The Murid Herpesvirus-4 gL Regulates an Entry-Associated Conformation Change in gH

The glycoprotein H (gH)/gL heterodimer is crucial for herpesvirus membrane fusion. Yet how it functions is not well understood. The Murid Herpesvirus-4 gH, like that of other herpesviruses, adopts its normal virion conformation by associating with gL. However, gH switched back to a gL-independent conformation after virion endocytosis. This switch coincided with a conformation switch in gB and with capsid release. Virions lacking gL constitutively expressed the down-stream form of gH, prematurely switched gB to its down-stream form, and showed premature capsid release with poor infectivity. These data argue that gL plays a key role in regulating a gH and gB functional switch from cell binding to membrane fusion.


Introduction
Enveloped viruses deliver their capsids to the cytoplasm by fusing with either the plasma membrane or the limiting membranes of endocytic vesicles. Many viruses use a single glycoprotein for both binding and membrane fusion. Herpesviruses, by contrast, use at least four [1,2]; their entry mechanisms are implicitly more complex. Proteins specific to individual herpesviruses can modulate fusion, for example the Herpes simplex virus gD brings together gB and gH/gL after ligand engagement to trigger fusion [3,4], and the Epstein-Barr virus gp42 binds gH/gL to promote B cell over epithelial cell fusion [5]. But the core fusion machinery of gB, gH and gL is conserved. Down-stream fusion events are therefore likely to be common to all herpesviruses.
We have analyzed gB and gH/gL using the MHV-68 isolate of Murid Herpesvirus-4 (MuHV-4) [6][7][8]. This gamma-herpesvirus has subtle differences from the more extensively studied Herpes simplex virus, for example it lacks an obvious homolog of gD and fuses after endocytosis rather than at the plasma membrane [9]. Nevertheless, the conservation of gB and gH/gL implies that they operate similarly in these and other herpesviruses once the fusion mechanism is engaged. Incoming MuHV-4 virions first bind to cell surface heparan sulphate proteoglycans via gH/gL or gp70 [10,11]. Gp150 displacement then allows further, heparan sulphateindependent binding, probably via gB [10,12]. Virions are endocytosed and transported to LAMP-1 + late endosomes [13]; membrane fusion then releases virion capsids into the cytoplasm; and these move to the nuclear margin for genome delivery [13].
Both gB and gH have hydrophobic loops that could participate directly in fusion [14,15]. Structural analysis of the Herpes simplex virus gB has [14] shown how these might be made available.
Comparison with the Vesicular stomatitis virus glycoprotein G [16] suggests that the crystallized form of gB, with its distinct crown formed mainly from C-terminal domains, is a down-stream form. Consistent with this, the Murid Herpesvirus-4 (MuHV-4) gB switches during entry from having readily accessible N-terminal epitopes and poorly accessible epitopes involving C-terminal domains to the converse [13]. The N-terminal domains are involved in cell binding [10]. Thus, gB on extracellular MuHV-4 virions adopts a conformation that allows cell-binding, then following endocytosis undergoes a pH-dependent conformation switch that leads to membrane fusion. Whether this conformation change itself drives membrane fusion, as for Vesicular stomatitis virus, or whether further conformation changes are required along the lines suggested by Heldwein et al. [14], remains unclear.
While quite a lot is known about gB, much less is known about gH/gL. Understanding gH/gL therefore presents the major obstacle to understanding herpesvirus entry. Unlike gH, gL lacks obvious fusion loops or a trans-membrane domain. Its contribution to membrane fusion is therefore likely to be indirect. So far, gL has been defined only as a chaperone for gH [9,17,18]. Some herpesviruses need gL to get gH into virions [17,18]; Pseudorabiesvirus does not, but remains non-infectious without gL [19]; MuHV-4 lacking gL both incorporates gH into virions and remains infectious [20]. These differences may reflect how gB and gH associate. The Herpes simplex virus gB and gH come together only when gD is ligated [3,4], so their interaction is likely to involve extracellular domains; for gH, this would imply a requirement for gL. The MuHV-4 gB and gH associate constitutively and independently of gL, probably via an intramembrane interaction [21]. Because MuHV-4 does not require gH/gL for viability, it provides a convenient tool with which to explore gL-dependent down-stream events in viral entry. gL 2 MuHV-4 virions do show some attenuation relative to the wild-type. One deficit is reduced binding to BHK-21 fibroblasts [20]. This reflects that gH/gL binds to GAGs, whereas gH alone does not [11]. However, gL-deficient virions infect NMuMG epithelial poorly even though they bind to them quite well [20]. gL must therefore have another function that is important for NMuMG cell infection. Here we identify an entry-associated conformation change in gH that was consistent with gL dissociation in late endosomes, the site of MuHV-4 capsid release. When virions lacked gL, gH and gB both adopted their downstream forms prematurely. This led to premature capsid release and poor infectivity. An important function of gL is therefore to regulate the gB and gH conformation changes that lead to viral membrane fusion.

