Fast type I interferon response protects astrocytes from flavivirus infection and virus-induced cytopathic effects

C57BL/6 (wild-type (WT)) mice and IFNAR^−/− mice on C57BL/6 background were bred at Umeå Transgene Facility.

The mice were sacrificed between postnatal day 1 and 4 for astrocyte isolation. Cerebral cortices were isolated, and cells were seeded in poly-d-lysine (Sigma) coated T75 tissue culture flasks as previously described [40]. Monolayers of astrocytes were shaken at 200 rpm for 1 h to remove microglia and oligodendrocyte precursors before seeding for experiments. Primary cortical neurons were derived from cerebral cortices of embryonic day 17 mice. The cortices were isolated, and cells were seeded in Dulbecco’s modified Eagle’s medium (DMEM, Sigma D5648-10L) containing 10 % heat-inactivated foetal bovine serum (FBS, Gibco) and 0.1 U/mL penicillin and 0.1 μg/mL streptomycin (Gibco) in poly-d-lysine (Sigma) coated 96-well plates as previously described [41]. After 3 h, the medium was replaced with Neurobasal medium (Gibco) containing B27 (Gibco), 0.1 U/mL penicillin, 0.1 μg/mL streptomycin, and 2 mM L-glutamine (Gibco). The neurons were infected at day 7 post seeding.

VeroB4 cells were cultured in medium 199/EBSS (HyClone) containing 10 % FBS, 0.1 U/mL penicillin, and 0.1 μg/mL streptomycin (Gibco). Mouse embryonic fibroblasts (MEFs) were grown supplemented with 10 % FBS, 50 μM β-mercaptoethanol (AppliChem), and 2 μg/mL tetracycline DMEM (Sigma). TBEV strains Hypr 71 (isolated in 1953 from blood of a patient in the Czech Republic), Aina (isolated in 1963 in Irkutsk from the blood of a patient), and Sofjin (isolated in 1937 from patient in Russia and showed 99 % sequence identity to strains Sofjin-Chumakov and SofjinKSY with BLAST) were a kind gift of G. Dobler (the Bundeswehr Institute of Microbiology, Munich, Germany). JEV (Nakayama strain) was a kind gift of S. Vene (Folkhälsoinstitutet, Stockholm, Sweden). WNV (isolated in 2003 in Israel WNV_0304h_ISR00, passage number 5) is a kind gift of S. Vene. Stocks were generated in VeroB4 cells. Vesicular stomatitis virus (VSV)-eGFP was propagated in VeroB4 cells. ZIKV MR 766 (isolated in 1947 in Zika Forest, Uganda) was propagated in VeroB4 cells and originally provided by Robert Shope (Yale Arbovirus Research Unit, New Haven, CT, USA) as a reference strain to Jürgen Pilaski (University of Düsseldorf, Germany), who kindly transferred the strain to the Bundeswehr Institute of Microbiology, Munich, Germany. TBEV strains, JEV, and ZIKV are cell culture-adapted reference strains with an unknown passage history. Cell monolayers were infected with virus for 1 h at 37 °C. The virus inoculum was then removed and replaced with DMEM supplemented with 2 % FBS, 0.1 U/mL penicillin, and 0.1 μg/mL streptomycin. Viral titers were determined by focus forming assay as previously described [42].

Total RNA was isolated at the indicated time points using Nucleospin RNA II kit (Macherey-Nagel), and cDNA was synthesized from 200 to 600 ng RNA as previously described [42]. mRNA expression of GAPDH, IFNβ, IFNα2, viperin and tripartite motif 79α (TRIM79α) were detected by QuantiTect primer assay (Qiagen) and the KAPA SYBR FAST qPCR kit (KAPA Biosystems) using a StepOnePlus fast real-time PCR system (Applied Biosystems). TBEV RNA was quantified using primers previously described [43] and the KAPA PROBE FAST qPCR kit (KAPA Biosystems). Gene expression was normalized to the endogenous GAPDH expression.

