The viral-attachment glycoprotein gD mediates strong interactions with cell-surface receptors. Subsequently, the viral-fusion machinery (gB and gH/gL) promotes fusion between the viral and cellular membranes [6]. Several cellular gD receptors have been characterized: herpesvirus entry mediator (HVEM), nectin-1 and -2, as well as modifications of heparan sulfate introduced by 3-O-sulfotransferases [7–9]. Murine nectin-2 is used for PRV entry, but not for HSV entry [10]. HVEM and nectin-1 are the principal gD-binding entry receptors allowing PRV entry into mouse cells, and allowing both serotypes of HSV entry into human cells. In addition, nectin-1 is the mediator of HSV entry into cultured primary rat and mouse sensory neurons [11]. Recently, a putative gB receptor, paired immunoglobulin-like receptor-α, has been reported [12].

Following amplification of infection in the epithelial cells, virions invade the nerve terminals innervating the site of infection, and spread by retrograde transport to the cell bodies, where a lifelong latent infection is established [13]. It was recently determined that the deubiquitinase domain of the viral tegument protein UL36 is required for spread of infection from the periphery to the innervating axons [14]. Sensory neurons are readily infected, but if deeper layers of the epithelium are exposed, efferent and autonomic nerve terminals can also be infected. Upon entering the axon termini of PNS neurons, nucleocapsids and their inner tegument components are released and engage the retrograde-directed, microtubule-associated cellular motor dynein for delivery to the neuronal cell body [15,16]. Studies of fluorescently tagged PRV capsids in cultured chick dorsal root ganglia (DRG) neurons reveal that the retrograde transport of the nucleocapsid is accomplished at the average velocity of 1.17 ± 0.03 μm/s. The movement was bidirectional and saltatory, with average retrograde run lengths of 7.38 ± 0.47 μm [17].

The structure of the PRV particle engaged in retrograde transport, as revealed by transmission electron microscopy, has no membrane surrounding it [18,19]. Furthermore, only the inner tegument proteins remain associated with the capsid during retrograde transport, while the outer tegument is released upon entry [20–22]. Therefore, viral proteins that are likely to interact with the dynein motor complex include capsid components and the inner tegument, as has been shown by in vitro motility assays for HSV [23]. While several herpesvirus proteins are capable of binding dynein subunits in vitro, these proteins either are absent from the retrograde-trafficking particle, or play no role in retrograde transport, as determined by live-cell imaging of fluorescently labeled capsids and other techniques [24,25]. Our current understanding is that the essential innermost tegument protein, UL36, engages the retrograde transport machinery. However, this hypothesis has proven exceedingly difficult to test because UL36 is essential for particle assembly [24].

Following arrival at the neuronal cell body, the capsid (with inner tegument proteins) docks at nuclear pores and delivers the viral DNA to the nucleus. Viral replication may ensue with subsequent death of the cell, or a quiescent, latent infection may be established. The number of neurons in a peripheral ganglion that can be infected varies substantially, depending on many variables (e.g., virus strain, animal species, site of infection and dose of virus). Latency has been well studied for HSV and poorly studied for PRV. Interestingly, at least for HSV-1, the copy number of quiescent genomes in latently infected mouse neurons ranges from one to 1000 or more. In murine systems, following experimentally induced stress, reactivation of productive infection does not occur in all latently infected neurons. As few as one to five of the latently infected neurons will reactivate while the others remain quiescent [26]. An obvious question asks why so few neurons are activated at any given time, even when all neurons should receive the activation signal? One recent study suggests that stochastic derepression of the viral VP16 promoter leads to gradual accumulation of the VP16 tegument protein [26]. This trans-activator forms a complex with two cellular gene products, the transcription factor Oct-1, and the cell-cycle regulator HCF-1, which initiates transcription of viral immediate–early genes [27,28]. After sufficient quantities of VP16 are synthesized, the VP16-induced complex binds to a particular sequence in the immediate–early promoters (ICP0 and ICP4 genes) to initiate the viral gene-expression cascade and mediate exit from latency [26]. Interestingly, the majority of the VP16 protein does not remain associated with capsids during retrograde transport in neurons [21], and the same is true for the HSV VP16 protein [Smith GA, Pers. Comm.]. Therefore, the absence of this transactivator in the cell body may favor the establishment of latency.

