T-Cell Tropism of Simian Varicella Virus during Primary Infection

Varicella-zoster virus (VZV) causes varicella, establishes a life-long latent infection of ganglia and reactivates to cause herpes zoster. The cell types that transport VZV from the respiratory tract to skin and ganglia during primary infection are unknown. Clinical, pathological, virological and immunological features of simian varicella virus (SVV) infection of non-human primates parallel those of primary VZV infection in humans. To identify the host cell types involved in virus dissemination and pathology, we infected African green monkeys intratracheally with recombinant SVV expressing enhanced green fluorescent protein (SVV-EGFP) and with wild-type SVV (SVV-wt) as a control. The SVV-infected cell types and virus kinetics were determined by flow cytometry and immunohistochemistry, and virus culture and SVV-specific real-time PCR, respectively. All monkeys developed fever and skin rash. Except for pneumonitis, pathology produced by SVV-EGFP was less compared to SVV-wt. In lungs, SVV infected alveolar myeloid cells and T-cells. During viremia the virus preferentially infected memory T-cells, initially central memory T-cells and subsequently effector memory T-cells. In early non-vesicular stages of varicella, SVV was seen mainly in perivascular skin infiltrates composed of macrophages, dendritic cells, dendrocytes and memory T-cells, implicating hematogenous spread. In ganglia, SVV was found primarily in neurons and occasionally in memory T-cells adjacent to neurons. In conclusion, the data suggest the role of memory T-cells in disseminating SVV to its target organs during primary infection of its natural and immunocompetent host.


Introduction
Varicella-zoster virus (VZV) is a ubiquitous human neurotropic alphaherpesvirus that causes varicella (chickenpox) as a primary infection and herpes zoster (shingles) upon reactivation of latent virus [1]. Primary VZV infection is acquired via the respiratory route and varicella occurs 2-3 weeks after exposure [2,3]. The pathogenesis of varicella is largely unknown, mostly due to the prolonged incubation period and restricted host range of the virus. VZV is detected in lymphocytes of varicella patients [4], suggesting that the virus spreads to susceptible organs including skin and ganglia via a cell-associated viremia [4]. However, the low number of VZV-infected lymphocytes has precluded their identification during natural infection in humans [5].
Most of the current understanding of VZV pathogenesis is based on experimental infection of human fetal tissue transplanted in severe combined immunodeficient mice (SCID-hu model) [6,7]. In this model, VZV has a tropism for T-cells within thymus and liver xenografts [8]. It has been postulated that VZV initially replicates in respiratory epithelial cells and is transferred to T-cells within tonsilar lymphoid tissue contacting the upper respiratory tract [9,10]. Virus transport to human fetal skin and ganglia explants in SCID-hu mice can be mediated by T-cells [11,12], most likely activated memory CD4 T-cells expressing the skin homing markers C-C type chemokine receptor type 4 (CCR4) and cutaneous lymphocyte antigen (CLA) [10]. However, the VZV SCID-hu mouse model does not reproduce the complex and dynamic virus-host interactions involved in the dissemination of VZV to its target organs during primary infection in its natural and immunocompetent host [6,7].
Simian varicella virus (SVV) produces a naturally occurring disease in non-human primates with clinical, pathological and immunological features that parallel human VZV infection [13,14]. The prevalence of SVV in free-ranging non-human primates is largely unknown. However, SVV outbreaks in primate centers have been associated with the introduction of monkeys captured from the wild into the colony [15]. The genomes of SVV and VZV are similar in size, structure and genetic organization, with an estimated 70-75% DNA homology [16]. SVV causes varicella, becomes latent in ganglionic neurons and reactivates after stress and immunosuppression to cause herpes zoster [17,18]. A cell-associated viremia is detected from 3 days post-infection (dpi), with the highest number of infected lymphocytes just before the onset of skin rash [14,19]. SVV reaches the ganglia before skin rash [20,21], indicating viremic spread to ganglia.
The aim of the present study was to characterize the kinetics of virus infection and the cell types involved in the dissemination of SVV during primary infection. We have previously shown that infection of macaques with recombinant measles virus expressing EGFP (rMV-EGFP) facilitated the identification of the cell types involved in MV pathogenesis with unprecedented sensitivity [22,23,24,25]. To detect SVV-infected cells at the low frequencies expected in blood and lungs, we infected African green monkeys (AGMs) with recombinant SVV expressing EGFP (SVV-EGFP) and, as a control, wild-type SVV (SVV-wt) to study SVV pathogenesis at the whole organism, tissue and cellular levels in its natural and immunocompetent host. The data presented suggest a crucial role for memory T-cells in the dissemination of SVV during primary infection.

