we inspected theamino acid sequence of the membrane-proximal ectodomain of G2.This region, within 35 amino acids of the membrane (D390 toD424 in JUNV), includes 12 charged residues in both JUNV andLASV (Fig. 1). Nine of these (D394, D400, E405, K409, E410,D413, R414, K417, and D424) are identical between the two viruses.Other positions vary conservatively among New World and OldWorld arenaviruses, and all species retain $≥$ 30% charged residues(Fig. 1). Only R414 and D424 are invariant among all of thearenaviruses.

The role of these charged residues in membrane fusion was examinedby replacing them individually with alanine. In brief, mutationswere introduced into a recombinant construct, CD4sp-GPC, encodingthe G1-G2 precursor of the pathogenic MC2 isolate of JUNV (21)preceded by the conventional signal peptide of CD4 (48). Whentransiently coexpressed with SSP (amino acids 1 to 58) in Verocells, the GPC complex is functionally reconstituted (15, 48).The membrane fusion activity of the mutant GPC was determinedusing a β-galactosidase fusion reporter assay (31) in whichsyncytium formation was initiated by exposing the culture ofGPC-expressing effector cells and fusion reporter target cellsto medium adjusted to pH 5.0 (44, 48). As illustrated in Fig.2, most of the alanine mutants in G2 were able to mediate pH-dependentcell-cell fusion at levels approaching ( $≥$ 50%) that of the wildtype (also see reference 44). Only R392A, E405A, and R414A weresignificantly diminished in their membrane fusion activity ( $≤$ 20%of the level of wild-type GPC).

View larger version (28K):
[in this window]
[in a new window]

FIG. 2. Membrane fusion activity of G2 mutants. The wild-type (wt) and G2 mutants in CD4sp-GPC were expressed in trans with wild-type SSP (open bars) or K33Q SSP (filled bars). The ability of the single and double mutants to mediate cell-cell fusion at pH 5.0 was determined as detailed in Materials and Methods. Error bars representing ±1 standard deviation are calculated for all points (n = 6) and may not be visible on the scale of the graph. These data represent one complete study (of at least three to six complete and partial experiments). All comparisons are reliably reproducible between experiments, although absolute percentages may vary. The average for K33Q GPC (6.1% ± 2.5%) differs from that previously cited (47) due to technical changes in the assay. The D400A, E405A, E410A, K417A, and D424A mutants have been reported previously (44).

Mutations at charged residues complement fusion defects in SSP. To specifically inquire whether these charged residues in G2interact with SSP, we screened the panel of mutants for theirability to rescue a fusion deficiency at SSP position 33. Thisgenetic method of complementation is frequently used to studyfunctional interactions between and within viral proteins (5,39). Of the previously characterized K33 mutants, only K33QGPC is severely debilitated yet retains sufficient activityat pH 5.0 (6% of the wild type) for experimental measurementsof complementation. Mutations in G2 that restored fusion activitybeyond this 6% baseline level were judged to interact geneticallywith SSP. We identified four such mutations (Fig. 2). Doublemutants with D400A, E410A, R414A, and K417A were able to mediatefusion at 20, 30, 55, and 20% of wild-type levels, respectively.In contrast, little or no increase in fusion activity was observedin double mutants containing D390A, R392A, D394A, E398A, E405A,K409A, D413A, or D424A. The presence or absence of complementationdid not correlate with the magnitude of the fusion defect inthe single alanine mutant. Nor could we discern significantdifferences in the extent of SSP association or proteolyticmaturation (45) among the mutant GPC complexes (not shown).Selective complementation by the D400A, E410A, R414A, and K417Amutations, all of which cluster toward the membrane-spanningdomain of G2, points to specificity in the rescue. These findingsindicate a genetic interaction between SSP and the most membrane-proximalregion of the G2 ectodomain.

Molecular basis for complementation of K33Q. To investigate whether complementation involves pH-sensitiveinteractions, we examined a series of charge mutants at thestrongly complementing R414 position. As anticipated, R414Ewas somewhat reduced in membrane fusion activity relative tothat of R414A and with respect to R414K, R414H, R414Q, and R414Smutants (Fig. 3). However, none of the charged, polar, or neutralmutations had major effects on the membrane fusion activityor complementation of K33Q. A difference in the extent of rescueby the R414 mutation generally correlated with its effect onfusion activity. In contrast to the systematic effects of mutationat K33, we were unable to discern a charge-dependent interactionbetween K33Q and the R414 side chain. Thus, complementationby K33Q may arise through effects other than those that aredetermined primarily by charge.

View larger version (17K):
[in this window]
[in a new window]

FIG. 3. Membrane fusion activity of charge mutants at R414. These studies were performed in the same fashion as those described in the legend to Fig. 2. The amino acid changes at R414 are indicated by single-letter abbreviations. Note that the vertical axis is truncated. wt, wild type.

