Characterization of the Herpes Simplex Virus (HSV) Tegument Proteins That Bind to gE/gI and US9, Which Promote Assembly of HSV and Transport into Neuronal Axons

Herpes simplex viruses 1 and 2 and varicella-zoster virus cause significant morbidity and mortality. One basic property of these viruses is the capacity to establish latency in the sensory neurons and to reactivate from latency and then cause disease in peripheral tissues, such as skin and mucosal epithelia. The transport of nascent HSV particles from neuron cell bodies into axons and along axons to axon tips in the periphery is an important component of this reactivation and reinfection. Two HSV membrane proteins, gE/gI and US9, play an essential role in these processes. Our studies help elucidate how HSV gE/gI and US9 promote the assembly of virus particles and sorting of these virions into neuronal axons.

ABSTRACT

The herpes simplex virus (HSV) heterodimer gE/gI and another membrane protein, US9, which has neuron-specific effects, promote the anterograde transport of virus particles in neuronal axons. Deletion of both HSV gE and US9 blocks the assembly of enveloped particles in the neuronal cytoplasm, which explains why HSV virions do not enter axons. Cytoplasmic envelopment depends upon interactions between viral membrane proteins and tegument proteins that encrust capsids. We report that tegument protein UL16 is unstable, i.e., rapidly degraded, in neurons infected with a gE−/US9− double mutant. Immunoprecipitation experiments with lysates of HSV-infected neurons showed that UL16 and three other tegument proteins, namely, VP22, UL11, and UL21, bound either to gE or gI. All four of these tegument proteins were also pulled down with US9. In neurons transfected with tegument proteins and gE/gI or US9, there was good evidence that VP22 and UL16 bound directly to US9 and gE/gI. However, there were lower quantities of these tegument proteins that coprecipitated with gE/gI and US9 from transfected cells than those of infected cells. This apparently relates to a matrix of several different tegument proteins formed in infected cells that bind to gE/gI and US9. In cells transfected with individual tegument proteins, this matrix is less prevalent. Similarly, coprecipitation of gE/gI and US9 was observed in HSV-infected cells but not in transfected cells, which argued against direct US9-gE/gI interactions. These studies suggest that gE/gI and US9 binding to these tegument proteins has neuron-specific effects on virus HSV assembly, a process required for axonal transport of enveloped particles.

IMPORTANCE Herpes simplex viruses 1 and 2 and varicella-zoster virus cause significant morbidity and mortality. One basic property of these viruses is the capacity to establish latency in the sensory neurons and to reactivate from latency and then cause disease in peripheral tissues, such as skin and mucosal epithelia. The transport of nascent HSV particles from neuron cell bodies into axons and along axons to axon tips in the periphery is an important component of this reactivation and reinfection. Two HSV membrane proteins, gE/gI and US9, play an essential role in these processes. Our studies help elucidate how HSV gE/gI and US9 promote the assembly of virus particles and sorting of these virions into neuronal axons.

INTRODUCTION

Alphaherpesviruses, such as herpes simplex virus (HSV), varicella-zoster virus (VZV), and the pig herpesvirus pseudorabies virus (PRV), replicate in mucosal epithelium or skin but then reside in neurons for the life of the infected individual. Reactivation of alphaherpesviruses from latency leads to the transport of virus particles into axons and anterograde transport from neuron cell bodies to axon tips (reviewed in reference 1). Anterograde axonal transport of HSV and PRV requires two viral membrane proteins, namely, gE/gI and US9 (2–4). gE/gI is a heterodimer of two type I membrane glycoproteins, gE and gI, which have substantial extracellular domains as well as relatively large cytoplasmic domains. Our early studies of the HSV gE/gI showed that there is very little gI not bound to gE and vice versa in infected cells, gE leaves the endoplasmic reticulum (ER) poorly without gI, and both gE and gI as a complex are required for all known functions of gE/gI, including anterograde transport (5–9). The HSV US9 protein is a type II membrane protein consisting of a transmembrane domain and a cytoplasmic domain without a substantial extracellular domain (10). HSV US9 is required for anterograde transport in neurons but has no obvious effects on virus replication or spread in cultured laboratory cells or in nonneuronal tissues in mice (4, 9, 11). The PRV US9 protein also appears to be neuron specific and mediates anterograde transport (12–16). PRV similarly relies on gE/gI for anterograde transport (15, 17–19). Our studies on how HSV gE/gI and US9 promote anterograde transport relied on double mutants lacking both gE and US9 because the loss of gE or US9 alone reduced transport by approximately 50%, but the loss of both gE and US9 reduced transport by 95% (4). With PRV, the loss of US9 had a more profound effect on anterograde transport, compared with the loss of gE/gI (2, 3, 16).

One model for how HSV and PRV gE/gI and US9 promote anterograde transport in neuronal axons proposes that gE/gI and US9 function in axons to tether vesicles containing virus particles onto kinesin motors moving anterograde along microtubules (3, 20–22). PRV enveloped particles are transported in axons in conjunction with kinesin-3 cargo molecules, and a dominant negative form of the KIF-1A (kinesin-3) protein reduced PRV anterograde transport (20). There was evidence for direct interactions between the PRV US9 and KIF1A (20). In contrast, HSV particles are transported by a different family of kinesins, kinesin-5 proteins (23, 24). HSV transport was blocked by expressing microRNAs (miRNAs) targeting kinesin-5 motors (24). There was a report that suggested that HSV US9 also binds directly to kinesins, but these studies involved in vitro pulldowns involving US9 and kinesins produced in bacteria and did not provide evidence that this occurs in HSV-infected neurons or evidence for the functional importance of this interaction (23).

However, this model does not fit with two sets of observations that suggest that US9 or gE/gI play no important role during the transport of HSV and PRV particles within axons. First, we constructed an HSV double mutant lacking both gE and US9, and this mutant transported very few virus particles into axons, approximately 5% of that transported into axons by wild-type HSV (4). Importantly, the few HSV particles that were transported into axons were transported normally, with similar kinetics of transport and without increased stalling of particles. Similarly, a PRV US9− mutant with major defects in anterograde transport displayed virus particles that moved with normal kinetics within axons (25). Thus, the few HSV and PRV particles that enter axons from neuron cell bodies are transported normally without gE/gI and US9.