SVV-infected African green monkeys develop transient fever and skin rash
Five SVV-seronegative adult AGMs were infected intratracheally with SVV-wt (n = 2) or SVV-EGFP (n = 3). A transient increase in body temperature was seen between 6 and 11 dpi ( Fig. 1A and B). SVV-wt2infected animals developed skin rash starting at 6 dpi, which increased in severity until 9 to 10 dpi and resolved thereafter (Fig. 1C). Macroscopic EGFP fluorescent lesions were detected on the skin and lips of all SVV-EGFP2infected animals starting at 7 dpi, which increased in severity until 9 dpi and resolved by 13 dpi (Fig. 1D and E). EGFP fluorescent lesions were also detected on the tongue of SVV-EGFP2infected monkeys, coinciding with appearance of skin rash (Fig. 1F). No lesions were observed on the lips and tongues of SVV-wt2infected animals, demonstrating the increased sensitivity of using SVV-EGFP to study varicella pathogenesis. Skin rash was more severe in SVV-wt2 compared to SVV-EGFP2infected monkeys. Collectively, the findings indicate the close resemblance of the clinical signs associated with experimental SVV-EGFP infection of AGMs and those of primary VZV infection in humans.

SVV infection of alveolar myeloid cells and T-cells in the lung
All SVV-infected animals became dyspneic at the time of skin rash. Macroscopic examination of lungs showed multifocal pulmonary consolidation and hemorrhage affecting at least one lobe in all animals euthanized 9 or 13 dpi ( Fig. 2A). Diffuse EGFP fluorescence was detected in an SVV-EGFP2infected monkey at 9 dpi ( Fig. 2B and C). Combined immunohistochemical (IHC) and immunofluorescence (IF) analyses for SVV antigens and EGFP on consecutive sections of lung showed that EGFP expression was restricted to SVV antigen-positive cells (Fig. 2D-G), demonstrating that EGFP is a valid marker to identify SVV-infected cells in the monkeys. To investigate SVV-infected cell types in situ, lung tissue sections were analyzed by dual-IF staining with SVV-specific antiserum and anti-keratin, -CD3, -CD68 and -CD11c mouse monoclonal antibodies (mAbs). SVV-infected cells were readily detected in lungs at 9 dpi, but not at later times (data not shown). At 9 dpi, abundant SVV pos keratin pos lung epithelial cells were observed (Fig. 2H), as well as SVV pos CD3 pos T-cells (Fig. 2I). In addition, SVV antigens were found in intra-alveolar cells that coexpressed CD68 and/or CD11c, consistent with alveolar macrophages (AM), some of which appeared to have phagocytosed SVVinfected cells (Fig. 2J). Occasionally, SVV pos CD11c pos dendritic cell (DC)-like cells displaying multiple branched projections were observed adjacent to bronchi (Fig. 2K).
To define the kinetics of virus replication and the cell types infected in the respiratory tract during primary SVV infection, bronchoalveolar lavage (BAL) cells were obtained at 5 dpi, 9 dpi and at necropsy (9, 13 or 20 dpi). SVV DNA load and infectious virus titers in BAL cells peaked at 5 dpi and declined rapidly thereafter ( Fig. 3A and B). Infectious virus was not recovered from BAL cells at 13 and 20 dpi ( Fig. 3B and data not shown). The viral DNA load and infectious SVV titer in BAL samples were similar in SVV-wt2 and SVV-EGFP2infected monkeys at 5 dpi, indicating a similar level of replication of both viruses in lung.
