Alphaherpesvirus Latency and Reactivation with a Focus on Herpes Simplex Virus

We are at an interesting time in the understanding of alpha herpesvirus latency and reactivation and their implications to human disease. Conceptual advances have come from both animal and neuronal culture models. This chapter focuses on the concept that the tegument protein and viral transactivator VP16 plays a major role in the transition from latency to the lytic cycle. During acute infection, regulation of VP16 transactivation balances spread in the nervous system, establishment of latent infections and virulence. Reactivation is dependent on this transactivator to drive entry into the lytic cycle. In vivo de novo expression of VP16 protein is mediated by sequences conferring pre-immediate early transcription embedded in the normally leaky late promoter. In vitro, alternate mechanisms regulating VP16 expression in the context of latency have come from the SCG neuron culture model and include the concepts that (i) generalized transcriptional derepression of the


caister.com/cimb
The availability of animal models that support HSV latency and reactivation, and the work of many laboratories over many years, have contributed to a collective understanding of the HSV life cycle. Infection occurs at the body surface, usually a mucosal epithelium, where the virus replicates, spreads, and gains access to the innervating neuron axonal endings. At this point, nucleocapsids deposited into the axonal cytosol use retrograde axonal transport to reach neuronal cell bodies that are housed within peripheral ganglia (a collection of sensory or autonomic nerve cell bodies innervating specific tissues). The trigeminal ganglia (TG), which innervate the oral mucosa and other regions of the face, is the typical site of HSV1 latent infection, whereas the sacral ganglia (SG) innervates the genital mucosa and serves as the predominant site of latent HSV2 infection.

Overview
Our current understanding of HSV latency and reactivation arises out of a blend of human studies and findings from various animal and neuronal culture models. It is important to recognize the limitations of these models when summing up what we know. Importantly, new technologies and approaches with increased sensitivity have driven new waves of data acquisition which requires assimilation into what we think we know about an extremely complex process. There is increasing evidence for the concept that HSV1 latency/ reactivation can be a risk factor in the development of neurodegenerative disorders. Also arising from the human model, the concept of "continuous" ganglionic reactivation has emerged as an explanation for the unexpectedly high frequency of viral DNA in human genital swabs in some patients in the absence of detectable infectious virus.
This article focuses on an unexpected observation that the tegument protein and viral transactivator, VP16, plays a major role in the transition from latent into the lytic cycle and is required for balancing the latent/lytic transition ( Figure 1). During acute infection, regulation of VP16 transactivation balances spread into the nervous system, establishment of latency, and virulence. In caister.com/cimb turn, reactivation is dependent on this transactivator to drive entry into the lytic cycle. A potential mechanism for regulating VP16 expression in the context of in vitro latency has come from the superior cervical ganglion (SCG) neuron culture model (Cliffe et al., 2015;Kim et al., 2012). The concepts that have Figure 1. Regulation of VP16 function balances latent/lytic programs. Entering the neuron: The VP16 tegument protein, which is a potent transactivator of viral immediate-early gene expression, is not available to perform this function in neurons infected via axonal endings because of inefficient transport of VP16 via axon to the cell body during acute infection. The viral DNA enters the nucleus where it forms a circular episome and becomes associated with nucleosomes, which promotes the latent transcriptional program.
emerged from this work are as follows: (i) generalized derepression of the viral genome (Phase I) precedes VP16-dependent reactivation and is required for it (Phase II), and (ii) a histone methyl/phospho switch during Phase I is required for Phase II reactivation. Notably, in vitro neuronal models of latency vary in this regard and some do not display this dual phase reactivation mechanism (Edwards and Bloom, 2019). It will be important to reconcile the findings in diverse in vitro and in vivo models of alpha herpesvirus latency and reactivation to develop testable hypotheses and determine relevant regulatory molecular mechanisms. To frame the present, we start by discussing the early seminal findings. Preserving the latent reservoir: During latent infection, modification in the chromatin associated with the viral genome evolves into an increasingly repressive state. Despite this repression, there is a low level of transcription from the genome related to lytic genes. The significance of this transcription is not known but could reflect a repressive mechanism working in tandem with caister.com/cimb the LATs and microRNAs. Replicating virus and viral proteins are not detected during latency in the TG. Long-term studies in the mouse and rabbit reveal the latent reservoir appears to be stable. Expression from the LAT locus is important to maintain reactivation competent latent infections.
Reactivation. Stress results in de novo expression of VP16 which coordinates activation of the viral immediate-early (IE) genes. A productive lytic cycle ensues in 0.05% of latently infected neurons per event although more generalized changes in the chromatin associated with the latent genomes occurs. Thus, HSV has devised a complex regulatory strategy that maintains the vast majority of its latent genomes in latency, while allowing the release into a productive lytic cycle in an extremely rare number of neurons. During reactivation, virus does not spread to neighboring neurons in the ganglia but is transported back the body surface where subsequent replication in epithelial cells amplifies virus output and facilitates transmission to new hosts.
Spontaneous reactivation in sensory ganglia has been documented in mouse models. Adapted from (Thompson et al., 2009).