If the loss of gE/gI and US9 does not affect alphaherpesvirus transport within axons, then to explain the very few particles observed in proximal segments of axons, viruses must be unable to enter axons. Our electron microscopic studies of the cell bodies of neurons infected with the HSV gE−/US9− double mutant found defects in virus assembly in the cytoplasm of neurons and sorting of virus from sites of assembly into axons (26). The gE−/US9− mutant accumulated large quantities of unenveloped and partially enveloped capsids in the cytoplasm, which were mostly adhered onto cytoplasmic membranes. These defects in secondary envelopment with the gE−/US9− HSV mutant can largely explain the defects in axonal transport because enveloped virions are the major form of transport in these neurons. An HSV mutant lacking only gE produced almost normal quantities of enveloped virions, but they accumulated in the cytoplasm of neurons and not on cell surfaces, which was the case with wild-type HSV (26). Thus, there was also missorting of enveloped particles in the neuronal cytoplasm with the loss of gE/gI, which might also explain some of the reduced anterograde transport and might involve kinesin motors. These studies suggest that a major function of HSV gE/gI and US9 in neurons is to orchestrate assembly of virus particles and cytoplasmic sorting so that enveloped virions are produced and can enter axons. What is particularly interesting is that US9 is largely or entirely neuron specific (4, 9, 11) and, thus, apparently participates in this assembly of viral proteins in a neuron-specific manner (26).

Secondary envelopment involves tegument proteins found on the surfaces of cytoplasmic HSV capsids that bind onto other tegument proteins and membrane proteins that are present on cytoplasmic membranes causing the wrapping of membranes around the capsids (reviewed in references 27 and 28). Previous studies that were all performed in nonneuronal cells demonstrated that HSV and PRV membrane glycoproteins gB, gD, gE/gI, and gM are involved in secondary envelopment (29–34). In every case, it required virus mutants lacking two of these glycoproteins to uncover substantial defects in envelopment or assembly; mutants lacking just one membrane protein showed more minor defects. There are numerous tegument proteins involved in bridging capsids onto the cytoplasmic domains of these glycoproteins. They include a complex of the following four tegument proteins: VP22, UL11, UL16, and UL21. UL16 interacts with gE/gI (34–39). VP22 interacts with gE and gM (32, 36, 40), and UL21 and UL11 interact with UL16 and gE (34, 41, 42).

As a first step toward understanding how HSV gE/gI and US9 promote assembly in neurons, we chose to characterize whether the loss of gE and US9 negatively impacted any of the tegument proteins listed above. In cells lacking gE and US9, UL16 was unstable, i.e., more rapidly turned over. Given that UL16 interacts with UL11, UL21, and VP22 as well as gE, we extended these studies by determining which of these tegument proteins interacted with the cytoplasmic (CT) domains of gE, gI, and US9. Little or nothing is known about whether gI and US9 bind tegument proteins; yet, gI and US9 have relatively large cytoplasmic domains enriched in cell sorting motifs similar to that of gE. In HSV-infected cells, there was good evidence that VP22, UL16, and UL11 interacted with the CT domain of gE and UL21 interacted with the gI CT domain. In transfected cells, there was evidence that VP22 and UL16 bound directly to gE. All four tegument proteins UL16, UL11, UL21, and VP22 were pulled down from infected cells with US9, and there was evidence for direct binding of VP22 and UL16 to US9 in transfected cells. Observations that both gE- and US9-bound UL16 might explain the instability of UL16 in cells infected with a gE−/US9− double mutant. We observed larger quantities of all of the tegument proteins bound to gE/gI and US9 in infected cells than those in transfected cells. This finding suggested that the assembly of a web of tegument proteins in infected cells leads to more of the tegument proteins pulled down with gE/gI and US9. Similarly, we found evidence that gE/gI and US9 do not directly interact in transfected cells but are apparently coprecipitated by this web of tegument proteins.

RESULTS

Characterization of tegument protein expression and membrane association in cells infected with the gE−/US9− double mutant.

One hypothesis to explain the defects in secondary envelopment associated with the loss of gE and US9 (26) suggests that certain tegument proteins do not associate with cytoplasmic membranes at sites of envelopment. To test this hypothesis, we infected CAD neurons with F-gE−/US9−, a virus unable to express gE and US9, or F-gE−/US9−R, a repaired version of this virus (26). CAD cells are mouse catecholaminergic central nervous system cells that can be differentiated into neurons and are readily infected with HSV (26, 43). Infected neurons were either harvested as whole-cell extracts reflecting the total protein in cells by solubilizing cells with 2% SDS and boiling the extracts or the cells were swollen in hypotonic buffer and subjected to Dounce homogenization. Membranes from the disrupted cells were mixed with sucrose producing 60% sucrose and placed at the bottom of ultracentrifuge tubes that are overlaid with 45% and 20% sucrose and then centrifuged for 16 to 18 h at 200,000 × g. Membranes floated up in the tubes through 45% sucrose and were found at the interface with 20% sucrose as described (34). In cells infected with the gE−/US9− double mutant, there was ∼10% the UL16 found in the membrane fraction, compared with the repaired virus (Fig. 1, bottom right panel). VP22, UL11, glycoprotein gD, and the major capsid protein VP5 were all found at normal levels in the membrane fraction, comparing mutant with repaired virus. However, the level of UL16 observed in whole-cell extracts of cells infected with F-gE−/US9− was also reduced to ∼10%, compared with UL16 in cells infected with the repaired virus (Fig. 1, bottom, left lanes). No defects in the expression of VP22 or UL11 were observed. This finding suggested that UL16 was not expressed well or was not stable in cells infected with the HSV gE−/US9− double mutant.

FIG 1.

