Characterization of the Host Immune Response in Human Ganglia after Herpes Zoster

Abstract

Varicella-zoster virus (VZV) causes varicella (chicken pox) and establishes latency in ganglia, from where it reactivates to cause herpes zoster (shingles), which is often followed by postherpetic neuralgia (PHN), causing severe neuropathic pain that can last for years after the rash. Despite the major impact of herpes zoster and PHN on quality of life, the nature and kinetics of the virus-immune cell interactions that result in ganglion damage have not been defined. We obtained rare material consisting of seven sensory ganglia from three donors who had suffered from herpes zoster between 1 and 4.5 months before death but who had not died from herpes zoster. We performed immunostaining to investigate the site of VZV infection and to phenotype immune cells in these ganglia. VZV antigen was localized almost exclusively to neurons, and in at least one case it persisted long after resolution of the rash. The large immune infiltrate consisted of noncytolytic CD8⁺ T cells, with lesser numbers of CD4⁺ T cells, B cells, NK cells, and macrophages and no dendritic cells. VZV antigen-positive neurons did not express detectable major histocompatibility complex (MHC) class I, nor did CD8⁺ T cells surround infected neurons, suggesting that mechanisms of immune control may not be dependent on direct contact. This is the first report defining the nature of the immune response in ganglia following herpes zoster and provides evidence for persistence of non-latency-associated viral antigen and inflammation beyond rash resolution.

Varicella-zoster virus (VZV) is a highly species-specific human alphaherpesvirus that infects a majority of the world's population. VZV causes two clinically significant diseases; varicella (chicken pox) and herpes zoster (shingles) (5, 8, 19). Varicella is characterized by widespread cutaneous vesicular lesions and is a consequence of primary VZV infection in VZV-naïve individuals. While varicella is a relatively mild disease in immunocompetent children, it can cause significant morbidity in healthy adults and is frequently life threatening in immunocompromised individuals (3, 4, 22). The innate and adaptive immune responses act to eliminate replicating virus during varicella, but not all virus is cleared during this time, with some presumed to access nerve axons in the skin, enabling transport to neurons in sensory ganglia, where the virus is able to establish a lifelong latent infection (5, 8, 12, 13, 20, 32). An alternative possibility is that virus is transported to ganglia via hematogenous spread (36). The ability of VZV to establish latency in the host is critical to the success of this virus as a human pathogen.

VZV reactivation from latency (herpes zoster) causes serious disease in older and immunocompromised individuals and is characterized by vesicular skin rash in a dermatomal distribution with preceding and concomitant pain (7, 10, 21, 68). During reactivation, sensory ganglia are sites of viral replication, where an intense inflammatory response is induced and widespread necrosis of glial cells and neurons ensues (14, 19, 27, 71, 72). Before the appearance of the zoster rash, VZV travels along the affected sensory nerves to the skin, where it produces the characteristic rash (10, 53) and neural and dermoepidermal inflammation. Clinically, herpes zoster is associated with severe, acute pain, as well as often prolonged severe pain, or postherpetic neuralgia (PHN), that often requires follow-up medical care for months or even years after the initial attack (29, 62, 73). PHN is estimated to occur in 40% of herpes zoster cases in individuals older than 50 years and 75% of adults older than 75 years (15, 43, 56). It is estimated that 1 million or more individuals are afflicted by herpes zoster each year in the United States (54). Herpes zoster pain, and especially PHN, can be disabling and can have a major negative impact on patients' quality of life (15). In the coming years, the number of individuals suffering from herpes zoster is predicted to rise, concomitant with the increasing number of patients who are elderly or who are receiving immunosuppressive therapies for cancer or transplantation.

