Differential Chemokine Responses in the Murine Brain Following Lyssavirus Infection

J. Comp. Path. 2013, Vol. 149, 446e462

Available online at www.sciencedirect.com

www.elsevier.com/locate/jcpa

EXPERIMENTALLY INDUCED DISEASE

Differential Chemokine Responses in the Murine Brain Following Lyssavirus Infection D. J. Hicks*, A. N
un ~ ez*, A. C. Banyard†, A. Williams‡, A. Ortiz-Pelaezx, A. R. Fooks†,{ and N. Johnson† * Pathology Unit, Department of Specialist Scientific Support, † Wildlife Zoonoses and Vector-Borne Diseases Research Group, Department of Virology, Animal Health and Veterinary Laboratories Agency, Woodham Lane, New Haw, Addlestone, Surrey KT15 3NB, ‡ Department of Veterinary Medicine, Cambridge University, Wolfson College, Barton Road, Cambridge CB3 9BB, x Epidemiology, Surveillance and Risk Group, Animal Health and Veterinary Laboratories Agency, Woodham Lane, New Haw, Addlestone, Surrey KT15 3NB and { The National Consortium for Zoonosis Research, University of Liverpool, Leahurst, Chester High Road, Neston CH64 7TE, United Kingdom

Summary The hallmark of lyssavirus infection is lethal encephalomyelitis. Previous studies have reported distinct lyssavirus isolate-related differences in severity of cellular recruitment into the encephalon in a murine model of infection following peripheral inoculation with rabies virus (RABV) and European bat lyssavirus (EBLV)-1 and 2. In order to understand the role of chemokines in this process, comparative studies of the chemokine pattern, distribution and production in response to infection with these lyssaviruses were undertaken. Expression of CCL2, CCL5 and CXCL10 was observed throughout the murine brain with a distinct caudal bias in distribution, similar to both inflammatory changes and virus antigen distribution. CCL2 immunolabelling was localized to neuronal and astroglial populations. CCL5 immunolabelling was only detected in the astroglia, while CXCL10 labelling, although present in the astroglia, was more prominent in neurons. Isolatedependent differences in the amount of chemokine immunolabelling in specific brain regions and chemokine production by neurons in vitro were observed, with a greater expression of CCL5 in vivo and CXCL10 production in vitro after EBLV infection. Additionally, strong positive associations between chemokine immunolabelling and perivascular cuffing and, to a lesser extent, virus antigen score were also observed. These differences in chemokine expression may explain the variation in severity of encephalitic changes observed in animals infected with different lyssavirus isolates. Crown Copyright Ó 2013 Published by Elsevier Ltd. All rights reserved. Keywords: chemokine; encephalitis; European bat lyssavirus; rabies

Introduction The lyssavirus genus is comprised of neurotropic viruses including rabies virus (RABV) and 13 other viruses (Kuzmin et al., 2010; Freuling et al., 2011, 2012). Lyssavirus infection of the central nervous system (CNS) results in non-suppurative meningoencephalomyelitis, with variable degrees of neuronal necrosis, neuronophagia, focal gliosis and lymphocytic perivascular infiltration in the presence or Correspondence to: A. N
un ~ez (e-mail: [email protected]. uk). 0021-9975/$ - see front matter http://dx.doi.org/10.1016/j.jcpa.2013.04.001

absence of Negri bodies (Charlton, 1984; Charlton et al., 1987; Fekadu, 1988; Hooper et al., 1999). The severity of inflammatory changes after lyssavirus infection can vary depending on the incubation period (Murphy, 1977; Hemachudha et al., 2006), host (Fekadu et al., 1982, Fekadu, 1988) and virus strain (Sugamata et al., 1992; Yan et al., 2001; Roy et al., 2007; Hicks et al., 2009). These inflammatory changes are a non-specific response and are observed commonly in other virus infections of the CNS including West Nile virus (Kelley et al., 2003), Japanese encephalitis virus (Johnson et al., 1985), tick-borne encephalitis virus (Gelpi et al., 2005), Crown Copyright Ó 2013 Published by Elsevier Ltd. All rights reserved.

Chemokine Responses in Murine Lyssavirus Infection

herpes simplex virus (Yen et al., 2009) and measles virus (Garg, 2002). In a previous study, a murine model has been used to demonstrate lyssavirus isolate-related differences in the level of cellular recruitment, largely comprised of T cells, to the perivascular spaces surrounding blood vessels in the brain and cellular infiltration of the brain parenchyma itself (Hicks et al., 2009). The importance of T cells to the host immune response to RABV infection had been established previously using a number of T-cell deficient (Weiland et al., 1992; Galelli et al., 2000; Lafon, 2005) and B-cell deficient (Hooper et al., 2009) mouse models to investigate abortive lyssavirus infection. Despite the clear role for T cells in the clearance of abortive lyssavirus infection (Weiland et al., 1992; Galelli et al., 2000; Lafon, 2005) and in conferring resistance to lyssavirus infection to SJL/J and BALB/cByJ mice (Perry and Lodmell, 1991), the contribution of infiltrating T cells to the pathogenesis observed in acute lyssavirus infection is uncertain. People who are naturally infected with pathogenic RABV exhibit limited CNS inflammation (Murphy, 1977; Hemachudha et al., 2006), while the severity of encephalitis after infection with RABV in mice appears dependent on the virus isolate that is inoculated (Sugamata et al., 1992; Yan et al., 2001; Roy et al., 2007). The pathology of European bat lyssavirus (EBLV) infection has not been studied as intensively as that of RABV infection and preliminary data suggest that EBLV infection is associated with more severe encephalitic manifestations in experimental models (Hicks et al., 2009). Chemokines are low molecular weight proinflammatory soluble mediators vital to homeostasis, response to injury, synaptic transmission and development and the cellular recruitment process observed in disease-associated neuroinflammation (De Haas et al., 2007). Functional criteria and N-terminal structural motifs are used to categorie chemokines into four different groups: CXC, CC, C and CX3C. The CXC subfamily, including CXCL10, is defined by an amino acid separating two conserved cysteine residues and is primarily involved in the recruitment of neutrophils. CXCL10 is a chemoattractant associated with monocyte/macrophage, T cell, natural killer (NK) cell and dendritic cell recruitment. It also aids T-cell adhesion to endothelial cells (Dufour et al., 2002) and has been associated with T-cell recruitment in response to CNS infection by measles virus (Patterson et al., 2003) and West Nile virus (Klein et al., 2005). Chemokines, including CCL2 and CCL5, belonging to the CC subfamily are defined by the presence of conserved adjacent cysteine residues, and are principally involved in the recruitment of

447

monocytes, T cells and macrophages (Banisadr et al., 2005). CCL2 is chemotactic for monocytes, memory T cells and dendritic cells and is produced in response to injury or infection (Carr et al., 1994). CCL2 can be produced by resident CNS cell populations including neurons, astrocytes and microglia (Banisadr et al., 2005; Hickman and El Khoury, 2010) and, consequently, CCL2 is thought to mediate neuroinflammatory processes including T-cell extravasation into the brain (Carrillo-de Sauvage et al., 2012). CCL5, a chemoattractant for T cells and monocytes (Amaral et al., 2011), has an established role in neuroinflammatory responses after infection with viruses including Dengue fever virus (Amaral et al., 2011), herpes virus (Savarin et al., 2010) and West Nile virus (Klein et al., 2005). Experimental infection of mice with an abortive RABV demonstrated that cellular recruitment and virus clearance was preceded by the up-regulation of a variety of inflammatory mediators including chemokines such as CCL2, CCL5 and CXCL10 (Phares et al., 2006). Similarly, chemokine mRNA transcripts were up-regulated in mice after infection with acute RABV isolates, including the highly pathogenic silver-haired bat virus (Roy et al., 2007; Johnson et al., 2008), and EBLV-2 (Mansfield et al., 2008). The combination of mRNA transcript up-regulation and the presence of a predominately T cell inflammatory infiltrate observed in mice after infection with EBLV isolates (Mansfield et al., 2008; Hicks et al., 2009) indicates that CCL2, CCL5 and CXCL10 appear to be important candidates for involvement in cellular recruitment during encephalitis. Furthermore, because of the pivotal role of chemokines in cellular recruitment and the up-regulation of chemokine mRNA transcripts after lyssavirus infection, the variable encephalitis occurring after lyssavirus infection may be attributable to virus-dependent differences in chemokine expression. Comparison of the host innate immune response, including chemokine up-regulation, after the infection of mice with either the abortive RABV strain CVS-F3 or the acute silver-haired bat virus, have highlighted that significant difference in chemokine expression levels do not exist between these two species of lyssaviruses, despite the extensive cellular recruitment observed after infection with CVS-F3 (Roy et al., 2007). However, differences in chemokine expression levels may account for the variation in the severity of encephalitis observed between virus isolates belonging to different lyssavirus species. The aims of this study were to establish whether the greater cellular recruitment into the brain observed in mice following infection with EBLVs when compared with RABV (Hicks et al., 2009) resulted from

D.J. Hicks et al.

