Antiviral CD8 T cells induce Zika-virus-associated paralysis in mice.

Zika virus (ZIKV) is an emerging, mosquito-borne RNA virus. The rapid spread of ZIKV within the Americas has unveiled microcephaly 1 and Guillain-Barré syndrome2,3 as ZIKV-associated neurological complications. Recent reports have also indicated other neurological manifestations to be associated with ZIKV, including myelitis 4 , meningoencephalitis 5 and fatal encephalitis 6 . Here, we investigate the neuropathogenesis of ZIKV infection in type I interferon receptor IFNAR knockout (Ifnar1 -/- ) mice, an infection model that exhibits high viral burden within the central nervous system. We show that systemic spread of ZIKV from the site of infection to the brain requires Ifnar1 deficiency in the haematopoietic compartment. However, spread of ZIKV within the central nervous system is supported by Ifnar1-deficient non-haematopoietic cells. Within this context, ZIKV infection of astrocytes results in breakdown of the blood-brain barrier and a large influx of CD8+ effector T cells. We also find that antiviral activity of CD8+ T cells within the brain markedly limits ZIKV infection of neurons, but, as a consequence, instigates ZIKV-associated paralysis. Taken together, our study uncovers mechanisms underlying ZIKV neuropathogenesis within a susceptible mouse model and suggests blood-brain barrier breakdown and T-cell-mediated neuropathology as potential underpinnings of ZIKV-associated neurological complications in humans.

ZIKV inhibits type I interferon (IFN) receptor (IFNAR) signaling in human ZIKV infections, but not mice [7]. Thus, mouse models of ZIKV infection necessitate use of genetic deficiencies in IFNAR or antibody blockade of IFNAR [8-10]. To examine immune responses that control ZIKV neuropathogenesis, we first probed the cellular compartment responsible for the type I IFN-dependent block of ZIKV infection using irradiation-induced bone marrow chimeric mice between ZIKV-resistant C57BL/6 CD45.1 (WT) and ZIKV-susceptible interferon α/β receptor-deficient (Ifnar1−/−) mice. This approach allowed us to differentiate type I IFN-dependency of brain non-hematopoietic cells (i.e. endothelial, astrocyte or neuron) versus hematopoietic cells (i.e. lymphocytes or leukocyte) in mediating ZIKV spread and neuropathology. Ifnar1−/− mice reconstituted with Ifnar1−/− bone marrow (Ifnar1−/− → Ifnar1−/−) were the only chimeric combination to lose weight (Figure 1a) and present with the symptoms of hindlimb paralysis (Figure 1b). WT mice reconstituted with Ifnar1−/− bone marrow (Ifnar1−/− → WT) had significantly levels of viremia (Figure 1c), but had very low brain viral load (Figure 1d). It is possible that the lower rates of infection in the brain within Ifnar1−/− → WT mice could be due to lower levels of virus within the blood. Additionally, the low levels of ZIKV RNA detected within the brain of the other chimeras may be representative of virus within residual blood. Nevertheless, these data indicated that, while systemic spread of ZIKV requires Ifnar1-deficiency within the hematopoietic compartment, Ifnar1-deficiency within hematopoietic cells is not sufficient to cause neurological viral amplification and disease.

Figure 1

Systemic spread of ZIKV is dependent upon Ifnar1−/− deficiency in the hematopoietic compartment

a–d, Ifnar1−/− and CD45.1 C57BL/6 WT bone marrow chimeric mice (n=7 or 8 per group in two separate experiments) were infected with 106 plaque-forming units (p.f.u.) of ZIKV intra-footpad. Daily weight (a), development of hindlimb paralysis (b) and viremia on days 2, 5 and 7-post infection (c) were monitored. d, Seven days post infection, viral load in tissue homogenates including the brain, spleen and inguinal lymph node was measured. Data are presented as mean ± s.e.m. *P < 0.05; **P < 0.01 (two-tailed unpaired Student’s t-test).