At 5 dpi, 7.2% of BAL cells from SVV-EGFP2infected monkeys were EGFP pos (Fig. 3C). In EGFP pos BAL cells, equal numbers of CD45 pos (leukocytes) and CD45 neg cells, most likely bronchial and alveolar epithelial cells, were detected ( Fig. 3D and Fig. S1). CD45 pos BAL cells could be categorized as T-cells, B-cells and alveolar myeloid cells, i.e. large granular cells expressing high levels of both CD14 and HLA-DR (Fig. S1). These myeloid cells could be alveolar macrophages (AM) and/or DC. At 5 dpi, 82% of CD45 pos BAL cells were alveolar myeloid cells, 17% were Tcells and only 1% were B-cells (data not shown). Most EGFP pos CD45 pos BAL cells were alveolar myeloid cells and Tcells (Fig. 3E). CD4 pos , CD8 dim and CD8 bright T-cells were infected at equal frequencies (Fig. 3F). The number of BAL-derived T-cells was too low to determine their differentiation status unequivocally (data not shown). At 9 dpi, frequencies of EGFP pos CD45 pos BAL cells were too low to conclusively identify the SVV-infected leukocyte subsets (Fig. 3C).

SVV infection of memory T-cells in blood during viremia
To determine the kinetics of virus infection and identify the blood lymphocyte subsets infected during the viremic phase of varicella, peripheral blood mononuclear cells (PBMC) isolated at multiple dpi from SVV-infected monkeys were analyzed. SVV DNA was detected in PBMC from 2 dpi until necropsy (Fig. 4A). Viral DNA load in PBMC peaked at 7 dpi and was higher in

Author Summary
Varicella-zoster virus (VZV) causes varicella, establishes lifelong latent infection in ganglia and reactivates later in life to cause zoster. VZV is acquired via the respiratory route, with skin rash occurring up to 3 weeks after exposure. The cell types that transport VZV to skin and ganglia during primary infection are unknown. Simian varicella virus (SVV) infection of non-human primates mimics clinical, pathological and immunological features of human VZV infection. African green monkeys were infected with recombinant SVV expressing enhanced green fluorescent protein (SVV-EGFP) or wild-type SVV (SVV-wt) as a control. By visualizing SVV-EGFP2infected cells in the living animal and in tissue samples, we identified the virus-infected cell types in blood, lungs, skin and ganglia during primary infection. Our data demonstrate that during viremia, SVV predominantly infects peripheral blood memory T-cells. Detection of SVV-infected memory T-cells in lungs, in early varicella skin lesions and also, albeit to a lesser extent, in ganglia suggests a role for memory T-cells in transporting virus to these organs. Our study provides novel insights into the cell types involved in virus dissemination and the overall pathology of varicella in a non-human primate model.
SVV-wt2 compared to SVV-EGFP2infected monkeys (Fig. 4A). Infectious virus was isolated from PBMC of both SVV-EGFP2 and SVV-wt2infected monkeys until 9 and 11 dpi (Fig. 4B). EGFP pos lymphocytes were detected from 5 to 11 dpi, peaking at 7 dpi (Fig. 4C). Together, the data indicate that the kinetics of viral DNA load and infectious virus titer represent the temporal change in the number of circulating SVV-infected lymphocytes, but not in the replication of SVV in blood lymphocytes during viremia. The rapid loss of SVV-infected lymphocytes from the circulation could be caused by virus-induced apoptosis [26] or, alternatively, infected lymphocytes may be cleared from the circulation by the SVV-specific adaptive immune response [14,21,27].
At 5 dpi, EGFP pos cells were detected at similar frequencies in all major PBMC subsets (i.e., T-cells, B-cells, natural killer cells, monocytes and dendritic cells) ( Fig. 4D and Fig. S2). However, given that most PBMC are T-cells (Fig. S2), T-cells were identified as the main SVV-infected lymphocyte subset in blood (Fig. 4D). Moreover, at the peak of viremia (7 dpi) T-cells were the only SVV-infected cells demonstrated in blood. Unlike humans and macaques, AGMs have three distinct T-cell subsets: CD8 bright , CD8 dim and CD4 pos T-cells (Fig. S2) [28]. While CD8 bright T-cells correspond to classical human CD8+ T-cells, CD4 pos T-cells and CD8 dim T-cells are considered dynamic populations of AGM T-helper cells functionally equivalent to human CD4+ T-cells [28]. Similar levels of CD8 bright , CD8 dim and CD4 pos T-cells were SVV-infected, most of which were memory T-cells ( Fig. 4E and F). Importantly, at 5 and 7 dpi, predominantly central memory (CM) T-cells and effector memory (EM) T-cells, respectively, were infected (Fig. 4F). The apparent dual phase of SVV-infected CM and EM T-cells may reflect the organ in which the T-cells have been infected. CM Tcells are preferentially found in lymphoid tissues, whereas EM Tcells are migratory memory T-cells that home to peripheral tissues to orchestrate local immune responses and may ultimately function as tissue-resident T-cells to sense the cognate antigen locally for extended periods of time [29,30].  In vitro infection studies on human tonsil-derived lymphocytes showed that VZV preferentially infects T-cells expressing the activation marker CD69 and skin-homing markers CCR4 and CLA [10]. To address this issue in SVV-EGFP2infected monkeys, peripheral blood-derived EGFP pos T-cells obtained at 5 and 7 dpi were analyzed for expression of both CCR4 and the T-cell activation marker CD137, the latter marker is selectively expressed by T-cells early after recognition of their cognate antigen [31,32]. No preference of SVV for memory T-cells expressing CCR4 or CD137 was seen in vivo ( Fig. S3), suggesting that SVV did not infect virus-specific T-cells that recognized SVV-infected antigen presenting cells like macrophages or DCs.