Compared to that of wild-type GPC, K33Q requires a lower pHto activate membrane fusion (47). With the extraordinary stabilityof the G1 and G2 association in the K33A mutant (44), we concludedthat mutations at K33 stabilize the native GPC complex withrespect to pH-induced activation. Therefore, we hypothesizedthat K33Q-complementing mutations in G2 act in an opposing mannerto destabilize the complex and raise the pH for membrane fusion.Initial experiments with three of these mutations (D400A, E410A,and K417A) had not revealed significant differences (44), sowe extended the analysis to include the strongly complementingR414A mutation.

The pH dependence of membrane fusion was determined by varyingthe pH of the acidic medium used to initiate cell-cell fusion(44, 47). Fusion activity by the wild-type GPC (Fig. 4) wasfirst observed at pH 5.5, attained a maximum level at pH 5.0,and subsequently declined (gray circles). As noted, fusion byK33Q GPC (gray squares) was minimal at pH 5.0 and reached itsmaximum level at pH 4.5. Single mutants with complementing (E410Aand R414A) and noncomplementing (E405A and D424A) mutationsbehaved much like the wild-type GPC, with maximal fusion activityat pH 5.0 to 5.5 (black circles). Although small differencesrelative to the wild type could be noted, these did not correlatewith the presence or absence of complementation and were notstatistically significant in our studies. Thus, the single alaninemutations in the membrane-proximal ectodomain of G2 did notby themselves increase the pH at which membrane fusion is activated.

View larger version (35K):
[in this window]
[in a new window]

FIG. 4. pH sensitivity of membrane fusion. Cells expressing either wild-type SSP (black circles) or K33Q SSP (black squares) with the G2 mutant were pulsed at the indicated pH. The lightly shaded curves, repeated in all panels, depict fusion of the wild-type CD4sp-GPC with wild-type (gray circles) and K33Q (gray squares) SSP. The results shown in this study are representative of at least three independent experiments. Fusion activity is normalized to the maximum of each mutant for this comparison. Error bars represent ±1 standard deviation.

Complementing mutants abolish pH suppression by K33Q. However, all of the complementing mutations, and only the complementingmutations, were able to mitigate the K33Q-induced reductionin the pH of activation (Fig. 4). In the highly complementingR414A mutant, maximal membrane fusion was shifted to pH 5.0(black squares), similar to that of wild-type GPC and distinctfrom that of the single K33Q mutant (pH 4.5). The weakly complementingE410A mutant showed a partial shift toward pH 5.0. In contrast,optimal fusion by noncomplementing double mutants (E405A andD424A) remained at pH 4.5, similar to K33Q GPC. Thus, the rescueof the membrane fusion activity of K33Q GPC by second-site mutationsin G2 can be accounted for entirely by their suppression ofthe K33Q-induced decrease in the pH of membrane fusion.

Role of the G2 membrane-spanning domain in complementation. Several lines of evidence suggested that interactions betweenSSP and G2 extend to the membrane-spanning region. In particular,amino acid determinants of resistance to small-molecule fusioninhibitors recently discovered by SIGA Technologies (Corvallis,OR) are localized not only in the ectodomain loop of SSP andthe membrane-proximal ectodomain of G2 but also in the adjoiningtransmembrane domain (3, 27, 44). We therefore extended ourcomplementation analysis to include transmembrane residues inG2 known to affect sensitivity to these inhibitors. As shownin Fig. 5, mutations at two major determinants of resistance(F427A and F438I) were able to rescue the K33Q defect to 50and 75% of wild-type levels, respectively. A435I, another resistancemutation, did not complement K33Q, whereas the inhibitor-sensitivemutant W428A did. Thus, the interactions responsible for thecomplementation of K33Q extend into the membrane-spanning domainof G2 and are not restricted to charged residues.

View larger version (22K):
[in this window]
[in a new window]

FIG. 5. Complementation by transmembrane domain G2 mutants. The G2 mutants were expressed in trans with wild-type SSP (open bars) or K33Q SSP (filled bars), and membrane fusion activity was determined. These data represent one complete study (of at least three complete and partial experiments).

Interestingly, the mechanism for complementation at these hydrophobicpositions (F427 and W428) appeared identical to that at chargedpositions in the membrane-proximal ectodomain. Neither of theseuncharged mutations alone affected the pH of membrane fusion,yet both restored the pH of the maximal fusion activity of K33QGPC to that of the wild type (Fig. 4). Complementation at thesehydrophobic positions is unlikely to be mediated by direct interactionwith K33. Rather, the restoration of the wild-type phenotypemay reflect indirect structural effects emanating from interactionsof the membrane-spanning domains of G2 and SSP or from potentialoligomerization determinants in G2 (16, 32). Thus, perturbationsin the membrane may couple to the pH-sensing ectodomain.