448

increased chemokine production, and to corroborate these observations with in vitro determination of chemokine production by infected neurons.

Materials and Methods Mouse and Virus Strain

Genetically outbred, 5-week-old CD-1 mice (Charles River Laboratories, L’Arbresis, France) were inoculated (n ¼ 25 mice per virus isolate) by the footpad route with 30 ml E-MEM cell culture media (Invitrogen, Paisley, UK) containing equivalent infectious virus particles (approximately 8  102) of RABV (isolate RV61), EBLV-1 (isolate RV1423) or EBLV2 (isolate RV1332), as described previously (Hicks et al., 2009). The acronyms RABV, EBLV-1 and EBLV-2 were used to simplify the nomenclature used throughout the manuscript and the work undertaken in the course of the study refers to the isolates (RV61, RV1423 and RV1332, respectively) used and not the virus species as a whole. Control animals (n ¼ 5) were inoculated with 30 ml of E-MEM. Mice exhibiting clinical signs (RABV n ¼ 13, EBLV-1 n ¼ 8 and EBLV-2 n ¼ 7) were killed and their brains were sampled for histopathology. Only mice that exhibited bipedal paralysis, a clinical phase less than 24 h and appropriate morphological preservation of the target brain regions were used for the subsequent analysis to allow comparability. Experimental procedures were conducted under UK Home Office License regulations and with the appropriate ethical approval (AHVLA Ethics Committee). Animals were provided with food and water ad libitum. Histopathology and Immunohistochemistry

Haematoxylin and eosin (HE) staining and viral antigen immunohistochemistry (IHC) (HAM 5DF123B0) was undertaken on tissues sampled and prepared for histopathological analysis as described by Hicks et al. (2009). For chemokine immunolabelling, slides were dewaxed, dehydrated and endogenous peroxidase activity was quenched prior to the antigen retrieval. Slides then received either microwavemediated antigen retrieval (AR) for 10 min at 100 C using high pH AR buffer (DAKO, Ely, Cambridgeshire, UK) to enable the detection of CCL2 and CCL5 antigen or 0.05% trypsin/0.05% a-chymotrypsin (SigmaeAldrich, Poole, Dorset, UK) enzyme pretreatment for 10 min at 37 C for the detection of CXCL10 antigen. After cooling at room temperature (RT) and washing in tap water, slides were assembled into Shandon Sequenza coverplates for immunolabelling. Endogenous biotin activity was blocked (DAKO biotin blocking system) before the sequential applica-

tion of normal rabbit immune serum (Vector Laboratories, Peterborough, UK); 1 in 66 dilution in Tris Buffered saline with 0.05% Tween 20 (TBST, VWR, Leicestershire, UK) for 20 min at RT and anti-CCL2, CCL5 and CXCL10 primary antibodies or protein concentration matched normal goat immunoglobulin isotype controls (R&D Systems, Abingdon, UK) for 18e20 h (overnight) at 4 C. A secondary rabbit antigoat immunoglobulin antibody (Vector Laboratories) was applied and the IHC signal amplified and ‘visualized’. Finally, sections were counterstained and mounted in preparation for light microscopy (Hicks et al., 2009). Assessment of Pathological and Immunohistochemical Findings

Pathological changes and perivascular cuff counts were determined using a Nikon Eclipse E400 microscope (Hicks et al., 2009). Specific virus antigen immunolabelling was assessed using a non-linear qualitative numerical scoring system from 1 (minimal immunolabelling) to 8 (widespread immunolabelling), which took into account the relative amounts of immunolabelling, intensity of the signal and distribution in comparison with known positive and negative controls. To quantify chemokine expression by IHC, the percentage area of chemokine immunolabelling in seven representative brain regions, of the 14 originally evaluated, from mice meeting the selection criteria (RABV: n ¼ 13, EBLV-1: n ¼ 8 or EBLV-2: n ¼ 7) was determined using an image analysis system (NIS Elements BR 3.0, Nikon UK Limited, Kingston-upon-Thames, UK). Images of single fields of view were knitted together using NIS Elements imaging software and the composite image was subdivided into regions of interest (ROIs), which were analyzed for specific immunolabelling to produce percentage area of positive immunolabelling. Statistical Analysis

The non-parametric Wilcoxon exact sum rank test was used to analyze virus-dependent differences in chemokine production in vitro and total perivascular cuff counts per animal in vivo between viral species (formally genotypes) and brain regions. The KruskaleWallis analysis of variance by rank test was used to establish whether the medians of the chemokine immunolabelling distributions were equal, while pair-wise analysis using one-way ANOVA, in conjunction with the post-hoc Tamhane test, was used to establish which brain regions exhibited significant differences in chemokine immunolabelling in comparison with the others. A square root transformation was performed on the data to normalize it for use in linear regression models. The

Chemokine Responses in Murine Lyssavirus Infection

Wald test, for simple and composite linear hypotheses, was undertaken to test pair-wise whether there were differences in the percentage area of chemokine immunolabelling within the different brain regions. Finally, three linear regression models, one for each chemokine assessed, were fitted to assess whether linear associations existed between the square root of the percentage area of specific chemokine immunolabelling and the amount of perivascular cuffing (PC), viral antigen IHC score and virus isolate. The PC and viral antigen IHC data were transformed into three category variables. The PC data were subdivided into no pathology (baseline, PC ¼ 0), low pathology (PC >1 and <5) and high pathology (PC >5) and the viral antigen IHC values were subdivided into no viral antigen (baseline, antigen score ¼ 0), low viral antigen (antigen score >1 and <3) and high viral antigen (antigen score >3) for analysis (StataCorp. 2007, Stata Statistical Software: Release 10. College Station, Texas). Primary Neuron Culture

Suspensions of CD-1 mouse cortex neurons (Lonza, Cambridge, UK) were resuscitated from liquid nitrogen and cultured in 24-well plates with primary neuron growth medium (PNGM, Lonza), supplemented with a PNGM SingleQuot kit (Lonza) containing NSF-1 (2%), GA-1000 (0.01%) and L-glutamine (1%), for 7 days at 37 C in a 5% CO2 incubator. PNGM was replaced with fresh, prewarmed (37 C) PNGM (50% change) after 7 days and again 10 days following cell resuscitation. After 10 days, the primary neuron cultures were challenged with 1 MOI (multiplicity of infection, 1 virus particle per cell) of virus (RV61, RV1423 or RV1332) for 24 h before the supernatant was removed (leaving approximately 250 ml to prevent the cells from drying out) and the cells were washed twice with prewarmed (37 C) PNBM culture medium. This was also discarded and replaced with fresh, prewa rmed (37 C) PNBM and the plates were then cultured for 5 days in a 37 C/5% CO2 incubator, without further media change. All culture steps were undertaken in accordance with the manufacturer’s instructions (Lonza). After 5 days, the neuron plates were transferred to a freezer (80 C for 24 h) and underwent a single freeze/thaw cycle. Once the plates had defrosted, the well contents were scraped and the cellular debris and supernatant was recovered for Luminex analysis. Luminex 100Ô xMAPÒ Technology Fluorescence System Analysis

Quantities of CCL2, CCL5 and CXCL10 protein present in cell culture supernatants and lysates derived

449

from primary neuron cultures infected with RABV, EBLV-1 or EBLV-2 were quantified using a threeplex chemokine kit (MilliporeÒ MAP mouse chemokine panel, Millipore, Billerica, Massachussetts, USA) on the Luminex 100Ô xMAPÒ technology fluorescence system (Luminex Corporation, Austin, Texas, USA). The immunoassay procedure for the preparation of the three-plex plate for LuminexÒ analysis was undertaken in accordance with manufacturer’s recommendations. Briefly, the microtitre filter plate was prewet with 200 ml of wash buffer (kit component) and was sealed and shaken for 10 min at RT. After vacuum and blot removal of excess wash buffer, 25 ml of each standard or control (kit supplied) was added to the appropriate wells. Assay buffer (25 ml; kit component) was added to all wells before the addition of 25 ml of cell lysate/culture supernatant to the sample wells. Antibody immobilized beads were premixed with assay buffer, vortex mixed and added (25 ml/well) to the filter plate, before the plate was sealed, covered with a lid secured with a rubber band and shaken on a plate shaker for 2 h at RT. After bead incubation, fluid was removed by vacuum and the plates were washed twice with wash buffer. Detection antibodies were then added (25 ml/well) and the plate sealed before being returned to the plate shaker for a further hour at RT. Following this incubation period, 25 ml of streptavidin-phycoerythrin was added to each of the wells and the plate was incubated on the plate shaker for a further 30 min at RT. All well contents were then gently removed from the plate under vacuum and the plate was washed twice with wash buffer, as before. Excess buffer was blotted from the bottom of the plate with absorbent paper before 150 ml of sheath fluid (kit component) was added to all wells. The beads were re-suspended on the plate shaker (5 min) before the plate was run through the Luminex 100 Ô fluorescence system for analysis.