Cells were grown on 96-well plates (Greiner CELLSTAR®⁾, fixed in 4 % formaldehyde and permeabilized in PBS containing 0.5 % Triton X-100 and 20 mM glycine. The cells were stained with primary antibodies; a mouse monoclonal TBEV anti-E antibody (Hypr, 1493 1:1000 [44], Sofjin and Aina, 1786 1:1000 [44]), flavivirus anti-E antibody (JEV, WNV, and ZIKV, HB112 1:1000, ATCC D1-4G2-4-15 [45]), and rabbit polyclonal anti-GFAP antibody (1:1000 Abcam, ab7260 [46]). Secondary antibodies were as follows (Thermo Fisher Scientific); donkey anti-rabbit Alexa Fluor 488 (A21206), donkey anti-mouse Alexa Fluor 555 (A31570), and goat anti-mouse Alexa Fluor 647 (A21236) which were diluted 1:500. Nuclear counterstaining was performed using DAPI (Life Technologies, D1306) 1 μg/mL. Viral spread was quantified using a TROPHOS Plate RUNNER HD® (TROPHOS SA, Marseille, France). Images of immunofluorescence staining were acquired under a Zeiss Axiovert 25 microscope using an infinity3 luminera® camera.

Supernatants from virus-infected astrocytes were inactivated using β-propiolactone (β-PL) (Acros Organics, 269040050), diluted in water to 0.96 % and used at a final concentration of 0.05 %, incubated at 4 °C for 16 h followed by 2 h incubation at 37 °C for hydrolysis, performed in plates (VWR, 734-2325) to avoid acidification of the samples [47, 48]. The cells (MEFs, astrocytes, or neurons) were seeded in 96-well plates (Greiner CELLSTAR®) and treated with β-PL-inactivated supernatants or serially diluted IFNαB/D [49] (a kind gift of Peter Stäheli, Virology, University of Freiburg, Freiburg, Germany). The cells were infected 24 h post treatment with VSV-eGFP (multiplicity of infection (MOI) 0.01) (MEFs) or TBEV (astrocytes or neurons) and fixed with 4 % formaldehyde at 16 h post infection (hpi). The number of infected cells was determined using a TROPHOS Plate RUNNER HD®, and the IFN levels were quantified using a standard curve (calculated from 1, 5, 10, 50, 100, and 500 U of IFNαB/D).

The cells were lyzed, proteins were separated, and Western blot was performed as previously described [50]. The primary antibodies directed against TBEV E (1493), actin (rabbit polyclonal anti-actin, 1:2500, Sigma [50]), and viperin (mouse monoclonal anti-viperin, 1:500 Abcam [50]) were used.

Primary WT and IFNAR^−/− astrocytes were seeded at 200,000 cells per 12-well. Upon reaching confluence, WT cells were either stimulated with β-PL-inactivated supernatant from WT astrocytes (24 hpi), 5000 U IFNαB/D [49], or left untreated. Six hours after stimulation, total RNA was isolated using Nucleospin RNA II kit (Macherey-Nagel) according to the manufacturers’ instructions. One thousand to three thousand nanograms of RNA was sent to GATC Biotech (Konstanz, Germany) for RNAseq (InView^TM Transcriptome Explore).

For analyses of the RNASeq data, all treatments were normalized to their respective control and gene names were converted to human HGNC gene names before bioinformatical analyses were performed. Ingenuity Pathway Analysis (IPA; Ingenuity Systems, http://​www.​ingenuity.​com/​) was used to analyze activated pathways and predicted upstream activators as previously described [51, 52]. In brief, activation of Z values were used to determine activation of pathway or regulators (−2 ≥ Z, significant inhibition; 2 ≤ Z, significant activation), p values were calculated using right-tailed Fisher’s exact representing the significance for the overlap between dataset and pathway or regulator. Gene set enrichment analysis (GSEA; http://​www.​broadinstitute.​org/​gsea) was performed to evaluate functional gene sets that are associated with each of the treatments compared to controls. The fold-change regulation of the genes after treatment was used to perform pre-ranked GSEA. The BIOCARTA, KEGG, and REACTOME databases were queried; p values were estimated by 100 permutations and a null-distribution of the enrichment scores. P values were adjusted using the Benjamini-Hochberg method. A normalized enrichment score (NES) was computed by mean normalization [53]. The Connectivity Map database (http://​www.​lincscloud.​org/​; [54]) was used to analyze upstream regulators based on regulated genes after IFNαB/D or supernatant treatment. The top 50 expressed genes and top 50 downregulated genes were selected from, each treatment, per analysis, and a connectivity score >90 or <−90 was considered significant.