Additional transmembrane domain residues that might be involvedin complementation were also examined. The possibility of atransient disulfide bond at C426 and the potentially ionizableH440 side chain were probed by replacements with serine andalanine, respectively. Neither of these mutations disruptedfusion activity or complemented the K33Q defect (Fig. 5). TheH440A mutation eliminated the sole histidine in the membraneand membrane-proximal sequences of G2.

Controlled activation of pH-induced membrane fusion. Our observation that complementing mutations, which alone donot affect the pH of membrane fusion, are able to abrogate thepH effect of the K33Q mutation presents us with a conundrum.In the simplest model for complementation, second-site mutationsmight generate a corresponding increase in pH sensitivity tocounterbalance the reduction by K33Q. Indeed, compensatory changesin pH sensitivity are well described in the influenza virusHA. Fusion inhibition by tert-butyl hydroquinone (TBHQ) compoundsthat stabilize the HA complex against low pH in the endosomecan be overcome by second-site mutations that render it moresensitive to pH (23, 36). Similarly, viral resistance to amantadine,an inhibitor of endosomal acidification, arises through mutationsin HA that specifically enable fusion at an elevated pH (11,40). Mutations affecting the pH of HA activation include bothcharged and uncharged residues that, although widely distributed,tend to cluster near the fusion peptide as well as at monomerinterfaces (11, 23, 36, 39-41). Despite atomic resolution knowledgeof HA, however, the primary site of pH-induced activation remainsunresolved (22, 42). In general, the compact and metastablenature of the native complex makes the dissection of primaryand secondary events in the response to pH difficult.

The molecular mechanism whereby mutations in G2 restore wild-typepH sensitivity to K33Q GPC without themselves altering the pHof activation is unclear. We suggest that the original K33Qmutation affects the prefusion complex in two opposing directions.First, by decreasing repulsive charge interactions through K33,the glutamine mutation may directly stabilize the pH-sensingmechanism against activation. Second, by replacing the naturallysine residue, the mutation also may contribute to destabilizationof the complex. In the single G2 mutants, the pH-sensing mechanismremains intact and responds in the same fashion as the wild-typeGPC, despite the G2 mutation. However, in the double mutant,the effects of the G2 mutation may combine with the destabilizingcomponent of the K33Q mutation to overcome the otherwise stabilizingeffect of the uncharged glutamine. Nonlinear interactions betweenmutations affecting pH sensitivity in HA also have been documented(39). It remains possible that the primary pH-responsive interactionof K33 includes charged residues elsewhere in G2. As illustratedin the recently published structure of the acid-sensing ionchannel ASIC1, pH-sensing sites can be highly distributed networksincluding multiple-charge, proton-sharing, and hydrogen-bondinginteractions (25). It is important to note that the geneticresults reported here do not necessarily connote direct chemicalinteractions. In the absence of the atomic resolution structure,however, our studies provide initial insights into the complexstructure-function relationships that promote the transitionfrom the native GPC complex toward membrane fusion.

Without an effective treatment or vaccine, the hemorrhagic feverarenaviruses remain a pressing public health and biodefenseconcern. In this regard, it is noteworthy that the process ofthe pH-induced activation of membrane fusion has been shownto offer a viable target for antiviral intervention. In a manneranalogous to that of the TBHQ inhibitors of HA (23, 36), small-moleculearenavirus-specific compounds described by SIGA (3, 27) inhibitGPC-mediated fusion by stabilizing the prefusion complex againstacidic pH (44). A further understanding of the molecular determinantsof inhibition at the SSP-G2 interface may lead to the identificationof improved antiviral compounds. The genetic manipulation ofpH sensitivity also may generate stable forms of the nativeGPC complex that can be used to elicit protective immunity aswell as in structural studies. The detailed characterizationof GPC-mediated membrane fusion will advance the developmentof effective vaccines and therapeutic agents against arenaviraldisease.

ACKNOWLEDGMENTS

We are grateful to Meg Trahey (The University of Montana) andMin Lu (Weill Medical College of Cornell University) for theirthoughtful comments and editorial assistance.

This work was supported by NIH research grant R01 AI074818 andby the Rocky Mountain Regional Center of Excellence for Biodefenseand Emerging Infectious Diseases at Colorado State University(U54 AI065357).

FOOTNOTES

* Corresponding author. Mailing address: Montana Biotechnology Center, The University of Montana, Science Complex, Room 221, Missoula, MT 59812. Phone: (406) 243-6421. Fax: (406) 243-6425. E-mail: jack.nunberg{at}umontana.edu

Published ahead of print on 18 February 2009.

REFERENCES

Top