New antiviral drugs and a vaccine for herpes zoster have been only partially successful, indicating the need for continuing studies of VZV immunopathogenesis to understand the reasons for this partial success and to provide the foundation for developing new immunotherapeutics and vaccines (38, 39, 65). Antiviral therapy, while effective against the rash and pain of acute herpes zoster, appears at best to prevent only 50% of PHN (16, 23, 24, 45, 75, 76). The zoster vaccine was demonstrated to prevent 51% of herpes zoster and 60% of postherpetic neuralgia in patients over the age of 60, although it appeared to be less effective against zoster in the older age group (54). Remarkably, despite the importance of ganglionic infection to the pathogenesis of herpes zoster and PHN, there have been no reports defining the immune response in human ganglia following natural VZV reactivation. Until now, the lack of available ganglia from patients following an episode of herpes zoster has limited these studies. We have overcome this hurdle by obtaining rare naturally infected human ganglia at autopsy from three donors who, near the time of death, had evidence of herpes zoster but who did not die from herpes zoster. The aim of this study was to undertake a comprehensive immunohistological examination of human ganglia following herpes zoster. Specifically, we utilized immunohistochemical (IHC) and immunofluorescent (IF) staining to characterize the infiltrating immune cell subsets and to assess the presence of VZV antigen within ganglia following herpes zoster. This study provides the first detailed examination of the types and distribution of immune cells present following natural VZV reactivation in human ganglia and provides new insights into the immunological mechanisms that may be important in controlling virus infection following the reactivation of a human herpesvirus infection in human ganglia in vivo.

MATERIALS AND METHODS

Human tissue samples.

Paraffin-embedded dorsal root ganglia (DRG) and trigeminal ganglia (TG) (Table 1) and human tonsil and skin tissue were obtained with approval by the Western Sydney Area Health Service and/or University of Sydney Ethics Committee.

TABLE 1.

Patient ganglion samples used in this study^a

Case history	Patient 1 (sample C1)	Patient 2 (sample C2)	Patient 3 (sample C3)	Patient 4 (sample Z1)	Patient 5 (sample Z2)	Patient 6 (samples Z3a, Z3b, Z3c, Z3d, and Z3e)
Age (yr)	83	86	16	54	82	36
Gender	Male	Male	Female	Male	Male	Female
Cause of death	Squamous cell carcinoma	Alzheimer's, mild tangles	Epilepsy, cerebral palsy	Non-small cell carcinoma	Bronchpneumonia	Vasculopathy
HZ details (time after onset of herpes zoster rash that ganglia were collected)	None	None	None	Trigeminal (not known)	Geniculate, thoracic (30 days)	Cervical, thoracic (4.5 mo)
Ganglia examined	Right TG	TG	DRG	Right TG	Left L1	Z3a, right C1; Z3b, right C2; Z3c, right C3; Z3d, right TG; Z3e, left TG

^aC, cervical ganglia; L, lumbar ganglia; HZ, herpes zoster.

Antibodies.

The following primary antibodies (Ab) were used: mouse anti-human CD3, CD8 (clone 1A5), CD20, CD68, CD57, CD1a (clone MTB-1), and CD25 (clone 4C9) (1:10; Novocastra, Melbourne, Australia); anti-human CD4 (clone IF6; 1:10; ID Laboratories, London, Canada); anti-VZV glycoprotein E (gE) mouse monoclonal antibody (1:100 for IHC and 1:10 for IF staining; Chemicon, Sydney, Australia); anti-T-cell intracellular antigen (TIA-1) (clone 2G9; 1:10; Immunotech, Sydney, Australia); anti-granzyme B (clone GrB7; 1:10 [Monosan, Uden, Netherlands]; isotype control mouse IgG2a (clone G155-178; 1:10; BD Pharmingen, Sydney, Australia); rabbit anti-human CD8 (clone EP1150Y; 1:10; Abcam, Cambridge, United Kingdom); rabbit anti-human β₂-microglobulin (β₂m) (1:200; Novocastra, Melbourne, Australia); and isotype controls, including rabbit IgG, and goat IgG (R&D Systems, Minneapolis, MN). Secondary fluorescent antibodies included donkey anti-mouse Alexa Fluor 488 and donkey anti-rabbit Alexa Fluor 594 (1:200; Invitrogen, Sydney, Australia).

Immunohistochemistry.

Deparaffinized, rehydrated 5-μm sections were incubated in 3% hydrogen peroxide (Fronine) for 10 min at room temperature (RT) prior to high-temperature antigen unmasking, which was carried out by boiling the sections in citrate buffer, pH 6, or EDTA buffer, pH 8, at 95°C for 15 min. The sections were blocked with 10% goat serum in Tris-buffered saline (TBS) for 20 min before addition of primary Ab for 1 h at 25°C, washing in TBS, and a 10-min incubation with a secondary polyvalent Ab conjugated to biotin, followed by a 10-min incubation with streptavidin-horseradish peroxidase (HRP) (ID Laboratories). Incubation and development with diaminobenzidine (DAB) were carried out according to the manufacturer's instructions (Novocastra). The sections were counterstained with Mayer's hematoxylin (Fronine), dehydrated in ethanol, and mounted using Ultramount 4 (Fronine).