Our data are consistent with recent work showing viremia, but limited brain infection and lack of neuropathology in LysMCre+Ifnar1fl/fl mice[11]. Moreover, our data indicate that Ifnar1-dependent block of ZIKV within brain non-hematopoietic cells controls the viral replication and associated neurological disease manifestation of a ZIKV infection. To identify the cell types that support ZIKV replication in the CNS, we conducted immunofluorescence staining of ZIKV-infected adult Ifnar1−/− brains 7 days post infection (DPI), a time when the majority of infected mice present with neurological complications and a high viral load is measured. Despite multiple cell types evaluated (microglia and oligodendrocytes (Supplemental Figure 1a and 1b)), we found that ZIKV infection is largely confined to astrocytes and to a lesser extent, within neurons of the cerebral cortex (Figure 2a). This tropism was evident in various regions within the CNS including the cerebellum, brain stem and spinal cord (Supplemental Figure 2). Despite a large number of inflammatory cells within the meninges, we did not find many positive ZIKV cells within this region 7 DPI (Supplemental Figure 1c).

Figure 2

ZIKV infection of astrocytes is associated with breakdown of the blood-brain barrier

a–d, Ifnar1−/− (n=5) and WT (n=3) mice were infected with 106 plaque-forming units (p.f.u.) of ZIKV or PBS (mock) (n=4) intra-footpad in two separate experiments. Frozen sections of the brain, specifically the cerebral cortex, were stained with antibodies against astrocytes (GFAP), neurons (NeuN) and ZIKV (scale bars, 50 μm) (a). Brain sections, specifically the cerebral cortex, were further stained with IgG (scale bars, 100 μm) (b), GFAP, ZIKV, CD31 (scale bars, 10 μm with 7.5 μm inset) (c). White arrows indicate blood vessels surrounded by ZIKV-infected astrocytes. d, Seven days post infection with ZIKV or PBS (mock), Oregon green 488-conjugated dextran (70 kDa) (5 mg ml−1, 200 μl per mouse) was injected intravenously in two separate experiments. Forty-five minutes later, these mice were sacrificed for immunohistochemical analysis (n=4 per infection condition). Nuclei are depicted by 4′,6-diamidino-2-phenylindole (DAPI) stain (blue) (scale bars, 10 μm).

Astrocytes extend their end foot processes to provide integral support for the blood brain barrier (BBB)[12]. Indeed, we observed prominent infection of the astrocyte end feet surrounding blood vessels in the CNS (Figure 2b). We predicted that ZIKV infection of astrocytes might lead to increased vascular permeability. To probe BBB breakdown, we stained for the presence of mouse IgG, which is normally blocked from accessing the CNS parenchyma by the BBB. We found bright IgG staining within infected brains of Ifnar1−/− mice, but not the Ifnar1-competent, wildtype ZIKV infected mice or the mock-infected Ifnar1−/− controls (Figure 2c). Consistently, intravenous injection of fluorescently-labeled dextran at the peak of ZIKV-disease presentation provided functional demonstration of a ZIKV-induced BBB disruption (Figure 2d, quantified in Supplemental Figure 3c). Further, areas where increased vascular permeability was detected (by evidence of leaky dextran) corresponded to regions where astrocyte foot process lining was fragmented (Supplemental Figure 3). Analysis of the brain sections of bone marrow chimeric mice corroborated stromal dependence in eliciting BBB breakdown, as Ifnar1−/− → Ifnar1−/− bone marrow chimeric mice were the only combination to exhibit positive IgG staining within the brain (Supplemental Figure 4). Our data therefore suggest that type I IFN signaling within astrocytes block their infection by ZIKV and prevent BBB breakdown. The neuroprotective role of the type I IFN cascade within astrocytes has recently been ascribed during the infection of another flavivirus, West Nile virus [13]. Astrocyte-specific Ifnar1-dependence for BBB leakage in ZIKV infection should be examined in future studies.

