Zika virus infection of retinal cells and the developing mouse eye induces host responses that contrasts to the brain and dengue virus infection

Zika virus (ZIKV) and dengue virus (DENV) are closely related mosquito-transmitted, flaviviruses, and both are associated with a febrile illness (Guzman and Harris 2015; Musso and Gubler 2016; White et al. 2016). DENV infection can result in serious outcomes—dengue with warning signs and severe dengue, associated with a vascular leak syndrome which is seen in both adults and children. In contrast, ZIKV infection in adults is largely asymptomatic, with generalised rash, headache, and fever (Halani et al. 2021). ZIKV, however has been recognised as a pathogen with damaging effects on development of the brain and the retina, with a diverse clinical presentation termed congenital zika syndrome (CZS) (Agrawal et al. 2018; de Paula Freitas et al. 2017; Musso and Gubler 2016; Musso et al. 2019).

One commonality between the diseases caused by ZIKV and DENV is eye involvement, although the specifics of this are quite different, with the clinical characteristics of these infections reviewed previously (Merle et al. 2018; Oliver et al. 2019). ZIKV is commonly associated with a conjunctivitis in adults (de Paula Freitas et al. 2017; Halani et al. 2021) or anterior uveitis, both of which are infections and inflammation in the anterior eye. In contrast, DENV has been well described to cause a retinopathy, including retinal vascular disease in the posterior eye (Li et al. 2017; Lim et al. 2004; Teoh et al. 2006). Our laboratory has demonstrated that DENV can infect cell types from the retina (Carr et al. 2017), and DENV can infect the eye in animal models of adult disease (Norbury et al. 2020). ZIKV has been demonstrated in the eye in models of in utero or post-natal mouse development (Li et al. 2021; Zhao et al. 2017), infection in adult mice when the IFN-response is lacking (Garcia et al. 2020; Miner et al. 2016; Singh et al. 2019), and CZS can be studied in a number of animal models in the laboratory (Caine et al. 2018; Narasimhan et al. 2020). In relation to infection of the brain, DENV can cause central nervous system disease, such as encephalitis in the adult, although rarely, (Li et al. 2017) but ZIKV is a well-recognised cause of microcephaly following infection in utero (de Paula Freitas et al. 2017; Musso and Gubler 2016; Musso et al. 2019). In the laboratory, DENV can infect the brain of adult mice and replicate in microglia, neurons, oligodendrocytes, and endothelial cells (Amorim et al. 2019; Velandia-Romero et al. 2012) and can move from the brain to the eye (Norbury et al. 2020), while ZIKV can infect the developing mouse brain (Noguchi et al. 2020) and isolated or organoid cultured neuronal progenitor cells (NPC) (Qian et al. 2017).

Here, we have extended these studies to define ZIKV infection and inflammatory responses in retinal cells in vitro and the developing neonatal mouse eye and brain early in infection and prior to the onset of neurological deficits. Further, viral infection and host responses are compared and contrasted to a laboratory strain of DENV. Results show that both viruses can infect eye cells in vitro and the brain and eye in vivo, with ZIKV replication comparable in both tissues during development. Both conserved and distinct virus, cell-line and tissue specific responses are observed reflecting differences in host antiviral responses, pro-inflammatory and developmental pathways that may be important for the specific ZIKV-induced changes in neuronal and retinal developmental dysfunction in the brain and eye.

DENV and other flaviviruses such as WNV and arboviruses such as CHIKV can infect the adult eye, with DENV associated with a number of different retinal pathologies (Merle et al. 2018; Oliver et al. 2019). DENV can cause infection in the adult brain, and although evidence for an effect of DENV on the central nervous system in patients is growing, DENV is generally not considered a neurotrophic flavivirus (Li et al. 2017). In contrast, ZIKV can cause major pathology in the brain and retinal defects following in utero infection of the developing fetus. Thus, definition of infection and host responses to both DENV and ZIKV in the eye and brain is of importance to help understand how these related viruses cause these different disease presentations. In this study, infection and host responses to ZIKV have been compared in vitro in adult retinal cell lines and in vivo in the developing brain and eye, alongside a laboratory strain of DENV.