Expression of tegument proteins in cells infected with a gE−/US9− double mutant and membrane association of these tegument proteins. CAD neurons were infected with HSV gE−/US9− lacking both gE and US9 or gE−/US9−R (repaired virus) for 18 h. Whole-cell extracts of the cells were made using Laemmli buffer (left panels). Other cells were pelleted and suspended in hypotonic buffer (10 mM Tris-HCl and 0.2 mM MgCl₂ [pH 7.4]) so that cells swelled for 20 min on ice. The cells were Dounce homogenized. Sucrose was added to cell extracts to 60% sucrose, and then these extracts were placed at the bottom of ultracentrifuge tubes and overlaid with 45% sucrose and 20% sucrose. The tubes were centrifuged at 200,000 × g for 16 h, and then membranes that floated to the interface between 40% and 25% sucrose were harvested. These membranes were pelleted and solubilized in Laemmli buffer (membrane floats, right panels). Whole-cell extracts and membrane float extracts were subjected to SDS-PAGE and then Western blotted with antibodies specific for VP5, gD, VP22, UL11, or UL16. Numbers along the bottom of the figure indicate the quantities of UL16 immunoprecipitated compared with that in the whole-cell extract for F-gE/US9/GFP-R.

To characterize whether there was reduced expression of UL16 or whether the protein was unstable, we performed pulse-chase experiments. Tegument proteins were labeled with [³⁵S]-methionine/cysteine for 40 min (pulse), and then the label was chased for 75 (C1) or 160 min (C2). Approximately 60% of UL16 was lost over 160 min in cells infected with the gE−/US9− double mutant (Fig. 2, bottom lanes). In two other experiments there was a 48% to 68% loss of UL16 in the 160-min chase sample. There was only 17% loss of UL16 in cells infected with the repaired virus. In the other experiments, the loss of UL16 in cells infected with repaired virus varied from 11% to 23%. While there was a more minor loss of UL21 observed over 160 min of chase, this did not differ when comparing cells infected with the gE−/US9− mutant to cells infected with the repaired virus (Fig. 2). VP22 was not obviously lost during the chase samples. Note that UL11 did not label or immunoprecipitate well in these experiments. We concluded that UL16 is less stable and degraded more rapidly in cells lacking gE and US9, while other teguments were not degraded more rapidly.

FIG 2.

Pulse-chase radiolabeling of tegument proteins in neurons infected with gE−/US9− HSV. CAD cells at 18 h postinfection with F-gE−/US9− or F-gE−/US9−R cells were labeled with [³⁵S]-methionine/cysteine for 40 min and chased for 75 (C1) and 160 min (C2). Cell extracts were made using RIPA (1% NP-40, 0.5% DOC, and 0.5% SDS), and cell extracts were immunoprecipitated using rabbit polyclonal antibodies specific for VP22, UL21, or UL16. Numbers along the bottom of the UL16 panel indicate the percentage of total UL16 remaining in the chase samples compared with that in the pulse samples.

Binding of tegument proteins to gE and gI in HSV-infected cells.

The instability of UL16 in neurons suggested that the loss of gE and US9 might alter the multiprotein matrix that is HSV tegument. To investigate this further, we chose to study the tegument proteins that bind to gE, gI, and US9 using immunoprecipitation of cell extracts derived from HSV-infected cells. We believe that this approach produces a more accurate picture of protein-protein interactions that exist in infected cells, compared with in vitro assays (adding proteins derived from bacteria into cell extracts) or immunofluorescence colocalization studies. One difficulty in showing that gE/gI binds tegument proteins is that gE/gI is an IgG Fc receptor (5). In most cases, rabbit antibodies (Abs) specific for tegument proteins were produced to characterize these proteins (34, 37, 38). Rabbit IgG binds well to the gE/gI Fc receptor (6, 44). Mouse IgG does not bind to the gE/gI Fc receptor in immunoprecipitation experiments (36). Therefore, here, we used two mouse monoclonal Abs (MAbs) 3114, specific for gE, and 3104, specific for gI, to immunoprecipitate gE/gI. These studies substantially extended the previous studies that focused on tegument proteins that bind to gE by including gI and US9.

CAD neurons were infected with wild-type (WT) HSV or a mutant lacking the entire CT domain of gE, F-gEdeltaCT (45). Without the CT domain, gE should be incapable of binding tegument proteins. However, since gE is entirely or largely complexed with gI, gEdeltaCT binding of tegument proteins will involve the gI CT domain. gEdeltaCT is able to complex with gI, is incorporated into virions, and traffics to cell junctions (45). Cell extracts were made using Nonidet-P40 (NP-40)/sodium deoxycholate (DOC) lysis buffer, which contains 1% NP-40 and 0.5% DOC, and then gE/gI was immunoprecipitated using a mixture of MAbs 3114 and 3104. As a control, NP-40/DOC extracts were also immunoprecipitated with MAb 14-4b, which is specific for human cytomegalovirus gH and does not react with HSV gH or bind to the gE/gI Fc receptor. The immunoprecipitated proteins were subjected to Western blotting with rabbit antibodies specific for VP22, UL16, UL21, or UL11. For comparison, whole-cell extracts were made by boiling the cells in buffer containing 2% SDS. VP22 was precipitated from WT HSV-infected neurons that express the CT domains of both gE and gI but not from neurons that were infected with F-gEdeltaCT, which expresses only the gI CT domain (Fig. 3). Similarly, UL16 and UL11 precipitated with WT gE/gI but not with gEdeltaCT/gI. In contrast, UL21 was precipitated with both WT gE/gI and gEdeltaCT at similar levels (Fig. 3). In two other experiments, we observed similar results (not shown). We concluded that VP22, UL16, and UL11 bind primarily to the gE CT domain or require this for binding to gE/gI. UL21 was different, as it was precipitated with gE/gI and gEdeltaCT/gI, suggesting that UL21 either binds directly to gI or binds to gE/gI indirectly in a manner that involves the gI CT domain.

FIG 3.

Tegument proteins that interact with gE/gI and gE deltaCT/gI in HSV-infected cells. CAD neurons were infected with HSV-F or F-gEdeltaCT, which expresses a form of gE/gI lacking the cytoplasmic domain of gE for 18 h. The cells were harvested in NP-40/DOC lysis buffer or in buffer containing 2% SDS and boiled for whole-cell extracts. NP-40/DOC extracts were immunoprecipitated using a combination of 2 MAbs, 3114 specific for gE and 3104 specific for gI, or with MAb 14-4b, an irrelevant Ab specific for HCMV gH/gL. Gel electrophoresis of immunoprecipitated proteins or whole-cell extracts was performed followed by Western blotting probing with rabbit polyclonal antibodies specific for UL16, VP22, UL21, or UL11. Numbers along the bottom of the panels indicate the percentage of immunoprecipitated proteins compared with that in WT whole-cell extracts.