Histological analysis of ZIKV-infected Ifnar1−/− brains as compared to Ifnar1-competent, wildtype infected and mock, uninfected Ifnar1−/− mice indicated a vast infiltration of leukocytes on 7 DPI (Figure 3a), which were CD45+ (Figure 3b). Flow cytometry analysis of single-cell suspensions of infected and mock Ifnar1−/− brains further confirmed a significant increase in CD45+ cells 7 DPI when compared to both mock, uninfected, but also within ZIKV-infected brains 5 DPI (Figure 3c). Notably, the population of infiltrating immune cells largely comprised of T cells (CD3+) at 7 DPI, which only served as a minority population at 5 DPI (Figure 3d). Further analysis indicated that the majority of CD3+ cells were CD8+ effector T cells as analyzed by flow cytometry (Figure 3e) and immunofluorescence staining revealed that many of these cells are perforin-positive (Figure 3f). This result is consistent with a previous work showing CD8+ T cell infiltration within the brains of ZIKV-infected immune competent neonate C57Bl/6 mice without Ifnar1 manipulation [14]. Since we find the vast infiltration of effector T cells to coincide with CNS pathogenesis, our data suggested that either the effector T cells are reaching the brain too late to control virus replication or that their antiviral actions may be immunopathogenic.

Figure 3

ZIKV replication within the brain causes a large influx of CD8+ effector T cells

a–f, Ifnar1−/− (n=5) and WT (n=3) mice were infected with 106 plaque-forming units (p.f.u.) of ZIKV or PBS (mock) (n=4) intra-footpad in two separate experiments. Seven days post infection frozen sections of fixed brain, specifically the cerebral cortex, were stained with H&E (scale bars, 50 μm) (a) and antibodies against CD45 (scale bars, 100 μm) (b) perforin and CD45 (scale bars, 10 μm) (f). Nuclei are depicted by 4′,6-diamidino-2-phenylindole (DAPI) stain (blue). Single-cell suspensions of ZIKV-infected and mock Ifnar1−/− brains (n=6) were analyzed for quantification of CD45 positive cells by flow cytometry five (n=7) and seven days post infection (n=9) in three separate experiments. Data are presented as mean ± SD from an analysis of 1.5 million whole brain cells. (c). CD45+ cells were further analyzed for percentage of CD3+ and F4/80+ cells in 1.5 million whole brain cells (d) and quantification of CD4, CD8 through flow cytometry of isolated leukocytes from the brain (e). *P < 0.05; **P < 0.01 (two-tailed unpaired Student’s t-test).

In order to understand the effects of T cell infiltration on viral load and neuropathology of a ZIKV infection, we depleted T cells with antibodies. Notably, depletion of CD8+ T cells, but not CD4+ T cells, resulted in a significant increase in survival (Figure 4a). Yet, viral load at 7 DPI indicated a significantly higher viral load in the brain of the CD8+ T cell-depleted animals as compared to those intact in CD8+ T cells (Figure 4d). The higher viral load in CD8 T-depleted mice at 7 DPI was corroborated by ZIKV immunofluorescence within CD8+ T cell intact vs. depleted brain tissues (Figure 4f). These data are consistent with recent studies demonstrating elevated ZIKV replication in CD8-depleted or CD8-knockout anti-Ifnar1 treated animals [11] or in Rag1−/− mice [15]. In contrast, hindlimb paralysis was strikingly abrogated in CD8-depleted mice (Figure 4b), which were euthanized due to weight loss greater than 20% rather than onset of paralysis as was the case within the undepleted or CD4-depeleted mice (Figure 4c and Supplemental Figure 5). The CD8-depleted mice had, in addition to significant weight loss, symptoms of illness including ruffled fur, hunched backs and/or tremors late in infection.