It has been previously shown that DENV can infect RPE and exert functional changes (Carr et al. 2017). Additionally, DENV RNA and inflammatory responses can be detected in the mouse eye following either a systemic or intracranial infection (Norbury et al. 2020). Other studies have demonstrated that ZIKV can infect retinal cell types such as the retinal pigment epithelium (RPE), both primary RPE, the cell line ARPE-19, Muller cells and retinal endothelial cells in vitro (Singh et al. 2017; Zhao et al. 2017), and in a human fetal RPE cell line (Garcia et al. 2020) or iPSC-derived RPE (Simonin et al. 2019). The significance of this in particular is proposed to be due to the roles of the latter two cell types in forming a blood retinal barrier (BRB), where a breach of this barrier can allow the entry of ZIKV into the eye (Nelson et al. 2020; Roach and Alcendor 2017; Singh et al. 2017, 2018). In the study herein, ARPE-19 were most susceptible to both ZIKV and DENV, HREC became infected, but at low levels, and high levels of viral RNA with respect to the number of antigen positive cells were produced in Muller cells for both viruses. HREC demonstrated strong induction of type I IFN and inflammatory responses, consistent with our prior infection studies of DENV in human umbilical vein endothelial cells (Calvert et al. 2015) while Muller cells demonstrated a relatively poor type I IFN response. This confirms the literature and suggests that with our DENV and ZIKV strains, multiple cell types of the retina are susceptible to infection with the potential for a highly productive infection in Muller cells, as previously suggested to be a primary target in the retina (Zhao et al. 2017). Furthermore, all these cell types are potential contributors of their own unique inflammatory and antiviral responses that may be linked to restricting viral replication in the eye or promoting retinal inflammatory disease.

Using these same viral strains, infection in a 1-day old mouse demonstrated movement of DENV and ZIKV from systemic site of administration to the brain and eye. DENV infected the brain of all mice, but viral RNA was only detected unilaterally in the eye and not in all mice. The DENV strain used here is a laboratory clone of the New Guinea C DENV isolate, passaged in the mouse brain. This strain can infect the adult brain (Al-Shujairi et al. 2017) and eye following an intracranial challenge (Norbury et al. 2020) and replicate in eye cells in vitro (Carr et al. 2017), and as shown in vitro herein. Thus, this DENV strain has tropism for the eye but does not infect the developing mouse eye well. In contrast, ZIKV infected the brain and both eyes in all animals at day 3 pi with increased viral RNA at day 6 pi, at comparable levels in the brain and eye. From this, we suggest that DENV infection of the developing eye is restricted, either by the immune response or access to the eye itself. Consistent with this, ZIKV but not DENV infects the mouse eye in an intrauterine model of viral challenge (Shi et al. 2018).