To determine whether the predominant infection of memory Tcells in vivo reflects viral tropism for a specific lymphocyte subset, PBMC from SVV-naive AGMs were infected in vitro with SVV-EGFP. Expression of EGFP was restricted to lymphocytes that expressed SVV antigens (Fig. S4A), supporting the use of EGFP as a surrogate marker for SVV-infected cells in flow cytometry. While all major PBMC subsets appeared to be equally susceptible to SVV infection, T-cells were the prominent SVV-infected PBMC subset in vitro (Fig. S4B), with similar infection levels in CD4 pos , CD8 dim and CD8 bright T-cells (Fig. S4C). In particular, significantly more memory T-cells were infected compared to naive Tcells (p,0.05; Mann-Whitney test) (Fig. S4D). Thus, SVV preferentially infects memory T-cells rather than naive T-cells both in vivo (Fig. 4) and in vitro (Fig. S4).

Detection of SVV in lymphoid organs
Alveolar macrophages and lung-resident DC transport antigens to lung-draining lymph nodes for presentation to T-cells [33,34], and VZV-infected human DCs can transfer infectious virus to Tcells in vitro [35]. We hypothesized that SVV-infected alveolar myeloid cells transport SVV to draining lymph nodes for subsequent virus transfer to memory T-cells. High SVV DNA loads were detected in lymph nodes, tonsils and spleens of SVVinfected monkeys at 9 dpi, declining rapidly thereafter (Fig. 5A). Cells in lymph nodes and tonsils of SVV-infected monkeys contained intranuclear inclusions bodies and SVV antigen ( Fig. 5B and C). Tracheobronchial lymph nodes showed more pronounced SVV-induced histopathology compared to peripheral lymph nodes (data not shown). However, SVV DNA loads were comparable in different lymph nodes collected at 9 dpi (Fig. 5A), emphasizing the need to investigate lymph nodes at earlier times after infection. In addition, detection of SVV-infected memory T-cells in blood may represent lung-resident T-cells involved in SVV dissemination. SVV infects alveolar epithelial cells leading to alveolar wall damage (data not shown) [19,27,36], which may result in egress of SVV-infected T-cells into the circulation.

SVV-infected perivascular lymphocytes in early varicella lesions implicate hematogenous spread of SVV to the skin
Detailed in situ analysis was performed to identify the SVVinfected cell types in varicella skin lesions. Macroscopic detection of EGFP fluorescence corresponded to SVV infection of the skin, as demonstrated by the co-localization of SVV protein and EGFP in consecutive skin sections obtained from an SVV-EGFP-infected monkey ( Fig. 6A and B). In vesicular skin lesions, SVV predominantly infected keratinocytes ( Fig. 6C and D). In deeper skin layers, SVV protein was frequently detected in hair follicles ( Fig. 6E and F) and sebaceous glands ( Fig. 6G and H).
Analysis of skin biopsies from SVV-EGFP2infected monkeys allowed investigation of the early stages of varicella, as evidenced on the skin by the appearance of EGFP fluorescent areas in the absence of lesions visible to the naked eye. In these biopsies, SVV protein expression was consistently located within perivascular lymphocytes ( Fig. 6I-K). Dual-IF staining for EGFP and specific lymphocyte markers identified SVVinfected perivascular cell subsets as CD68 pos macrophages (Fig. 6L), CD11c pos DCs (Fig. 6M) and CD3 pos T-cells (Fig. 6N). The remaining SVV-infected cells, which stained negative for lymphocyte markers, phenotypically resembled dendrocytes (data not shown) [37]. Interestingly, SVV-infected T-cells were also observed in the epidermis of SVV-wt infected monkeys at 9 dpi (Fig. 6O). Flow cytometric analysis of skinresident T-cells showed exclusively memory T-cells, mostly EM T-cells (data not shown).