Binding of tegument proteins to US9 in HSV-infected cells.

To assess whether these tegument proteins bound to US9, we infected CAD neurons with an HSV recombinant, denoted F-US9-HA, expressing an epitope-tagged US9 protein modified with a 9-amino acid influenza virus hemagglutinin (HA) epitope inserted at the N terminus (9). Previous studies showed that F-US9-HA replicated normally in neurons and the US9-HA functioned in anterograde transport just as WT US9 (9). Cells were infected with WT HSV or F-US9-HA, cell extracts were produced using NP-40/DOC lysis buffer, and then US9 was immunoprecipitated with an HA epitope-specific mouse MAb. As a positive control, gE/gI was immunoprecipitated with MAbs 3114 and 3104. The irrelevant mouse MAb 14-4b was used to characterize nonspecific precipitation. Again, VP22, UL16, UL21, and UL11 all coprecipitated with gE/gI derived from both WT- and F-US9-HA-infected neurons (Fig. 4). Immunoprecipitation of US9 with the HA MAb produced relatively strong signals for UL16 and UL21. Weaker signals were observed with VP22 and UL11 blots following US9 immunoprecipitation. Nevertheless, the observed levels of VP22 and UL11 precipitated with US9 were above background levels observed with the irrelevant MAb 14-4b in this experiment (Fig. 4) and several other experiments (not shown). We concluded that UL16, UL21, VP22, and UL11 all precipitate with US9 from infected neurons.

FIG 4.

Tegument proteins that interact with US9 in HSV-infected cells. CAD neurons were infected with HSV-F or HSV-US9-HA for 18 h. The cells were harvested in NP-40/DOC lysis buffer or in buffer containing 2% SDS and boiled (whole-cell extracts). NP-40/DOC extracts were immunoprecipitated using mouse antibodies against gE and gI (MAb 3104 and 3114), a MAb specific for HA, or with MAb 14-4b, an irrelevant Ab specific for HCMV gH/gL. Immunoprecipitated proteins or whole-cell extracts were subjected to electrophoresis, and then Western blots were probed with rabbit polyclonal antibodies against UL16, VP22, UL21, or UL11. Numbers along the bottom of the panels indicate the percentage of immunoprecipitated proteins compared with that in WT whole-cell extracts.

Interactions between gE/gI and tegument proteins in transfected cells.

To examine whether individual tegument proteins bound to gE/gI in the absence of other HSV proteins, we transfected CAD neurons with plasmids encoding VP22-GFP, UL16, UL21-HA, or UL11-GFP and gE/gI. There is evidence that VP22-GFP (46), UL11-GFP (47, 48), and UL21-HA (34, 42) all function normally during infection. gE/gI was immunoprecipitated with MAbs 3114 and 3104, and then Western blots were performed using rabbit antibodies specific for VP22, UL16, UL21, or UL11. The quantities of VP22 and UL16 immunoprecipitated along with gE/gI (9% and 11% of the total in whole-cell extracts) (Fig. 5) were significantly lower (about one-third) than those observed in HSV-infected cells (Fig. 3). However, the irrelevant Ab control 14-4b showed very little or no VP22 and UL16, suggesting that VP22 and UL16 observed in the immunoprecipitated gE/gI were specific. In two additional experiments, 8% and 11% of VP22 and 10% and 12% of UL16 were found associated with gE/gI and the 14-4b controls were approximately 1%. Only 4% of UL21 bound to gE/gI in transfected cells, although again the 14-4b control showed less UL21 (1%). Approximately 14% of UL11 was precipitated with gE/gI, but 8% of UL11 was observed following precipitation with MAb 14-4b (Fig. 5). The difference between 14% and 8% for UL11 binding to gE/gI was not as clear-cut as with the other three tegument proteins, although other experiments showed similar results. Note that we previously characterized UL11 as a relatively “sticky protein” (36). We concluded that there was evidence for direct binding of VP22 and UL16 to gE/gI in transfected cells suggesting direct interactions. There was also some evidence that UL21 and UL11 might directly bind gE/gI, but binding of both of these tegument proteins was less extensive and there was a larger nonspecific component with UL11.

FIG 5.

Interactions between tegument proteins and gE/gI in transfected neurons. CAD neurons were transfected with plasmids encoding gE/gI and also with either UL16, UL21-HA, VP22-GFP, or UL11-GFP for 24 h, and cells were harvested in NP-40/DOC buffer or whole-cell extracts made using buffer containing 2% SDS. NP-40/DOC extracts were subjected to immunoprecipitation with a combination of MAbs, namely, 3114 specific for gE and MAb 3104 specific for gI or MAb 14-4b that is an irrelevant antibody specific for HCMV gH/gL. This was followed by gel electrophoresis and Western blots probed with rabbit polyclonal antibodies against UL16, VP22, UL21, or UL11. Numbers along the bottom of the panels indicate the percentage of immunoprecipitated proteins compared with that in whole-cell extracts.

Interactions between US9 and tegument proteins in transfected cells.

To determine whether individual tegument proteins could bind to US9 in the absence of other viral proteins, we transfected CAD neurons with plasmids containing VP22-GFP, UL16, UL11-GFP, or UL21-HA. To deliver US9, the cells were infected with a nonreplicating (E1A−) adenovirus (Ad) vector expressing an HA epitope-tagged US9 protein for an additional 24 h after transfection. Expression of US9-HA using the Ad vector was superior to that observed by transfection (not shown). Following immunoprecipitation of US9-HA with the HA-specific MAb, VP22 (14% of the total VP22 in cells) and UL16 (7% of the UL16 in cells) were observed (Fig. 6A). In two additional experiments, 11% and 14% of VP22 and 7% and 10% UL16 were found associated with US9. There was also evidence for binding of UL11 to US9, but again the background with irrelevant MAb 14-4b was higher with UL11 than that of other tegument proteins (Fig. 6A). Because both US9 and UL21 contain HA tags, we could not use the mouse HA-specific MAb to pull down US9. Instead, we used UL21-specific rabbit polyclonal antibodies to immunoprecipitate UL21 and then blotted with US9 rabbit antibodies. There was no obvious UL21 bound to US9 (Fig. 6B). This finding suggested that US9 interacts directly with VP22 and UL16.