How 150 years of research have shaped our ideas about herpesvirus pathogenesis
The earliest scientific investigations on alpha herpesvirus latency were primarily clinical observations. Herpes, an ancient Greek word meaning "to creep", was used by Hippocrates to describe diverse skin diseases. The association between fever and blisters around the mouth and nose was recognized nearly 2500 years ago by the Roman physician Herodotus who described herpes febrilis (Wildy, 1973). The terms "fever blisters" and "cold sores" are still in use today. Through the ensuing millennia the recurrent nature of these types of diseases was appreciated.
caister.com/cimb Why do herpetic lesions recur? Are they from an infectious agent? What is the nature of the agent? 1863-1930s. Available Toolbox: • Knowledge of transmission of infection and infectious agents including bacteria and "filterable agents" (the latter the earliest indication of viruses) • Koch's famous postulates delivered in 1890 (Koch, 1890) • Identification of a "neutralizing" substance in the blood that countered infections (Behring and Kitasato, 1890) • Generation of these neutralizing substances, including those to herpes simplex • Methods to store or serially passage herpes in animal brains (Perdrau, 1925) • Light microscopy combined with histological staining to study tissue.
Neurons distinguished from other cells used in conjunction with agents transmitted through nerves contributed to knowledge of brain structure and innervation from the periphery (Doer, 1920;Friedenwald, 1923;Goodpasture and Teague, 1923) The association of herpetic lesions with the nervous system was appreciated as early as the mid-19th century (Von Barensprung, 1863). However, it was not until the beginning of the 20th century that clinicians and scientists began to appreciate a relationship between herpetic lesions on the body surface and the sensory nerve endings innervating the skin. In 1892 von Bokay observed that children often developed varicella after exposure to an adult suffering from zoster (von Bokay, 1909). In 1900 Head and Campbell deduced from the pattern of lesions on their patients that herpes zoster must be related to sensory ganglia. Their tour-de-force study (Head and Campbell, 1900), reprinted in part in 1997(Head et al., 1997, was the first to show how herpetic viral diseases can be employed to help understand the anatomy of the nervous system, a practice that continues today (Sarno and Robison, 2018).
Relying heavily on their work, Howard noted that human herpetic lesions of the face were associated with trigeminal ganglionitis and pneumonitis early in the 20th century (Howard, 1903). At about the same time Cushing reported that caister.com/cimb some individuals treated for trigeminal neuralgia by surgical resection of the trigeminal ganglion and nerve roots developed herpetic lesions in areas innervated by the contralateral (opposite side) nerve but not on the ipsilateral (resected) side (Cushing, 1905), supporting the hypothesis that herpetic lesions were associated with stress or damage to the peripheral nervous system.
Vidal first demonstrated the infectious nature of herpes (Vidal, 1873), but of greater significance to those interested in latency were later studies involving transmission of herpetic stromal keratitis. At the end of World War I the transmission of human herpetic stromal keratitis to rabbit corneas by Gruter (Gruter, 1920;Kraupa, 1920) and Loewenstein (Loewenstein, 1919(Loewenstein, , 1920) and subsequently transmission back to a human (reviewed in (Holden, 1932)) cemented the idea that herpes was an infectious agent, and provided an animal model for study. During this same time, varicella was transmitted to naive children using vesicle fluid from children with varicella lesions (Kundratitz, 1925); however, attempts to infect laboratory animals failed (Rivers and Tillett, 1924). The lack of an animal model is a difficult challenge for the study of VZV pathogenesis and latency that persists to the present day (Mahalingam et al., 2019;Ouwendijk and Verjans, 2015). A creative approach to overcome the species specificity of VZV is the use of human tissue xenografts in mice with severe combined immunodeficiency (SCID). This model allows the analysis of VZV infection in differentiated human cells in the context appropriate tissue microenvironments, providing many insights into VZV pathogenesis (Moffat et al., 1995;Zerboni and Arvin, 2015;Zerboni et al., 2014). More recent studies using dissociated human TG have provided insight into VZV reactivation (Cohrs et al., 2017).
Several groups noted that herpes infection on the rabbit eye led to infection of the central nervous system (CNS) and changes in rabbit behavior (e.g. turning the head to a particular side), both suggesting a neural route of transmission (Doer, 1920;Friedenwald, 1923;Goodpasture and Teague, 1923).
Goodpasture began a series of histological studies delineating herpes spread caister.com/cimb within rabbits (Goodpasture, 1925a, b;Goodpasture and Teague, 1923) that culminated in the astounding and prescient conclusion that: "Following a primary infection, it seems quite probable that the virus remains in a latent state within the ganglia after the local lesion has healed. A second cutaneous eruption may occur as a result of injury…or the disturbed physiological states... which sets in activity a latent virus" (Goodpasture, 1929). Despite the seemingly overwhelming evidence of nerve involvement in herpes infection this idea was not universally accepted.
1930s-1980s. Toolbox additions include: • The nature of viruses (Stanely, 1935) • Tissue and virus culture as well as plaque assays for animal viruses (Dulbecco and Vogt, 1953) • Ultra-low freezers for virus storage • Electron microscopy • Refined histological approaches, fixation methods and immunohistochemistry; • Conditionally lethal viral mutants • The "phage Church" and Lambda phage latency, DNA structure and code (Watson and Crick, 1953) • The central dogma of DNA to RNA to protein • Liquid DNA/RNA hybridizations • Bacterial restriction-modification system (S.E. and M.L., 1952) and commercial restriction enzymes • In situ hybridization for DNA and RNA (Gall and Pardue, 1969) • Radiolabeling of proteins and nucleic acids and various separation methods including chromatography, electrophoretic gel systems, and blotting and probing (Southern, 1975) • Cloning and cloning vectors (Bolivar et al., 1977) • Engineered viral mutations • Genetically modified and engineered animal models, especially mice • Maxim and Gilbert, then Sanger sequencing caister.com/cimb Good and Campbell employed Perdrau's rabbit model of herpetic encephalitis (Perdrau, 1925(Perdrau, , 1938 to demonstrate that anaphylactic shock could "precipitate" herpetic encephalitis in previously infected rabbits pre-sensitized to egg albumin (Good and Campbell, 1948). Rabbits were infected with HSV and one to three months after they appeared to be disease free, anaphylactic shock was induced by exposure to egg albumin. Encephalitis occurred in 19/44 tests and virus was recovered from the brains of all rabbits that died and many that recovered. Virus was not detected in the brains of rabbits prior to anaphylaxis. Arguably this was the first well controlled evidence of latency and reactivation of HSV in the nervous system in vivo. We now know that intramuscular injection of HSV most likely resulted in virus replication in the skin at the site of injection and subsequent spread to the innervating sensory ganglia. Anaphylactic shock presumably caused virus reactivation within sensory neurons with subsequent spread to the brain resulting in fatal encephalitis but virus reactivation within brain neurons was not ruled out.
Schmidt and Rasmussen explored alternate methods to "precipitate" herpes leading them to favor HSV latency occurs in the skin. Of many methods tried, only intramuscular injections of adrenalin "precipitated" encephalitis (Schmidt and Rasmussen, 1960). Encephalomyelitis was "precipitated" in 60% of the rabbits given intramuscular adrenalin injections and, importantly, herpesvirus was detected in all six of these rabbit brains. A few years earlier at the Wisconsin meeting on Latency and Masking in Viral and Rickettsial Infections (Andrewes, 1957), six types of "latent" infections were described. Herpes was thought to be a unique agent because it could not be cultured from skin between episodes. Herpes latency was defined as the period of time in which skin was negative between outbreaks. Rasmussen speculated "that temporary vasoconstriction resulting from increased adrenalin output, could produce a local anoxia {e.g. local reducing conditions} in the skin and consequent 'reactivation' of residual herpesvirus" (Schmidt and Rasmussen, 1960). This hypothesis harkened back to the early observations of Perdrau. Early virologists stored their virus stocks as bits of infected rabbit brains in glycerin in ice boxes. Exclusion of air greatly increased the time such a stock remained caister.com/cimb infectious. Perdrau discovered that oxidation destroyed herpesvirus activity, but the virus stocks could be "reactivated" by subsequent reduction (Perdrau, 1931).
Rasmussen replaced the commonly employed term "precipitate" with "reactivate" directing thought toward the concept that an active agent has been muted to a latent agent, pre-existing in toto, that is somehow reactivated (presumably by reducing conditions in vasoconstricted skin) to become infectious. Latent herpes was thought by most to be present in the skin and this was an early example of a "skin trigger" hypothesis: an idea that has recently regained some attention and is further discussed below.
The early neural work by Goodpasture was supported by the formation of cutaneous herpes lesions following surgery in the trigeminal nerve tract. For example Carton and Kilbourne noted that within a few days after section (axotomy) of the fifth cranial nerve, oral or facial lesions occurred in 90% of patients if the ganglion and nerve were not destroyed (Carton and Kilbourne, 1952). The existence of a dormant form of herpes in the TG was supported by the failure of several groups to isolate the infectious agent from human TG (Burnet and Williams, 1939;Carton and Kilbourne, 1952). However, observation of HSV lesions in skin following "blowout" fractures that destroyed the nerve tract innervating the site of recurrence suggested that the virus must have already been in the skin (Hoyt and Billson, 1976). We now know the complexity of innervation of the facial skin provides a plethora of alternate routes for virus transport to the skin.

HSV-2 is discovered and VZV is proposed to be a virus that goes latent in sensory ganglia
Plummer determined that there were two serotypes of herpes simplex virus (Plummer, 1964), with different biological properties (Plummer et al., 1968), and was the first to show reactivation of HSV in the peripheral nervous system (albeit in the absence of a critical control as detailed below) (Plummer et al., 1967). But, in the absence of a tractable animal model for varicella zoster virus caister.com/cimb (VZV) infection, the idea that zoster was the result of reactivation of latent VZV arose instead from a long-term epidemiological study (Hope-Simpson, 1965).
Formal proof that the same strain of VZV could cause both chickenpox and, subsequently, shingles in the same patient was obtained 20 years later (Straus et al., 1984) While HSV had been isolated from skin, saliva, tears and mucosal surfaces (especially genital mucosa) by many different groups in the absence of frank lesions, the source of this virus was unknown (reviewed in (Stevens, 1975a, b)). Transplantation of facial skin from sites where lesions occurred to other parts of the body did not result in virus recurrence at those sites, and most efforts to isolate virus from such skin either directly or following explant into culture failed (reviewed in (Finlay and MacCallim, 1940;Hill, 1985)). Notably, in the absence of lesions, very low titers are usually found (Agyemang et al., 2018). And yet, viral DNA is detectable by PCR in genital swabs much more frequently than is infectious virus as discussed below. A potential role for extraneuronal sites of HSV latency in these phenomena should not be discounted.
During the 1950s to late 1960s laboratory animal models of herpetic disease were further developed, notably in rabbits, guinea pigs, and mice. The use of these models has led to several seminal observations, including the discovery that the two serotypes of HSV (HSV1 and HSV2) caused different pathologies in mice (e.g. HSV-2 is much more virulent than HSV-1 and causes larger pox (plaques) on liver following intraperitoneal injection) (Plummer, 1964;Plummer et al., 1974;Plummer et al., 1968). In 1970 the first experimental evidence that HSV2 remained in neural tissues for many months was presented. HSV2 was isolated in vivo after adrenalin injection, and also by cultivation of trypsinized rabbit neural tissue with indicator cells that presumably resulted from reactivation in vitro (Plummer et al., 1970).

What is the cellular site of HSV latency and reactivation?
Cook and Stevens confirmed and extended the observation that HSV remained latent in neural tissues by developing a method to reactivate HSV-1 from dorsal root ganglia (DRG) of mice infected via the rear footpad. Following infection, virus could be isolated from feet, the PNS, and CNS with replication ceasing about 8 days post infection (dpi). Thirty dpi DRG were either excised and homogenized immediately, or explanted and co-cultivated on monolayers of susceptible cells. Infectious virus was not detected in DRG that were directly homogenized, but virus was produced by the great majority of explanted ganglia 7 to 14 days post explant. This report is often considered the first definitive proof that HSV resides within sensory ganglia in a non-infectious or latent state and is reactivated following the stress of axotomy and explant into culture (Stevens and Cook, 1971). Numerous groups subsequently applied this approach to diverse human tissues and detected reactivated HSV1 in sensory and autonomic ganglia (Baringer and Swoveland, 1973;Bastian et al., 1972;Plummer, 1973).
Stevens and colleagues went on to show that latent HSV1 could be reactivated only from central and peripheral neural tissues, including the adrenal medulla, following intravenous injection of virus to induce an artificial viremia. This solidified the idea that nervous tissues are the predominant sites of HSV1 latency (Cook and Stevens, 1976). This group also sought to demonstrate that the sensory neuron was the site of viral reactivation that led to skin lesions through the use of immunohistochemical and electron microscopic techniques (Cook et al., 1974), but their data cannot be unambiguously interpreted since tissue was examined at 48 hours and later following axotomy, a time now known to be confounded by secondary spread of reactivated virus within the explanted ganglia. The virus could have reactivated in cells other than neurons and then spread to them. The earliest solid evidence for virus reactivation in neurons was provided by the studies McLennan and Darby who employed temperature sensitive (ts) mutants in mice. At a core body temperature of 38.5 degrees, replication of the ts mutants was restricted and viral proteins in mice in which nerves were resected to induce reactivation in vivo were confined to neurons (McLennan and Darby, 1980). These neurons "appeared" to die.

caister.com/cimb
Are any essential viral gene products required for the establishment or maintenance of latency? Watson et al. and Lofgren et al. explored whether viral proteins essential for the viral lytic cycle were required for the establishment of latency by intracranially inoculating mice with ts viral mutants and asking if the mutants could later be reactivated from brain tissue at the permissive temperature in vitro (Lofgren et al., 1977;Watson et al., 1980). The major viral immediate early transactivator ICP4 was initially thought to be essential for latency establishment, a hypothesis disproved much later by Stevens (Sedarati et al., 1993). Therefore, even viral proteins essential for viral replication like ICP4 are not absolutely required for the establishment or maintenance of latency.