Collectively, these data suggest that SVV reaches the skin hematogenously. Since the skin vasculature is composed of an upper horizontal superficial vascular plexus just beneath the epidermal surface and a deep vascular plexus that supplies the hair bulbs and sweat glands [38], it seems likely that SVVinfected memory T-cells transfer the virus to skin-resident perivascular macrophages, DCs or dendrocytes, which in turn transfer SVV to adjacent epidermal or hair follicle keratinocytes via cell-to-cell spread. Alternatively, epidermal SVV-infected Tcells may transfer the virus directly to skin epithelial cells (Fig. 6O). Neurons are the main SVV-infected cell types in ganglia The hallmark of primary SVV and VZV infection is the capacity of virus to infect and establish latency in ganglionic neurons along the entire neuraxis [1,13,[39][40][41][42]. Virus may reach ganglia hematogenously or by retrograde axonal transport along axons innervating varicella lesions [12,20,21,43,44]. We determined the kinetics of virus infection and the cell types infected in ganglia during primary SVV infection. The SVV DNA load in ganglia was significantly higher in SVV-wt2 compared to SVV-EGFP2infected monkeys (p,0.01; Mann-Whitney test) (Fig. 7A), peaking at 9 dpi and decreasing thereafter (Fig. 7A), as might be expected during the establishment of latency. Despite high SVV DNA loads, no virus-mediated cytopathology was seen in ganglia (data not shown). Virus antigen was more abundant at 9 dpi than at 13 and 20 dpi (data not shown). SVV-infected cells in ganglia were detected in situ by IHC using SVV-specific antiserum ( Fig. 7B-D). Dual-IF staining for SVV and the neuron-specific marker NCAM (neural cell adhesion molecule) showed that most SVV pos cells were neurons (Fig. 7E). Occasionally, SVV antigens were seen at the neuronal cell surface or potentially within satellite glial cells (SGC) (Fig. 7D). SGC form a sheet that completely enwraps neuronal cell bodies, providing physical and metabolic support to the neurons and contributing to regulation of the immune response in the peripheral nervous system [45,46]. Virusinfected cells located in vicinity to neurons did not express the SGC-specific marker glial fibrillary acidic protein (GFAP) [45], implicating that SGC were not infected with SVV at 9, 13 and 20 dpi (Fig. 7F and data not shown). To address the possibility of T-cell-mediated transfer of SVV to neurons, ganglia were examined using dual-IF staining for SVV antigens and CD3. In an SVV-wt2infected monkey euthanized at 9 dpi, SVV-infected T-cells were detected in close proximity to neurons (Fig. 7G). Notably, this animal also had the highest SVV DNA load in blood and ganglia. Flow cytometric analysis of ganglion-derived singlecell suspensions demonstrated that ganglion-resident T-cells were memory T-cells, predominantly EM T-cells (Fig. 7H).
Our findings in ganglia contrast with the pronounced VZVinduced histopathology of both SGCs and neurons found in VZVinfected human fetal ganglia xenografts in the SCID-hu mouse model [12,47]. Most likely, these differences are due to the use of fetal human ganglia and the lack of adaptive immune responses in the SCID-hu mouse model. The absence of SVV-induced histopathology in ganglia is consistent with previous studies [19,27] and the inability to recover infectious virus from ganglia [15] at 10 dpi. Nonetheless, virus-induced cytopathology of ganglia may have occurred during the peak of viremia (5-7 dpi), which will be considered in future studies. The detection of SVV protein in the cytoplasm of neurons, but not in the interacting SGC ( Fig. 6B-F), supports the notion of retrograde axonal route of virus entry into ganglia [43,44,48,49]. In contrast with this hypothesis, the SVV DNA load did not differ among ganglia, including those that innervated the dermatomes showing varicella rash ( Fig. 7A and data not shown). The alternative scenario is that virus traffics to ganglia during viremia within lymphocytes. Indeed, both SVV and VZV enter ganglia before the onset of rash, arguing for hematogenous virus spread [1,20,21]. VZVinfected T-cells infiltrate human ganglion xenografts and transmit VZV to neurons in the VZV SCID-hu mouse model [12]. The occasional detection of neuron-interacting, SVV-infected memory T-cells within ganglia (Fig. 7G) supports the role of memory Tcells in virus dissemination to ganglia. Further studies on ganglia from SVV-EGFP2infected monkeys euthanized at earlier times after primary infection are warranted to test this hypothesis.