FIG 6.

Interactions of tegument proteins with US9 in transfected/transduced neurons. CAD neurons were transfected with plasmids containing UL16, UL21-HA, VP22-GFP, or UL11-GFP and subsequently infected with Ad-US9-HA (multiplicity of infection [MOI], 20) for 24 h. Cells were lysed using NP-40/DOC buffer for immunoprecipitation (IP) or whole-cell extracts made using 2% SDS. (A) gE/gI was IPed using a MAb specific for HA, followed by gel electrophoresis and Western blots probed with rabbit polyclonal antibodies specific for UL16, VP22, or UL11. 14-4B was used as a control nonrelevant antibody. (B) As both UL21 and US9 proteins contained an HA tag, cell extracts in lysis buffer were IPed using rabbit antibodies against UL21 followed by gel electrophoresis and Western blots probed with rabbit polyclonal antibodies against US9. 14-4B was used as a control nonrelevant antibody. Numbers along the bottom of the panels indicate the percentage of immunoprecipitated proteins compared with that in whole-cell extracts.

Interactions between US9 and gE/gI in HSV-infected cells.

There was a report that HSV US9 forms a complex with gE/gI that can be immunoprecipitated from HSV-infected cells but only after extraction in nonionic detergents (49). We investigated interactions between US9 and gE/gI in HSV-infected cells by infecting cells with WT HSV or F-US9-HA and performing immunoprecipitation followed by Western blotting. In Fig. 7A, we infected CAD neurons with WT HSV and then produced lysates using NP-40/DOC lysis buffer or buffer containing the nonionic detergent NP-40 and no DOC. Cell extracts were subjected to immunoprecipitation using MAbs 3114 (gE) and 3104 (gI) or irrelevant MAb 14-4b, and then proteins were separated by electrophoresis and Western blotted with rabbit polyclonal US9-specific Abs. There was no obvious US9 associated with gE/gI in these samples. In Fig. 7B, we infected neurons with F-US9-HA, derived cell extracts with either NP-40/DOC or NP-40 buffers, and then immunoprecipitated them with the HA MAb or MAb 14-4b. In this case, there were substantial quantities of gE (25% of the total) detected in the proteins immunoprecipitated with HA antibody that pulled down US9, which were significantly higher than that observed with the irrelevant MAb 14-4b (6%). Note that only 13% of US9 was precipitated with the total amount of US9 which is common in immunoprecipitations involving tags. We observed more gE coprecipitated with US9 with the NP/DOC buffer perhaps because there was more extensive solubilization of gE/gI or other proteins. In two other experiments 25% and 29% of the gE was associated with US9, and the irrelevant antibody showed 3% and 6% of gE associated with US9. Unlike the previous studies extracts that only observed gE/gI and US9 interactions in a nonionic detergent (49), we observed that a combination of ionic and nonionic detergents produced better coprecipitation of gE/gI and US9.

FIG 7.

Interaction of gE/gI and US9 proteins. CAD neurons were infected with HSV-F (A) or HSV-US9-HA (B) for 18 h. The cells were subsequently harvested in lysis buffer consisting of NP-40/DOC or in NP-40 lacking the DOC and boiled in buffer containing 2% SDS for whole-cell extracts. (A) HSV-F-infected cells were extracted in lysis buffer NP-40/DOC or NP-40 and immunoprecipitated using mouse antibodies against gE and gI (3104 and 3114), followed by gel electrophoresis and Western blots probed with rabbit polyclonal antibody against US9. (B) HSV-US9-HA-infected cells were extracted in lysis buffer as above and were immunoprecipitated using mouse antibodies against HA, followed by gel electrophoresis and Western blotting. The blot was cut in half and the top probed with mouse anti-gE (3104) and bottom probed with rabbit polyclonal antibody against US9 to verify US9-HA pull down by the HA antibody. 14-4B was used as a control nonrelevant antibody. Numbers along the bottom of the panels indicate the percentage of immunoprecipitated proteins compared with the whole-cell extract.

Interactions of US9 and gE/gI in transfected cells.

To address whether gE/gI and US9 could interact in the absence of tegument and other HSV proteins, we transduced cells with gE/gI and US9. CAD neurons were transfected with plasmids expressing gE and gI for 24 h and then transduced with the Ad vector expressing US9-HA for an additional 24 h. US9 was immunoprecipitated with the HA-specific MAb or the irrelevant MA 14-4b, and then precipitated proteins were subjected to Western blotting for gE or US9. As in Fig. 7, the HA MAb precipitated about 12% of the total US9 (Fig. 8). There was 6% of the total gE present in the US9-precipitated sample, but this was not different from that precipitated with 14-4b. Several other experiments showed similar results, with little or no differences in the levels of gE pulled down with US9-HA versus the irrelevant antibody. In comparison, in Fig. 7, we observed 25% of the total gE precipitated with US9-HA and 6% of the gE pulled down with MAb 14-4b. We concluded that gE/gI and US9 do not directly interact in transfected cells. It is likely that in infected cells, gE/gI and US9 bind to tegument proteins that form matrix or web so that gE/gI and US9 coprecipitate with this matrix and appear to interact with each other.

FIG 8.

Interaction of gE/gI with US9 in transfected cells. CAD neurons were transfected with plasmids encoding gE/gI for 12 h and then subsequently infected with Ad-US9-HA (MOI, 20) for an additional 24 h. Cells were lysed using NP-40/DOC buffer for immunoprecipitation or whole-cell extracts were made using 2% SDS. US9 was immunoprecipitated using a MAb specific for HA, and 14-4B was used as a control nonrelevant antibody. Proteins were subjected to gel electrophoresis; the gel was cut in half and the top section was probed with mouse MAb 3114 specific for gE and the bottom section probed with rabbit polyclonal antibodies specific for US9. Numbers along the bottom of the panels represent the percentage of immunoprecipitated gE or US9 compared with that in the whole-cell extract. Whole-cell extract and IP lanes are shown with lighter and darker exposures, respectively.