Results Histopathology

Clinical features indicative of lyssavirus infection (i.e. piloerection, unsteady gait, progressive paralysis, body spasms and a hunched appearance) observed in infected CD-1 mice were similar to those observed previously in OF-1 mice (Hicks et al., 2009). Limb paralysis, body spasms and a hunched appearance were the most prominent features, irrespective of the lyssavirus isolate inoculated. Histopathological examination of brain sections revealed a mild to moderate nonsuppurative meningoencephalitis, characterized by the infiltration of perivascular regions with inflammatory cells accompanied by localized gliosis, formation

450

D.J. Hicks et al.

of neuronophagic nodules and meningitis. This was observed in mice exhibiting clinical disease after peripheral inoculation with all three lyssavirus species. These pathological changes were more prominent in the caudal brain regions, including the pons and medulla, when compared with the frontal brain regions including the cortex, hypothalamus, hippocampus and thalamus. The severity of encephalitis was more pronounced after infection with EBLV-2, which led to the greatest amount of PC, while other features of encephalitis, including gliosis, formation of neuronophagic nodules and pyknotic debris, were also more abundant after EBLV infection (Supplementary Fig. S1). The caudal bias in distribution and the differences in mean PC between different lyssavirus inoculation groups were shown to be statistically significant (P <0.05) by the Wilcoxon exact sum rank test (Supplementary Table S1) and similar to that reported previously (Hicks et al., 2009). Immunohistochemical Distribution of Viral Antigen

Lyssavirus antigen was widely distributed throughout the brains of CD-1 mice exhibiting clinical disease and demonstrated a bias toward the caudal regions of the brain, irrespective of the virus isolate inoculated (Supplementary Fig. S2 and Table S2). Viral antigen immunolabelling was prominent in the cortex, colliculi, pons, cerebellum and medulla and comparatively scarce in the olfactory lobe, cortex level 4 and the hippocampus after RABV infection. A similar pattern of distribution, abundance of viral antigen and mean immunolabelling score was observed after EBLV-1 infection. The mean viral antigen score was considerably lower after EBLV-2 infection when compared with both RABV and EBLV-1 infection, with EBLV-2 infection producing the lowest viral antigen score in 11 of the 14 brain regions assessed. CCL2 Immunolabelling

CCL2 immunolabelling was only present in challenged animals and appeared in either a fine granular form primarily observed in non-neuronal cell populations (Fig. 1ai) or in an intense, homogenous intracytoplasmic form in neurons (Fig. 1aii). CCL2 immunolabelling was commonly associated with inflammatory changes including perivascular cuffing in the white matter, where endothelial and infiltrating cells appeared labelled, and areas of gliosis in the cerebellum. The molecular, cellular and Purkinje cell layers of the cerebellum, the spinal and medial vestibular nuclei and gigantocellular nuclei regions in the rostral medulla (Fig. 1) and ventrolateral regions, including the spinal five nuclei and the

gigantocellular reticular nuclei regions, of the caudal medulla were also prominently immunolabelled for CCL2 (Fig. 1). The distribution of CCL2 immunolabelling through the brain regions assessed was consistent, irrespective of the viral isolate inoculated, although virus-specific variation in the amounts of immunolabelling was detected. CCL2 immunolabelling was restricted to sporadic cellular labelling in the hippocampus, but was more abundant in the posteriolateral thalamic nuclei of the thalamus and dorsoventral hypothalamic nuclei regions of the hypothalamus, in mice exhibiting clinical disease after peripheral inoculation with RABV, EBLV-1 or EBLV-2. CCL2 immunolabelling was most frequently observed in the caudal brain regions (colliculi, pons, cerebellum and medulla) of mice exhibiting RABV infection, with labelling being particularly prevalent in the pons and medulla (Supplementary Table S3 and Fig. S3). Cellular labelling, mostly neuronal in nature, was detected scattered throughout the colliculi, while a mixture of both non-neuronal and neuronlocalized CCL2 immunolabelling was observed in the ventrolateral regions, including the pontine and pontine reticular nuclei, and, to a lesser extent, the central regions, including the deep mesencephalic nuclei (DpMe) and the red magnocellular nuclei, of the pons. A similar distribution of CCL2 immunolabelling was observed in the caudal brain regions of EBLV-1- and EBLV-2-infected mice, although CCL2 immunolabelling within the DpMe featured more prominently after inoculation with EBLV-2 (Supplementary Fig. S3 and Table S3). CCL5 Immunolabelling

CCL5 immunolabelling was present throughout the brains of mice exhibiting clinical disease following peripheral inoculation with RABV, EBLV-1 or EBLV2. CCL5 expression, visualized as granular labelling of cellular processes and associated intense neuropil labelling by IHC, only appeared to be present in non-neuronal cell types (e.g. astrocytes and microglia) and could be found, like CCL2 immunolabelling, in association with PC and areas of diffuse or focal gliosis (Figs. 2aec). CCL5 immunolabelling appeared most abundant after footpad inoculation with EBLV-1 (Figs. 2bi, 2bii), but was also present, albeit at lower levels, after inoculation with RABV (Figs. 2ai, 2aii) and EBLV-2 (Figs. 2ci, 2cii). CCL5 cellular labelling was observed in the olfactory region, but featured more prominently in other frontal brain regions including the motor, sensory and lateral cortical regions after RABV infection (Supplementary Fig. S3 and Table S3). CCL5

Chemokine Responses in Murine Lyssavirus Infection

451

Fig. 1. Immunohistochemical demonstration of CCL2 in the rostral medulla after infection with different lyssavirus isolates. Endothelial cell (closed arrows), perivascular cuffing (angle-headed arrows)-associated (ai) and neuron-localized (closed arrowheads) CCL2 immunolabelling (aii) after RABV infection. CCL2 immunolabelling in the spinal and medial vestibular nuclei (bi), the gigantocellular reticular nuclei and the facial nuclei/subglia limitans ventral regions (bii) of the rostral medulla after EBLV-1 infection. Cellular labelling; neurons (closed arrowheads), non-neuronal (angled arrowheads) and endothelial cells (closed arrows) in the spinal and medial vestibular nuclei (ci) and gigantocellular regions (cii) of the rostral medulla after EBLV-2 infection. Mock challenged animals did not show significant CCL2 immunolabelling (di and dii). Images were taken at 100 magnification cropped from composite images produced using Elements software.

452

D.J. Hicks et al.

Fig. 2. Distribution of CCL5 immunolabelling in the rostral medulla after peripheral inoculation with lyssaviruses. Sporadic cellular (arrowhead) and perivascular cuff-associated (arrows) CCL5 immunolabelling in the spinal and vestibular nuclei regions (ai) and gigantocellular regions (aii) of the rostral medulla after peripheral inoculation with RABV. Inoculation with EBLV-1 (b) and EBLV-2 (c) produced a similar distribution of CCL5 immunolabelling in the rostral medulla, with CCL5 labelling particularly frequent in both the spinal and vestibular nuclei (bii and ci) and gigantocellular regions (bi and cii). CCL5 immunolabelling was absent from these regions in the mock-infected controls (d). Images were taken at 100 magnification and cropped from composite images produced using Elements software.

Chemokine Responses in Murine Lyssavirus Infection

expression was extremely rare in the hippocampus, more abundant in the posteriolateral thalamic nuclei and the lateral hypothalamic nuclei, but most abundant in the caudal brain regions, particularly the pons and medulla, after RABV infection (Supplementary Table S3 and Fig. S3). In RABVinfected animals, where only weak CCL5 immunolabelling was observed in the pons (Fig. S3), distribution was limited to those areas adjacent to the periaqueductal grey matter (Fig. 2b), otherwise labelling was found extending ventrally and laterally through the deep mesencephalic nuclei and the pontine nuclei. Scattered cellular labelling was observed in the superior and inferior colliculi regions and was present throughout the central and lateral regions of the caudal medulla and the medial and lateral vestibular nuclei regions in the more rostral medulla in RABV-infected mice (Fig. 2). A similar distribution of CCL5 immunolabelling was observed throughout the forebrain, hippocampus, pons and medulla after EBLV-1 and EBLV-2 infection in comparison with RABV-infected mice. However, there was a discernible increase in the abundance of CCL5 immunolabelling in all these brain regions after peripheral inoculation with EBLV-1. CCL5 immunolabelling was also more abundant in the medulla of mice exhibiting clinical disease after inoculation with EBLV-2 when compared with RABV-infected mice. Higher labelling of CCL5 was also present in the cerebellum of EBLV-1- and EBLV-2-infected mice when compared with RABV-infected mice. The localization of labelling to areas of focal gliosis, extending from the Bergman layer and through the molecular layer, white matter-localized PC and the interposed and lateral cerebellum nuclei, was the same irrespective of the virus isolate inoculated. CXCL10 Labelling