For quantification, cell counts were performed for 15 randomly chosen fields per section, using a 1-mm² graticule (Leica Microsystems, Sydney, Australia), and the percentage of viral-antigen- and/or immune cell marker-positive cells was determined. Quantitative data for the normal sections were represented as mean ± standard error of the mean (SEM) unless otherwise noted.

Dual-immunofluorescence staining.

Slides were deparaffinized and rehydrated, and antigen retrieval was performed as outlined above. Following a 30-min block with 20% normal donkey serum (NDS) (Sigma-Aldrich, Sydney, Australia) in TBS, primary antibodies were applied for 1 h at room temperature or 37°C. Following this, the slides were washed in TBS for 5 min, and then a 0.3% solution of Sudan Black B (Sigma-Aldrich, Sydney, Australia) in 70% ethanol was applied to the sections. Sudan Black B binds lipofuscin present in neurons to reduce autofluorescence (59). After 5 min, the Sudan Black B solution was rinsed off by dipping the slides in 70% ethanol, and this was followed by a further 5-min wash in TBS. Secondary antibodies were then applied for 30 min at 37°C. The slides were then washed in TBS twice for 5 min each time and mounted using Prolong Gold antifade reagent containing 4′-6-diamidino-2-phenylindole (DAPI) (Invitrogen, Sydney, Australia).

RESULTS

Description of cases and histological analyses of human ganglia following herpes zoster.

A major limiting factor in advancing our understanding of the interaction of VZV with human ganglionic cells and the role of the immune response following VZV reactivation in vivo has been obtaining human ganglion samples. We sourced autopsied paraffin-embedded ganglia from three individuals (Z1, Z2, and Z3) with evidence of herpes zoster before death but who did not die from herpes zoster (Table 1). Additional details derived from the available clinical notes are as follows.

Case Z1 was a 54-year-old man with underlying disseminated undifferentiated small cell carcinoma of the lung, who presented with pneumonia and pleurisy, confusion, headaches, vomiting, and later right lower facial vesicular rash with no focal neurologic signs. He deteriorated over 3 weeks and died 4 days after admission to the hospital. At postmortem, there was histologic evidence of right trigeminal ganglion herpes zoster but not of VZV encephalitis or arteritis. Case Z2 was an 82-year-old man admitted with a second episode of pseudobowel obstruction who developed extensive evolving typical unilateral vesicular rash in left T10-L1 dermatomes. Herpesvirus particles were observed in cutaneous vesicles by electron microscopy. Right nonvesicular otitis externa and right facial nerve palsy developed, suggesting Ramsay Hunt syndrome. The rash resolved over 3 weeks, but he died in the hospital from bronchopneumonia. At postmortem, there were no viral inclusions in the lungs, but the L1 dorsal root ganglion showed inflammation consistent with herpes zoster. Case Z3 was a 36-year-old woman treated with prednisone and azathioprine following a renal transplant at the age of 11 and then changed to prednisone and mycophenylate mofetil (MMF) 2 years prior to death. She presented with right C1, C2, and T2 herpes zoster with scattered extradermatomal abdominal and limb lesions 4.5 months prior to death, followed by a new upper lip lesion (not treated with antivirals). Two months later (after healing of the rash), she developed right-limb and facial weakness, fluctuating expressive aphasia, and dysarthria. Imaging studies showed tight stenosis in the left middle cerebral artery and infarcts in the left basal ganglia, pons, and left frontotemporal lobe. Her serum was VZV IgM positive, cerebrospinal fluid (CSF) was VZV IgG antibody positive and IgM negative, and CSF PCR for VZV was negative. The patient was diagnosed with herpes zoster cerebral arteritis and treated with intravenous acyclovir but developed suspected pulmonary embolus, treated with heparin, and then staphylococcal septicemia. She died of acute subdural hematoma despite surgery.