DISCUSSION

Our previous studies demonstrated that an HSV mutant lacking both gE and US9 was largely unable to transport virus particles into the most proximal sections of neuronal axons (4). This could be largely or entirely explained by major defects in the assembly of enveloped virus particles in neurons (26). These effects were primarily observed in neurons, and there were only minor defects in secondary envelopment in epithelial cells infected with the gE−/US9− double mutant (26). These neuron-specific defects in virus assembly are remarkable and are likely explained by neuron-specific proteins that the virus uses in assembly or sorting of viral proteins. One such protein is HSV US9, which exclusively functions in neurons in vivo and in cultured cells (4, 9, 11).

Given that it is well established that secondary envelopment requires interactions between membrane-anchored HSV proteins and tegument proteins that together anchor membranes onto capsids, we examined whether the loss of both US9 and gE/gI might reduce tegument protein association onto cytoplasmic membranes. Membrane float experiments showed that VP22, UL11, and the major capsid protein VP5 were all bound to cytoplasmic membranes normally in neurons infected with the HSV gE−/US9− mutant. However, the quantity of the UL16 protein associated with membranes in neurons was reduced by ∼90% comparing the gE−/US9− mutant with a repaired HSV. This was not related to changes in UL16 membrane binding but instead was because UL16 was less stable or degraded by proteolysis in cells infected with the double mutant, so that there was a 90% reduction in the quantity of UL16 over time. This instability of UL16 was a striking observation because UL16 is an important component of tegument, binding to both capsids and viral membrane proteins and bridging to other tegument proteins, such as UL11, UL21, and VP22 (32, 34, 36, 38, 41, 50).

In order to further characterize how the loss of gE and US9 reduced secondary envelopment, we examined the binding of tegument proteins to gE, gI, and US9. There have been extensive studies showing that gE binds at least four tegument proteins, namely, UL11, UL16, UL21, and VP22 (32, 34, 36, 37, 50, 51). Many of the previous studies describing tegument-tegument and tegument-gE interactions involved in vitro binding of proteins produced in bacteria. These in vitro studies have difficulty avoiding nonspecific binding or artifacts. Proteins produced in bacteria were often mixed into cell extracts at nonphysiologic concentrations. There are highly charged domains in the CT domain of gE and US9, e.g., six consecutive arginine residues (R56 to R61) in US9 (10, 36, 52, 53). These domains have the potential to act as ion exchangers and bind multiple proteins nonspecifically. Moreover, there have been multiple studies of tegument protein interactions with gE but no studies characterizing binding to gI or US9. It seemed likely that both gI and US9 are also involved in tegument binding and assembly because both have relatively large CT domains that contain numerous cell sorting motifs like the gE CT domain (10, 36, 52–54).

We chose to characterize interactions between tegument proteins and gE/gI and US9 initially by immunoprecipitating gE/gI and US9 from infected CAD neurons. To address binding of tegument proteins to gI versus gE, we used HSV F-gEdeltaCT that expresses a form of gE lacking the CT domain. Significant quantities of VP22, UL16, and UL11 were pulled down with WT gE/gI (28.5% to 37% of the total of these proteins in cells) but not with gEdeltaCT/gI. Much lower levels of these tegument proteins were pulled down with an irrelevant MAb. In contrast, UL21 bound to both WT gE/gI and gEdeltaCT/gI at about equal levels. We concluded that VP22, UL16, and UL11 can be immunoprecipitated with gE/gI and that requires the gE CT domain, but UL21 pulldown does not require the gE CT and likely involves the gI CT domain. We also found that all four of the tegument proteins, namely, VP22, UL16, UL21, and UL11, were immunoprecipitated with US9 from HSV-infected cells, although the VP22 and UL11 pulldowns were weaker.

The data from infected cells do not prove that there are direct interactions between the gE, gI, and US9 CT domains and any or all of these tegument proteins. We know that these tegument proteins assemble into a matrix or web of proteins that can assemble onto membranes and capsids, and our data suggest that gE, gI, and US9 all contribute to that assembly in infected cells. To address whether any or all of these tegument proteins directly interact with gE/gI, i.e., in the absence of other HSV proteins, we transfected cells and performed pulldown experiments. There were generally lower quantities of each of the tegument proteins that precipitated with gE/gI in transfected cells (∼10% the total of the proteins observed in whole-cell extracts) than ∼25% to 38% of these proteins pulled down from extracts of infected cells. However, it was clear that ∼10% of the total VP22 and UL16 bound to gE/gI, compared with the irrelevant Ab 14-4b that precipitated ∼1% of the VP22 and UL16. There were smaller quantities of UL21 (∼4%) pulled down with gE/gI. But again ∼1% was observed with the irrelevant MAb. UL11 was also precipitated with gE/gI (14%), but the control Ab 14-4b precipitated 8% of the total UL11. It is possible that tegument proteins bind to gE/gI through interactions with cellular proteins, although clearly there is a gE/gI-specific component. Therefore, we believe that it is likely that UL16 and VP22 bound directly to gE/gI.

One might argue that the binding of UL11 and UL21 to gE/gI in transfected cells is weak. It is difficult to address the affinity of these protein-protein interactions in pulldown experiments, and thus, we cannot comment on the weakness of UL11 or UL21 binding. We performed multiple experiments, each internally controlled by using an irrelevant antibody, and in every case, we found similar results, as follows: there was always more UL21 and UL11 bound to precipitates containing gE/gI than precipitates with an irrelevant antibody. UL11 has been characterized by us as sticky in experiments showing that UL11 bound to gD lacking the cytoplasmic domain about half as well as to wild-type gD (36). In addition, UL11 and other tegument proteins may fall out of solution and be pelleted (even at low speeds) from cell extracts during immunoprecipitation. The nature of these proteins is to aggregate and form matrices. Antibodies, especially the monoclonal antibodies used to precipitate gE/gI and US9 here, do not always efficiently precipitate antigen in cell extracts. We routinely obtain only 10% to 25% of the total gE/gI expressed in cells by transfection measuring what was pulled down in Western blots. Thus, obtaining 4% to 11% of the total tegument protein precipitated with gE/gI is not insignificant. One should also consider that these experiments were performed in transfected cells where the proteins are not all wrapped up with one another. Even so, it remains that there was always more UL11 and UL21 seen in gE/gI immunoprecipitates than the irrelevant antibody control. That said, the lower levels of UL21 and higher levels of nonspecific immunoprecipitation of UL11 casts doubt on whether these tegument proteins bind directly to gE/gI.