Figure 4

Antiviral activity of effector CD8+ T cells within the brain limits ZIKV replication within neurons, but instigates ZIKV-associated paralysis

a–e, Ifnar1−/− were injected with depletion antibodies for CD8 (n=9), CD4 (n=9), CD4 and CD8 (n=8) or with isotype control (n=14) −1, 4 and 9 days post infection with 106 plaque-forming units (p.f.u.) of ZIKV intra-footpad in three separate experiments. Survival (a) development of hindlimb paralysis (b) were monitored with weight at time of sacrifice due to development of moribund state (onset of paralysis or >20% weight loss) was determined (c). d, Seven (and eleven for CD8 depleted) days post infection, viral load in tissue homogenates of brain and spleen were determined. Data are presented as mean ± SD (d). e, Intracellular Foxp3+ cells were analyzed via flow cytometry within ZIKV-infected (n=8) and mock (n=4) brains in two separate experiments. f, Frozen sections of the brain, specifically the cerebral cortex, seven days post infection were stained with antibodies against NeuN and ZIKV with nuclei are depicted by 4′,6-diamidino-2-phenylindole (DAPI) stain (blue) (scale bars, 50 μm). (g) ZIKV positive neurons within the cortex and distal spinal cord were quantified in CD8 intact and depleted murine tissues. *P < 0.05; **P < 0.01 ***P < 0.001 (two-tailed unpaired Student’s t-test or log-rank test for survival curve).

Dual depletion of CD4+ and CD8+ T cells resulted in an intermediate phenotype with a significant increase in survival, but not to the same extent as seen within solely CD8 T-depleted mice. Consistent with the intermediate increase in survival, only 1 of the 8 dual-depleted mice developed hindlimb paralysis. Interestingly, all CD4-depleted mice developed hindlimb paralysis, suggestive of a potential regulatory role of CD4+ T cells. Within this vein, we probed for the presence of Foxp3+ regulatory T cells within the CNS. Increased number of Foxp3+ regulatory T cells were found within ZIKV-infected animals compared to mock-infected Ifnar1−/− mice within the brain, while comparable levels of Tregs were found in the spleen (Figure 4e). These data suggested that while CD8 T cells cause neuropathology, the CD4+ Tregs dampen the immunopathology caused by CD8 T cells.

We further investigated ZIKV-infected cell types within the brain and found that the percentage of ZIKV-positive neurons increased drastically within the CD8-depleted animals as compared to the CD8+ T cell intact mice at 7 DPI (Figure 4e). This difference was determined to be quantitatively significant in both the cerebral cortex and the distal spinal cord (Figure 4g). Further we found positive activated caspase 3 staining within the NeuN+ neurons with fragmented nucleus (per DAPI staining) in the ZIKV-infected brains and spinal cords 7 DPI within CD8 T cell intact mice (Supplementary Figure 6b and c, respectively, white arrow). Some of the dying neurons expressed low levels of NeuN (Supplementary Figure 6b, inset, yellow arrow). We also observed activate caspase 3 staining in rare non-neuronal cells (Supplementary Figure 6b and c, grey arrow). In contrast, activated caspase 3 staining was completely absent in the CNS of the CD8-depleted mice at 7 DPI (Supplementary Figure 6a). Lastly, we were able to visualize an immunological synapse of cytolytic cells (perforin positive) with neurons within the brain of ZIKV-infected mice (Supplemental Figure 6d). These data reveal that CD8+ T cell-mediated lysis of ZIKV-infected neurons, while controlling the viral load, leads to hindlimb paralysis.

In summary, our study in Ifnar1−/− deficient mice demonstrates that ZIKV infects circulating leukocytes in order to become blood borne. Once in circulation, ZIKV-infected, bone marrow derived cells facilitate viral spread to various tissues and assume organ-specific virus replication. Within the CNS, Ifnar1-signaling within brain non-hematopoietic cells is able to block ZIKV spread. In the absence of Ifnar1, astrocytes become infected with ZIKV, resulting in a loss of the BBB. BBB breakdown is accompanied by a large influx of infiltrating cells into the CNS. In particular, CD8+ T cells infiltrate into the brain and are able to limit ZIKV replication within neurons. However, this antiviral process comes at the cost of neuropathogenesis leading to hindlimb paralysis.