Is the latent HSV genome integrated into the host genome?
Alpha herpesvirus genomes are linear double-stranded DNA that circularize upon entry into the nucleus (Sheldrick and Berthelot, 1975). The physical state of the viral genome during latency was the subject of much speculation. Fraser and colleagues first demonstrated "endless" HSV DNA in latently infected mouse brain tissue by restriction endonuclease digestion and Southern blotting. Later, this and other groups confirmed that circularized genomes are also the predominant form found in latently-infected mouse TG (Efstathiou et al., 1986;Fraser et al., 1981;Rock and Fraser, 1983). We now know that the great majority of HSV and other human alpha herpesvirus latent genomes exist as extra-chromosomal circular episomes (Azarkh et al., 2010). While integration events cannot be ruled out, they are not thought to be biologically relevant. Curiously, pseudorabies virus (PRV), an alpha herpesvirus of swine, is reported to persist largely in a linear form along with some circularized genomes (Rziha et al., 1986).

Spontaneous reactivation from latency in mice
HSV is commonly stated to not spontaneously reactivate in the mouse. This misconception is argued against by the consistent levels of spontaneous reactivation that have been reported for over three decades. In the late 70s and early 80s, Hill and Blyth developed an informative model of in vivo caister.com/cimb reactivation, recurrence, and recurrent disease (recrudescence) in mice.
Infection of the mouse ear pinna resulted in latent infections in superior cervical ganglia (SCG) that could be reactivated by diverse stimuli including UV light, DMSO on the pinna, or stripping of the pinna with cellophane tape (Hill et al., 1975). In this model, virus could be routinely isolated from the pinna in about 10% of the animals on any given day, consistent with later estimates of spontaneous productive reactivation seen within mouse TG (Blyth et al., 1984;Blyth et al., 1981). They showed that some procedures to induce in vivo reactivation resulted in detectable virus within ganglia and in skin with no evidence of recurrent disease, whereas other procedures induced virus reactivation within ganglia and recurrent lesions that contained infectious virus.
This led them to postulate that there were "ganglionic triggers" and "skin triggers", the latter leading to recurrent lesions with the possible involvement of prostaglandins (known to enhance the replication of HSV in cultured cells), which were found to be elevated in the skin selectively by "skin triggers" and not "ganglionic triggers" (Harbour et al., 1978;Harbour et al., 1977), for review see (Hill, 1985). This hypothesis may still be relevant today as it might explain why infectious virus, and more frequently viral DNA, can be detected in the absence of a frank lesion.
The development of a sensitive procedure to detect viral protein expression in the whole ganglion allowed extensive analysis of latently infected ganglia for the expression of viral proteins at a single cell level (Sawtell, 2003). Whole ganglion IHC (WGIHC) allowed for detection of single neurons positive for viral proteins. A comprehensive long-term study revealed that in mice, as in humans (Agyemang et al., 2018), the frequency of spontaneous HSV reactivation (based on detection of TG neurons positive for infectious virus and viral proteins) declined during the time period between 15 and 40 days and then remained stable over the 240 days examined (Sawtell, 2003). The frequency of spontaneous reactivation in the TG was similar to the frequency observed in SCG, which also corresponded to the frequency of spontaneously positive mouse ear pinnas as reported by Hill and colleagues (Blyth et al., 1981) (reviewed in (Hill, 1985). Analyzing sectioned ganglia for HSV proteins, caister.com/cimb Feldman et al. similarly found that 10% of mouse TG pairs contained neurons positive for HSV proteins and called this "molecular reactivation" (Feldman et al., 2002). Later, the same group determined that 10% of unstressed mice were positive for infectious virus in TG (Margolis et al., 2007) demonstrating that molecular reactivation is consistent with the well-recognized spontaneous reactivation rate of mice as previously reported by others. This rate is likely dependent upon the viral strain, latent reservoir, and mouse strain. The animal models noted for spontaneous reactivation, rabbit and guinea pig, also have a time-dependent reduction in the frequency of reactivation, with the highest rate of spontaneous reactivation observed being 20-35 dpi. In rabbits, reactivation is quantified by virus shedding in tear films. The source of this virus is not known but thought to be neurons in TG. In guinea pigs, vaginal lesions are scored as reactivation events. Viral DNA can be recovered from these lesions . Thus, spontaneous reactivation of HSV occurs in diverse animal models including mice, rabbits and guinea pigs.

The timing of viral reactivation in neurons in vitro or in vivo following stress.
The kinetics of induced reactivation, both in vivo and in explanted ganglia, remained a major gap in our understanding of alpha herpesvirus infections.
Although knowing when a triggering event occurred that resulted in spontaneous reactivation is not possible, an increased frequency of ocular shedding occurred in rabbits over a period of two weeks following a three-day procedure of iontophoresis of epinephrine into corneas to induce reactivation (presumably initiated in the innervating neurons) (Toma et al., 2008). Likewise reactivation in explanted ganglia was thought to take several days or even weeks (Stevens, 1975a). However, the rate of reactivation was faster in subsequent studies. While a variety of stimuli caused virus reactivation and spread to the mouse ear pinna in 3 to 5 days, Harbour et al. found that virus could be isolated from a few SCG as early as one day post treatment (Harbour et al., 1983). Furthermore, using a mouse model of in vivo HSV1 reactivation induced by a 10 minute hyperthermic stress (42.5 C), virus was detected in TGs of mice as early as 14 hrs with 70% of mice positive by 22 hrs (Sawtell and Thompson, 1992a) (reviewed in (Webre et al., 2012)).

caister.com/cimb
The same timing of primary reactivation events is evident in axotomized and explanted TG with virus being detectable by 14 hours post explant. However, unlike the in vivo situation where virus does not spread from the original reactivating neurons, virus rapidly spreads within the explanted ganglia (Doll and Sawtell, 2017;Pesola et al., 2005;Sawtell and Thompson, 2004). The potential for reactivation to occur in explants within 14 hrs and the rapid spread of virus within the ganglion is inconsistent with conclusions drawn from earlier studies that reactivation requires several days to occur in explanted sensory ganglia. However, it should be noted that many factors could affect the levels of latency established in animal models and in turn the subsequent frequency and perhaps timing of viral reactivation in vitro and in vivo. These parameters include the method of inoculation, the virus inoculation titer, the strain of virus employed, and strain of animals employed.
What is the role of the immune system in maintaining latency?
As is the case today, the role of the host immune system in alpha herpesvirus latency and reactivation was of great interest to investigators. However, immunosuppression of latently infected mice failed to induce reactivation or produce recurrent disease (Blyth et al., 1981;Hurd and Robinson, 1977;Stevens and Cook, 1973), with only a low incidence of recurrence in hairless mice treated with the immunosuppressant drug, prednisone, being observed (Underwood and Weed, 1974). Conflicting with these findings is the more recent hypothesis that CD8 T-cells directed against a specific epitope on glycoprotein B may play a role in the maintenance of HSV1 latency (Bourne et al., 2018;Held and Derfuss, 2011;Knickelbein et al., 2008;Lahmidi et al., 2017;Liu et al., 2000;Treat et al., 2017). In contrast to HSV, depletion of CD4 cells induced reactivation of simian varicella virus in Rhesus Macaques (Traina-Dorge et al., 2019). The role of the host immune system in HSV and VZV latency and reactivation (Gershon et al., 2015) is a topic worthy of a its own review article and is not covered in further depth here.