One ganglion each was obtained from cases Z1 and Z2, whereas 5 separate ganglia were obtained from case Z3. The five ganglion samples from case Z3 were labeled Z3a, Z3b, Z3c, Z3d, and Z3e. Normal control TG and DRG were also obtained (C1 to C3) (Table 1). The underlying conditions in the healthy control cases were noninflammatory in DRG or TG. Consecutive 5-μm sections from each paraffin-embedded ganglion were used for subsequent analyses.

Initial histological examination of hematoxylin- and eosin-stained sections revealed in all cases intact sensory neurons surrounded by satellite cells. A significant number of infiltrating cells were observed in all the human ganglia following herpes zoster (Fig. 1A and B) but were absent in the normal control ganglia (Fig. 1C and D). The extent of the immune cell infiltrate varied among the post-herpes zoster ganglionic samples, with ganglia from herpes zoster case Z3 having the most prominent cellular infiltrate.

FIG. 1.

Hematoxylin- and eosin-stained ganglion sections from herpes zoster cases (A and B) and healthy cases (C and D). Higher-power images of the boxed insets from herpes zoster case Z1 (A) and normal-ganglion case C1 (C) are shown in panels B and D, respectively. In normal ganglia, the large neurons (arrows) are surrounded by satellite cells (arrowheads). Note the large inflammatory infiltrate following herpes zoster.

Presence and distribution of VZV antigen in human ganglia following herpes zoster.

VZV latency in human ganglia is characterized by the transcription and, in some cases, translation of several VZV open reading frames (ORFs), including ORFs 4, 21, 29, 62, 63, and 66 (11, 33, 48, 49). In contrast, VZV gene expression during productive replication in a fully permissive cell type follows an ordered cascade, with the expression of immediate-early, early, and late genes in a sequential manner (28, 35). VZV glycoproteins (late gene products) are not expressed during latent VZV infection (6), and thus, their detection in neurons provides a useful means to exclude true latency. We therefore examined ganglionic sections from cases following herpes zoster for VZV gE by IHC staining. Bound gE-specific antibodies were visualized with an avidin-HRP/DAB peroxidase substrate. In addition to using isotype control antibodies on all sections from herpes zoster cases, sections from healthy control cases were also stained for VZV glycoproteins. Recently, it has been demonstrated that VZV antigen within neurons can be difficult to distinguish from lipofuscin and neuromelanin, which accumulate in neurons with age, by IHC staining (78). In order to differentiate VZV antigen staining from these neuronal pigments, azure B was substituted for hematoxylin as a counterstain. This resulted in nuclei and lipofuscin being stained blue and neuromelanin green (78) and confirmed that the neurons were VZV gE antigen positive (data not shown).

Viral glycoproteins were detected in the majority of sections following VZV reactivation, with neurons exhibiting punctate cytoplasmic VZV gE staining scattered throughout the ganglia (Fig. 2A). In Z1, approximately 2% of the neurons were gE positive. No VZV antigen-positive neurons were detectable in case Z2. In ganglia Z3a and -b, approximately 25% of the neurons stained positive for gE, while 8%, 3%, and 5% of the neurons were stained positive for VZV gE in Z3c, Z3d, and Z3e, respectively (Fig. 2D). No glycoprotein-specific staining was observed in the three normal control ganglia or in the ganglia from herpes zoster cases stained with isotype control antibodies (Fig. 2B and C). These results demonstrate that neurons are the prominent cell type expressing VZV glycoproteins following natural VZV reactivation in vivo. Furthermore, in Z3, VZV gE was detected in sensory neurons even as late as 4.5 months after the onset of the rash.

FIG. 2.

Immunohistochemical detection of VZV gE in human ganglia following herpes zoster. (A and B) Representative images of ganglionic sections from herpes zoster case Z1 (A) and uninfected control case C1 (B) stained with a VZV gE-specific antibody. (C) Representative image of a ganglionic section from herpes zoster case Z1 incubated with isotype control antibody. (D) Graph depicting the percentages of VZV gE-positive neurons from ganglia from individual control and herpes zoster cases.

Characterization and distribution of immune infiltrating cells in human ganglia following herpes zoster.

To characterize the phenotype of the marked inflammatory cell response observed in human ganglia following herpes zoster, serial 5-μm sections were immunohistochemically stained with a panel of antibodies specific for T cells (CD3), NK cells (CD57), macrophages (CD68), B cells (CD20), and dendritic cells (CD1a). Sections from three normal ganglia were stained in parallel. In addition, isotype control antibodies for each cell-type-specific antibody were included on adjacent sections from all cases.