Results
The MuHV-4 gH changes conformation after endocytosis MuHV-4 infects adherent cells via endocytosis. Capsid release is pH-dependent and occurs when virions reach the late endosomes/ lysosomes marked by lysosome-associated membrane protein-1 (LAMP-1) [13]. Capsid release is associated with a conformation change in gB. To test whether gH also changes, we examined its antigenicity before and after endocytosis ( Fig. 1) using mAbs that distinguish gL-dependent and gL-independent gH conformations [9,20]. The gH on wild-type virions is mostly bound to gL [20]. MAbs requiring both gH and gL for recognition, such as 7E5 and T2C12, accordingly stained incoming virions strongly at the cell surface (Fig. 1A). However, this staining was lost after endocytosis. In contrast, the gH-only-specific mAb MG-9B10 [20] gave weak staining at the cell surface and much stronger staining after endocytosis. Other gH-only-specific mAbs such as MG-1A2 and MG-2E6 [20] gave similar results to MG-9B10. The corresponding gH and gH/gL epitopes are all conformational, as the mAbs do not recognize denatured virions, and can all be expressed in the absence of other virion proteins on transfected cells. Therefore gH did not appear to be denatured or disguised, but rather switched from a gL-dependent to a gL-independent conformation.
The MuHV-4 gH changes conformation in LAMP-1 + endosomes The disappearance of gH/gL epitopes coincided with virions reaching LAMP-1 + late endosomes; any residual gH/gL-specific staining after incubation at 37uC was confined to LAMP-1 2 compartments (Fig. 1B, arrow). In contrast, gH-only staining colocalized with LAMP-1. Most virion gH therefore switched from gH/gL on extracellular virions to gH-only in late endosomes. Figure 1. Glycoprotein H changes its antigenicity after endocytosis. A. NMuMG cells were exposed to MuHV-4 virions (3 p.f.u./cell, 2 h, 4uC), washed 63 with PBS and either fixed immediately (4uC) or after a further 2 h incubation at 37uC (37uC). The cells were then stained for gH (MG-9B10), gH/gL (T2C12, 7E5) or gN (3F7) as an invariant control. Glycoprotein staining is green; nuclei are counterstained with DAPI (blue). The data are representative of 5 experiments, and similar results were obtained with 3 other gH-specific/gH/gL-specific mAb pairs. Single cells are shown for optimal resolution. In this as in subsequent figures, each individual cell is fully representative of at least 75% of the total examined (n.100). B. Cells were infected as in A, then stained for LAMP-1 (red) plus either gH (mAb MG-9B10) or gH/gL (mAb 7E5) (green). Co-localization is yellow. The arrow indicates residual gH/gL staining that does not colocalize with LAMP-1. doi:10.1371/journal.pone.0002811.g001 Thus, the gB conformation switch, the gH conformation switch and capsid release all coincided in time and place.
The gH/gL conformation changes are pH-dependent The localization of the gH conformation switch to LAMP-1 + late endosomes suggested that it required low pH. This was confirmed with inhibitors of lysosomal acidification ( Fig. 2A). After incubating virions with cells at 37uC without drugs, gH-only staining predominated over gH/gL. But when the cells were treated with bafilomycin or Concanamycin A, although virions were still endocytosed and transported to LAMP-1 + endosomes, gH/gL staining was preserved and gH-only staining correspondingly reduced. In particular, gH/gL staining now co-localized with LAMP-1, something never observed without drug treatment.
Blocking endocytosis with chlorpromazine also blocked the gH conformation change, as did ammonium chloride treatment (Fig. 2B). Although the latter classically blocks endosomal acidification, it also blocked MuHV-4 endocytosis. Such an effect has been observed before [22]. Thus, in contrast to bafilomycin and Concanamycin A, the effect of NH 4 Cl could not be attributed to a rise in endosomal pH. However, it seemed clear from the bafilomycin and Concanamycin A treatments that the MuHV-4 gH conformation switch required virion delivery to low pH endosomes rather than just endocytosis.