Both neuronal and non-neuronal cell populations exhibited CXCL10 immunolabelling, although most labelling was associated with neurons in mice exhibiting clinical disease after peripheral inoculation with RABV, EBLV-1 and EBLV-2 (Figs. 3aec). Three general distribution patterns of CXCL10 immunolabelling were observed in the brains of RABV, EBLV-1- and EBLV-2-infected mice: (1) a caudal distribution pattern similar to that observed for CCL2 and CCL5 (e.g. vestibular and gigantocellular nuclei in the rostral medulla; Figs. 3aec), (2) widespread throughout the brain region assessed (e.g. caudal medulla), and (3) restricted to the peripheral regions (medulla: pyramidal, olivary and ventral gigantocellular regions; pons, lateral and ventral regions; Fig. 3bi, 3bii). The distribution of CXCL10

453

immunolabelling, which featured more prominently in the caudal brain regions, was similar irrespective of the virus isolate inoculated (supplementary Table S3 and Fig. S3), although some isolatedependant regional variation in abundance was evident. All brain regions assessed exhibited CXCL10 immunolabelling after inoculation with RABV. CXCL10 immunolabelling in the forebrain ranged from sporadic cellular labelling to extensive labelling of the motor and frontal association cortex, with more widespread immunolabelling in the dorsal regions of the cortex. Sporadic CXCL10 immunolabelling was also observed in the hippocampus and ventral regions of both the thalamus and hypothalamus (Supplementary Table S3 and Fig. S3). In the caudal brain regions, the diffuse pattern of cellular labelling was accompanied by more focused areas of CXCL10 immunolabelling after inoculation with RABV. This was particularly evident in the vestibular nuclei and central areas of the medulla and also the central and ventral regions of the deep mesencephalic and pontine nuclei of the pons. CXCL10 immunolabelling was also observed occasionally in both Purkinje cells and neurons localized within the lateral nuclei of the cerebellum of RABV and EBLV-1 infected mice. Apart from the olfactory region, where immunolabelling was not observed, CXCL10 immunolabelling was similarly localized and distributed in the brains of mice exhibiting clinical disease after peripheral inoculation with EBLV-1 (Supplementary Table S3 and Fig. S3). CXCL10 immunolabelling was absent from the caudal regions of the cortex and hippocampus and less abundant overall in EBLV-2infected mice when compared with RABV and EBLV-1-infected mice (Supplementary Fig. S3). When present, CXCL10 immunolabelling was restricted to scattered cellular labelling in similar regions to those described in RABV- and EBLV-1-infected mice (Supplementary Table S3 and Fig. S3). Statistical Analysis Intragroup Analysis. To evaluate the significance of differences in chemokine immunolabelling between brain regions in each of the challenged groups (intra-group analysis for RABV-, EBLV-1- or EBLV-2 infected mice), the amount of chemokine (CCL2, CCL5 and CXCL10) immunolabelling (mean percentage area) per region was analyzed by one-way ANOVA using the post-hoc Tamhane method and summarized in Table 1. Seven of the 14 brain regions, representative of the overall distribution and abundance of chemokine immunolabelling, were selected for the intragroup comparison.

454

D.J. Hicks et al.

Fig. 3. Distribution of CXCL10 immunolabelling in the caudal medulla after footpad inoculation with different lyssavirus isolates. Neuronal (arrowheads), astromicroglial (open arrowheads) and cuffing (arrow)-associated CXCL10 immunolabelling in the hypoglossal region (ai) and lateral reticular nuclei region (aii) after inoculation with RABV. Similar immunolabelling was observed in the medial/spinal vestibular and solitary tract nuclei region (bi) and the ventral spinal trigeminal tract and lateral reticular nuclei region (bii) after inoculation with EBLV-1 and also in the ventral medulla reticular nuclei (ci) and the reticular and dorsal medulla reticular nuclei regions (cii) of the caudal medulla after peripheral inoculation with EBLV-2. Similar immunolabelling was absent from the intersection between the solitary tract and medulla reticular nuclei (di) and in areas adjacent to artefactual damage (dii) in the mock-infected negative controls. Images were taken at 100 magnification and cropped from composite images produced using Elements software.

The regional differences in CCL2 immunolabelling observed after RABV infection were shown to be statistically significant (P <0.05), with the rostral medulla demonstrating a greater mean percentage area of CCL2 immunolabelling than the other six brain re-

gions assessed (Table 1). Other caudal brain regions, including the caudal medulla (CM), pons and hypothalamus, were shown to exhibit significantly (P <0.05) more CCL2 immunolabelling than more rostral brain regions including the cortex,

455

Chemokine Responses in Murine Lyssavirus Infection Table 1 Summary of one-way ANOVAs with post-hoc Tamhane test statistical analysis (P values) of brain region-related differences in chemokine expression after lyssavirusinoculation †

BR

CCL2 1 2 3 4 5 6

RABV

EBLV-1

EBLV-2

1 2 3 4 5 6 7 1 2 3 4 5 6 7 1 2 3 4 5 6 7

e e * * * *

* * * * *

e e * * * *

* * * * *

e e e * * e

e * * * *

CCL5

CXCL10

1 2 3 4 5 6 1 2 3 4 5 6

* e * *

+ + e e e * * e e e + *

e * * * *

e * * *

+ e e * e * * e * * e e

e * * * *

e * * e

e e e e e e * e e e e e

e * * * *

* * * *

e e * e e e e * * e e *

e * * * *

* * e * e * * * * *

e e * *

e e e e e * * * * * e *

e * * * *

* * e * * * * * * *

e * * *

e e e e e e e e * * e e

e * * * *

* * e * e e * e e e

* Denotes that the percentage area of chemokine immunolabelling was significantly higher in the column region when compared with the row brain region (P < 0.05). e Signifies that any differences observed were not significant. + Denotes that the percentage area of chemokine immunolabelling was significantly higher in the row brain region when compared with the column region. † Brain region: (1) cortex 3, (2) hippocampus, (3) thalamus, (4) hypothalamus, (5) pons, (6) rostral medulla and (7) caudal medulla.

hippocampus and thalamus (Table 1). Similarly, after EBLV-1 infection, CCL2 immunolabelling was significantly greater (P <0.05) in the medulla (caudal and rostral) when compared with all other regions and in the pons when compared with the higher brain regions (Table 1). The rostral medulla (RM) exhibited significantly (P <0.05) more CCL2 immunolabelling than the pons, hypothalamus and forebrain regions and the pons exhibited more than the forebrain regions after EBLV-2 infection (Table 1). The mean percentage area of CCL5 immunolabelling was significantly (P <0.05) higher in the CM when compared with all other brain regions, bar the pons and RM, and in the RM, pons, hypothalamus and cortex level 3 when compared with the hippocampus and thalamus after RABV infection (Table 1 and Supplementary Fig. S3). Similarly, CCL5 immunolabelling was significantly (P <0.05) higher in the medulla (rostral and caudal) when compared with the frontal brain regions (apart from the cortex), after

EBLV-1 infection (Table 1). The CM also exhibited significantly (P <0.05) more CCL5 immunolabelling than all other brain regions assessed, apart from the cortex and the RM, and the RM and pons demonstrated significantly more CCL5 immunolabelling than the hippocampus and thalamus after EBLV-2 infection (Table 1 and Supplementary Fig. S3). The percentage area of CXCL10 immunolabelling was found to be significantly higher (P <0.05) in the CM when compared with all other brain regions and in the RM, pons and hypothalamus when compared with the hippocampus and thalamus in mice exhibiting clinical disease after inoculation with RABV (Table 1 and Supplementary Fig. S3). A similar significance (P <0.05) in the distribution of CXCL10 immunolabelling to the caudal brain regions was observed in EBLV-1-infected mice where the CM and RM exhibited more immunolabelling than all other brain regions (Table 1 and Supplementary Fig. S3). Although regional variation in CXCL10 immunolabelling was observed after EBLV-2 infection, these differences were only found to be significant (P <0.05) when caudal brain regions, including the caudal medulla, rostral medulla and pons, were compared with rostral brain regions including the hippocampus and thalamus (Table 1 and Supplementary Fig. S3). Challenge Group Analysis

The chemokine immunolabelling dataset was also analyzed using linear regression models, in combination with the Wald test, to determine whether the differences within specific brain regions were due to the challenge virus (Table 2). Comparison of mockinfected and lyssavirus-infected groups demonstrated that CCL2 chemokine immunolabelling was significantly higher after virus inoculation (P <0.05) in all Table 2 Summary of the Wald test for virus-related differences in the percentage area of chemokine immunolabelling †

Mock

RABV

RABV *1, 3, 4, 5, 6, 7 e EBLV-1 1, 3, 4, 5, 6, 7 e EBLV-2 1, 3, 4, 5, 6, 7 e CCL5 RABV 5, 6, 7 e EBLV-1 1, 4, 5, 6, 7 1, 3, 4, 5, 6, 7 EBLV-2 5, 6, 7 6, 7 CXCL10 RABV 4, 5, 6, 7 e EBLV-1 4, 5, 6, 7 7 EBLV-2 4, 5, 6, 7 4, 6, 7

CCL2

†

EBLV-1 EBLV-2 4, 6 e 4, 7 e e e 7 e 6, 7

3, 6 e e e 1, 3, 5, 6 e 4, 6, 7 6, 7 e

*Numbers denote brain regions: (1) cortex level 3, (2) hippocampus, (3) thalamus, (4) hypothalamus, (5) pons, (6) rostral medulla and (7) caudal medulla, where the mean percentage area of chemokine immunolabelling observed was significantly higher (P <0.05) in the † column inoculation group than the †row inoculation group.