In vitro primary neuronal quiescence/latency and the NGF depletion reactivation model
The application of cultured primary neuron models of HSV latency led to several new concepts in reactivation. Establishment of a quiescent/latent like state in cells including neurons requires the use of antiviral compounds, or the use of replication deficient viral mutants (Harris and Preston, 1991;O'Neill et al., 1972;Wilcox and Johnson, 1987). During the late 1980s Wilcox and colleagues developed an in vitro model of HSV latency in cultured primary neurons. Quiescent/latent infections were established by infecting the neurons at very low MOI and maintaining them in the presence of human IgG (which contains significant amounts of anti-HSV antibodies). The antibodies were removed after fourteen days and the cultures remained latently infected for at least five weeks. These autonomic neurons derived from SCG were dependent on nerve growth factor (NGF) for their survival and deprivation of NGF (by anti-NGF antibody depletion) resulted in reactivation of latent virus (Wilcox and Johnson, 1987;Wilcox et al., 1990). These cultures were interesting in that they mimicked some of what was known about latency in vivo, including expression of the LATs (stable LAT introns) in a subset of neurons (Doerig et al., 1991). However, they did not parallel the in vivo situation in all cases. For example, thymidine kinase negative mutants established latency and reactivated within these cultures (Wilcox et al., 1992), which is not the case for either in vivo reactivation in mice, or reactivation in explanted latently-infected mouse ganglia (Field et al., 1982;Field et al., 1979;Field and Wildy, 1978;Pyles and Thompson, 1994b). Reactivation in these cultures could be induced by cyclic AMP repressors and was dependent on caspase three (Hunsperger and Wilcox, 2003). This model was eventually abandoned in part because latent infections (identified by LATs expression) and reactivation was rare at the cellular level and it was not thought likely that widespread death of neurons caused by NGF deprivation would be relevant biologically except with a few exceptions such as nerve resection caused by "blowout" fractures or surgery (Hunsperger and Wilcox, 2003). Rodent fetal and adult neuronal culture models are also being used and are discussed further below.

Discovery of the latency-associated transcripts (LATs) or latency related RNAs (LRRs)
There was great interest in determining whether any viral gene expression was required to maintain latency. The first report of abundant RNA transcription during HSV latency in neurons was published by Tenser and colleagues.
However, the entire viral genome was employed as a probe and so the genomic location of this transcription was not known (Tenser et al., 1982). In 1986 two abstracts describing significant RNA transcription from the terminal repeat of HSV-1 during latency were submitted to the 11 th annual International Herpesvirus Workshop. Stevens and colleagues described the transcription of ICP0 during latency but employed a double stranded probe. Rock and Nesburn also detected transcription but from the strand opposite of that of ICP0. Stevens et al. became aware of Rock's strand specific finding, confirmed that the transcription they detected was from the strand opposite of that of ICP0 and first described the latency associated transcripts (LATs) (Stevens et al., 1987). Arguably, credit for this discovery might be shared.
The discovery of the LATs led to speculation that they might encode a protein or a function important for the establishment, maintenance and/or reactivation of latent infections. However, the earliest reports suggested that the LATs played no role in the establishment or reactivation of HSV latency (Sedarati et al., 1989) (reviewed in (Fraser et al., 1992)). Subsequent early studies led to caister.com/cimb confusion and controversy, largely due to the complex nature of transcription from this region that was not appreciated at the time. We now know that long, short, and micro non-coding RNAs from both DNA strands as well as unusually stable introns are generated from the LAT locus, and that most of these can be detected during both latency and during acute viral replication. Presumably expression of the LATs introns and miRNAs during the lytic cycle are the result of read through transcription late in the infection cycle. Activity from the LAT promoter is restricted during acute infection (Batchelor and O'Hare, 1990).
Additionally, there are reports noting potential open reading frames (ORFs) in some of these RNAs (Bloom et al., 1996;Thomas et al., 1999;Thomas et al., 2002), but no latency related protein has yet been convincingly described. In contrast, the recently discovered VZV LAT RNA does encode a protein (Depledge et al., 2018a). The role of the LAT locus will be discussed in greater detail below.

Quantification of the number of genomes in the latently infected ganglion by the polymerase chain reaction (PCR)
Katz et al. first employed PCR to detect latent viral genomes, which laid the groundwork for the use of qPCR to address questions in HSV latency (Katz et al., 1990). These investigators confirmed the early finding of Stevens and colleagues that latent infections in sensory neurons could be established even with HSV mutants that did not replicate efficiently, including TK-and ICP4deficient mutants, and added to this list mutants lacking ribonucleotide reductase, ICP27, and DNA polymerase (Katz et al., 1990;Lofgren et al., 1977;Watson et al., 1980). More startling was the high numbers of viral genomes Katz et al. detected in latently infected ganglia (Katz et al., 1990). While many groups had observed only a few to several hundred neurons expressing LATs, PCR detected hundreds of thousands to millions of viral genomes (reviewed in (Phelan et al., 2017)). This suggests that either neurons contain multiple copies of the viral genome or that many neurons harboring latent genomes do not express detectable LATs, or both. We now know that there are 1,000s of latently infected neurons in the TGs of mice and humans, each containing 10-100s of viral genomes and most of these do not express caister.com/cimb detectable LAT stable introns. (Sawtell, 1997(Sawtell, , 1998Sawtell et al., 2001;Wang et al., 2005b).

How many genomes are in individual latently infected neurons? Latency at the single neuron level
While the qPCR method provided unprecedented sensitivity for detecting and quantifying viral genomes in latently infected TG, methods to examine viral DNA in single cells were needed to address outstanding questions. For example, how many neurons were latently infected, were other cells also harboring viral genomes, how many viral genomes were within individual latently infected neurons, and was this genome number uniform or variable?
Towards that goal, an approach termed contextual analysis (CXA) was develop. TGs were perfusion-fixed to stop all metabolic processes and to preserve DNA, RNA and protein, and the cells were dissociated and purified on Percoll gradients. Following treatment with DNase linked to beads to remove DNA from the outside of the cells, the intracellular viral genome content of individual neurons and other cells was analyzed by qPCR. Viral genomes were detected in about 1/4 th of the neurons, only in neurons and with copy numbers ranging from ~1 to >10,000 (Sawtell, 1997). Similar percentages of positive neurons and ranges of viral genome copies were later detected in human TG neurons using laser capture technologies combined with qPCR (Wang et al., 2005b). Other single-cell approaches addressing these questions were applied to VZV in human TG (LaGuardia et al., 1999;Levin et al., 2003).
The viral genome copy number in individual latently-infected neurons varies across HSV1 strains . Neurons latently infected with the virulent HSV1 strains McKrae or 17Syn+ contained an average of 81 ± 36 or 50 ± 12 genome copies, respectively. However, neurons latently infected with the avirulent strain KOS contained only 7 ± 2 viral genome copies, which may explain why it reactivates less well than the other strains ).
Yet, the percentage of latently infected neurons was similar with all three strains, varying between 26 and 32%. The reason for the low genome copy number in KOS latently-infected TG neurons is not known but may be a reflection of the mutations in the KOS strain that disrupt the US9 gene TATA box, truncate US9 at aa58, and eliminate the native stop codon of pUS8A (Negatsch et al., 2011). pUS9 is required for efficient anterograde axonal transport (Howard et al., 2013), and pUS8A is a neurovirulence factor (Kato et al., 2016). Thus, the multiple "round trips" of virus from the body surface to nerve cells and back again, which contributes to the higher viral genome copy number latency (Thompson and Sawtell, 2000), may not occur with the HSV1 KOS strain and its derivatives.

Do any viral proteins enhance the establishment of latency?
A number of viral genes contribute to the efficiency with which latency is established. However, many of these are required for efficient lytic replication at body surfaces or in the nervous system. Surface replication efficiency is correlated with the number of latent sites established and, therefore, viral proteins important for lytic replication such as thymidine kinase appear to increase latent infections indirectly (Katz et al., 1990;Thompson and Sawtell, 2000). Other virally-encoded enzymatic functions such as dUTPase and uracil glycosylase enhanced establishment of latency and reactivation frequency likely by promoting replication within the trigeminal ganglia (TG) (Pyles et al., 1992;Pyles and Thompson, 1994a). The lack of these enzymes may also have led to a higher mutation frequency in the latent viral genomes that could exert a deleterious effect on reactivation from latency (Pyles and Thompson, 1994b). In aggregate these and similar studies on diverse viral proteins (i.e. TK, ribonucleotide reductase, virion host shut off, and various glycoproteins) demonstrate that no viral proteins are absolutely essential for the establishment of latency (Aggarwal et al., 2012;Diefenbach et al., 2008;Field et al., 1982;Field and Wildy, 1978;Izumi and Stevens, 1990;Jacobson et al., 1989;Johnson et al., 1986;Lam et al., 1996;Meignier et al., 1988;Smith, 2012;Strelow and Leib, 1995;Wang et al., 2005a;Wang et al., 2010).
However, all viral proteins likely contribute to the efficiency with which latency is established indirectly by enhancing virus replication at body surfaces and in the nervous system.