Similar experiments were performed transfecting CAD neurons with VP22, UL16, UL11, and UL21 and transducing cells with a nonreplicating Ad vector expressing HA-tagged US9. As with gE/gI, there was good evidence that VP22 and UL16 bound to US9. Once again, UL11 precipitated with US9, but the background with the irrelevant antibody was high. We observed no US9 precipitated with an UL21 antibody in these experiments. The observations that UL16 binds directly to both gE/gI and US9 may explain the instability of UL16 in cells infected with the gE−/US9− double mutant. Perhaps, if UL16 is not anchored onto membranes via US9 and gE/gI or sorted to specific cytoplasmic membranes where assembly occurs, the protein is less stable.

It was telling that the quantities of tegument proteins that were pulled down with gE/gI or US9 from infected cells was substantially higher than those observed from transfected cells. For example, 37% of the VP22 found in infected cells was precipitated with gE/gI, but only 9% of the VP22 in transfected cells was associated with gE/gI. We observed 21% of the total UL21 precipitated with gE/gI and 32% with US9 from infected cells, but only 4% of UL21 (4%) with gE/gI and no UL21 with US9 in transfected cells. Tegument apparently forms a dense web of interwoven proteins anchored onto membranes via gE/gI, US9, gD, and gM and other membrane proteins. We hypothesize that the larger quantities of tegument proteins pulled down with gE/gI or US9 from infected cells reflects this cross-linked web composed of different tegument and membrane proteins. As with the extracellular matrix of cells, this matrix of cross-linked tegument proteins survives extraction in nonionic and milder ionic detergents.

It was reported that HSV gE/gI and US9 form a complex that can be immunoprecipitated from HSV-infected cells (49). Similarly, coimmunoprecipitation of PRV US9 and gE/gI from infected cells was also described (22). We found similar results, in HSV-infected neurons; there was ample gE (likely in the form of gE/gI) precipitated along with US9-HA. However, in transfected cells, we found no evidence that gE was precipitated with US9-HA. We concluded that US9 and gE/gI can be pulled down together from extracts of infected cells, but these two membrane proteins apparently do not interact directly with one another in transfected cells. These observations strengthen the conclusions that there is a web of tegument proteins bound to both gE/g and US9 so that immunoprecipitation with US9 antibodies brings down gE/gI that is connected to US9 by a tegument protein bridge. This gives the impression that gE/gI and US9 interact when these interactions are actually indirect.

In summary, these studies help to further explain how HSV gE/gI and US9 promote axonal transport of virus particles by promoting the assembly of HSV enveloped particles in the cytoplasm of neurons. The gE (US8), gI (US7), and US9 genes likely coevolved with large cytoplasmic domains that function in assembly but also to sort viral proteins and virus particles in polarized cells and into axons (27). US9 is particularly fascinating in this regard because this protein appears to specifically function in neurons (11, 26). Previous studies had shown that gE binds some of the tegument proteins described here, but we extended this analysis to US9 that, like gE, binds VP22 and UL16, as well as gI that apparently binds UL21. There was also evidence for some binding of UL11 to both gE/gI and US9, but there was also a substantial nonspecific component to UL11 binding to these membrane proteins. We also found evidence that this contradicts the previous conclusions (22, 49) that gE/gI binds directly to US9. Together, these studies support the hypothesis that the gE, gI, and US9 CT domains orchestrate envelopment of HSV particles in neurons, a process necessary for egress into axons.

MATERIALS AND METHODS

Viruses.

HSV-1 mutant F-gE/US9-GFP, in which the gE (US8) and US9 genes were replaced with the enhanced green fluorescent protein (EGFP), and the repaired virus F-gE/US9-GFP-R were constructed from a bacterial artificial chromosome copy of HSV-1 strain F as previously described (4). HSV-1 recombinant HSV-gEdeltaCT, which lacks the entire cytoplasmic domain of gE, was described previously (45). HSV-1 recombinant F-US9-HA, which expresses a US9 protein modified with the addition of an N-terminal 9-amino acid influenza virus hemagglutinin (HA) epitope was described in reference 9. A nonreplicating (E1−) adenovirus expressing a HSV-1 F strain US9 protein modified with the N-terminal HA tag was constructed as described before (55). In brief, the US9-HA open reading frame was inserted into the plasmid pDC316(io) containing an Ad type 5 (Ad5) E1 region with a murine cytomegalovirus (MCMV) immediate early promoter separated from transgenes by a lac repressor binding site which allows downregulation of the transgene in 293 IQ cells that express the Lac repressor protein. This plasmid was cotransfected with plasmid pBHGloxΔE1,3Cre, which contains the remainder of the Ad5 genomic plasmid, into 293 IQ cells (56). Expression of US9-HA was verified by Western blotting using both HA- and US9-specific antibodies; Ad vectors were then propagated in 293 IQ cells, and virus particles were purified using CsCl gradient centrifugation.

Plasmids.

Plasmids containing UL16, UL21-HA (34, 42), VP22-GFP (57), or UL11-GFP (47) were constructed into vector pEGFP-N2 under the control of the CMV promoter from the KOS strain of HSV-1 and were kind gifts from John W. Wills (Pennsylvania State University College of Medicine, Hershey, PA) and have been described previously. Plasmids expressing HSV-1 strain F gE and gI (pCA3gE and pCA3gI) were previously described (58). In brief, full-length HSV-1 strain F gE and gI sequences from plasmids pSV2X3gE and pSV2X3gI (6) were subcloned into plasmid pCA3 and under the control of the CMV promoter with a SV40 polyA sequence.

Neuronal cell cultures.

