Oligodendrocytes are susceptible to Zika virus infection in a mouse model of perinatal exposure: Implications for CNS complications

Abstract Some children with proven intrauterine Zika virus (ZIKV) infection who were born asymptomatic subsequently manifested neurodevelopmental delays, pointing to impairment of development perinatally and postnatally. To model this, we infected postnatal day (P) 5–6 (equivalent to the perinatal period in humans) susceptible mice with a mammalian cell‐propagated ZIKV clinical isolate from the Brazilian outbreak in 2015. All infected mice appeared normal up to 4 days post‐intraperitoneal inoculation (dpi), but rapidly developed severe clinical signs at 5–6 dpi. All nervous tissue examined at 5/6 dpi appeared grossly normal. However, anti‐ZIKV positive cells were observed in the optic nerve, brain, and spinal cord; predominantly in white matter. Co‐labeling with cell type specific markers demonstrated oligodendrocytes and astrocytes support productive infection. Rarely, ZIKV positive neurons were observed. In spinal cord white matter, which we examined in detail, apoptotic cells were evident; the density of oligodendrocytes was significantly reduced; and there was localized microglial reactivity including expression of the NLRP3 inflammasome. Together, our observations demonstrate that a clinically relevant ZIKV isolate can directly impact oligodendrocytes. As primary oligodendrocyte cell death can lead later to secondary autoimmune demyelination, our observations may help explain neurodevelopmental delays in infants appearing asymptomatic at birth and commend lifetime surveillance.

The CNS is susceptible to congenital infection through vertical transmission during the entire gestational period (Brasil et al., 2016); however, infection in the earlier weeks of the embryonic phase generally results in more severe malformations (Chimelli et al., 2017 andreviewed in Saad et al., 2018). This likely reflects the susceptibility of neural progenitor cells (NPCs), resulting in changes in gene expression, impaired proliferation and migration, and apoptotic cell death (Cugola et al., 2016;Garcez et al., 2017;Li et al., 2016;Souza et al., 2016;Tang et al., 2016). Nonetheless, CNS abnormalities have been reported following infection as late as 39 weeks of gestation, demonstrating that later developmental processes are also susceptible (Brasil et al., 2016).
To determine if oligodendrocytes are susceptible to ZIKV during myelination in vivo, we infected P5/6 mice with ZIKV, some days after CNS myelination commences at P1. We mainly used mice lacking the type I interferon receptor (Ifnar1 knockout mice), which recapitulate aspects of human ZIKV infections and disease (Miner & Diamond, 2017), to approximate ZIKV antagonism of the type I interferon (IFN) response in humans (Serman & Gack, 2019). Immunohistochemistry and cell quantification revealed that oligodendrocytes were particularly vulnerable. The functional outcome could not be determined, as mice had to be euthanized due to rapid development of severe clinical disease. However, in other contexts, oligodendrocyte death is followed after some delay, by loss of compact myelin (Pohl et al., 2011;Traka et al., 2010). Further, our data may explain neurodevelopmental delays in some congenitally infected infants and warn of susceptibility to later autoimmune mediated demyelination (Traka, Podojil, McCarthy, Miller, & Popko, 2016).

| Mice
Ifnar1 knockout (KO; type I interferon receptor deficient) and wild type (WT) mice of both sexes, on a 129S7/SvEvBrdBkl-Hprtb-m2 background (B&K Universal) were maintained in Tecniplast 1284L Blue line IVC cages, in a 12 hr light/dark cycle and provided with sterile food and water ad libitum. All animal studies were approved by the Ethical Committee of the University of Glasgow and licensed by the UK Home Office (Project Licence numbers PPL P78DD624O and P9722FD8E).
Genomic DNA was extracted from ear biopsies using a modified protocol (Truett et al., 2000). Briefly, ear notches were heated to 95 C for 90 min in 50 mM NaOH. Following neutralization with 10% v/v 1 M Tris pH 5, the resultant solution was vortexed to release DNA and 2 μl was used for PCR as described previously (Cumberworth et al., 2017).