In all the herpes zoster cases, infiltrating immune cells expressing CD3, CD20, CD68, or CD57 were detected (Fig. 3A). No CD1a staining was identified in any of the herpes zoster cases. In the control cases, occasional CD3-, CD57-, and CD68-positive cells were detected, whereas no CD20 or CD1a staining was observed (Fig. 3B). VZV DNA, but not herpes simplex virus (HSV) DNA, was detected by PCR from DNA extracted from each of the herpes zoster samples utilized in this study, demonstrating that this infiltrate was not a consequence of HSV infection (data not shown).

FIG. 3.

Immunohistochemical staining for a panel of immune cell markers in ganglion sections following herpes zoster. (A) Immunohistochemical staining was performed with antibodies specific for T cells (CD3), B cells (CD20), macrophages (CD68), NK cells (CD57), and dendritic cells (CD1a) and visualized with DAB substrate. Representative images from herpes zoster ganglia (case Z1) are shown. No specific staining was seen using isotype control antibodies, as shown. Human skin (positive control) was stained with an anti-CD1a specific antibody. (B) Quantification of each of the immune cell markers was performed for each ganglion from the herpes zoster cases and expressed as a percentage of the total number of nonneuronal cells. The immune cell counts from the three normal control ganglion cases were averaged, with standard errors of the mean shown by error bars.

The immune cell counts from the three different normal control ganglia were averaged, and there was limited or no staining for the various immune cell markers. In control ganglia, CD3⁺ T cells were detected sporadically. In contrast, in Z1, Z3a, and Z3b, the immune infiltrate was characterized by a predominance of CD3⁺ T cells, followed by CD20⁺ B cells and CD68⁺ macrophages. In case Z1, 31% of infiltrating cells expressed CD3 compared to 20% expressing CD20, indicating an abundance of T cells. In three ganglia (Z2, Z3d, and Z3e), CD68⁺ macrophages were the predominant immune infiltrating cell type (Fig. 3B).

To further define the nature of the CD3⁺ T cells in the immune infiltrate observed in ganglia following herpes zoster, we determined the proportion of CD4⁺ and CD8⁺ T cells by immunostaining. The majority of the infiltrating T cells (>75%) observed in ganglionic sections following herpes zoster were CD8⁺ T cells (Fig. 4A and B). Furthermore, both the CD4⁺ and CD8⁺ T cells did not appear to be activated, as they did not stain for T cell activation markers CD25 (Fig. 4A and B) and CD69 (data not shown).

FIG. 4.

Characterization of T cell subsets and proximity to VZV antigen-positive neurons in human ganglia following herpes zoster. (A) Immunohistochemical staining was performed with antibodies for CD8, CD4, and CD25 T-cell subsets and visualized with DAB substrate. Representative images of staining for CD8, CD4, and CD25 T cells following herpes zoster (case Z1) are shown. No specific staining was seen using isotype control antibodies, as shown. (B) Quantification of each of these T-cell subsets and CD25 was performed for each ganglion and is expressed as a percentage of the total number of nonneuronal cells. (C) Dual-immunofluorescence staining for VZV gE and CD8 in ganglia following herpes zoster (case Z1) and normal control ganglia. gE was detected by AF488-tagged secondary antibody (green), CD8 by AF594-tagged secondary antibody (red), and cell nuclei with DAPI (blue). An image with all three targets overlaid is shown.

To more closely examine the distribution of CD8⁺ T cells with respect to VZV antigen localization, we performed dual immunofluorescent staining so that the presence of VZV gE and CD8⁺ T cells could be examined in the same ganglionic section. The staining observed in cases Z1 and Z3 revealed that CD8⁺ T cells extended throughout the ganglia and were not localized exclusively to VZV antigen-positive neurons (Fig. 4C). Interestingly, despite the significant influx of CD8⁺ T cells into ganglia, no dual VZV gE-CD8⁺ or gE-CD4⁺ T cell staining was observed. No gE or CD8 staining was observed in the normal control ganglia (Fig. 4C). Taken together, these data suggest that, unlike in natural cutaneous VZV infection, where productively infected T cells are observed (41, 42), in ganglia after herpes zoster, VZV-infected T cells are not detectable, although we cannot exclude the possibility that T cells in ganglia do become infected during the time of acute herpes zoster.