Viral lytic gene transcripts are present in latently infected ganglia in the absence of detectable reactivation
In a landmark study, Kramer and Coen found transcripts for ICP4 and TK in latently infected mouse TG. Quantification of these transcripts revealed that they were present at extremely low levels. These transcripts did not necessarily initiate at recognized promoters and may have been the products of random transcriptional activity (Kramer and Coen, 1995). Thompson and Sawtell also reported RNAs related to ICP4, ICP0, ICP22, ICP27 and ICP47 in TG latently infected with HSV1. Transcription was found upstream of normal mRNA initiation sites. Furthermore, the ICP0 transcripts were not spliced as is the case during productive infection, again suggesting these might not be properly processed mRNAs, but rather random transcriptional events (Thompson and Sawtell, 2006). The presence of HSV related RNAs during latency has complicated the interpretation of results that utilize viral transcriptional activity as a marker of reactivation. This is especially true in vivo where reactivation occurs in only one or a very few neurons per ganglion and the neuronal distribution of the RNAs detected is not known. Sensitive whole ganglion approaches to detect viral transcription at the single neuron level during latency and in vivo reactivation are needed.

In depth considerations of latency models and mechanisms
The role of VP16 transactivation function in the switch between latent and lytic infection Virion protein 16 (VP16, α-TIF, pUL48) is produced as a leaky late protein and is packaged into the viral tegument. Upon viral fusion into cultured cells and concomitant loss of the envelope, VP16 is released from the nucleocapsid/ tegument and is thought to complex with the host cell factor 1 (HCF-1) in the cytosol. The HCF-1/VP16 dimer is transported into the nucleus where it forms the trimeric VP16-induced complex (VIC) with Octamer-binding protein-1 (Oct-1). The VIC binds to the consensus sequences, TAATGARAT, which are present in the promoters of the five immediate early (IE) genes of HSV. The acidic carboxy terminal domain of VP16 is a strong transcriptional activator, which then initiates the virus lytic gene transcription program by transactivating the IE genes (Ace et al., 1988;Campbell et al., 1984;Kristie and Roizman, 1987;Mackem and Roizman, 1982;O'Hare, 1993), reviewed in (Wysocka and Herr, 2003). The importance of VP16 in initiating the viral lytic cycle is revealed at low moi. In its absence, entry into the lytic cycle (plaquing efficiency) is reduced 100 to 1,000-fold (Ace et al., 1989;Smiley and Duncan, 1997).
Based on what is known about the critical role played by VP16 in initiation of the viral lytic cycle at low moi, its absence or presence would also be expected to play central roles in the establishment of, or reactivation from latency.
Indeed, in cultured neurons, VP16 deposited into axons as part of the tegument complex is not transported retrograde to the neural soma with the nucleocapsid, and in the absence of its delivery to neuronal nuclei latency establishment may be favored (Aggarwal et al., 2012;Antinone and Smith, 2010). Importantly, cultured neurons infected via axons favors the entry of virus into a quiescent/latent state (Hafezi et al., 2012) and the addition of pseudorabies virus tegument proteins to the neuronal soma can shift the outcome to the lytic cycle (Koyuncu et al., 2017), a result similar to early studies of HSV light particles containing tegument proteins including VP16 (Dargan et al., 1995). As discussed further below, VP16 and its unique regulation as a pre immediate early gene in neurons orchestrates the choice between lytic and latent viral programs.
The generation of a transactivation-deficient mutant (VP16TF), in1814, provided the critical tool needed to test the role of VP16TF during latency and reactivation (Ace et al., 1989). This mutant contains a 12 bp insertion at aa379 in VP16 that retains the protein's essential contribution to virion structure but selectively disrupts the interaction of VP16 with Oct-1, thus preventing the formation of the VIC (Ace et al., 1988;Ace et al., 1989;Campbell et al., 1984;Wysocka and Herr, 2003). This and other VP16TF mutants are severely deficient in replication at low moi (McFarlane et al., 1992;Preston and McFarlane, 1998;Smiley and Duncan, 1997 from latently-infected ganglia explanted into culture (Steiner et al., 1990).
Moreover, attempts to artificially-induce the expression of VP16 in vivo did not disrupt the balance between latent and lytic infection (Sears et al., 1991).
These two influential reports seemingly disproved the hypothesis that VP16 played an important role in regulating HSV latency and reactivation.
However, whether the VP16TF studies truly ruled out an initiating role of VP16 in reactivation requires a nuanced consideration of experimental design and interpretation. First, the method used to evaluate reactivation is a central factor. The most widely used approach to evaluate the ability of latent virus within a ganglion to produce infectious virus is to dissect the infected animal, sever the ganglionic neurons by axotomy, and explant the ganglion into culture (Stevens and Cook, 1971), sometimes also dissociating or mincing the ganglion prior to explantation (Blyth et al., 1981;Leib et al., 1989a;Nicholls and Blyth, 1989). Using this method, virus was detected five to ten days post explant, which was not different than the five to six days required for wildtype and rescued virus reactivation (Steiner et al., 1990). At the time, these studies and others led to the conclusion that VP16 transactivation was not involved in reactivation from latency, and as a consequence the IE protein ICP0 superseded VP16 as the major contender for this important role. However, whether explant-based reactivation accurately models in vivo reactivation is an important consideration (Sawtell and Thompson, 2004;Thompson et al., 2009).
Although there is general consensus that ICP0 is required for efficient reactivation from latency. In vivo, ICP0 is required for efficient viral replication and establishment of latency (Cai et al., 1993;Cai and Schaffer, 1992;Everett, 2000;Halford and Schaffer, 2001;Leib et al., 1989b) Sawtell, 2006). Thus, while ICP0 is essential for amplification of infectious virus production during reactivation in vivo this protein is not required, and appears to play no major direct role, in the initiation of reactivation in vivo.
Whether the contribution of ICP0 is to the initiation or progression of reactivation may seem to be a hairsplitting issue, but as discussed above the goal of these studies is to identify the interfaces between the critical host cell factors and their mechanisms of regulating essential viral functions. An early effort to achieve this used a viral mutant, ΔTFI, that was designed to gain insight into the role of cis-acting sequences in the ICP0 promoter that potentially regulate ICP0 expression during acute infection, establishment of latency, and reactivation (Davido and Leib, 1996). A deletion removed a number of sites in the proximal ICP0 promoter including Sp1, NF-kB, C/EBP, F2, and four TAATGARAT motifs. Strikingly, although there were reduced levels of ICP0 protein expression at early times post-infection in Vero cells and in vivo, and reduced levels of blepharitis in CD1 mice and death in SCID mice, neither establishment of latency nor explant reactivation were different from the wild type or the marker rescue viruses (Davido and Leib, 1996).
Our group was also focused on ICP0 and its role in reactivation, but we were utilizing an in vivo reactivation model triggered by hyperthermic stress to the intact animal (Sawtell and Thompson, 1992b (Ace et al., 1989), in the assays, reactivation in vivo (infectious virus production) was not detected in assays in the presence of HMBA.
We hypothesized that some yet unknown regulatory mechanism altered the normally leaky late expression of VP16 to that of a pre-immediate early gene in a neuron destined to reactivate, thereby permitting it to coordinate the expression of the five important IE genes and initiate the lytic cascade (Thompson et al., 2009). If this were the case, then in contrast to ΔTFI, in which viral proteins were expressed in neurons "starting" to reactivate, neurons expressing viral proteins in in1814 infected TG would be absent following stress. Using a whole ganglion approach that allows quantification of viral protein expression at the individual neuron level, the lack of neurons expressing viral proteins was confirmed (Thompson et al., 2009). This extended our understanding of the reactivation process now to include VP16 transactivation of ICP0 (and presumably the other IE genes as well but this has not been directly tested) as a key and absolutely required step in the earliest stages of viral protein expression from the latent genome. Of great caister.com/cimb importance, mutations in the VP16 promoter, including a four base substitution in a G+C rich region scoring as a putative overlapping Egr-1/SP1 site ablated the ability of the virus to exit the latent state (Sawtell and Thompson, 2016;Thompson et al., 2009;Thompson and Sawtell, 2019). These findings are consistent with the hypotheses that both a functional ICP0 and VP16 transactivation were required for efficient virus production, but de novo preimmediate early expression of VP16 is required to initiate viral gene expression during reactivation.
Additional supporting data comes from a series of refined viral mutants containing single or double amino acid changes in the core domain critical for VIC formation (Sawtell and Thompson, 2016;Sawtell et al., 2011), including the first viral mutant to specifically disrupt the HCF-1 interaction. The in vivo reactivation phenotypes observed for these mutants align with predictions based on in vitro biochemical analyses of amino acids critical for VIC formation (Stern et al., 1989;Wysocka and Herr, 2003). Together, there is now strong evidence that VP16 interactions with Oct-1 and HCF are critical for the earliest stages for reactivation in vivo (Figure 1). That the core domain mutant phenotype was a result of a deficit in VP16 transactivation was supported further by the in vivo reactivation impairment of a mutant in which the core domain was intact, but the acidic transactivation domain (TAD) was lacking (Thompson et al., 2009). Importantly, all of these mutants reactivate following ganglion explantation, emphasizing the importance of the reactivation model used and the associated neuronal metabolic state. Efforts to develop and characterize additional in vivo reactivation triggers to ask whether hyperthermic stress is representative of other stressors (i.e., do diverse stressors flow into a common interface on the viral genome) is in progress. In vivo reactivation by additional triggers, for example, local skin trauma (see below), is dependent upon VP16 transactivation.