| Infection of mice with Zika virus
At P5/6, pups of both sexes were removed from the dams and administered an intraperitoneal (ip) injection with 7.5 × 10 4 to 7.5 × 10 5 (Table S1) plaque forming units (PFU) of ZIKV per animal, or with an equivalent volume of vehicle only (cell culture media), using a 1 ml syringe or 5 μl Hamilton syringe. Each litter received both virus and vehicle, providing littermate controls for virally infected animals. Pups were randomly assigned to each of the two groups. Six Ifnar1 knockout litters and two wild type litters were used (Table S1).

| Examination and assessment of mice
Mice were examined twice daily following inoculation. Their behavior was assessed visually by one or other of two experienced observers while the pups were in the cage together with the dam, or out of the cage without the dam. Videos of the mice were examined by TJA (veterinary neurologist) who provided clinical descriptions. Pups were humanely killed at timepoints indicated in Table 1, usually being when clinical signs were observed, or as age-matched controls to sick animals.
Most pups were immersion fixed in approximately 100 ml 8% paraformaldehyde (PFA) in phosphate buffered saline (PBS) for 2-3 days before dissection. To facilitate impregnation with fixative, pups were decapitated, and a vertical incision was made in the skin on the back and front of the body and the skin pared away from the muscles. Two litters of Ifnar1 knockout mice were perfusion fixed in 8% paraformaldehyde or 4% paraformaldehyde and 5% glutaraldehyde (the latter for electron microscopy) at 4 dpi, as described (Edgar, Smith, & Duncan, 2020).

| Tissue preparation and immunohistochemical staining
The brain, spinal cord, and optic nerve were dissected and placed for further fixation in 4% PFA in PBS overnight. The tissue was then transferred into 20% sucrose in PBS until it sank (usually overnight), embedded in Tissue-Tek OCT medium (Sakura Europe), and rapidly frozen in liquid nitrogen chilled isopentane. Optic nerves were frozen flat between discs of frozen OCT (approximately −23 C). Transverse sections of forebrain and cervical and lumbar spinal cord, and longitudinal sections of optic nerves, were cut at 12-μm thickness and collected on plus charged slides (VWR or Waldemar Knittel) at 96-μm intervals. Sections were stored in a −20 C freezer until required.
Slides were allowed to reach room temperature before they were removed from their storage box, then washed in PBS to remove OCT.

| Microscopy and cell quantification
To quantify cells across spinal cord transverse sections, images of immunostained nervous tissue were captured, as illustrated in T A B L E 1 Clinical signs in wild type (WT) and Ifnar1 −/− mice after infection with mammalian cell-propagated ZIKV PE243 or ZIKV MR766 (i.p.) Figure 5, at ×40 magnification using an Olympus IX70 microscope with standard epifluorescence optics and Image Pro Plus 6 software.
Cell counts were made in areas of interest (AOI) of 44,835 μm 2 (celltype specific marker and anti-ZIKV envelope protein) or 5,400 μm 2 (DAPI +ve nuclei). The experimenter was blinded to the conditions and the field of view was selected in the blue channel (DAPI) to avoid biased selection. One to four sections per spinal cord, at least 50-μm apart, were analyzed (2-10 images/section). To quantify the proportion of ZIKV +ve cells that were CC1 +ve, or cleaved caspase 3 +ve cells that were ZIKV and/or CC1 +ve, images were captured in four channels at sites where ZIKV +ve cells were present, at ×20 magnifi-

| Quantification of myelin and axon volumes
Fluorescence images of myelin basic protein (MBP; labeled in green) and phosphorylated H-neurofilament and M-neurofilament (NF; labeled in red) were captured of the cervical and lumbar cord white matter (input images). These were transformed in Cell Profiler (Jones et al., 2008) to binary images (output images) and the total pixels per AOI, red pixels per AOI and green pixels per AOI were quantified to provide a readout of the relative volumes occupied by myelin or axons; being red or green pixels as a percentage of all pixels, as described previously (Bijland et al., 2019).

| Resin sections and electron microscopy
At 4 dpi, mice were rapidly perfusion fixed in 4% paraformaldehyde, 5% glutaraldehyde in cacodylate buffer and tissue was processed, stained, and imaged as described (Edgar et al., 2020). infected with mammalian cell-propagated ZIKV MR766, a mouse brain-passaged isolate (Dick, 1952) from a sentinel monkey in the Zika Forest (Table S1).