ZIKV infection in the eye is supported by a number of studies in mouse models of systemic ZIKV infection in immunodeficient adult mice (Miner et al. 2016) or following direct inoculation into the eye (Singh et al. 2017; Zhao et al. 2017). These models result in high levels of ZIKV replication within choroidal endothelial cells and cells of the retina, identified as retinal pigment epithelial cells and Muller cells, consistent with in vitro studies (Singh et al. 2017; Zhao et al. 2017). ZIKV can induce an ocular pathology in the developing fetus following maternal infection (Mohr et al. 2018), and post-natally in a model of ZIKV infection during pregnancy in Rhesus macaques (Yiu et al. 2020). An alternative ZIKV developmental model is challenge of the 0–1 day post-natal (P0/1) mouse (Li et al. 2021), where mouse development from E15-P10 reflects third trimester gestation in humans (Chen et al. 2017; Workman et al. 2013). In this developing mouse model at P6-7, there is little replication as detected by direct staining of the eye or morphological change (Li et al. 2021). By P14-21 days of infection, there is significant morphological effect on brain and eye, with ZIKV causing apoptosis and destruction of the retinal ganglion layer, with progressive inflammation, loss of neurons, and vascular damage in the retina (Li et al. 2021). This is accompanied by neurological deficits and hind limb paralysis suggesting, similarly, a significant impact on the brain. The study here has analysed a similar model of infection at P0-1 and analysis at P6/7, where the developing pups are still healthy with no measurable clinical score or neurological defect. The retina is still immature, eyes are closed, the retinal epithelium is not yet pigmented, but there is detectable and increasing ZIKV-infection and a substantial inflammatory response. Importantly, this demonstrates responses and changes in the retina concurrent with responses in the brain and prior to major neurological deficits, suggesting the impact on the retina is not secondary to destruction of the brain. Although a prior study described little effect of ZIKV-infection on the retina at P6/7 (Li et al. 2021), in the study here, a clear retinal defect was observed with altered formation of OPL, INL and IPL layers, and morphological differences in the cells of the INL. Generation of the OPL has been staged in humans (Prameela Bharathan et al. 2021) and is important for development of the retina, where the OPL and IPL are regions for synaptic interactions of cells. Results here are thus consistent with an early defect in synaptic development in the retina of ZIKV-infected mice and would be expected to have significant impact on later vision outcomes. This has also been observed in the intrauterine ZIKV infection model, where by P7 the retina of ZIKV-infected mice is thinned, with loss of the OPL and starburst amacrine cells that are important for retinal circuitry (Shi et al. 2018).

Here, our study has also assessed ZIKV and DENV infection in the brain with results showing comparable replication for ZIKV and DENV and host responses detected with a NanoString inflammatory panel, segregating clearly from those of mock-infected mice. There were numerous conserved responses induced by ZIKV and DENV, with inflammatory and antiviral responses, as expected. Interestingly, CXCL9 and CXCL10 were the highest induced mRNA following ZIKV infection and was also induced but to a lesser degree by DENV. Since CXCL9 and CXCL10 have an important role in T-cell recruitment and favour a Th1 response, this suggests that ZIKV induces a stronger stimulus for a cellular recruitment to the brain. In contrast, CfD was the highest mRNA induced by DENV, suggesting a more local acting, complement driven response to infection. Both DENV and ZIKV induced a major upregulation of Myl2, a gene typically associated with cardiac growth, previously reported to be upregulated in ZIKV-infected myeloblasts (Riederer et al. 2022), and the relationship of this to non-muscle myosin and growth processes in the brain remains to be defined. Markers such as CD40, CD86, and CCL5 were increased, while induction of IFN-γ was not detected by NanoString, and CD4 and CD8 mRNA could not be detected by RT-PCR (data is not shown). This is suggestive of APC activation and monocyte responses. In contrast in the adult brain, infection with DENV induced CD8 + T-cell infiltration (Al-Shujairi et al. 2017), and the lack of T-cell responses here in the neonate is consistent with altered functions of neonatal T-cells (Rudd 2020). Interestingly, results show induction of components of the complement system that were common to both DENV and ZIKV in the brain such as CfB and C3, but C2 and C4a—components of the classical and lectin pathways, were only induced by ZIKV infection. The lack of induction of MASP1 and MASP2 suggests that the lectin pathway is not induced and that this increase in C2 and C4a reflects an increase in the classical pathway by ZIKV infection. Importantly, C1q, the starting substrate for the classical pathway, and C3b, the cleaved form of C3 following complement terminal pathway activation, both have roles in neuronal development and synaptic pruning in the brain (Warwick et al. 2021). C1q and C3 upregulation has also been described during ZIKV infection of the adult mouse brain and proposed to influence synapse formation (Figueiredo et al. 2019). Similarly, here, the increase in C1q and C3 combined with C2 and C4a following ZIKV infection may result in increased C1q cleavage and formation of excessive C3b and hence have a greater impact on neuronal development than DENV, which does not induce C2 and C4a and thus may maintain more C1q and less C3b. Only a few mRNA’s were downregulated, but notably, these included Rac1/RhoA, which are GTPases with known roles in maintaining a neural progenitor cell pool and neuronal development, that we have previously proposed may be relevant to ZIKV infection in the developing brain (Norbury et al. 2022).