The current study is the first to present experimental evidence (summarized in Fig. 8) that supports the role of memory T-cells in the inter-organ dissemination of varicella virus in its natural and immunocompetent host. Our current hypothesis on the pathogenesis of primary SVV infection is presented in Figure 9. We hypothesize that upon intratracheal inoculation, SVV replicates in the respiratory tract and infects epithelial cells, alveolar myeloid cells (AM and/or DC) and T-cells in the lungs. Subsequently, the virus enters the circulation as cell-associated virus predominantly within memory T-cells, first within CM and subsequently within EM T-cells. Most likely, virus-infected alveolar myeloid cells   transport SVV to lung-draining lymph nodes, with subsequent transfer of SVV to memory T-cells. Peak viremia coincided with onset of fever and appearance of skin rash. SVV reached the skin by the hematogenous route, most likely via virus-infected memory T-cells. SVV may enter ganglia by retrograde axonal transport from the infected epithelia and/or by the hematogenous route. In addition to memory T-cells, other lymphocyte subsets may also contribute to the viremic spread of SVV. Virus-infected DC, NK cells, B-cells and monocytes were detected in peripheral blood at 5 dpi, albeit at low frequencies compared to memory T-cells. The contribution of each lymphocyte population in transfer of SVV to its target organs will be addressed in future studies by analyzing virus-infected lymphocytes in tissues of animals euthanized during peak viremia at 5-7 dpi.
Like VZV, SVV is considered to spread to naive monkeys via aerosols and therefore most likely targets mucosal epithelial cells of the upper respiratory tract, although -depending on the size of the aerosols -some virus may also directly reach the lower in a cervical ganglion at 9 dpi. Squares indicate corresponding tissue areas shown at higher magnification in (C) and (D). (E) Dualimmunofluorescence (IF) staining of a thoracic ganglion at 9 dpi for SVV proteins (green) and neural cell adhesion molecule (NCAM; red). Arrows indicate SVV-positive neurons. (F) Dual-IF staining of a thoracic ganglion at 9 dpi for SVV protein (green) and glial fibrillary acidic protein (GFAP; red). Arrow indicates neuron-adjacent SVV-positive cell. (G) Dual-IF staining of a thoracic ganglion from a monkey at 9 dpi for SVV protein (green) and CD3  respiratory tract [1,13,14,25]. In the current study, we have used intratracheal inoculation of monkeys with SVV, bypassing the putative initial site of local SVV replication in the upper respiratory tract or tonsils [1,13,14]. Primary VZV infections in adults are more severe than in children and frequently complicated by varicella pneumonia [1]. Consequently, the adult status of SVV-infected AGM may have enhanced disease severity, although pneumonia is a common feature in SVV-infected monkeys due to the intratracheal route of inoculation [19,27]. Recombinant SVV-EGFP was attenuated in vivo compared to SVV-wt, possibly due to insertion of the EGFP gene between open reading frames (ORFs) 66 and ORF67 [50]. Recombinant VZV lacking ORF67 is severely impaired for growth in cell culture [51]. Although attenuated in severity, SVV-EGFP-induced disease resembled that of a SVV-wt infection and attenuation did not alter the cell tropism of SVV-EGFP. Both SVV-wt and SVV-EGFP infected the same cell types in lung, lymph nodes, ganglia and skin in vivo, and identical PBMC types in vitro. The recent cloning of the SVVwt full-length genome in a bacterial artificial chromosome facilitates the generation of a potentially less attenuated recombinant EGFP-expressing SVV by inserting the EGFP gene adjacent to SVV genes dispensable for growth in vitro [51,52].