Cells were incubated with DMEM, 10 % heat-inactivated FBS, and 2 mM glutamine with MAR1-5A3 antibody (Affymetrix eBioscience, 16-5945-85 [19]) or IgG1 κ isotype control (eBioscience, 14-4714-85) at a concentration of 10 μg/mL for 2 h at 37 °C. After 2 h, the antibodies were removed, and cells were infected with TBEV for 1 h. Following removal of the viral infection inoculum, the medium containing 10 μg/mL antibody was returned to the cells.

Three hours before the indicated time point, the cells were treated with 40 μM resazurin; after 3 h of incubation, fluorescent signal was quantified using a plate reader (Paradigm, Beckman Coulter).

Data from quantitative reverse transcription PCR (qPCR), bioassay, focus forming assays, and viral spread assays were analyzed with unpaired t test using GraphPad Prism software. Statistical analyses of RNASeq data were performed by GATC Biotech using Cufflinks [55, 56].

Astrocytes are the most abundant glial cell type in the brain and an important source of type I IFN during various neurotropic viral infections [18, 22, 23]. To investigate the importance of type I IFNs for astrocyte function in neurotropic flavivirus infection, primary astrocytes were isolated from wild-type (WT) and IFNAR^−/− mice. These were infected with TBEV strain Hypr at a MOI of 0.1 and analyzed by immunofluorescence (Fig. 1a). Both WT and IFNAR^−/− infected astrocytes expressed the characteristic marker glial fibrillary acidic protein (GFAP) and were found positive for TBEV E (Fig. 1a). Viral replication was analyzed over time, and low viral titers and replication were observed in WT astrocytes. In contrast, TBEV titers and RNA replication increased dramatically in IFNAR^−/− astrocytes over time (Fig. 1b, c). Infection was also analyzed in MEFs where an unrestricted growth profile was observed in both WT and IFNAR^−/− cells (Fig. 1d, e). Our results indicate that IFNAR restricts TBEV replication in a cell type-dependent manner.

The type I IFN response could limit viral replication by different mechanisms: (i) restricting the spread of progeny viruses to neighboring cells, (ii) reducing overall viral replication, or (iii) both. To investigate this, WT and IFNAR^−/− astrocytes and MEFs were infected by TBEV at a MOI of 0.1, stained by immunofluorescence, and the numbers of infected cells were quantified (Fig. 2a shows a representative picture of infected wells). Even though the initial infections (24 hpi) in WT and IFNAR^−/− astrocytes were comparable, the outcome of infection differed dramatically. Infection of WT astrocytes was low and never exceeded 20 %; however, infection of IFNAR^−/− astrocytes was significantly higher at the late time points, reaching an average of 69 % at 72 hpi (Fig. 2b). In MEFs, IFNAR expression failed to restrict TBEV spread, and no difference were observed between WT and IFNAR^−/− cells (data not shown). These results suggest that IFNAR restricts the spread of TBEV in astrocytes.

TBEV can be categorized into European, Siberian, and Far Eastern subtypes according to phylogenetic differences. The Siberian and Far Eastern subtypes have been associated with more severe disease compared to the European subtype [57, 58]. Increased pathogenicity could be due to several factors, e.g., ability to interfere with the type I IFN response. Therefore, viral growth of two reference strains; the Siberian Aina and Far Eastern Sofjin, was analyzed in WT and IFNAR^−/− astrocytes (Fig. 2c, d). Similar to the European subtype of TBEV, an interferon response also restricted the spread of Siberian and Far Eastern TBEV subtypes (Fig. 2c, d). Although the type I IFN response plays an important role in mosquito-borne neurotropic flavivirus (WNV, JEV, and ZIKV) infections in vivo [12‐14, 59, 60], the specific role of IFNAR signaling in restricting viral growth in astrocytes is unknown. WT and IFNAR^−/− astrocytes were infected with JEV, WNV, and ZIKV, and the numbers of infected cells were determined (Fig. 2e–g). For all three mosquito-borne viruses, the IFN response was required for the restriction of viral spread in astrocytes. Our data shows that the type I IFN response in astrocytes efficiently restricts the spread of both tick- and mosquito-borne neurotropic flaviviruses.