456

D.J. Hicks et al.

brain regions apart from in the hippocampus, where no significant difference was found (Table 2). RABV infection resulted in significantly (P <0.05) more CCL2 immunolabelling than EBLV-1 infection in the hypothalamus and the rostral medulla and more CCL2 immunolabelling than EBLV-2 infection in the thalamus and rostral medulla (P < 0.05). CCL2 immunolabelling was significantly higher in both the hypothalamus and caudal medulla of EBLV-2-infected mice when compared with the amount observed in EBLV-1-infected mice (Table 2). RABV- and EBLV-2-infected mice demonstrated substantially more CCL5 immunolabelling in the pons, rostral medulla and caudal medulla when compared with mock-infected mice (P <0.05). CCL5 immunolabelling in the cortex level 3, hypothalamus, pons, rostral medulla and caudal medulla was significantly (P <0.05) higher in EBLV-1 infected mice when compared with mock-infected mice (Table 2). EBLV-1 infection resulted in significantly (P <0.05) more CCL5 immunolabelling than RABV infection in all the brain regions assessed bar the hippocampus (P <0.05) and more CCL5 labelling than EBLV-2 in all brain regions apart from the hippocampus, hypothalamus and caudal medulla (Table 2). In contrast, CCL5 immunolabelling was significantly (P <0.05) higher in both the caudal and rostral medulla of EBLV-2-infected mice when compared with RABV-infected mice (Table 2). RABV, EBLV-1 and EBLV-2 infection all resulted in significantly (P <0.05) more CXCL10 immunolabelling than mock-infection in four of the seven brain regions assessed; hypothalamus, pons, rostral medulla and caudal medulla (Table 2). RABV infection produced significantly (P <0.05) more CXCL10 immunolabelling than EBLV-1 infection in the caudal medulla, and significantly more CXCL10 expression than EBLV-2 infection in the hypothalamus, rostral medulla and caudal medulla (P <0.05). EBLV-1 infection also produced significantly more CXCL10 immunolabelling than EBLV-2 infection in both the rostral medulla and caudal medulla (Table 2). Associations between Chemokine Labelling, Pathology and Viral Antigen Score

For the purpose of analyzing the relationship between the different variables, three linear regression models were established, one for each chemokine. These models sought to detect linear associations between perivascular cuff count, viral antigen IHC (viral antigen), virus isolate genotype and the % of chemokine immunolabelling across the brain regions assessed. A summary of the model coefficients, standard errors,

P values and 95% confidence intervals for the coefficients are shown in Table 3. A strong positive association (P <0.05) was found in the CCL2 regression model between the PC counts and the % area of CCL2 immunolabelling, when adjusted by viral antigen and virus type, indicating that an increase in PC count (as an indication of encephalitic changes) corresponded with an increase in CCL2 immunolabelling. The strength of this association was more pronounced, a 50% increase in effect, when PC counts were >5 (Table 3). There was also a strong positive association between the amount of viral antigen immunolabelling and the percentage area of CCL2 immunolabelling, indicating that an increase in viral antigen also corresponded with an increase in CCL2 immunolabelling. The effect of this association was more pronounced (a three-fold increase) when high viral antigen scores were compared with low viral antigen scores (Table 3). The CCL5 and CXCL10 linear regression models also showed strong positive associations (P <0.05) between the PC counts and the percentage area of chemokine immunolabelling, indicating that an increased PC count also corresponded with an increase in CCL5 and CXCL10 immunolabelling. A 23% increase in the effect of this association was observed in the CCL5 model when the high PC count category was compared with the low PC count category and a 50% increase in effect was observed when the same PC categories were compared in the CXCL10 model (Table 3). In both the CCL5 and Table 3 Summary of chemokine regression model coefficient outputs Coefficient

Standard error

P value

95% confidence interval

CCL2 Low PC count High PC count Low viral antigen High viral antigen Constant (a)

0.256 0.381 0.104 0.294 0.112

0.021 0.024 0.028 0.029 0.016

<0.05 <0.05 <0.05 <0.05 <0.05

0.215e0.297 0.333e0.429 0.047e0.161 0.236e0.352 0.079e0.145

CCL5 Low PC count High PC count Viral antigen Constant (a)

0.167 0.218 0.005 0.190

0.011 0.013 0.002 0.015

<0.05 <0.05 0.024 <0.05

0.145e0.190 0.192e0.244 0.0007e0.009 0.159e0.221

CXCL10 Low PC count High PC count Viral antigen Constant (a)

0.198 0.307 0.078 0.093

0.023 0.027 0.015 0.022

<0.05 <0.05 <0.05 <0.05

0.152e0.244 0.253e0.360 0.047e0.108 0.048e0.138

*Viral antigen remained in this model as an ordered categorical variable and the coefficient refers to the increase in the % of chemokine for an increase in a unit of viral antigen. PC, perivascular cuff.

457

Chemokine Responses in Murine Lyssavirus Infection

icantly greater (P < 0.05) by the Wilcoxon exact sum rank test, when lyssavirus-infected cell culture supernatants and lysates were compared with culture supernatants and lysates derived from mock-infected (cell culture media only) primary neuron cultures. Lyssavirus-related differences in chemokine protein expression levels were observed for all three chemokines assessed. CXCL10 was the most abundant chemokine and CCL5 the least abundant chemokine produced by lyssavirus-infected neurons for all three viruses (Fig. 4). No significant difference in mean CXCL10 protein expression levels were observed between EBLV-1 and EBLV-2 infection, although both produced significantly (P <0.05) more CXCL10 than RABV infection (Fig. 4). RABV infection produced the highest mean CCL2 protein concentration, which was significantly higher than the mean amount of protein produced after either EBLV-1 or EBLV-2 infection (P <0.05). EBLV-1 infection produced the smallest amount of CCL2, which was significantly less than the mean amount of CCL2 produced by EBLV-2 infection (Fig. 4). CCL5 was produced in the smallest amounts after lyssavirus infection, with RABV infection producing significantly more than both EBLV-1 and EBLV-2 infection (P <0.05). EBLV-1 infection resulted in the production of significantly more CCL5 than EBLV-2 infection (P <0.05, Fig. 4).

CXCL10 models a positive association was found between the viral antigen score and the percentage area of chemokine immunolabelling; however, in both cases, particularly in the CCL5 model, the strength of the association was weaker than the association between chemokine expression and PCs (Table 3). Overall, significant (P <0.05) positive associations were shown between PC count and the percentage area of chemokine immunolabelling (CCL2, CCL5 and CXCL10), adjusted by viral antigen score and virus type, and therefore with increasing PC counts a higher percentage area of chemokine immunolabelling would be expected. When the PC values were >5 a stronger effect was observed for all three chemokines. The positive association between viral antigen score and percentage area of chemokine immunolabelling was also significant (P <0.05), although the effects in the CCL5 and CXCL10 models were more subtle (i.e. an increase in viral antigen resulted in a very small increase in the percentage area of CCL5 and CXCL10 immunolabelling). This association between viral antigen score and chemokine immunolabelling was observed to be much stronger in the CCL2 regression model. In vitro Quantification of Chemokine Expression by Infected Neurons

Chemokine production by neurons infected with lyssaviruses was also investigated in an in-vitro model. The concentration of CCL2, CCL5 and CXCL10 in cell culture supernatants and lysates derived from primary neuron cultures infected with RABV, EBLV-1 or EBLV-2 was evaluated using a three-plex MILLIPEXÒ MAP mouse chemokine kit in conjunction with the LuminexÒ system. CCL2, CCL5 and CXCL10 (Fig. 4) protein levels were shown to be signif-

Chemokine protein (pg/ml)

3180

CCL2

***

600

**

2650

300

1060

200

*

0

12000

500

1590

530

The present study describes the distribution of chemokine immunolabelling in association with histopathological changes and lyssavirus antigen distribution in CD-1 mice after peripheral inoculation with RABV, EBLV-1 or EBLV-2. It has also measured differential chemokine production by neurons in vitro after

CCL5

RABV

1 EBLV-1

EBLV-2

0

**

**

8000

**

6000

*

100

Mock

CXCL10

10000

***

400

2120

Discussion

*

4000 2000

Mock

RABV

1

EBLV-1 EBLV-2

0

Mock

RABV

1

EBLV-1 EBLV-2

Fig. 4. Mean CCL2, CCL5 and CXCL10 protein (pg/ml) expression levels in cell lysates and supernatants derived from primary neuron cultures infected with RABV, EBLV-1 or EBLV-2 (error bars denote standard deviation from the mean). Asterisks denote significant differences (P <0.05, by the Wilcoxon exact sum rank test) in the pg/ml of chemokine produced when the different viruses were compared. CCL2: * significantly higher than mock infection, ** significantly higher than EBLV-1 and mock infection and *** significantly higher than EBLV-1, EBLV-2 and mock infection, CCL5: * significantly higher than mock infection, ** significantly higher than EBLV-2 and mock infection and *** significantly higher than EBLV-1, EBLV-2 and mock infection and CXCL10: * significantly higher than mock infection, ** significantly higher than RABV and mock infection.