| Statistical analysis
Ifnar1 knockout mice infected with ZIKV PE243 developed clinical signs and died or had to be euthanized at 5 or 6 dpi. Signs included wide stance and failure to bear body weight, flaccid tail, urinary retention, weakness or paralysis of hindlimbs, or failure of righting ( Figure 1).

| The CNS appears grossly normal in clinically affected ZIKV-infected animals
To determine whether CNS changes might contribute to clinical signs, we used immunohistochemical staining of nervous tissue of clinically affected mice (5 or 6 dpi) to examine localization and frequency of infected cells and overall tissue integrity. As at 4 dpi, ZIKV +ve cells were distributed heterogeneously throughout the nervous system, particularly as clusters in white matter of the spinal cord (Figure 3a,b and Figure S1) and cerebellum ( Figure S1). Due to these observations and because clinical signs (urinary retention and hindlimb weakness) suggested spinal cord involvement, we examined the cervical and lumbar cord in more detail. Grossly, the overall appearance of the spinal cord at 5/6 dpi, including axons, myelin and astrocytes was comparable to that of mock-infected animals (Figure 3a,b). Furthermore, quan-

| Zika virus-infection leads to cell death
Although the CNS appeared grossly normal despite the severe clinical signs, we next asked whether ZIKV-infection caused localized cell death. At 4 dpi, when mice appeared clinically unremarkable, we found cleaved caspase 3 +ve cells localized to regions containing ZIKV +ve cells (magenta arrowheads Figure 4a). To determine if the dying cells were oligodendrocytes, which are highly abundant in white matter, we co-labeled tissue with antibody clone CC1 [a marker that labels oligodendrocytes but not oligodendrocyte progenitor cells

| Oligodendrocytes are particularly vulnerable following Zika virus infection
As shown in Figures 2-4, most ZIKV +ve cells appeared to be in white matter. We confirmed this by quantifying the proportion of ZIKV +ve cells in cervical and lumbar cord white or gray matter (Table 2). Next, we used co-labeling with anti-ZIKV and cell type specific markers to determine which cell types support productive ZIKV infection. We Single Z plane confocal images from dorsal horn or ventrolateral columns of two more animals. Cells positive for cleaved caspase 3 +ve alone (magenta arrowheads) were observed. Some cleaved caspase 3 +ve cells co-labeled with antibody CC1 (white arrowheads), which labels mature oligodendrocytes, and some cells were co-labeled with anti-cleaved caspase 3, CC1, and anti-ZIKV (yellow arrow). Some ZIKV +ve cells did not co-label with either cleaved caspase 3 or CC1 (green arrowheads). Pyknotic or karryhexic nuclei (white arrows) were also observed in close proximity to ZIKV +ve cells.