Future studies on juvenile African green monkeys, infected with less-attenuated SVV-EGFP strains and via alternative inoculation routes (e.g., via the nose or throat), are warranted. Particularly, analysis of tissues obtained from infected animals euthanized shortly after primary infection are needed to unequivocally determine the early target cell types of SVV, their role in virus dissemination to the target organs affected during primary infection and the route of SVV entry into sensory ganglia [25]. Our current SVV-EGFP/AGM model, which largely covers the clinical and pathological features seen in both SVV-wt2infected monkeys and human varicella patients, provides novel opportunities to elucidate the virus-host cell interactions involved in varicella pathogenesis. This will open new avenues to develop and test new VZV vaccination and therapeutic interventions that limit viremic spread, while inducing long-lasting adaptive VZV-specific immunity.

Ethics statement
This study was performed in strict accordance with European guidelines (EU Directive on Animal Testing 86/609/EEC) and Dutch legislation (Experiments on Animals Act, 1997). The protocol was approved by the independent animal experimentation ethical review committee DCC in Driebergen, the Netherlands (Erasmus MC permit number EMC2374). Animals were housed in groups, received standard primate feed and fresh fruit daily, and had access to water ad libitum. Cages also contained sources of ''environmental enrichment'' such as hiding places and hanging ropes, tires and other toys. During infection, study animals were housed in HEPA-filtered, negatively pressurized BSL-3 isolator cages. Animal welfare was monitored daily and all animal handling was performed under light anesthesia (ketamine) or deep anesthesia (ketamine and medetomidine) to minimize animal discomfort. After deep anesthesia, atipamezole was administered to antagonize the effect of medetomidine. Animals were euthanized by sedation with ketamine and medetomidine followed by exsanguination.

Viruses
Low-passage clinical isolates of the Delta herpesvirus strain of SVV-wt and SVV-EGFP were obtained from PBMC of acutely infected AGM and propagated less than 5 times in AGM-kidney epithelial cell line BSC-1 (American Tissue Type Culture no. CCL-26) to generate virus stocks as described [53]. Virus stocks were confirmed as Mycoplasma-free. SVV-EGFP was generated by insertion of the EGFP gene downstream from a Rous sarcoma virus promoter between SVV ORF66 and ORF67 [20,50].

SVV infection of PBMC in vitro
PBMC from SVV-naive AGM were infected by co-cultivating PBMC (5610 5 ) with SVV-EGFP-infected Vero cells (0.5-1610 5 ), showing 70% virus-induced cytopathic effect (CPE), in 0.5 ml DMEM supplemented with antibiotics and 10% heat-inactivated fetal bovine serum (FBS) for 24 hr in 24-well plates at 37uC in a CO 2 -incubator. Mock-infected PBMC were similarly generated by co-cultivating PBMC with uninfected Vero cells. SVV-EGFP2infected PBMC were stained and analyzed by flow cytometry or spotted on microscope slides, fixed and stained by immunofluorescence for SVV as described below.

Experimental SVV infection of AGM, necropsy and collection of tissues
Five adult (10-to 12-year-old) SVV-seronegative AGMs (Cercopithecus aethiops) with intraperitoneal implanted temperature transponders were inoculated intratracheally with ,10 6 plaqueforming units (pfu) of SVV-EGFP (n = 3 animals; 1 male and 2 females) or SVV-wt (n = 2 animals; 1 male and 1 female) diluted in 5 ml of phosphate-buffered saline [21]. Just before infection, animals were sedated with ketamine and medetomidine. The abdomen and back of the animals were shaved to allow careful examination for skin rash every other day until necropsy. Heparinized blood samples were collected under light ketamine sedation at 0, 2, 7, 11, 13, 17 and 20 dpi. Bronchoalveolar lavage (BAL) samples and peripheral blood (PB) samples were collected under deep anesthesia at 5 and 9 dpi. Three punch biopsies (3 mm) of varicella rashes and EGFP fluorescent skin tissue, while showing no characteristic varicella-like skin rash by the naked eye, were obtained from anesthetized animals at 9 dpi under anesthesia. SVV-EGFP2infected animals were checked for macroscopic EGFP fluorescence using a custom-made lamp containing 6 LEDs (peak emission 490-495 nm) mounted with D480/40 bandpass filters [22]. Fluorescence was detected by an amber cover of a UV transilluminator used for screening DNA gels [22]. Photographs were taken using a Nikon D80 SLR camera. SVV-infected animals were euthanized at 9 dpi (n = 2; one SVV-wt-and one SVV-EGFP2infected animal), 13 dpi (n = 2; one SVV-wt2 and one SVV-EGFP2infected animal) and 20 dpi (one SVV-EGFP- Figure 9. Model of the pathogenesis of primary SVV infection. Upon intratracheal inoculation of African green monkeys, SVV replicates in the lower respiratory tract and infects lung epithelial cells, alveolar macrophages (AM), dendritic cells (DC) and T-cells. SVV-infected AM and DC may transport the virus to draining lymph nodes and subsequently transfer SVV to local lymphocytes resulting in a cell-associated viremia. Memory T-cells are the predominant SVV-infected lymphocyte subset during viremia and may play a central role in dissemination of SVV to its target organs. SVV reaches the skin by the hematogenous route, presumable via virus-infected memory T-cells, which results in the infection of perivascular macrophages, DC and dendrocytes. Subsequently, SVV may infect epidermal and hair follicle keratinocytes via cell-to-cell spread and cause vesicular skin lesions. SVV may enter ganglia by (a) retrograde axonal transport and/or (b) by viremic spread via virus-infected lymphocytes. doi:10.1371/journal.ppat.1003368.g009 infected animal). Multiple tissues including lung, lymph nodes, spleen, tonsils, skin and ganglia were collected at necropsy and either snap-frozen or fixed and paraffin-embedded.