458

D.J. Hicks et al.

infection with the same viruses. The features of clinical disease, caudal bias in brain distribution and the virus isolate-dependent differences in the abundance of inflammatory changes in CD-1 mice were consistent with findings reported previously in OF-1 mice, where pathology was most severe in EBLV-2infected animals and least prominent in RABVinfected mice (Hicks et al., 2009). Although chemokine and viral antigen immunolabelling was present throughout all the brain regions after lyssavirus infection, the immunolabelling featured more prominently in the caudal regions of the brain. The strong positive associations found between PC count and chemokine immunolabelling and viral antigen score and chemokine immunolabelling corroborated this apparent synchronicity in distribution, with an increase in PC or viral antigen corresponding to an increase in chemokine immunolabelling. This highlights the role of chemokines in the development of encephalitic changes in viral infections of the CNS (Savarin and Bergmann, 2008). The demonstration of CCL2, CCL5 and CXLC10 immunolabelling in lyssavirus-infected mice and chemokine production by neurons in vitro confirmed previous reports describing the upregulation of chemokine gene transcripts in the murine CNS after experimental infection with RABV (Johnson et al., 2008) and EBLV-2 (Mansfield et al., 2008). The localization of CCL5 immunolabelling to non-neuronal cell populations was consistent with other studies that described the upregulation of CCL5 mRNA transcripts in mixed glia, astrocyte and microglia, but not neuronal cell cultures, after Japanese encephalitis virus infection (Chen et al., 2004) and also by microglia after RABV infection (Nakamichi et al., 2005). Similarly, the presence of CCL2 and CXCL10 immunolabelling in both neuronal and non-neuronal cell populations after lyssavirus infection was consistent with previous in-vitro experiments investigating chemokine production by microglia (Nakamichi et al., 2005) and neurons (Klein et al., 2005) in response to RABV and West Nile virus infection, respectively. The contribution of chemokines to the ensuing encephalitis has been investigated in several studies comparing the silver-haired bat rabies virus (SHBRV), a species 1 RABV, with pathogenic and attenuated lyssavirus strains, challenge virus standard F3 (CVS-F3) and CVS-B2C, respectively (Wang et al., 2005; Roy et al., 2007). Infection with SHBRV produces a muted inflammatory response in comparison with the attenuated CVS-B2C and, subsequently, distinct differences were observed in the level of immunopathology produced (Wang et al., 2005). However, comparison of pro-inflammatory gene expression in the cerebellum and other regions after infection with

either SHBRV or the pathogenic CVS-F3 lyssavirus revealed similar levels of CCL2, CCL5 and CXCL10 being produced, demonstrating that the variation in encephalitis observed was not attributable to differences in gene expression (Roy et al., 2007). However, these studies have focused on RABV strain comparisons and evaluated mRNA expression, but not measured protein production. The present study compared lyssaviruses from different species and demonstrated that distinct isolate-related differences in chemokine immunolabelling were present in the brains of RABV-, EBLV-1- and EBLV-2 infected mice. Lyssavirus isolate-dependent differences in the amount of CCL2, CCL5 and CXCL10 immunolabelling were observed both in vivo and in vitro. Consequently, due to their apparent paradoxical role in disease pathogenesis, these differences could contribute to the isolate-dependent differences in the severity of encephalitis observed. CCL2 has been shown to be integral to cellular recruitment in Dengue-3 feverassociated viral encephalitis (Amaral et al., 2011) and instrumental for T-cell migration through the glia limitans during herpes virus infection (Savarin et al., 2010) in murine infection models. However, overproduction and accumulation of CCL2 has been shown to exacerbate the human immunodeficiency virus-associated encephalitis through the manipulation of tight junction protein expression and distribution, which ultimately disrupts bloodebrain barrier (BBB) integrity and functionality (Eugenin et al., 2006; Dhillon et al., 2008). Upregulation of neuronal CXCL10 mRNA expression has been reported to be fundamental for the recruitment of T cells, which ultimately resolve West Nile virus infection (Klein et al., 2005). A similar, initially beneficial, response was triggered in neurons after mouse hepatitis virus infection (Liu et al., 2000, 2001) and in murine models of abortive RABV infection where CXCL10 affected the permeability of the BBB (Kuang et al., 2009). However, prolonged CXCL10 activity and sequestration of T cells to the CNS can cause demyelination through the release of other inflammatory mediators including CCL5 (Lane et al., 2000; Lui et al., 2001; Glass et al., 2004, 2005). In the present study, significant virus isolate-related differences in CCL2 and CXCL10 immunolabelling were limited to a few brain regions in vivo with RABV-infected mice producing greater chemokine immunolabelling than both EBLV-1 and EBLV-2 in these regions. However, neurons removed from the tissue context and challenged in vitro with RABV, EBLV1 and EBLV-2 produced results contrary to those observed in vivo. CCL2 production by neurons in vitro appeared consistent with that observed in vivo, with RABV infection producing the most CCL2 and

Chemokine Responses in Murine Lyssavirus Infection

EBLV-1 infection producing the least. Neurons infected in vitro with EBLV produced significantly more CXCL10 than those infected with RABV. Taking into consideration the neurotropic nature of lyssaviruses (Iwasaki and Clark, 1975; Tsiang et al., 1983) and the observation that CXCL10 was a primary response to neuronal infection in other examples of viral encephalitis (Liu et al., 2000, 2001; Klein et al., 2005), the in vitro results appear to support a crucial role for CXCL10 in the initial response by neurons to lyssavirus infection. Consequently, increased CXCL10 production and its associated ramifications (e.g. T-cell recruitment and neuronal functional change) could contribute to the severity of encephalitis observed when compared with RABV infection. In vivo, the difference in CXCL10 immunolabelling was not so striking. This could be attributed to the in vitro results reflecting an early response by neurons to infection without secondary activation of residential microglia (Nakamichi et al., 2005; Hosking and Lane, 2010). Furthermore, the in vitro system, unlike the infected brain, is not influenced by infiltrating cell populations capable of producing chemokines and other inflammatory mediators including macrophages (Nakamichi et al., 2004) and lymphocytes (Celis et al., 1986; Perrin et al., 1991). CCL5 has been shown to be critical for leukocyte adhesion during cellular recruitment in mice infected with herpes simplex virus (Vilela et al., 2009) and upregulated prior to both behavioural changes and death in mice infected with Dengue fever virus (Amaral et al., 2011). CCL5 has also been reported to feature prominently in Japanese encephalitis virus (Winter et al., 2004), Borna disease virus (Ovanesov et al., 2008) and West Nile virus (Shirato et al., 2004; Klein et al., 2005) infections of the CNS. In the present study, the most striking lyssavirus isolate-dependent differences in chemokine immunolabelling in vivo were also associated with CCL5 and, in particular, after EBLV-1 infection when compared with RABV infection. In contrast, CCL5 production in vitro was highest after RABV infection, although the amount of CCL5 was considerably less than that for both CCL2 and CXCL10. This could suggest that CCL5 production is supplemented in vivo, as a result of direct RABV (Nakamichi et al., 2005), EBLV-1 or EBLV-2 infection (data not shown) or through bystander activation of non-neuronal cell populations including the microglia, which were absent from the neuronal cultures. In vitro models of West Nile virus (Klein et al., 2005) and Borna disease virus (Ovanesov et al., 2008) infection have highlighted the sequential nature of chemokine production. In both cases, CCL5 production was secondary to the primary cellular recruitment signal, implying that bystander and recruited