A lack of gL causes a post-binding entry deficit
The apparent dissociation of gL from gH during MuHV-4 entry suggested that the gH-only conformation is functionally important, and therefore that gL has a more complicated role than merely folding gH for cell binding. One prediction of this would be that removing gL has consequences for post-binding entry events. We looked for evidence of a gL-dependent post-binding infection deficit by binding gL + and gL 2 virions to NMuMG cells for 3 h, washing off any unbound virions, and then comparing the initial virion uptake with subsequent viral eGFP expression (Fig. 3). We compared first a gL 2 mutant with the wild-type, and then an independent gL 2 mutant with its revertant. Cell binding was similar between the gL 2 and gL + viruses, but virus-driven eGFP expression was markedly different. After 6 h, eGFP expression by the gL 2 viruses was virtually undetectable: for an equivalent level of binding, the infectivity of gL + MuHV-4 for NMuMG cells at 6 h was at least 100 times greater than that of gL 2 mutants. Even after 18 h, gL + viruses showed at least 10-fold more eGFP expression for an equivalent level of binding. gL 2 virions therefore showed a marked post-binding infection deficit in NMuMG cells.

gL 2 virions show premature capsid release
We tracked the entry of gL 2 virions into NMuMG cells by immunofluorescence (Fig. 4A). MAb MG-12B8 recognizes an ORF65 capsid epitope that is inaccessible on virions until they have uncoated [23]. There was accordingly no MG-12B8 staining of either wild-type or gL 2 virions after binding to NMuMG cells at 4uC (Fig. 4A). After incubation at 37uC, the capsids of wild-type virions had become visible and reached the nuclear margin. In contrast, gL 2 virion capsids, although accessible, remained scattered throughout the cytoplasm. The same abnormal capsid MuHV-4 virions were bound to NMuMG cells (2 h, 4uC) with or without bafilomycin (500 nM) or concanamycin A (1 mM). The cells were then washed in PBS to remove unbound virions, further incubated with or without drugs (2 h, 37uC), fixed, permeabilized and stained for gH or gH/gL plus LAMP-1. Nuclei were counterstained with DAPI. B. In a similar experiment, MuHV-4 virions were bound to NMuMG cells (2 h, 4uC) with or without NH 4 Cl (100 mM) or chlorpromazine (3 mg/ml). Unbound virus was removed by PBS wash and the cells incubated further (37uC, 2 h), again with or without drug treatment, before fixing, permeabilizing and staining for gH or gH/gL. Although NH 4 Cl is generally considered to block endosomal acidification, here it also seemed to block virion endocytosis. doi:10.1371/journal.pone.0002811.g002 distribution after incubation at 37uC was observed for a range of different gL mutants, but not for their revertants or the wild-type (Fig. 4B). Since capsids are normally released only from perinuclear late endosomes, the gL 2 virions must have released their capsids prematurely in the entry pathway.
The glycoproteins of gL 2 virions were still internalized rather than incorporated into the plasma membrane. Surprisingly, infection by gL 2 virions also remained at least as sensitive to inhibitors of endocytosis and lysosomal acidification as that of the wild-type or gL + revertants (Fig. 5). This may reflect that low pH has a wider importance for MuHV-4 entry than switching the conformation of virion glycoproteins. For example, it could be required to reveal a key cellular ligand. Simply exposing cellbound virions to low pH triggered the gH conformation change at best poorly (data not shown). We have observed this also for gB [13]. Therefore the MuHV-4 gH and gB conformation switches were not simply pH-dependent as for Vesicular stomatitis virus glycoprotein G [16]: other features of the endosome environment were required. Nevertheless, low pH was essential for efficient gH and gB switching and for capsid release.
It is important to note that most gL 2 virions were non-infectious (Fig. 3). Therefore most of the prematurely capsids of gL 2 virions were destined not to reach nuclear pores. What gL 2 infection there was may have depended on a few of these capsids being transported, or on some virions still managing to follow the normal pathway to late endosomes without premature membrane fusion.

gL 2 virions show gB conformational instability
Since MuHV-4 membrane fusion, as defined by capsid release, is associated with a conformation change in gB [13], we further analyzed incoming gL 2 virions for changes in gB antigenicity. The gB of wild-type virions changed from BN-1A7 + MG-1A12 2 (prefusion) at the cell surface to BN-1A7 2 MG-1A12 + (post-fusion) in late endosomes (Fig. 4A). gL 2 virions attached to the plasma membrane were also BN-1A7 + MG-1A12 2 . However, unlike the wild-type, some acquired the MG-1A12 epitope before reaching LAMP-1 + endosomes. A premature gB switch to MG-1A12 + was also observed for gL 2 virions in NIH-3T3 cells (Fig. 6). The gLdependent conformational instability of gB was not marked, as considerable BN-1A7 staining remained outside LAMP-1 + endosomes. Nevertheless, as judged by capsid release this degree of instability was functionally important.
Flow cytometry of infected BHK-21 cells provided further evidence of gB conformational instability in the absence of gL   (Fig. 7). For an equivalent level of gN expression, BN-1A7 gB staining (pre-fusion) was weaker on cells infected with gL knockout viruses than with wild-type or revertant viruses, and MG-1A12 staining (post-fusion) was stronger. Thus, the premature capsid release of gL 2 virions reflected gH being in its down-stream gHonly form from the start, and gB engaging prematurely in membrane fusion after endocytosis.