Data from PRV and HSV infections have provided two different answers. Electron- and confocal-microscopy studies are consistent with the idea that mature virions produced from cell bodies move through the axons in transport vesicles [13,18,32–42]. However, other studies provide data that are more consistent with the concept that virions are not sorted fully assembled into axons, but rather enter and move as separate structural components, or subassemblies [43–46]. The static images demonstrated capsids with no surrounding membranes in axons, and no obvious colocalization of capsid and glycoprotein signals using antibody techniques [44–48]. In addition, electron microscopy demonstrated accumulations of enveloped capsids only at axonal varicosities and termini, the suggested sites of virion egress [49,50]. In this scenario, unenveloped capsids and viral glycoproteins are independently transported in axons, and secondary envelopment of these subassemblies occurs in distal sites of the axon and axon terminals [51]. Currently, the former model (fully assembled virions in vesicles move into axons for transport) is favored for PRV, while the latter (separate capsids and envelopes move into axons for independent transport) is favored for HSV.

The differences observed in the PRV and HSV data suggest that the two viruses use different mechanisms for axonal transport of newly made virions and distant egress. If true, it will be important to understand the basis of different evolutionary adaptation to such a fundamental process as axonal sorting and transport. However, other explanations should be explored as well. For example, many of the studies are not directly comparable. Differences in viral strains, protocols and preparations of neuronal tissue and cells (explants vs dissociated neurons, vs cell lines) all provide important variables that need to be reconciled. Recent work by Smith and Antinone, where HSV-1 and PRV were compared directly in chick DRG neurons, supports the idea that both viruses have much in common with respect to axonal sorting and anterograde transport [Smith GA, Pers. Comm.]. Certainly some important differences are obvious when one compares neuronal infections by HSV and PRV. For example, during PRV and HSV infection of cultured rodent PNS neurons, many more virions and virion components are observed in axons of PRV- than in HSV-infected neurons; moreover, in direct comparisons, HSV structural proteins enter axons substantially later than do the PRV structural proteins [Mettenleiter TC, Pers. Comm.] [Smith GA, Pers. Comm.] [46,48,51]. Consequently, the time of observation of infected axons coupled with different particle abundance in axons make it difficult to compare and interpret data from both viruses, particularly using the electron microscope.

Our own experience demonstrates some of the important confounding factors in analysis. Initially, we reported that axons of PRV-infected PNS neurons in culture contained capsids with no envelopes [43,52]. However, it became clear that at least two issues complicated and confounded these experiments (and any experiments that use dissociated cultures of neurons or neuron-like cell lines). First, it was difficult, if not impossible, to distinguish input virions from newly replicated virions in neurons. Virions adsorb to cultured neurons not only at cell bodies and axon terminals, but also along the axon shafts. Second, axons in dissociated neuronal cultures form synaptic connections with other cell bodies, as well as axons, such that a single axon can support both entering particles (retrograde movement) and egressing particles (anterograde movement) [53]. In fact, as time progresses, it becomes impossible to deduce by looking at fixed images of axons if the particles or subassemblies observed are ‘coming or going’.

The use of live-cell imaging to study PRV particle dynamics, as well as compartmented neuronal cultures to physically isolate the site of infection from the site of imaging, have been particularly useful in solving several problems. These techniques help define the direction of movement in axons (toward or away from the cell body), and reduce or eliminate the ambiguity between input inoculum and newly made virions. Obviously, imaging techniques rely on the use of recombinant viruses that produce fluorescent fusion proteins that assemble into virions. Such recombinants are, by definition, mutants and it is essential to demonstrate that the fusion proteins have wild-type function. This condition holds for many, but not all PRV recombinants that express GFP- and RFP-tagged virion proteins in neurons. Not as many studies have been reported with HSV-1 recombinants, primarily because many grow poorly in neurons for reasons that are not entirely well understood.

Electron microscopy of PRV-infected PNS neurons in compartmented cultures (where cell bodies are physically separated from the imaged axons) demonstrated that the vast majority of anterograde-trafficking PRV nucleocapsids in axons are enclosed in a double membrane – the viral envelope and, presumably, the trans-Golgi network (TGN) vesicle from which it is derived [35,54]. Similar results were obtained by examining axons in PRV-infected rat superior cervical ganglion explants [Mettenleiter TC, Pers. Comm.]. By tagging different virion components with fluorescent proteins (envelope, tegument and capsid), Antinone et al. demonstrated that the newly made PRV capsid fluorescent puncta undergoing anterograde movement are associated with viral membrane proteins [32]. By contrast, during retrograde movement, the fluorescent capsid puncta do not exhibit colocalization with viral membrane proteins signals. These results were recapitulated by Feierbach et al. in a compartmented neuronal culture system, where viral inoculum was applied to cell bodies and images were obtained of physically isolated axons, and by Liu et al. in fluidically isolated axons [54,55]. Coller and Smith first revealed another source of so-called ‘naked capsids’ in axons of infected neurons [22]. They proposed that partially enveloped virions were sorted into axons and moved in the anterograde direction. However, these structures are moderately unstable and can dissociate into capsid- and membrane-vesicle components. When this dissociation occurs, the capsid immediately moves back to the cell body (retrograde motion), while the glycoprotein-tagged membrane vesicle (sans capsid) continues in the anterograde direction. This finding was also supported by data from Feierbach et al. [54]. Coller and Smith demonstrated that the function of the two PRV-encoded protein kinases, products of the Us3 and UL13 genes, are important in the stability of these partially assembled complexes [22]. This phenomenon of dramatic reversals of anterograde moving capsid puncta was first reported by Smith et al. [17], and later documented further in a second paper [56]. Using a microfluidic chamber system, Liu et al. quantified the number of retrograde and anterograde moving particles in axons after sorting [55]. They recapitulated the observations that fluorescent puncta exhibiting capsid and tegument signals travel predominantly in the net anterograde direction, while nearly all puncta exhibiting only capsid fluorescence travel in the retrograde direction [55]. Therefore, acquisition of a viral envelope appears to dictate the direction of viral transport: enveloped capsids engage anterograde-directed motors, while naked capsids engage the retrograde-directed motor dynein [32]. These observations offer an explanation for the presence of occasional ‘naked’ capsids in axons of PRV-infected neurons, and highlight the importance of temporal dynamics in the study of anterograde trafficking.