459

cell expression of CCL5 function to amplify the initial host response to infection and may exacerbate encephalitis. The nominal CCL5 production by lyssavirusinfected primary neurons in vitro and the predominantly non-neuronal pattern of CCL5 expression and the association between CCL5 and areas of inflammation described in this in vitro model of lyssavirus infection in the present study appears to support bystander or recruited cell expression of CCL5. Despite exhibiting a similar disease course and clinical period to RABV infection (data not shown), EBLV-1-infected mice, in the present study, exhibited more CCL5 immunolabelling and more severe encephalitis. The most severe encephalitis was observed in EBLV-2-infected mice, but in these animals the abundance of CCL5 was found to be lower in many brain regions compared with that observed after EBLV-1 infection. This suggests that other factors, such as the longer disease course observed with EBLV-2 infection (data not shown) and the expression of other chemokines (Kuang et al., 2009) and inflammatory mediators, including inducible nitric oxide synthase (Koprowski et al., 1993; Komatsu et al., 1999; Hooper et al., 2001), tumour necrosis factor-a (Phares et al., 2006; Roy et al., 2007) and interferon-g (Johnson et al., 2008), which have been shown to be involved in CCL5 transcription pathways (Randall and Goodbourn, 2008; Homma et al., 2010), could contribute to the severity of lyssavirusassociated encephalitis. Therefore, although CCL5 production may not be the primary driver of cell recruitment, the level of CCL5 expression during clinical disease may influence the severity of the encephalitis observed in CD-1 mice after lyssavirus infection. In conclusion, isolate-related differences in chemokine production by neurons in vitro and by CNS resident cell populations in vivo were observed after lyssavirus infection. CXCL10 production was significantly higher after EBLV infection in vitro, and this direct response to lyssavirus infection could represent the primary cellular recruitment signal for T cells. Despite similarities in the cellular localization and in the distribution of chemokine immunolabelling throughout the brain, lyssavirus-dependent differences in the percentage area of chemokine immunolabelling were observed in vivo. Positive associations between chemokine immunolabelling and both perivascular cuffing and, to a lesser extent, viral antigen score were also demonstrated. These associations, taken in conjunction with the greater percentage area of CCL5 immunolabelling after EBLV infection, suggest that isotype-related differences in CCL5 expression may exacerbate the encephalitis observed after lyssavirus infection. The isolate-related differences in both chemokine expression and inflammatory changes

D.J. Hicks et al.

460

described in this study could be indicative of genotyperelated differences. Other viral isolates from each of the three species, RABV, EBLV-1 and EBLV-2, have produced inflammatory changes consistent with the representative isolates used in this study (data not shown). However, the investigation of chemokine immunolabelling produced during infection by these additional virus isolates would be required to confirm this hypothesis. Further work focusing on the contribution of other chemokines, the sequential expression of chemokines during clinical disease and the effect of virus dynamics on disease pathogenesis would also help to put the significance of these isolate-related observations into the greater context of virus-induced encephalitis.

Acknowledgments The authors would like to thank members of the Department of Specialist Scientific Support and the Wildlife Zoonoses and Vector-Borne Diseases Research Group for their technical assistance and members of the Animal Services Unit at the AHVLA for their animal husbandry expertise. This work was funded by the Defra project ROAME SE0421.

Conflict of Interest Statement None of the authors has any financial or personal relationships with other people or organizations that could have inappropriately influenced this work.

Supplementary data Supplementary data related to this article can be found at doi:10.1016/j.jcpa.2013.04.001.

References Amaral DC, Rachid MA, Vilela MC, Campos RD, Ferreira GP et al. (2011) Intracerebral infection with dengue-3 virus induces meningoencephalitis and behavioural changes that precede lethality in mice. Journal of Neuroinflammation, 8, 23. Banisadr G, Rostene W, Kitabgi P, Parsadaniantz SM (2005) Chemokines and brain functions. Current Drug Targets: Inflammation and Allergy, 4, 387e399. Carr MW, Roth SJ, Luther E, Rose SS, Springer TA (1994) Monocyte chemoattractant protein 1 acts as a T-lymphocyte chemoattractant. Proceedings of the National Academy of Sciences of the USA, 91, 3652e3656. Carrillo-de Sauvage MA, G
omez A, Ros CM, RosBernal F, Mart
ın ED et al. (2012) CCL2-expressing astrocytes mediate the extravasation of T lymphocytes in the brain. Evidence from patients with glioma and experimental models in vivo. PLoS One, 7, e30762.

Celis E, Miller RW, Wiktor TJ, Dietzschold B, Koprowski H (1986) Isolation and characterization of human T-cell lines and clones reactive to rabies virus: antigen specificity and production of interferon-g. Journal of Immunology, 136, 692e697. Charlton KM (1984) Rabies: spongiform lesions in the brain. Acta Neuropathologica, 63, 198e202. Charlton KM, Casey GA, Webster WA, Bundza A (1987) Experimental rabies in skunks and foxes. Pathogenesis of the spongiform lesions. Laboratory Investigation, 57, 634e645. Chen CJ, Chen JH, Chen SY, Liao SL, Raung SL (2004) Upregulation of RANTES gene expression in neuroglia by Japanese encephalitis virus infection. Journal of Virology, 78, 12107e12119. De Haas AH, van Weering HR, de Jong EK, Boddeke HW, Biber KP (2007) Neuronal chemokines: versatile messengers in central nervous system cell interaction. Molecular Neurobiology, 36, 137e151. Dhillon NK, Williams R, Callen S, Zien C, Narayan O et al. (2008) Roles of MCP-1 in development of HIV-dementia. Frontiers in Bioscience, 13, 3913e3918. Dufour JH, Dziejman M, Liu MT, Leung JH, Lane TE et al. (2002) IFN-g-inducible protein 10 (IP-10; CXCL10)-deficient mice reveal a role for IP-10 in effector T cell generation and trafficking. Journal of Immunology, 168, 3195e3204. Eugenin EA, Osiecki K, Lopez L, Goldstein H, Calderon TM et al. (2006) CCL2/monocyte chemoattractant protein-1 mediates enhanced transmigration of human immunodeficiency virus (HIV)-infected leukocytes across the blood-brain barrier: a potential mechanism of HIV-CNS invasion and NeuroAIDS. Journal of Neuroscience, 26, 1098e1106. Fekadu M (1988) Pathogenesis of rabies virus infection in dogs. Reviews of Infectious Diseases, 4(Supplement), S678eS683. Fekadu M, Chandler FW, Harrison AK (1982) Pathogenesis of rabies in dogs inoculated with an Ethiopian rabies virus strain. Immunofluorescence, histologic and ultrastructural studies of the central nervous system. Archives of Virology, 71, 109e126. Freuling CM, Beer M, Conraths FJ, Finke S, Hoffmann B et al. (2011) Novel lyssavirus in Natterer’s bat, Germany. Emerging Infectious Diseases, 17, 1519e1522. Freuling CM, M€ uller T, Marston D, Fooks AR, Rupprecht CE et al. (2012) Updates on the diversity of the Lyssavirus genus. Rabies Bulletin Europe, 35, 8e10. Galelli A, Baloul L, Lafon M (2000) Abortive rabies virus central nervous infection is controlled by T lymphocyte local recruitment and induction of apoptosis. Journal of Neurovirology, 6, 359e372. Garg RK (2002) Subacute sclerosing panencephalitis. Postgraduate Medical Journal, 78, 63e70. Gelpi E, Preusser M, Garzuly F, Holzmann H, Heinz FX et al. (2005) Visualization of Central European tickborne encephalitis infection in fatal human cases. Journal of Neuropathology & Experimental Neurology, 64, 506e512.