Cytolytic nature of infiltrating T cells in human ganglia following herpes zoster.

To examine the cytolytic nature of CD8⁺ T cells, we immunostained ganglionic sections for granzyme B, a protein known to be released by cytolytic CD8 cells (26, 50), and TIA-1, also a marker for cytolytic activity (2). As expected, neither TIA-1- nor granzyme B-stained T cells were observed in the normal control ganglia. Only a small number of the total infiltrating T cells stained for either of these two cytolytic-T cell markers in the zoster ganglion cases (Fig. 5A and B). Our capacity to detect cytolytic CD8⁺ T cells was confirmed by the staining of positive-control tonsil sections, which revealed extensive granzyme B- and TIA-1-positive T cells (Fig. 5A). Collectively, these data demonstrate that the vast majority of T cells present within ganglia following herpes zoster do not display the hallmarks of cytolytic CD8⁺ T cells.

FIG. 5.

Immunostaining for CD8⁺ T cells, cytolytic-T cell markers, and MHC class I in human ganglia following herpes zoster. (A) Immunohistochemical staining was performed with antibodies for CD8 T cells and the T cell cytolytic markers TIA-1 and granzyme B. Staining was visualized with DAB substrate. Tonsil sections included as positive controls for antibodies against CD8, TIA-1, and granzyme B are also shown. (B) Graph showing the percentage of nonneuronal ganglionic cells expressing CD8, TIA-1, and granzyme B in control ganglia (the three normal control ganglion cases were averaged) and ganglia following herpes zoster (zoster ganglia). (C) Dual-immunofluorescence staining for VZV gE and MHC class I β₂m in ganglia following herpes zoster (case Z1). gE was detected by AF594-tagged secondary antibody (red), β₂m by AF488-tagged secondary antibody (green), and cell nuclei with DAPI (blue). An image overlaying all three targets is shown (Merge). VZV antigen-positive neurons are indicated by white arrows.

Impact of neuronal VZV infection on major histocompatibility complex (MHC) class I expression following herpes zoster.

MHC class I molecules have been reported to be upregulated in mouse sensory neurons after HSV-1 infection (55). It is currently unknown if viral-antigen presentation to immune cells occurs within human ganglia following natural VZV infection. It is also not known if VZV-infected human neurons have the capacity to present viral antigens in the context of MHC class I molecules to T cells. Thus, dual-immunofluorescence staining for VZV protein and MHC class I was carried out to determine if VZV antigen-positive cells also express MHC class I molecules and have the potential to present viral proteins to CD8⁺ T cells. β₂m, an essential component of MHC class I molecules expressed on the cell surface, was identified using an anti-β₂m specific rabbit polyclonal antibody, which was detected with a secondary antibody, AF 488 (green), while VZV gE was detected with an anti-gE mouse monoclonal Ab and secondary antibody AF 594 (red) (Fig. 5C). β₂m staining was observed throughout the ganglia in both infiltrating cells and satellite cells. VZV gE was readily detected in human neurons, but neither viral-antigen-positive nor -negative neurons expressed detectable levels of β₂m (Fig. 5C). These data suggest that there is no detectable induction of MHC class I by either infected or uninfected neurons in human ganglia following herpes zoster.

DISCUSSION

Despite herpes zoster and PHN causing significant morbidity in the community, there is a striking paucity of studies that have reported any examination of human ganglia following herpes zoster. Thus, a review of the literature revealed only 7 publications that have reported any such examination over the past 100-plus years, and those studies were restricted to basic histological observations (14, 18, 27, 47, 52, 69, 70). Our study provides the first detailed characterization of the repertoire of immune cells that infiltrate human sensory ganglia following natural VZV reactivation (herpes zoster) and identifies noncytolytic CD8⁺ T cells as a prominent feature of this process. In addition, we provide evidence of persistent viral-antigen detection in neurons after resolution of the herpes zoster rash, despite the presence of a large immune infiltrate. The success of VZV as a human pathogen is largely dependent on its ability to establish a lifelong latent infection, reactivate sporadically, and transmit efficiently within the population. Although it is well documented that VZV establishes latency in sensory ganglia, much less is known about the critical immune events that occur in these ganglia following VZV reactivation, mainly due to the lack of any suitable animal model of herpes zoster (40, 60) and the inherent difficulty in obtaining sensory ganglia from individuals who had suffered from herpes zoster. Our ability to source human ganglia from such individuals provided a unique opportunity to perform this study.