Previous studies have indicated the importance of astrocyte-produced IFN in various viral infections [18, 22, 23]. The observed differences in the impact of IFNAR on astrocytes and MEFs could either be explained by the ability to respond to, or the capacity to induce, IFN following infection. To determine the impact of the type I IFN response in the overall antiviral response, astrocytes and MEFs were pretreated with IFNαB/D [49] Sixteen hours before TBEV infection (Fig. 3a, b). TBEV growth was inhibited to a similar extent showing that both cell-types were able to mount a strong antiviral response upon IFN treatment. Thus, the ability to respond to IFN (Fig. 3a, b) could not explain the lack of control of viral replication in IFNAR^−/− MEFs (Fig. 1d, e). To test whether astrocytes and MEFs differ in their capability to induce type I IFNs after infection, IFNβ and IFNα2 mRNAs were quantified by real-time qPCR (Fig. 3c, d). Interestingly, the IFN responses in the two cell types were quite different. Whereas astrocytes induced IFN mRNA after 6 h, the response in MEFs was delayed and IFNβ and IFNα2 mRNAs were detected only after 24 or 48 hpi, respectively. Next, a VSV-GFP-based bioassay on MEFs was applied to quantify secreted IFN, and astrocytes were found to secrete higher levels of IFNs 24 hpi compared to MEFs (Fig. 3e). These data indicate that astrocytes react to viral infection by the induction of IFNs much more rapidly than MEFs.

In order to define the time point after infection at which astrocytes initiate an antiviral response and limit viral spread in astrocytes, a TBEV-based bioassay on astrocytes was developed. Pretreated astrocytes were infected, and the numbers of infected cells were quantified at 48 hpi (Fig. 3f). The supernatants showed an inhibitory effect as early as 6 hpi on WT, but not on IFNAR^−/− cells, suggesting a very early type I IFN-mediated antiviral effect.

The antiviral effect of the supernatant was further investigated using qPCR as a more sensitive method of quantifying viral replication. Sixteen hours before TBEV infection, the cells were treated with either supernatant of 24-h TBEV-infected cells or IFNαB/D. The cells were infected with TBEV, and total RNA was extracted 48 hpi to detect TBEV RNA (Fig. 3g). Pretreatment with supernatant of infected cells and 5000 U IFNαB/D had a similar inhibitory effect on TBEV replication in WT astrocytes. However, no difference in viral replication was detected in IFNAR^−/− astrocytes treated with or without supernatant, indicating a type I IFN-dependent antiviral effect. These results further suggest that the rapid type I IFN response is responsible for limiting viral infection and replication in astrocytes.

Our data indicate that WT astrocytes are quite resistant to viral infection and that the main effectors in the supernatant from virus-infected astrocytes are type I IFNs. To characterize these responses in more detail, we analyzed transcriptional regulation by RNAseq. Gene expression levels from uninfected IFNAR^−/− astrocytes and IFNαB/D-treated, or supernatant-treated, WT astrocytes were compared to untreated WT astrocytes. Fold-change values of at least twofold over the uninfected control and a q value of <0.05 were considered indicative of up- or downregulation. Our data yielded 732 upregulated transcripts while 944 were found to be downregulated in IFNAR^−/− astrocytes. Treatment of primary astrocytes with IFNαB/D or supernatant led to 634 or 1092 upregulated and 149 or 423 downregulated genes, respectively (Additional file 1: Table S1).

The majority of all significantly regulated genes showed a unique pattern of regulation for each sample (Fig. 4a–c). Looking more closely at the differences between astrocytes treated with supernatant or IFNαB/D, relatively few genes were downregulated, and we detected 257 genes upregulated in both samples, out of total 894 upregulated genes in the supernatant and 550 in IFNαB/D (Fig. 4b, c). This indicated that although the main transcriptional response is type I IFN-dependent and overlaps with IFNαB/D treatment, virus infection induces numerous other IFN-independent responses in neighboring, uninfected astrocytes.

Only 112 transcripts were found to be significantly regulated between all three samples compared to WT (Fig. 4a). However, they showed a very different expression profile (Fig. 4d). Interestingly, IFNAR-deficient astrocytes showed lower levels of PRR, innate immune signaling, transcription factors, ISGs, and adaptive immune response compared to untreated WT astrocytes, whereas, IFNαB/D- and supernatant-treated WT astrocytes showed an upregulated profile (Fig. 4d, Additional file 1: Table S2).