Modeling reactivation in vitro
The development of an in vitro latency model that provides a platform for dissecting host cell signaling pathways involved in viral genome entry into caister.com/cimb latency, its maintenance, and its reactivation would be a valuable resource.
Ideally such models could yield hypotheses testable in the more complex in vivo setting. Defining the limitations to these models and the approaches used to probe outcomes is critical to interpreting results obtained from them.
There is a growing number of in vitro latency models for both HSV and VZV that are the topic of several recent reviews ( caister.com/cimb

Reactivation in the superior cervical ganglion neuron culture model
As discussed above a primary SCG neuronal culture model of HSV quiescence/latency and reactivation was reported to require nerve growth factor (NGF) to maintain HSV latency/quiescence Johnson, 1987, 1988;Wilcox et al., 1990). One current hypothesis is that the on the abilities of these growth factors to provide sustained signaling through PI3-K and Akt. More recently, the importance of neurotropic factors in adult sympathetic and sensory neurons was examined (Yanez et al., 2017). While NGF and GDNF withdrawal induced HSV1 reactivation in adult sympathetic neurons, in adult sensory neurons NGF deprivation had no effect. Neurturin (NTN) and GDNF withdrawal induced HSV1 and HSV2 reactivation, respectively. Thus, the nature of receptor tyrosine kinase (RTK) signaling appears to be a key host parameter that regulates the HSV1 quiescent to lytic activation in a subset of neurons even in this "homogeneous neuronal" model system. While the importance of neurotrophic factors during development is well established (Indo, 2018), in the adult nervous system a role for continuous NGF or other neurotropic support is less clear as is the relevance of neurotropic factor withdrawal as a trigger of reactivation (Skaper, 2017).

A two phase VP16TF dependent HSV reactivation model
In a second study using the SCG neuron culture model, the expression of 4-5 viral genes from IE, E, Lg1, Lg2 kinetic classes were profiled following LY294002 treatment . Two distinct phases of transcriptional activity were observed. During the first at 20-24hr (Phase I), extremely low levels of UL30 (658 RNA copies/sample*) and UL48 (3,470 RNA copies/ sample*) transcripts were detected when protein synthesis was inhibited (cycloheximide, which blocks all protein synthesis, was added 10 hrs after LY294002  (Sawtell, 1998(Sawtell, , 2003Sawtell et al., 1999;Sawtell et al., 2001). This is observed in vivo during the final stages of a reactivation event in reactivating neurons and late during the lytic cycle in ganglia (Doll et al., 2020;Goodpasture, 1929). However, if this were the case and such neurons were removed through time, a reduction in the size of the in vivo latent reservoir would be expected over time, which is not observed (Thompson and Sawtell, 2011). These differences could arise from differences between SCG and sensory neurons, and examination of latent infection and reactivation at the neuronal level in SCG in vivo may be informative.
An additional feature of the SCG neuron culture model is the hypothesis is that VP16 remains sequestered in the cytoplasm with HCF-1 until HCF-1 is transported to the nucleus, which is triggered by LY294002 treatment. Whether these HCF-1/ LY294002 studies were done in infected or uninfected cultures is difficult to interpret

A histone methyl/phospho switch is required for HSV reactivation
A third study using the SCG neuronal culture model revealed novel concepts regarding the mechanism of HSV reactivation with respect to Phase I and its requirement for Phase II. The conclusions drawn from this study are that (i) a methyl/phospho switch at H3S10 during Phase I is required for HSV reactivation and that (ii) the DLK/JIP-3 JNK pathway is directly linked to this switch. While this study has the potential to be quite important in providing insight into the molecular signaling underlying HSV reactivation, aspects of the experimental design raise questions and emphasize the challenges faced. The potential ability of HCF-1 to be recruited to the viral genome through binding partners other than VP16 raises the possibility that HCF-1 could play a caister.com/cimb role in initiating expression of the viral IE genes apart from its interaction with VP16. This hypothesis was proposed more than a decade ago (Kristie et al., 2010;Whitlow and Kristie, 2009)  caister.com/cimb

Figure 2. Comparison of latency mouse TG in vivo and in vitro in SCG rodent neuron model. The SCG neuron model appears to have hallmarks of a persistent infection
including viral protein expression in 10-20% of latently infected neurons/well and a "spontaneous reactivation rate 50-fold greater than observed in vivo. References (1)= for example (Sawtell, 1998); (2) = for example    Nonetheless, the ability of JQ1 to perturb viral latency in vivo is significant.
JQ1 is a broad BET domain inhibitor and not selective for Brd4. Thus, in the absence of additional studies, the mechanism underlying the observed phenotypes is yet to be illuminated.

Frequency of HSV2 genital shedding implies HSV2 is continuously "reactivating" in the ganglion. Corollary: the reactivating neuron must survive the production and release of infectious virus
Multiple studies have now solidified not only the shedding frequency but also the peak amounts of viral DNA and rate of decay with which HSV2 DNA is detected in human genital swabs collected from a specified regional map (Agyemang et al., 2018;Ramchandani et al., 2017;Sacks et al., 2004). These studies have substantially altered what had been established concepts of the frequency of viral presence at the genital mucosal surface and raise important caister.com/cimb questions regarding the transmission potential of these frequent PCR detectable shedding events. The results of these studies were best fit by a mathematical model that assumed a nearly constant release of small numbers of virions from ganglionic neurons (Schiffer et al., 2009). While a legitimate hypothesis, it is worth noting that the source of viral DNA detected by PCR in the genital swab is not known but rather assumed to be the innervating sensory ganglia. Thus, the concept is based on a mathematical model which assumes that the only source of viral DNA is (or could be) the sensory ganglion. More recent mathematical modelling from this group is focused on immune control in the mucosa (Gottlieb et al., 2017;Schiffer et al., 2018).
Based on the shedding frequency, neurons presumably survive reactivation and are able to undergo repeated reactivation events; otherwise, frequent shedding cannot easily be explained. However, the ability of HSV2 to persist in other cell types is well described in the older literature, including HSV in hair follicles, skin, T cells, B cells, and myeloid cells (Al-Saadi et al., 1988;Claoue et al., 1987;Easty et al., 1987;Nicholls et al., 1996;Scriba, 1977Scriba, , 1981Shimeld et al., 1986;Shimeld et al., 1982;Tullo et al., 1985). Additionally, the idea that the reactivating neuron must survive based on this level of shedding does not appear to consider the number of ganglia that could be potentially alternate sources for the viral DNA in genital skin or mucosa. Progress in developing animal models to gain insight into the contribution of other ganglia to vaginal reactivation (Bertke et al., 2007;Lee et al., 2015;Pieknik et al., 2019;Yanez et al., 2017) as well as developing the tools in HSV2 to test hypotheses in vivo (Kawamura et al., 2018;Pieknik et al., 2018) will undoubtedly deepen current understanding of HSV1 and 2 reactivation and the distinctions between them.
A recent report detailed the resolution of reactivation of HSV1, characterizing the surrounding cellular context and morphological changes in individual neurons undergoing reactivation in vivo (Doll et al., 2020).