Chemokine Responses in Murine Lyssavirus Infection

Glass WG, Hickey MJ, Hardison JL, Liu MT, Manning JE et al. (2004) Antibody targeting of the CC chemokine ligand 5 results in diminished leukocyte infiltration into the central nervous system and reduced neurologic disease in a viral model of multiple sclerosis. Journal of Immunology, 172, 4018e4025. Glass WG, Lim JK, Cholera R, Pletnev AG, Gao JL et al. (2005) Chemokine receptor CCR5 promotes leukocyte trafficking to the brain and survival in West Nile virus infection. Journal of Experimental Medicine, 202, 1087e1098. Hemachudha T, Sunsaneewitayakul B, Desudchit T, Suankratay C, Sittipunt C et al. (2006) Failure of therapeutic coma and ketamine for therapy of human rabies. Journal of Neurovirology, 12, 407e409. Hickman SE, El Khoury J (2010) Mechanisms of mononuclear phagocyte recruitment in Alzheimer’s disease. CNS and Neurological Disorders Drug Targets, 9, 168e173. Hicks DJ, Nunez A, Healy DM, Brookes SM, Johnson N et al. (2009) Comparative pathological study of the murine brain after experimental infection with classical rabies virus and European bat lyssaviruses. Journal of Comparative Pathology, 140, 113e126. Homma T, Matsukura S, Hirose T, Ohnishi T, Kimura T et al. (2010) Cooperative activation of CCL5 expression by TLR3 and tumor necrosis factor-a or interferon-g through nuclear factor-kB or STAT-1 in airway epithelial cells. International Archive of Allergy and Immunology, 152(Suppl. 1), 9e17. Hooper DC, Kean RB, Scott GS, Spitsin SV, Mikheeva T et al. (2001) The central nervous system inflammatory response to neurotropic virus infection is peroxynitrite dependent. Journal of Immunology, 167, 3470e3477. Hooper DC, Phares TW, Fabis MJ, Roy A (2009) The production of antibody by invading B cells is required for the clearance of rabies virus from the central nervous system. PLoS Neglected Tropical Diseases, 3, e535. Hooper PT, Fraser GC, Foster RA, Storie GJ (1999) Histopathology and immunohistochemistry of bats infected by Australian bat lyssavirus. Australian Veterinary Journal, 77, 595e599. Hosking MP, Lane TE (2010) The role of chemokines during viral infection of the CNS. PLoS Pathogens, 6, e1000937. Iwasaki Y, Clark HF (1975) Cell to cell transmission of virus in the central nervous system. II. Experimental rabies in mouse. Laboratory Investigation, 33, 391e399. Johnson N, Mansfield KL, Hicks D, Nunez A, Healy DM et al. (2008) Inflammatory responses in the nervous system of mice infected with a street isolate of rabies virus. Developments in Biologicals, 131, 65e72. Johnson RT, Burke DS, Elwell M, Leake CJ, Nisalak A et al. (1985) Japanese encephalitis: immunocytochemical studies of viral antigen and inflammatory cells in fatal cases. Annals of Neurology, 18, 567e573. Kelley TW, Prayson RA, Ruiz AI, Isada CM, Gordon SM (2003) The neuropathology of West Nile virus meningoencephalitis. A report of two cases and review of the literature. American Journal of Clinical Pathology, 119, 749e753.

461

Klein RS, Lin E, Zhang B, Luster AD, Tollett J et al. (2005) Neuronal CXCL10 directs CD8+ T-cell recruitment and control of West Nile Virus encephalitis. Journal of Virology, 79, 11457e11466. Komatsu T, Ireland DD, Chen N, Reiss CS (1999) Neuronal expression of NOS-1 is required for host recovery from viral encephalitis. Virology, 258, 389e395. Koprowski H, Zheng YM, Heber-Katz E, Fraser N, Rorke L et al. (1993) In vivo expression of inducible nitric oxide synthase in experimentally induced neurologic diseases. Proceedings of the Natural Academy of Sciences USA, 90, 3024e3027. Kuang Y, Lackay SN, Zhao L, Fu ZF (2009) Role of chemokines in the enhancement of BBB permeability and inflammatory infiltration after rabies virus infection. Virus Research, 144, 18e26. Kuzmin IV, Mayer AE, Niezgoda M, Markotter W, Agwanda B et al. (2010) Shimoni bat virus, a new representative of the Lyssavirus genus. Virus Research, 149, 197e210. Lafon M (2005) Modulation of the immune response in the nervous system by rabies virus. Current Topics in Microbiology and Immunology, 289, 239e258. Lane TE, Liu MT, Chen BP, Asensio VC, Samawi RM et al. (2000) A central role for CD4+ T cells and RANTES in virus-induced central nervous system inflammation and demyelination. Journal of Virology, 74, 1415e1424. Liu MT, Chen BP, Oertel P, Buchmeier MJ, Armstrong D et al. (2000) The T cell chemoattractant IFN-inducible protein 10 is essential in host defense against viralinduced neurologic disease. Journal of Immunology, 165, 2327e2330. Liu MT, Keirstead HS, Lane TE (2001) Neutralization of the chemokine CXCL10 reduces inflammatory cell invasion and demyelination and improves neurological function in a viral model of multiple sclerosis. Journal of Immunology, 167, 4091e4097. Mansfield KL, Johnson N, Nunez A, Hicks D, Jackson AC et al. (2008) Up-regulation of chemokine gene transcripts and T-cell infiltration into the central nervous system and dorsal root ganglia are characteristics of experimental European bat lyssavirus type 2 infection of mice. Journal of Neurovirology, 14, 218e228. Murphy FA (1977) Rabies pathogenesis. Archives of Virology, 54, 279e297. Nakamichi K, Inoue S, Takasaki T, Morimoto K, Kurane I (2004) Rabies virus stimulates nitric oxide production and CXC chemokine ligand 10 expression in macrophages through activation of extracellular signal-regulated kinases 1 and 2. Journal of Virology, 78, 9376e9388. Nakamichi K, Saiki M, Sawada M, Takayama-Ito M, Yamamuro Y et al. (2005) Rabies virus-induced activation of mitogen-activated protein kinase and NF-kB signalling pathways regulates expression of CXC and CC chemokine ligands in microglia. Journal of Virology, 79, 11801e11812. Ovanesov MV, Ayhan Y, Wolbert C, Moldovan K, Sauder C et al. (2008) Astrocytes play a key role in

462

D.J. Hicks et al.

activation of microglia by persistent Borna disease virus infection. Journal of Neuroinflammation, 5, 50. Patterson CE, Daley JK, Echols LA, Lane TE, Rall GF (2003) Measles virus infection induces chemokine synthesis by neurons. Journal of Immunology, 17, 3102e3109. Perrin P, Joffret ML, Zanetti C, Bourhy H, Gontier C et al. (1991) Rabies-specific production of interleukin-2 by peripheral blood lymphocytes from human rabies vaccinees. Vaccine, 9, 549e558. Perry LL, Lodmell DL (1991) Role of CD4+ and CD8+ T cells in murine resistance to street rabies virus. Journal of Virology, 65, 3429e3434. Phares TW, Kean RB, Mikheeva T, Hooper DC (2006) Regional differences in blood-brain barrier permeability changes and inflammation in the apathogenic clearance of virus from the central nervous system. Journal of Immunology, 176, 7666e7675. Randall RE, Goodbourn S (2008) Review: interferons and viruses: an interplay between induction, signalling, antiviral responses and virus countermeasures. Journal of General Virology, 89, 1e47. Roy A, Phares TW, Koprowski H, Hooper DC (2007) Failure to open the blood-brain barrier and deliver immune effectors to central nervous system tissues leads to the lethal outcome of silver-haired bat rabies virus infection. Journal of Virology, 81, 1110e1118. Savarin C, Bergmann CC (2008) Neuroimmunology of central nervous system viral infections: the cells, molecules and mechanisms involved. Current Opinions in Pharmacology, 8, 472e479. Savarin C, Stohlman SA, Atkinson R, Ransohoff RM, Bergmann CC (2010) Monocytes regulate T-cell migration through the glia limitans during acute viral encephalitis. Journal of Virology, 84, 4878e4888. Shirato K, Kimura T, Mizutani T, Kariwa H, Takashima I (2004) Different chemokine expression in lethal and non-lethal murine West Nile virus infection. Journal of Medical Virology, 74, 507e513.

Sugamata M, Miyazawa M, Mori S, Spangrude GJ, Ewalt LC et al. (1992) Paralysis of street rabies virusinfected mice is dependent on T lymphocytes. Journal of Virology, 66, 1252e1260. Tsiang H, Koulakoff A, Bizzini B, Berwald-Netter Y (1983) Neurotropism of rabies virus. An in vitro study. Journal of Neuropathology & Experimental Neurology, 42, 439e452. Vilela MC, Mansur DS, Lacerda-Queiroz N, Rodrigues DH, Lima GK et al. (2009) The chemokine CCL5 is essential for leukocyte recruitment in a model of severe herpes simplex encephalitis. Annals of the New York Academy of Sciences, 1153, 256e263. Wang ZW, Sarmento L, Wang Y, Li XQ, Dhingra V et al. (2005) Attenuated rabies virus activates, while pathogenic rabies virus evades, the host immune responses in the central nervous system. Journal of Virology, 79, 12554e12565. Weiland F, Cox JH, Meyer S, Dahme E, Reddehase MJ (1992) Rabies virus neuritic paralysis: immunopathogenesis of nonfatal paralytic rabies. Journal of Virology, 66, 5096e5099. Winter PM, Dung NM, Loan HT, Kneen R, Wills B et al. (2004) Proinflammatory cytokines and chemokines in humans with Japanese encephalitis. Journal of Infectious Diseases, 190, 1618e1626. Yan X, Prosniak M, Curtis MT, Weiss ML, Faber M et al. (2001) Silver-haired bat rabies virus variant does not induce apoptosis in the brain of experimentally infected mice. Journal of Neurovirology, 7, 518e527. Yen CF, Patel MS, Chu P, Yen Y, Chen W (2009) Case report: herpes simplex encephalitis in cancer patients. Bioscience Trends, 3, 38e40.

October 12th, 2012 ½ Received,  Accepted, April 6th, 2013