VZV reactivates from human sensory ganglia during herpes zoster, and virus is transported to the skin, followed by viral replication in the skin and lesion formation (5). However, the precise roles of the various ganglionic cell types (i.e., neurons and satellite cells) in the reactivation process have yet to be fully established. Our examination of human ganglia demonstrated viral antigen almost exclusively within neurons and not in satellite cells or other ganglion cells, suggesting that reactivation of virus in neurons is the major stimulus of the immune response in sensory ganglia following reactivation. It remains possible, however, that at earlier time points infected satellite cells, which may have been undetected in our study due to rapid death, could also contribute to the immune response in this tissue. VZV virions have been found by electron microscopy in satellite cells, as well as neurons, in ganglia following herpes zoster (18). Furthermore, experimental VZV infection of either intact human ganglion explants or human DRG xenografts in SCID mice, neither of which can mount an immune response, resulted in productive infection of satellite cells (25, 58). It is therefore possible that the host immune response that occurs during natural herpes zoster may restrict replication to neuronal cells. Interestingly, despite the large T-cell infiltrate in ganglia following herpes zoster and previous work demonstrating that both CD8⁺ and CD4⁺ T cells are permissive to VZV and play a role in dissemination of virus to cutaneous sites during varicella (41, 42, 77), we did not observe VZV antigen-positive T cells in these ganglia. However, it remains possible that T cells in ganglia do become infected during the time of acute herpes zoster, as our study did not include examination of ganglia from the acute phase of infection due to lack of availability of such material. Thus, examination of human ganglia sourced from cases with active herpes zoster at the time of death will be a very important component of future studies to compare and contrast the roles of different cell types in VZV pathogenesis in the context of herpes zoster.

Despite the large influx of CD8⁺ T cells in all affected ganglia, very few infiltrating cells expressed detectable TIA-1 and granzyme B cytolytic-T-cell markers. This finding contrasts with studies of infected skin samples from both varicella and herpes zoster, which showed significant numbers of TIA- and granzyme B-expressing cytotoxic T cells (44, 51, 74). Determination of whether the cytolytic nature of the immune cells observed in this study is the same as that which occurs immediately following natural VZV reactivation (i.e., within the first few days versus the first few months) or whether the CD8⁺ T-cell response differs depending on tissue factors specific to the skin or sensory ganglia awaits the acquisition and analysis of human ganglia retrieved early in the course of herpes zoster (i.e., the first few days following virus reactivation).

Although we did not examine latent VZV infection, it is interesting that very low numbers of CD8⁺ T cells have been reported in latently HSV type 1 (HSV-1)-infected human ganglia, and subsets of these cells selectively cluster around HSV-1 LAT⁺ neurons, but not around ganglia containing latently VZV-infected neurons (66). The CD8⁺ T cells that clustered around HSV-1 LAT⁺ neurons expressed granzyme B, but neuronal damage was not observed (66). Noncytolytic granzyme B⁺ CD8⁺ T cells have also been reported to inhibit HSV-1 reactivation from latency in a mouse model of latency (37). In our study, we did not observe CD8⁺ T cells juxtaposed exclusively to VZV antigen-positive neurons, but rather, CD8⁺ T cells were distributed throughout the ganglia. This finding, together with the noncytolytic nature of the majority of infiltrating CD8⁺ T cells and the possibility that even the granzyme B⁺ CD8⁺ T cells may not be cytotoxic, suggests either that CD8⁺ T cells do not play a major role in controlling VZV following herpes zoster or that they do so in a noncytolytic manner.