Consistent with our observations of viral RNA in the eye, NanoString analysis demonstrated major responses to ZIKV but only induction of a small group of mRNAs for key antiviral factors in the eyes of DENV-infected mice. This supports our suggestion above, that DENV-infection induces an effective antiviral response that restricts infection in the eye. Notably, although the mRNA’s downregulated in the brain was altered by less than 1-log fold, in the eye, DENV and ZIKV induced an approximately 3-log fold decrease in Arg1 and Retnla, respectively. Both Arg1 and Retnla are influenced by Th2 cytokines and can regulate Th2 responses. Retnla-/-mice have increased Th2 responses, and conversely, Retnla overexpressing transgenic mice have lower Th2 responses (Lee et al. 2014; Nair et al. 2009). Arg1 is increased and associated with a Th2 environment (Bronte et al. 2003; Muraille et al. 2014). This suggests that the Th1/Th2 environment may be different with DENV downregulating Arg1 and thus decreased Th2 environment but ZIKV downregulating Retnla and thus promoting a Th2 environment. Consistent with a less inflammatory environment in the eye, ZIKV induced various C1 complement components in eye but did not induce key downstream activators of the AP such as CfD or the substrate for the terminal pathway, C3. In addition to these unique inflammatory profiles induced by ZIKV in the eye in comparison to the brain, ZIKV also induced CSF-1 and Ddit3, (also known as CHOP). CSF-1 is a growth factor responsible for supporting microglia in the brain (Elmore et al. 2014), which is produced by activated microglia and important in the retina for promoting photoreceptor survival (Jones and Ricardo 2013). Use of CSF1R blockers can deplete microglia and prevent microglial driven inflammatory damage in the eye (Kokona et al. 2018; Okunuki et al. 2019; Tang et al. 2020; Todd et al. 2019), and hence, it is unclear if an increase in CSF-1 would benefit the maintenance of photoreceptors or support a damaging inflammatory response to ZIKV infection driven by microglia, as we have observed, and can occur in vitro in Mueller cells. Ddit3 has been shown in the adult to contribute to retinal ganglion cell death (Wang et al. 2021) and consistent with the increase in Ddit3 seen here, and the pathology seen at later stages of ZIKV-infection in a similar model is loss of the RGC layer of the retina (Li et al. 2021; Shi et al. 2018). These changes in mRNA levels are seen in our study in the absence of major morphological change in the retina, such as RGC layer loss, and thus may be preceding triggers for this damage. Importantly, our study has observed a difference in the formation of the layers of the retina (INL and OPL) that form areas of synaptic interactions of retinal cells, which similar to the above discussions regarding the brain, could be driven by the complement system and the C1q and C3b roles in neurogenesis and synapse formation.

In conclusion, DENV and ZIKV both have capacity to infect cells of the adult eye in vitro, but ZIKV has a greater propensity to infect the developing mouse eye than DENV. These viruses induce distinct antiviral and inflammatory responses. ZIKV infects both the developing brain and eye at comparable levels and drives a retinal pathology and hosts responses including effects on factors, such as the complement system, that may influence brain and/or retinal development. Defining these responses and ways to dampen developmental impact without loss of control of viral replication may be of future benefit to lessen the burden of CZS and comparison to DENV, which infects the developing eye poorly, may help define responses that prevent infection of the developing eye.