Functions of products from the LATU locus
As more knowledge concerning the rather subtle effects that non-coding RNAs exert on many different processes including transcription, translation, post translational modification, and chromatin structure have been described, the appearance that the non-coding RNAs generated in the LAT locus had no or minor effects on latency/reactivation can be attributed at least in part to the general lack of assays sensitive enough to measure their influences individually and in aggregate (Chen and Aravin, 2015; Dhanoa et al., 2018;He et al., 2014;Zheng et al., 2017). Space restrictions prevent presentation in detail of all of these earlier studies on the LAT locus and its functions here, and readers are directed to a recently published comprehensive review on this subject (Phelan et al., 2017). In brief, the latency associated transcript (LAT) is a ~8.5 kb mRNA that maps largely to the long terminal repeat of the virus and is antisense to the ICP0 and ICP34.5 genes (Wechsler et al., 1988). It is the only mRNA abundantly transcribed during latent infection of neurons and it expression is strongly repressed during acute infection by an ICP4 binding site that includes the transcriptional start site of LAT (Batchelor et al., 1994). The LAT mRNA is spliced, which produces two co-linear introns of 2.0 and 1.5 kb that share the same splice donner and acceptor sites, differing by an internal caister.com/cimb splice in the 5' end of smaller intron (Farrell et al., 1991;Phelan et al., 2017).
These introns are stable due to an unusual lariat structure that is not resolve efficiently and are partially antisense to the 3' end of the ICP0  Thompson first determined that LAT locus mutants establish latent infection less efficiently than wild type. While the mechanism(s) involved still remain controversial, this phenotype is one that has stood the test of time and is shared among the various mouse rabbit and guinea pig models (Phelan et al., 2017; (Krause et al., 1995;Perng et al., 2000b;Sawtell and Thompson, 1992a;Sawtell, 1997, 2001).
Two important questions remained. Is the LAT locus required solely for the efficient establishment and/or maintenance of latency, or does it also exert a caister.com/cimb positive effect on reactivation from latency? What is the mechanism underlying the reduced establishment of latent infections and/or the increase in reactivation from latency? Answering these questions required more sensitive quantitative assays. Using a quantitative single neuron approach, it was found that fewer latent infections were established in mice infected with LAT-null mutants. When methods were employed to reach equivalent latency between wild type and mutant strains, no evidence for a specific defect in viral reactivation from latency was detected (Thompson and Sawtell, 1997).
Counting the total number of neurons present in uninfected or latently infected mouse ganglia latently infected with wild type or LAT null mutants in strain 17syn+ revealed twice the number of neurons were killed in mouse TG infected with the LAT-null mutants compared to the rescuants and wild type.
Very low levels and similar numbers of neurons undergoing apoptosis were seen in both groups (Thompson and Sawtell, 2001). Whether transcription from the LATU locus exerts a negative effect on virus entry into the lytic cycle in neurons and in its absence viral lytic infection kills more neurons, or something made from the LATU protects neurons from cytopathic effects associated with HSV replication and thereby increases successful establishment and/or reactivation frequencies, is unknown. Whatever the mechanism, LATU expression reduces lytic cycle gene expression during acute infection of mouse TG neurons (Garber et al., 1997). Increased neuroinvasiveness was displayed by mutants lacking specific regions of the LAT locus (Jiang et al., 2016;Jiang et al., 2015;Thompson and Sawtell, 1997).
More recent findings demonstrate that the LAT transcription program is expressed early and first in TG neurons during acute in vivo infection. Thus, LAT is expressed at a time when it could negatively influence entry into the lytic cycle in neurons. (Sawtell and Thompson, 2016).

HSV LATs and neuronal apoptosis
The hypothesis that the LAT locus protects newly infected sensory neurons from apoptosis immediately after the lytic cycle has ended in the ganglia (hypothesized to be induced as the result of toxic viral virion proteins) is steeped in the lore of HSV1 latency (Perng et al., 2000a). As recently reviewed this may not be true . Significantly, Perng et al. reported extensive apoptosis (>25% of neurons per section) in 66% of TG sections infected with a LAT mutant and in 4% of sections from TG infected with the rescuant or from TG of uninfected rabbits on day 7 postinfection (Perng et al., 2000a). Unfortunately, this experiment has never been repeated. A manuscript that is often cited to support the idea that the LAT locus blocks apoptosis in the mouse model actually reported very low numbers of neurons undergoing apoptosis (<0.01 to <1%) 30 days postinfection with LAT mutants (Branco and Fraser, 2005), and the relationship of this minor amount of TUNEL staining at 30 day postinfection to the "extensive apoptosis" detected by (Henderson et al., 2004;Jin et al., 2005) and colleagues in rabbit TG only at day 7 postinfection, but not at earlier or later times is not obvious. Further studies with mutants in which anti-apoptotic proteins were inserted in place of LAT are considered to support this hypothesis (Jin et al., 2005;Perng et al., 2002).
However, only the downstream phenotypes of ocular shedding (rabbit) or explant reactivation (mouse) were examined, and not the apoptosis of neurons per se. That Wechsler and colleagues were unable to demonstrate more extensive apoptosis in mouse TG infected with LAT null mutants but did find suppression of the reactivation deficient phenotype in mice suggests other mechanisms may have been responsible. The pleomorphic functions of the inserted proteins and the absence of assessment of apoptosis precludes attributing the observed effect on the inhibition of apoptosis. Further experimentation is warranted to test this hypothesis, and experimentation might be productively extended to studies in human neuronal cultures as well.
Ectopic expression of portions of the LAT locus interferes with apoptosis induced with drugs like etoposide in cultured cells (Jin et al., 2003;Peng et al., 2003). Two small non-coding RNAs from this locus can inhibit viral replication and can also cooperate to reduce apoptosis in cultured cells (Shen et al., 2009). These functions may be important during productive infection by the virus, which encodes many functions that inhibit apoptosis including ICP0, ICP34.5, gJ, gD, US3, and ribonucleotide reductase (for review see (Yu and He, 2016)).

The LAT locus serves to maintain latent infections
Regardless of the mechanisms involved, it is clear that the latency associated transcript locus of HSV-1 is required for long-term maintenance of reactivation competent latent infections (Thompson and Sawtell, 2011). Does the LAT locus specifically enhance reactivation from latency in the rabbit model?
As discussed above, when the level of latency established by LAT null mutants in mouse TG is equivalent to that of wild type virus, stress induced reactivation from latency is also equivalent, and even more efficient on a per latently infected neuron basis (Thompson and Sawtell, 1997). This locus is also required for the long-term maintenance of latent infections (Thompson and Sawtell, 2011), emphasizing that in the mouse model it serves to reduce lytic infection. However in the rabbit a number of studies support the idea of a direct role for the LAT locus in promoting viral reactivation from latency (Bloom et al., 1996;Hill et al., 1990) as reviewed in (Toma et al., 2008). This may indeed be a difference between the rabbit and murine models as in the rabbit LAT locus transcription is associated with an increase in lytic gene transcripts such as ICP4, TK and gC, (Giordani et al., 2008) whereas the exact opposite, a decrease in these transcripts, was seen in the presence of an intact LAT locus (Garber et al., 1997).
A recent report suggests LAT null mutants may have a specific reactivation deficit. Using an interesting new approach that circumvents this problem, Bloom and colleagues expressed hammerhead ribozymes directed at the 5' exon of the primary LAT (or a control ribozyme) in TG neurons of rabbits latently infected with wild type HSV1. A significant reduction of induced reactivation was found in rabbits receiving the anti-LAT ribozyme (Watson et al., 2018). If this approach and result stand the test of time, a pro reactivation function of the LAT locus will have strong support. Similar approaches in the mouse model could reveal such a pro-reactivation function of LAT as well.
HSV is also well documented to establish latent infections that can be reactivated in a variety of non-neural tissues including denervated mouse footpads or footpads following long-term acyclovir treatment, various nonneural tissues from rabbit, mouse, rat, and guinea pig including cornea, conjunctiva, and hair follicles in humans and LAT expression has been detected in many of these ((Al-Saadi et al., 1988;Claoue et al., 1987;Easty et al., 1987;Nicholls et al., 1996;Scriba, 1977Scriba, , 1981Shimeld et al., 1986;Shimeld et al., 1982;Tullo et al., 1985) and see (Kennedy et al., 2011) for a relatively recent review). These extra-neural sites of latency are usually dismissed but may be biologically relevant, especially in recurrent ocular shedding and eye disease in the rabbit and humans. Whether the LAT locus plays a direct role in promoting reactivation from such sites deserves further study.