Also addressing the issue of whether viral peptides from infected neurons are able to be presented to CD8⁺ T cells is previous work demonstrating that VZV downregulates cell surface MHC class I expression by causing its retention in the Golgi compartment in certain cell types, such as fibroblasts, (1, 9, 17). In addition, neurons generally do not express MHC class I protein (31, 64). Our finding that MHC class I molecules were not detectable by immunofluorescence staining on either uninfected or VZV antigen-positive neurons following herpes zoster is consistent with the notion that MHC class I-restricted presentation of viral peptides by neurons does not play a major role in controlling VZV infection in this setting. This contrasts with an earlier study of acute HSV-1 infection of mice that reported upregulation of MHC class I molecules in primary mouse sensory neurons, with the implication that HSV-1-infected neurons could potentially present viral antigens to CD8⁺ T cells (55). In addition, it has been reported that despite MHC class I not being detectable by immunohistochemistry in latently HSV-1-infected mouse ganglia, HSV-1-specific CD8⁺ T cells still interact and form immunological synapses with the neurons, suggesting expression of some MHC class I (34). Thus, at least in the case of HSV-1, there is some evidence that neurons may have the capacity to be recognized by CD8⁺ T cells, and while it is not known if priming of VZV-specific T cells takes place in ganglia, it remains possible that neurons may play a role in presenting viral antigens to invading CD4⁺ or CD8⁺ T cells to aid in virus clearance. Due to the lack of tools to detect VZV-specific T cells in paraffin sections, it is not possible to determine if the invasion of T cells is specific to VZV antigens or is a generalized infiltration in response to a cytokine milieu created during reactivation. Indeed, the mechanisms by which the immune cells are recruited into the ganglia and how they are maintained following VZV reactivation remain to be elucidated.

Most T cells infiltrating ganglia were CD8⁺, but CD4⁺ T cells were also detected. The role of CD4⁺ T cells in VZV replication in human ganglia remains to be determined, but CD4⁺ T cells are important in clearing ganglia infected by other neurotropic viruses (30, 61, 63). We also detected NK cells, B cells, and macrophages, but not dendritic cells. NK cells have previously been reported to be important in the host response to VZV infection (67) and have been observed to infiltrate into TG during HSV-1 latent infection in a mouse model (46). Our detection of NK cell infiltration into ganglia implicates these cells in the control of VZV in ganglia following herpes zoster and represents an association of this cell type with VZV infection. B cells were absent in normal control ganglia, and the role of B cells in human ganglia following herpes zoster remains to be established. B cells and antibody production have been shown to be crucial for preventing viral reactivation in central nervous system (CNS) glial cells in mice infected with neurotropic mouse hepatitis virus (57), raising the possibility that a similar mechanism may exist in VZV-infected ganglia.

From case Z3, we obtained 5 ganglia 4.5 months after the onset of the herpes zoster rash (which healed after approximately 6 weeks). In these ganglia, we observed large numbers of neurons harboring gE, with up to 27% of neurons being VZV gE antigen positive. Although the donor of the Z3 ganglia had been under maintenance immunosuppressive therapy for 25 years, these ganglia still showed a very strong immune response, although we cannot exclude the possibility that the immunosuppressive regimen may have had some impact on the repertoire of infiltrate or the noncytolytic nature of the T cells. Similarly, the degree of any immunosuppression in donors Z1 and Z2 is not known, so we cannot determine whether their immunity may differ from that of individuals who develop zoster by aging alone (with no underlying illness). What is clear, however, is that like those of Z3, both Z1 and Z2 ganglia displayed robust immune infiltrates. Ganglia with the greatest inflammatory infiltrate correlated with ganglia with the highest number of infected neurons. The presence of viral antigen not expressed during latency in neurons after resolution of the herpes zoster rash raises the intriguing possibility that persistent VZV replication or long-lived viral proteins remaining after cessation of replication may continue to stimulate inflammatory cells, which results in ongoing damage to ganglia and causes pain that persists after the rash has healed, perhaps persisting long enough to be classified as PHN (i.e., pain persisting beyond 3 months after onset). If they are ever able to be sourced, examination of ganglia removed from nonimmunosuppressed donors long after resolution of a herpes zoster rash (but who did not die from herpes zoster) will provide confirmation of the capacity of VZV antigen not associated with latency to persist in neurons for an extended period after reactivation.

In summary, this is the first characterization of the immune response in human ganglia following natural VZV reactivation (herpes zoster). We identified and defined the nature of inflammatory cells in these ganglia and also identified persistence of viral antigen in neurons, perhaps providing the basis for long-term pain and other neurologic symptoms associated with herpes zoster. These findings also serve as a platform for future studies of the virus-host relationships that underpin the immunological mechanisms that attempt to control VZV infection in vivo and will inform future immunotherapeutic and vaccine strategies, and they may also provide paradigms for the disease-causing processes of other neurotropic viruses.