gL deficiency compromises host entry
A key point with MuHV-4 is that in vitro molecular deficits can be linked to in vivo host colonization phenotypes. We have shown before that MuHV-4 gL knockouts colonize mice remarkably normally after intranasal inoculation of 10 4 plaque-forming units [20]. However, the in vitro entry deficit of gL 2 virions (Fig. 3) suggested that initial host entry might be impaired. Since host colonization by MuHV-4 is relatively dose-independent [24], such a deficit might not be apparent except at very low inoculation doses. We therefore compared gL 2 (2 independent mutants) and gL + (revertant and wild-type) viruses for their capacity to establish infection after intranasal inoculation of 1-100 p.f.u. (Table 1). gL disruption reduced significantly the rate of infection after low dose inoculation. Thus, while gL deficiency has little impact on MuHV-4 intra-host spread, it compromised severely the efficiency with which cell-free virions first entered naive hosts.

Discussion
Sheer complexity has made herpesvirus membrane fusion hard to unravel, for example Herpes simplex virus can undergo endosomal or plasma membrane fusion in different cell types [25]. However, the underlying molecular events are likely to be similar. Endocytic entry is easier to dissect because binding and fusion remain anatomically distinct. MuHV-4 capsid release after endocytic entry was associated with conformation switches in both gB and gH. Low pH was a key trigger of these switches and gL was a key negative regulator. Thus, by associating with newly synthesized gH in the endoplasmic reticulum gL sets virions for cell binding, and by dissociating again in late endosomes of the next cell it triggers membrane fusion.
Analogy with the Vesicular stomatitis virus glycoprotein G [15,16] would suggest that the conformation changes in the MuHV-4 gB are intrinsic to its fusion mechanism, with the stable BN-1A7 + or MG-1A12 + forms representing the respective preand post-fusion states. This general model does not apply so easily to gH (gH/gL being pre-fusion and gH-only post-fusion), as gL 2 virions fail to express gH/gL yet fuse more readily than the wildtype rather than less. It is possible that gL 2 virions express another, as yet unidentified, pre-fusion form of gH. However, we have found no antigenic evidence of other forms (data not shown). Our working hypothesis is therefore that gH/gL and gH-only are both pre-fusion -gH/gL being up-stream of gH-only -and that further gH conformation changes accompany fusion itself. In support of this, gH-only epitopes tended to disappear after a 4 h 37uC incubation, whereas MG-1A12-type gB epitopes were maintained (data not shown). Thus, gL dissociation from gH appears to prime gH for membrane fusion rather than being a part of the fusion reaction per se.
A pre-fusion conformation change cannot be ruled out for gB either. Heldwein et al. [15] proposed that their gB conformationwhich is probably a down-stream form [26] -undergoes further changes to reveal fusion loops. The neutralization epitopes mapped onto this structure [15] would also be difficult to explain if it were simply a post-fusion form. Thus, although the default explanation has to be that gB functions like the Vesicular stomatitis virus glycoprotein G because of their structural homology, it is also possible that the herpesvirus membrane fusion mechanism is different. gB and gH conformation changes must drive fusion, but this does not necessarily mean that every conformation change is part of the fusion mechanism. It is possible that the conformation changes described for Vesicular stomatits virus have for herpesviruses acquired a different function. But whatever the precise fusion mechanism of herpesviruses, a central role for the MuHV-4 gL in stabilizing gH and gB against engaging in fusion seemed clear.
The influence of gL on the conformation of gB was presumably indirect. A loss of gL would change gH/gL to gH-only and would therefore disrupt the interaction [27] between gH/gL and the gB N-terminus (gB-NT). This would not separate gH from gB, since their association is independent of both gL [21] and gB-NT [27], but could promote a conformation change in the gB extracellular domain. Consistent with such a scheme, MuHV-4 lacking gB-NT shows a gB conformational instability much like that of gL 2 mutants [27]. The gB-NT-gH/gL interaction does not appear to play a significant role in stabilizing gH/gL, since gB-NT-deficient mutants maintain gH/gL expression and infect normally [27]. Therefore the gB conformation change alone is not sufficient for fusion: gL rather than gB-NT is the key fusion regulator. We envisage that the late endosomal milieu disrupts the gH/gL interaction, with low pH playing an important permissive role; gH then changes its conformation to allow hemi-fusion [28] and displace gB-NT; this then allows gB to switch to its pro-fusion form and complete the fusion reaction. In this way, the whole of membrane fusion can be regulated by gL. The neatness of this mechanism for regulating virion entry suggests that it will prove common to many herpesviruses.