ZIKV infection of the Ifnar1-deficent mouse allows us to model ZIKV pathogenesis under circumstance of high viral burden within the CNS, where innate resistance is inefficient or impaired. Such conditions might mimic the settings of neonates[16], elderly[17], and immunosuppressed individuals, where innate resistance is suboptimal resulting in overdrive of compensatory immune effector mechanisms. Consistent with previous work[11,15], our findings indicate an antiviral role for CD8 T cells in the ZIKV infection. Yet under conditions of high viral burden within the CNS through use of the Ifnar1−/− mouse model, this protection resulted in lethal immunopathology. A recent postmortem analysis of human neonates with congenitally acquired ZIKV found leukocyte aggregation near blood vessels or perivascular cuffing, as a common phenotype within the brains analyzed[18]. Our data are strikingly similar in that we observed perivascular infiltration of T cells within the brains of the Ifnar1-deficent mice (Figure 2A). Additionally, “red neurons”, a pathological finding of apoptotic neurons within histology slides, were found within congenitally infected neonatal brains indicative of neuronal apoptosis/necrosis[18], a phenotype found within the Ifnar1−/− mice (Supplemental Figure 6). Lastly, work with primary developing human brain tissue, found ZIKV to target astrocytes for viral replication[19], also consistent with our work, these pathology similarities suggest ZIKV infection of the Ifnar1-deficent mouse may serve to model some aspects of congenitally acquired ZIKV infection. However, whether CD8 T cells play a role in neonatal brain immunopathology following congenital ZIKV infection remains to be determined.

Our results suggest a potential involvement of adaptive immune response as a basis for apoptosis of infected neurons, leading to immunopathology in the setting of impaired innate resistance. Our data thereby provide a compelling basis to investigate the role of T cells and BBB breakdown within ZIKV-induced neurological complications that follow ZIKV infection in humans [20].

Methods

Mice

Ifnar−/− mice (# 32045-JAX) and CD45.1 mice (stock # 002014) were purchased from The Jackson Laboratory and C57BL/6 mice (strain code #027) from Charles River. Ifnar−/− and C57BL/6 mice were subsequently bred and housed at Yale University. Mice of both sexes were between 6–8 weeks of age for the initiation of all experiments conducted. No randomization protocol or blinding was used. Sample size was designed to provide statistical significance with a strict minimum of 3 mice per condition. All animal procedures were completed in compliance with approved Yale Institutional Animal Care and Use Committee protocols.

Virus and cells

ZIKV Cambodian FSS13025 strain (World Reference Center for Emerging Viruses and Arboviruses at University of Texas Medical Branch, Galveston) was used for all experiments. ZIKV stocks were grown in C636 cells and titered by plaque assay within Vero cells (ATCC CCL-81) as previously described [21]. C6/36 mosquito cells (Aedes albopictus) (ATCC CRL-1660) were maintained in Dulbecco’s Modified Eagle’s Medium (GIBCO) supplemented with 10% heat-inactivated fetal bovine serum, 1% penicillin/streptomycin (GIBCO) and 1% tryptose phosphate broth (Sigma) at 30 °C. Vero cells were maintained at 37 °C in Dulbecco’s Modified Eagle’s Medium (GIBCO) supplemented with 10% heat-inactivated fetal bovine serum, 1% penicillin/streptomycin (GIBCO). Cell lines were authenticated by morphology and are routinely tested for mycoplasma contamination.