Collection and processing of PB and BAL samples PBMC were isolated by density-gradient centrifugation and used for virus isolation, DNA isolation and flow cytometry or cryopreserved as viable cells as described [19]. Cells recovered from BAL samples were centrifuged, dissolved in RPMI-1640 medium supplemented with 10% FBS plus antibiotics (R10F medium), and used for virus isolation, DNA isolation and flow cytometry as described [22].

Virus isolation from PB and BAL samples
Infectious SVV was isolated from PB and BAL cells by incubating 1-2610 6 cells in 10-fold serial dilutions in R10F medium on confluent monolayers of BSC-1 cells in 6-well plates. Cells were monitored for SVV-induced CPE or EGFP expression after 7 days of co-cultivation and results were expressed as numbers of SVV-infected cells per 10 6 input PBMC and BAL cells.

Nucleic acid extraction and quantitative PCR (qPCR)
DNA was isolated from PBMC, BAL cells, pooled ganglia, pooled lymph nodes, tonsils and spleen using a QIAamp DNA Mini Kit (Qiagen). qPCR was performed in triplicate on a ABI Prism 7500 using Taqman 26 PCR Universal Master Mix (Applied Biosystems) with primers and probes specific for SVV open reading frame 21 (ORF21) and the pan-primate single-copy gene oncostatin-M (OSM) as described [14,21,54]. DNA dilutions obtained from uninfected PBMC were used to validate the OSM Taqman assay.
For immunofluorescence staining, sections were incubated with secondary Alexa Fluor 488 (AF488)-or AF594-conjugated goatanti-mouse and/or goat-anti-rabbit antibodies and mounted in Prolong Gold Antifade reagent with 49,6-diamidino-2-phenylindole (Invitrogen) [56]. Sections were analyzed on a Zeiss LSM 700 confocal laser scanning microscope fitted on an Axio Observer Z1 inverted microscope (Zeiss). Images were obtained using 2-46 frame averaging and the pinhole adjusted to 1 airy unit. ZEN 2010 software (Zeiss) was used to adjust brightness and contrast.  Figure S4 Memory T-cells were preferentially infected in vitro. (A) SVV-naive African green monkey peripheral blood mononuclear cells (PBMC) were infected with SVV-EGFP in vitro and stained 24 hr later for SVV proteins to show that EGFP fluorescence (green) co-localized with SVV proteins (red). Nuclei were counterstained with DAPI (blue). Magnification: 4006. (B) African green monkey PBMC were infected with SVV-EGFP in vitro and analyzed 24 hr later by flow cytometry for EGFP expression in the indicated lymphocyte subsets. Data are plotted as the frequency of EGFP pos cells within individual PBMC subsets (within subset) or as the percentage of EGFP pos cells within each lymphocyte subset relative to the total number of PBMC (absolute). (C, D) Percentage of EGFP pos cells in the indicated T-cell subsets as assessed by flow cytometry. The lymphocyte subsets were defined as described in Figure S2. Data represent means 6 SEM of three independent experiments performed on PBMC from three animals. * p,0.05 by Mann-Whitney test. (TIF)