How does the virus go latent in the first place?
The ubiquity of HSV1 in the human population is the result of the efficient establishment of latent infections that subsequently reactivate throughout the host's lifetime and transmit to new hosts. A long-lived question is how the virus evolved to interact with sensory neurons to promote the efficient establishment of latency, which requires viral replication within the sensory ganglia but also a self-limiting mechanism to prevent neurovirulence and transmission into the CNS. The discovery of the LAT (Stevens et al., 1987) and mapping of its promoter (Dobson et al., 1989) provided a powerful tool to examine latent phase transcription in TG neurons by in situ hybridization for the LATs or by using LAT promoter/beta-galactosidase reporter mutants (Margolis et al., 1992;Sawtell and Thompson, 1992a;Simmons et al., 1992;Speck and Simmons, 1992).
A somewhat unexpected result common to all these studies was that the acute stage of infection in the TG (examined at 48 hours postinfection and later) was characterized by neurons expressing viral proteins and others expressing only the LATs or the LAT promoter. A few neurons were detected that expressed both, but these were rare. These data were interpreted to suggest that latent and lytic pathways in neurons were district from each other (Lachmann and Efstathiou, 1997;Margolis et al., 1992;Sawtell and Thompson, 1992b;Simmons et al., 1992;Speck and Simmons, 1992), and that the lytic and latent viral transcriptional programs in neurons were regulated in ways that were not understood. Hypotheses were proposed to explain this transcriptional programmatic duality including differential lytic/latent expression in neuronal subtypes, (Cabrera et al., 2018;Yang et al., 2000), transition from some acute gene expression into the latent program (Proenca et al., 2008), failure of VP16 to be transported to neuronal nuclei (Sears et al., 1991;Steiner et al., 1990), alternative regulation of gene expression (Kosz-Vnenchak et al., 1993), strict nonnuclear compartmentalization of essential co-activators such as HCF-1 specifically in neurons (Kristie et al., 1999), or that its interaction with alternate neuronally expressed cofactors like HCF-2, Oct-2 or Brn3 might inhibit its function (Liang et al., 2009;Lillycrop et al., 1993;Nogueira et al., 2004). Any or all of these mechanisms may act to constrain virus entry into the lytic phase in neurons, although replication in TG neurons is required to promote the efficient establishment of latency (Field et al., 1982;Field and Wildy, 1978;Katz et al., 1990;Kramer and Coen, 1995;Thompson and Sawtell, 2000), for review see (Hill, 1985;. To avoid injury or death of the host, this entry into the lytic cycle in neurons must be tightly controlled. Several lines of investigation support the concept that VP16 and its unusual regulation in neurons plays a central role in the choice between latent and lytic caister.com/cimb infection (Sawtell and Thompson, 2016;Sawtell et al., 2011;Thompson et al., 2009). In the absence of the VP16 transactivation function latency is favored over acute viral replication in TG neurons in vivo (Sawtell and Thompson, 2016;Sawtell et al., 2011;Steiner et al., 1990;Thompson et al., 2009) Farrell et al., 1994;Margolis et al., 1992;Sawtell and Thompson, 1992a), we concluded that the viral protein expression observed at these early times postinfection occurred in neurons transitioning out of the acute stage latent program into the lytic cycle.
Mutants expressing a second copy of wild type VP16, the VP16TF deficient mutant, or a third copy of ICP0 driven by the LAT promoter were examined to determine whether the de novo expression of either ICP0 or VP16 was sufficient to push neurons out of the default latent transcription program into the lytic program. If so, viral protein expression and virus production in TG would be expected to occur about 14 hours earlier than that seen with wild type HSV1 infection. Early viral protein expression and infectious virus production was seen in TG of mice infected with the mutant expressing VP16 from the LAT promoter, but not in those infected with the LAT driven VP16TF mutant or ICP0 gene. Therefore, de novo expression of VP16, but not of ICP0, precipitated entry into lytic infection in TG neurons and the transactivation function of VP16 was required for this.
However, the requirement for the VP16 transactivation function is not absolute in so far as VP16TF viruses replicate to some extent in TG albeit at a ~100 fold reduction (Sawtell et al., 2011;Tal-Singer et al., 1999;Thompson et al., 2009). High genome copy number could potentially override a requirement for VP16TF and this is supported by the finding that strategies increasing surface caister.com/cimb replication also increase replication in the TG (Sawtell et al., 2011;Thompson et al., 2009).
A specific region of the VP16 promoter regulated the transition into the lytic transcription program The region of the VP16 promoter near and downstream of the VP16 TATA box is dispensable for normal leaky late kinetics (Lieu and Wagner, 2000).
Thompson and Sawtell found that 3 bp mutations in two putative factor binding sites in this region resulted in wild type virus replication in cultured cells and on the mouse eye, but nearly completely ablated the ability of the virus to transition out of the default latent pathway (2% transition vs. 30% at 44 hours postinfection (Sawtell and Thompson, 2016) Combined, the studies offer strong support for the hypotheses that: (i) VP16 is generally not transported retrograde in axons, favoring latency in neurons.
(ii) The VP16 promoter contains elements required for the de novo preimmediate early production of VP16, which drives exit from a default latent pathway into lytic viral replication in neurons.
(iii) Production of this VP16 is not the result of random "phase 1" type transcription seen in persistently infected cultures of primary neurons treated with AKT or protease inhibitors (Cliffe et al., 2015;Kim et al., 2012).
(iv) Latency is established by default in sensory neurons because VP16 does not arrive with the viral genome. Subsequent reactivation of HSV from the default, or consolidated latent state requires de novo VP16 production, and at least for exit from the default latent state this is mediated specifically by the VP16 promoter (Sawtell and Thompson, 2016;Thompson et al., 2009). caister.com/cimb

Events during the consolidation of HSV latency
Assembly of chromatin structures on latent viral genomes is thought to be an important aspect of silencing the genome. In cultured cells viral DNA is rapidly associated with histones and chromatin is rapidly assembled, occurring in the first thirty minutes post infection (Roizman et al., 2005). This appears to be a multi-step and curiously slow process in TG neurons in vivo with initial events detectable at 7 days postinfection and modifications associated with silencing of genes such as heterochromatin formation occurring on lytic promoters after two weeks post infection, whereas the LAT promoter is associated with euchromatin at this time (for review see (Conn and Schang, 2013;Knipe, 2015;Kristie, 2015;). Since viral gene expression ceases well before this time, this process may be more important for maintaining latency rather than the initial formation of latent infections. In support of this idea, recent studies suggest that chromatin insulator sequences and their binding proteins may maintain boundaries between regions of the latent HSV genome that are actively transcribed (e.g. the LAT locus) from the lytic phase genes that are silenced (Lee et al., 2018;Washington et al., 2018b;Watson et al., 2018).
Part of the difficulty in understanding the effects of chromatin modifications and boundaries is that latent viral genomes are diverse in these properties.
Approaches employed on whole tissues can only detect changes averaged over these millions of latent genomes. Lomonte and co-workers have developed methods to take such analyses to the next level in single neuronal nuclei. As expected, they detected multiple genomes in many neurons. What was not expected is that some of these are intimately associated with PML containing nuclear bodies, whereas others in the same nucleus are not. Also, of interest, some genomes express LAT locus RNAs, whereas others in the same nucleus do not. The silencing properties of these PML-NBs are thought to provide an intrinsic anti-viral defense mechanism that serves to promote latency (Cohen et al., 2018;Lomonte, 2016;Maroui et al., 2016).
Efstathiou and colleagues developed a method to historically interrogate viral promoter expression in individual neurons that survive the acute stage of infection. Employing the Cre reporter mouse strain ROSA26R in combination with viral mutants that express Cre from various promoters they discovered that promoter expression in neurons that survive is far more frequent than previously thought (Proenca et al., 2008;Proenca et al., 2011;Proenca et al., 2015). However, the extent of production of the relevant viral proteins is not yet known. Another group employed this assay to examine the potential expression of viral promoters for genes encoding proteins that are important targets for immune surveillance. Evidence for the expression of these promoters suggest periodic expression of these promoters might be one mechanism whereby long-term immunity is maintained (Russell and Tscharke, 2016).

A role for HSV 1 latency/reactivation in the development of neurodegenerative disorders
The neurotropic nature of HSV and VZV, the extensive data identifying these viruses in the human CNS at autopsy or identification of virus in the cerebrospinal fluid together with the concept of viral latency and reactivation in the trigeminal ganglion led to the hypothesis "that reactivation of the same dormant viral material travelling centripetally instead might be the cause of the "degenerative" lesions typical both of AD and of the normal aged brain." -MJ Ball (Ball, 1982). In 1997 Itzhaki et al. provided evidence that the combination of HSV1 in brain and carriage of an APOE-e4 allele is a strong risk factor for AD Lin et al., 1997). Evidence is now emerging that directly supports the concept that HSV1 infection is a major risk factor in the development of Alzheimer's Disease (reviewed in (Itzhaki, 2018;Lathe et al., 2019)). This concept has been met with much skepticism over the years for reasons including the high prevalence of HSV infection relative to AD and the correlative nature of the studies. Clinical trials are now ongoing to test the efficacy of valacyclovir in individuals with mild AD (ClinicalTrials.gov Identifier: NCT03282916). Animal models are under development, and in human-ApoE4targeted replacement mouse, long-term HSV1 infected but not mock-infected caister.com/cimb mice exhibited spatial memory deficits and CNS pathology consistent with AD.
HSV protein was detected in the hippocampus but was extremely rarely (work in progress). Such models will allow efficacy of antiviral treatments to be tested and investigation into the mechanisms of disease initiation and progression.

Conclusions
Increasingly sensitive methods to detect viral transcriptional activity from the latent viral genome have modified the notion of a strictly "silent" vs "active" While the animal models of HSV latency/reactivation are not perfect, they have proven to support the complex viral life cycle with reasonable fidelity and these models approximate the level of complexity encountered in the human host.
The simple act of axotomizing and explanting the ganglia results in changes that yield outcomes that do not align with the requirements for productive reactivation in the in vivo context such as the requirement of the VP16 caister.com/cimb transactivation function to initiate detectable viral protein synthesis (compare (Steiner et al., 1990) to (Thompson et al., 2009). It is becoming increasingly apparent that properties of a complex system cannot be reduced to the components of a system because they depend on many interactions.
Emergent properties of a complex system are lost when a system is stripped down to eliminate "extraneous" interactions or "complicating" interactions, with surrounding cells including satellite cells, resident glial and microglial cells, and infiltrating immune cells. In the pursuit of understanding HSV reactivation, a revival of reductionist approaches and models has arisen. The development of an in vitro latency model that provides a platform for dissecting host cell signaling pathways involved in viral genome entry into latency, its maintenance, and its activation would be an important tool. Ideally such a model could yield hypotheses testable in the more complex in vivo setting.
However, defining the limitations to all models and approaches used to probe outcomes in these models is a critical component of the scientific process.