In vivo experiments

Mice were infected with 106 PFU ZIKV via intra footpad injection. For bone marrow chimera studies, mice were irradiated at two doses of 475 cGy (total 950 cGy) and reconstituted with 1 × 106 WT or Ifnar1−/− bone marrow cells. The mice were used for experiments eight weeks after bone marrow transplantation. Chimeras were checked for efficiency of reconstitution through use of congenic bone marrow transfer. Reconstitution was determined to be 91–93% efficient in the Ifnar1−/− → WT mice, while it was 93–95% efficient within the WT → Ifnar1−/− mice. For BBB breakdown assay, at 7 DPI with ZIKV or PBS (mock), Oregon green 488-conjugated dextran (70 kDa) (5 mg ml-1, 200 μl per mouse) was injected intravenously. Forty-five minutes later, these mice were sacrificed for immunohistochemical analysis. For depletion studies, Ifnar−/− mice were injected intravenously with 300 ug of anti-CD4 (GK1.5;BioXCell) and/or anti-CD8 (YPT; BioXCell) antibody at days −1, 4, 9 post ZIKV infection. Depleted mice demonstrated 90–95% efficiency of depletion treatment on day 7 post ZIKV infection. All antibodies for depletion studies were purchased from BioXCell. Mice monitored for clinical illness and were euthanized before reaching the moribund state (onset of hindlimb paralysis (the inability to move hind legs) and/or >20% weight loss) due to humane concerns.

Virus quantification

To determine ZIKV viremia, 50–100ul of blood was collected via retro-orbital bleed at indicated time points post-infection. Total RNA was extracted using Trizol (Thermo Fischer Scientific) extraction followed by purification with RNeasy Mini Kit (QIAGEN). To determine levels of virus in tissues, mice were euthanized and indicated tissues were collected and extracted in Trizol. cDNA was synthesized using iScript cDNA synthesis kit (Bio-Rad) with quantitative PCR being performed using SYBR-Green (Bio-Rad) or PrimeTime master mix (Integrated DNA Technologies) on a CFX Connect instrument (Bio-Rad). ZIKV RNA was detected using primers designed to NS5 [21], with limit of detection determined as previously described [21].

Immunofluorescence

Harvested organs were fixed in 4% paraformaldehyde (PFA) for 24 hours or IHC zinc fixative (BD Pharmingen) for anti-perforin stained tissues. Fixed tissues were then sucrose sedimented in 10%, 20% and 30% steps prior to embedding within optimum cutting temperature (O.C.T.) medium (Tissue Tek). Frozen sections of 5–7 μm thicknesses were cut and left to dry at ambient temperature. These tissues were then stained with primary antibodies: Cleaved Caspase 3 (Cell Signaling Technology), CD31 (Novus Biologicals INC), CD45 (Novus Biologicals INC), glial fibrillary acidic protein (GFAP) (Agilent Technologies INC), rabbit-neuronal marker (NeuN) (Cell Signaling Technology), guinea pig-NeuN (EMD Millipore Corporation), rat anti-perforin (Novus Biologicals INC) and anti-ZIKV rat serum [22] after rehydration with PBS (2X for 10 minutes) and permeabilization with PBST (3X for 10 minutes) at RT overnight. Slides were washed and incubated with appropriate secondary prior to a final wash series and incubation with 4′,6-diamidino-2-phenylindole (DAPI) prior to mounting with SlowFade Diamond Antifade (Fisher Scientific Company LLC). Stained tissues were then analyzed by fluorescence microscopy (BX51; Olympus) or confocal microscopy (TCS SP2; Leica).

Flow Cytometry

Preparation of single-cell suspensions from brain tissues was completed as previously described [23] using the following e-Bioscience or Biolegend antibodies: CD45.2 (104 and 30-F11), CD45.1 (A20), CD4 (RM4-5 and GK1.5), CD8 (53-6.7), CD3 (145-2C11 and 17A2), F4/80 (BM8), CD25 (PC61), GITR (DTA-1), CD127 (A7R34), FoxP3 (FJK-165), and isotype (eBRG1). Aqua Live/dead stain (ThermoFisher) and FoxP3 intracellular staining kit (eBioscience) were utilized. Multiparameter analyses were performed on an LSR II flow cytometer (BD Biosciences) and analyzed using FlowJo software (Tree Star).

Statistical Analysis

Survival curves were analyzed using a log-rank test. Variable comparisons were completed using a two-tailed unpaired Student’s t-test within GraphPad Prism. We used a minimum of 3 animals per condition within any experiment. Data is represented as individual data points ± SD or ± SEM if greater than 7 data points, as stated in figure legends.