Lyman and colleagues discovered that the viral protein Us9 mediates targeting of PRV virions to axons [62]. This 98-amino acid type II membrane protein has a short (~two amino acid) ectodomain, and a relatively large 68-amino acid cytoplasmic tail [63]. Like other viral membrane proteins, Us9 localizes to the TGN-derived membrane that is the site of viral envelope acquisition. The predicted topology of viral membrane proteins in the TGN vesicle is opposite to that found in the viral envelope: the short ectodomain of the vesicle-incorporated type II Us9 protein faces the enclosed virion, while its tail domain is in the cytoplasm, well positioned for establishing interactions with cellular proteins. A recent study by Lyman et al. revealed that Us9 must associate with lipid-raft microdomains in order to mediate axonal sorting of PRV structural proteins [53]. In addition, putative phosphorylation sites in the Us9 tail, an acidic cluster region and two dityrosine motifs, are required for axonal sorting of PRV [53]. The current hypothesis is that the PRV Us9 protein localizes to lipid microdomains in TGN-derived vesicles, where its phosphorylated form engages the appropriate cell machinery to enable targeting of the virion-containing vesicle to the axon (Figures 1A & B). It is still unclear if Us9 is involved only in sorting to the axonal compartment or in movement of cytoplasmic vesicles to and through the cytoskeleton-rich proximal axonal segment.

The viral proteins gE and gI, which function as heterodimers, work with Us9 to facilitate axonal sorting of PRV virions. PRV mutants that do not express either of these proteins exhibit reduced anterograde spread of infection in vivo and in vitro [35,64,65]. Currently, the idea is that the PRV gE–gI complex may stabilize interactions between Us9 and cellular adaptor proteins, thereby facilitating the process of axonal targeting of viral cargo [53].

The presence of ‘viral highways’ – virus particles arranged in a linear fashion – has been noted on the surface of VZV-infected non-neuronal cells. These were observed on actin-rich filopodia via scanning electron microscopy [77,78]. While this arrangement of particles is so far unique to VZV among the α-herpesviruses subfamily [77], HSV-1 was recently proposed to induce filopodia in differentiated P19 cells, and ‘surf’ on the extracellular surface in a manner that facilitates viral spread [79]. It remains to be conclusively shown whether the proposed surfing particles are on the outside or the inside of the cell. However, viral surfing may represent another mode of α-herpesvirus transmission.

Similar to cell–cell spread in epithelial cells, transneuronal spread of PRV does not require gD, but does require the viral fusion machinery, gB and gH/gL [35,75,76,80,81]. Several experiments suggest that viral egress and subsequent spread to neighboring cells occurs at specific axonal locations. Using an in vitro co-culture system with swine sensory neurons and epithelial cells, De Regge et al. demonstrated that spread of PRV from an infected neuron to epithelial cells occurs primarily at axonal varicosities [82]. Other in vivo and in vitro studies also report that spread of PRV from an axon to the surrounding cells occurs at scattered sites along the axon, and at neuronal termini [35,83]. Similar conclusions have been made for HSV spread. For example, electron microscopy of human fetal DRG infected with HSV demonstrates accumulations of mature virions at varicosities and growth cones [49]. Therefore, it may be that both HSV and PRV virions are targeted to specific locations within the axon for egress. However, the mechanisms underlying virion targeting and neuronal egress remain uncharacterized (Figure 1C).

The authors thank members of the Enquist laboratory for advice and encouragement. GA Smith provided critical comments. Anonymous reviewers helped improve the review.

The authors acknowledge support from NIH grants R37 NS 033506, R01 NS060699, P40 RR 018604, AI 033063 and P50 GM071508. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.