| Perinatal Zika virus infection causes mild neuroinflammation
At the point of euthanasia, our mice showed no gross pathological changes that would explain the neurological signs. We therefore asked if these signs might reflect a neuroinflammatory process. Using antibody to IBA1, we demonstrated a propensity for microglia to cluster around ZIKV +ve cells (Figure 6a and Figure S2), however overall microglial densities in white matter were similar to control (Figures 5g,   h and 6a). Using anti-CD68 (a marker of activated microglia/macrophages) and anti-CD3 (a T lymphocytes marker), we observed a localized activation of microglial/macrophages (Figure 6b), but only very rare T cells in the CNS parenchyma ( Figure 6c). As the NLRP3 inflammasome was recently implicated in white matter injury in the perinatal and neonatal period in children (Holloway et al., 2021), we co-stained spinal cord and optic nerve sections with anti-NLRP3. We found NLRP3 +ve cells mainly in association with ZIKV +ve cells ( Figure 6d). NLRP3 is likely expressed by microglia, however, as antibodies to NLRP3 and IBA1 are raised in the same species, we were unable to co-stain. Comparing Figure 6a, (Familiar et al., 2020;Mulkey et al., 2020;Peçanha et al., 2020;Pimentel et al., 2021), is obscure. Here, using a mouse model of perinatal infection, we confirmed our earlier observations in cell culture (Cumberworth et al., 2017;Schultz et al., 2021), that postmitotic neurons are rather refractory to infection, whilst newly generated CNS glia are considerably more susceptible, leading to a reduction in the density of oligodendrocytes. Our data might help predict pathological changes in humans infected perinatally and explain the emergence of neurodevelopmental delays postnatally.
In our mouse model, we found most ZIKV +ve cells in the CNS were located in white matter, and the majority co-labeled with antibody CC1, which labels Quaking (QKI) 7, an RNA-binding protein highly upregulated in myelinating oligodendrocytes (Bin et al., 2016).
We found significantly reduced densities of CC1 +ve cells in white matter and confirmed by electron microscopy that oligodendrocytes were dying. However, by immunostaining, only approximately 16% of cleaved caspase 3+ ve cells, were positive for CC1. We speculate this reflects that QKI7 is downregulated in apoptotic oligodendrocytes.
Nonetheless, it is likely that some of the cleaved caspase 3 +ve/CC1 −ve cells are OPCs, although we were unable to confirm this as the fixation required to inactive ZIKV (8% paraformaldehyde) is not compatible with tissue staining with antibody to NG2, the classical OPC marker. Certainly, we previously showed in cell culture that OPCs are vulnerable to ZIKV infection (Figure 3b, Cumberworth et al., 2017). In the current study we showed that NLRP3 is expressed in proximity to ZIKV +ve cells. Recently, Holloway et al. (2021) showed that microglial activation of the NLRP3 inflammasome drives developmental hypomyelination through dysregulation of Activin A signaling, which promotes developmental myelination (Goebbels et al., 2017;Miron et al., 2013). Consequently, both cell death and impaired signaling are likely to impact developmental myelination following ZIKV infection.
Oligodendrocyte death is followed after some delay by loss of compact myelin (Pohl et al., 2011;Traka et al., 2010). However, the regenerative properties of CNS myelin are well known (Franklin & Ffrench-Constant, 2017) and impaired developmental myelination can recover in both humans (Yan et al., 2019 andreviewed in Malik, Muthusamy, Mankad, Shroff, &Sudhakar, 2020), and animal models (Câmara et al., 2009;Yool et al., 2001). Consequently, if the vascular and neuronal "scaffoldings" are intact, and microglia downregulate the NLRP3 inflammasome, then ZIKV-associated oligodendrocyte injury might represent a temporary pathology. Certainly, it is one that might be difficult to detect by clinical examination in new-borns.
Indeed, MRI observations in clinically unremarkable congenitally infected infants revealed high signal in T 2 weighted images of the white matter (Brasil et al., 2016), potentially reflecting otherwise silent myelin changes. Further, in a study of infected mothers in Rio de Janeiro, 29% of pregnancies with third trimester infection had adverse outcomes including dysphagia, clonus, hyperreflexia, hypertonicity, and irritability (Table S2, Brasil et al., 2016), suggesting potential white matter involvement, as in the hypomyelinating leukodystrophies (Adang et al., 2017). Thus, myelin changes should not be ruled out in apparently asymptomatic new-borns.
Earlier experimental studies also provided support for myelin involvement, including reports of ZIKV-related disruption to OPC development and myelin deposition in mouse models of direct CNS ZIKV inoculation, embryonically (E15.5) or postnatally (P0; Zhang et al., 2017;Li et al., 2018). In contrast to the current study, these studies involved ZIKV inoculation at timepoints prior to the differenti-  et al., 2016). As in the current study, these non-human primates were killed before the long-term consequences were known and it will be important for future studies to address later outcomes, within the limitations of animal welfare issues.
Whilst myelin can be restored following dysmyelination and/or demyelination (Franklin & ffrench-Constant, 2017), it has been shown using an experimental mouse model of primary oligodendrocyte cell death, that demyelination and repair are followed months later by fatal secondary disease characterized by extensive myelin and axonal loss (Traka et al., 2016). These data demonstrate that primary oligodendrocyte death is sufficient to trigger an adaptive autoimmune response against myelin (Traka et al., 2016)