CAD cells are a derivative of a mouse catecholaminergic central nervous system cell line and were a kind gift from Greg Smith at Northwestern University Medical School, Chicago, IL (43). CAD cells were maintained in Dulbecco’s modified Eagle medium (DMEM)/F12 containing 8% fetal bovine serum (FBS) and passaged using sodium citrate buffer (134 mm KCl and 15 mM Na citrate [pH 7.3 to 7.4]). Cells were differentiated by plating onto dishes coated with poly-d-lysine (30 μg/ml) and laminin (2 μg/ml) in differentiation media (DMEM/F12 containing 0.5% FBS and 1 ng/ml nerve growth factor [NGF; 2.5S; Invitrogen]). Additionally, 10 μM 3-isobutyl-1-methylxanthine (IBMX) and 150 μM dibutyryl-cAMP (dbcAMP) was added for 2 days. The cells were differentiated for 7 days before being infected with HSV using 20 PFU/cell for 18 h.

Antibodies.

UL11, UL21, UL16, and VP22 rabbit polyclonal antibodies were a kind gift from John W. Wills (The Pennsylvania State University College of Medicine, Hershey, PA). UL11 (48), UL21 (42), UL16 (39, 59), and VP22 (48) were made using GST-UL11, GST-UL16, GST-UL21, and GST-VP22, respectively. Mouse monoclonal antibodies specific for gE (3114) and gI (3104) were gifts from Anne Cross and Nigel Stow Institute of Virology, Glashow, UK, and were described previously (7, 36, 58). A mouse MAb specific for the HA epitope (HA.11 clone 16B12) was purchased from Covance Inc. Rabbit polyclonal antiserum specific for HSV-1 US9 has been described previously (9). MAb 14-4B, which is specific for human cytomegalovirus (HCMV) gH (60) was kindly provided by William Britt (University of Alabama, Birmingham, AL).

Transfections.

CAD cells were differentiated for 4 days as described above and then transfected with UL16-, UL21-HA-, VP22-GFP-, UL11-GFP-, gE-, and gI-expressing plasmids by using Lipofectamine 2000 (Thermo-Fisher Scientific, Rockford, IL) according to the manufacturer’s instructions for 24 h.

Immunoprecipitations and Western blotting.

Cells were harvested and solubilized in NP-40/DOC lysis buffer (1% NP-40, 0.5% DOC, 50 mM Tris-HCl [pH 7.4], and 100 mM NaCl) or NP-40 lysis buffer (1% NP-40, 50 mM Tris-HCl, pH 7.4, and 100 mM NaCl), which both contained protease inhibitors (cOmplete mini protease inhibitor cocktail, Roche, Germany). Cell extracts were centrifuged for 10 min at 12,000 × g at 4°C and then immediately incubated with primary antibodies, which were usually antibodies specific for gE/gI or US9 or the HA tag on US9, overnight at 4°C on a rotating platform. Protein A or protein G agarose beads (Goldbio, St. Louis, MO) were added for 4 h at 4°C on a rotating platform. Beads were spun down and washed 3 times; then, an equal volume of 2× Laemmli buffer (0.125 M Tris-HCl [pH 6.8], 20% glycerol, 4% SDS, and 4% beta-mercaptoethanol) was added to the beads. For whole-cell extracts, cells were scraped and solubilized in 1× Laemmli buffer and then the extracts boiled. Proteins were separated using SDS-polyacrylamide gel electrophoresis (SDS-PAGE) using 4% to 20% Mini-Protean TGX gels (Bio-Rad, Hercules, CA). Then, proteins were transferred onto polyvinylidene fluoride (PVDF) membranes using a Trans-Blot SD semi-dry transfer cell (Bio-Rad). These Western blots were probed with appropriate primary antibodies and either goat anti-rabbit or goat anti-mouse SuperClonal horseradish peroxidase (HRP)-conjugated secondary antibodies (ThermoFisher Pierce). The HRP-conjugated antibodies were detected using a chemiluminescent reagent (Thermo-Scientific) and imaged with an Imagequant LAS 4000 system (GE Healthcare). The density of Western blot bands was quantified using FIJI software (61).

Membrane flotation assay.

CAD neurons were infected with HSV for 18 h. Whole-cell extracts of the cells were made using 1× Laemmli buffer and boiled. Other cells were harvested into buffer containing 10 mM Tris-HCl, 100 mM NaCl, and 1 mM EDTA (pH 7.4), and then the cells were centrifuged at 1,000 × g for 10 min. Cell pellets were spun down and resuspended in 250 μl of hypotonic lysis buffer (10 mM Tris-HCl [pH 7.4] and 0.2 mM MgCl₂) and placed on ice for 20 min. Cells were Dounce homogenized using a tight pestle, and then nuclei were removed by centrifugation at 1,000 × g for 10 min. Sucrose was added to the cell extracts to make 60% sucrose, and this mixture was placed at the bottom of Beckman SW55Ti ultracentrifuge tubes. This 60% sucrose layer at the bottom of the tube was gently overlaid with 40% and 25% sucrose in hypotonic buffer, and the tubes were centrifuged 200,000 × g for 16 to 18 h at 4°C. Membranes that floated up, so that they were found at the interface of 40% and 25% sucrose, were harvested and pelleted. The pelleted membranes were solubilized in 1× Laemmli buffer, boiled for 5 min proteins separated on SDS-polyacrylamide gels, and subjected to Western blotting.

Pulse-chase ³⁵S-methionine/cysteine protein labeling.

CAD neurons were infected with HSV 16 h, were washed 3 times in DMEM/F12 cell culture media lacking cysteine and methionine and containing 1% dialyzed FBS and 10 mM HEPES (pH 7.3) and then were incubated in this medium for 2 h. This medium was removed and replaced with medium lacking methionine and cysteine and containing 250 μCi/ml ³⁵S-methionine/cysteine for 40 minutes. The radiolabeling medium was removed, the cells washed twice with DMEM growth medium containing methionine and cysteine, and incubated in this medium for 75 or 160 min. Cells were lysed in RIPA buffer (50 mM Tris-HCl [pH 7.4], 100 mM NaCl, 1% NP-40, 0.5% DOC, 0.5% SDS, and protease inhibitors), centrifuged for 10 min at 12,000 × g at 4°C, and then immediately immunoprecipitated using rabbit polyclonal antibodies against VP22, UL21, or UL16. Proteins were separated by SDS-PAGE; gels were dried using a Bio-Rad 583 gel dryer (Bio-Rad), exposed to phosphorimager plates, and scanned using a STORM phosphorimager (GE Healthcare).