The susceptibility of IFNAR^−/− astrocytes to flavivirus infection might be due to lowered basal expression of antiviral ISGs, such as viperin. Together with TRIM79α, viperin has recently been identified as an inhibitor of TBEV in vitro [50, 61]. Viperin has also been shown to have antiviral activity against WNV [62]. Viperin and TRIM79α mRNAs were quantified over time by qPCR in WT and IFNAR^−/− astrocytes and WT MEFs (Fig. 5a, b). ISGs were rapidly induced in WT astrocytes at early time points whereas induction was delayed in IFNAR^−/− astrocytes and WT MEFs. The basal levels of viperin and TRIM79α were higher in WT astrocytes compared to WT MEFs and IFNAR^−/− astrocytes (Fig. 5b) and might contribute to limit initial viral infection. Viperin protein was also strongly induced in WT astrocytes as early as 6 hpi whereas corresponding protein levels were only reached after 24 hpi in IFNAR^−/− astrocytes (Fig. 5c). Together, these data show that WT astrocytes express higher basal levels of a subset of antiviral genes and rapidly respond upon viral infection to express antiviral protein that can limit viral infection. MEFs and IFNAR^−/− astrocytes show lower basal expression of some antiviral ISGs and a delayed IFN response and ISG induction which is likely to render them more susceptible to viral infection.

Gene set enrichment analysis was used to identify classes of genes overrepresented in the dataset. IFNAR^−/− cells showed a negative enrichment compared to WT in immune pathways and interferon signaling, while the positively enriched gene sets were below our cut-off adjusted p value < 0.05 (Fig. 6). Both IFNαB/D and supernatant treatment lead to enrichment in “interferon signaling” and “interferon alpha beta signaling” (Fig. 6). Differentially expressed genes between WT and the three samples were analyzed using IPA to uncover predicted activation of canonical pathways and disease functions. Pathways involved in viral recognition and antiviral signaling as well as functions involved in viability of leukocytes and antiviral and inflammatory responses were found to be downregulated in IFNAR^−/− compared to WT astrocytes (Additional file 1: Tables S3 and S4), which can explain why the IFNAR^−/− cells responded poorly after viral infection. IPA analysis revealed activation of pathways involved in innate immune signaling and viral recognition, antiviral and immune response after treatment with IFNαB/D and supernatant (Additional file 1: Table S3 and S4). To extend our analyses and investigate what might be responsible for these transcriptomic changes, with focus on the difference between the IFNαB/D and the supernatant treatment, we performed an IPA upstream analysis, and a Connectivity Map (CMAP) analysis. Both of these analyses evaluate the relationship(s) that might have led to the pattern of transcript regulation that we found after treatment. In Fig. 7a, we have plotted all upstream regulators that are predicted by the IPA analysis to be either activated or inhibited after IFNαB/D or supernatant treatment. Upstream regulators that are predicted to be activated after both treatments include several interferon regulated factors (IRF3, IRF7, STAT1) as well as IFNAR and IFNγ as the most activated. Each dot in Fig. 7a represents a regulator e.g., (IFN alpha/beta) with a circle of markers of upregulated transcripts found in the dataset (Fig. 7b). Interestingly, five upstream regulators were predicted to be activated specifically after supernatant treatment (IL5, CD38, KMT2D, HIF1A, UCP1, and JAK2; Fig. 7a, y-axis >2). The CMAP database is an orthogonal bioinformatics tool that correlates specific gene-overexpression with their associated transcriptomic changes. Our transcriptional changes after IFNαB/D or supernatant treatments were compared to the CMAP database and the connectivity scores are compared to each other in Fig. 7c, and top scoring genes included IFNB1, IFNγ, and CD40. IFNγ came up in both gene set enrichment analyses, IPA and CMAP (Figs. 6, 7a, b), suggesting that TBEV induces IFNγ after infection in astrocytes. However, no differences in IFNγ mRNA expression levels were detected after infection in astrocytes (data not shown).

With this comprehensive bioinformatics analysis, we can show that the main gene sets enriched after supernatant treatment are IFN signaling gene sets very similar to the IFNαB/D treatment (Fig. 6), overall indicating that IFNs are the main response induced after infection. We show that IFNAR^−/− astrocytes not only express lower levels of key antiviral ISGs (Fig. 5) but also express lower levels of ISGs, PRRs, innate immune signaling, and transcription factors generally (Fig. 4d and Additional file 1: Table S2), which probably contributes to the susceptibility of these cells (Figs. 1 and 2).

Astrocytes resistance to viral infection might be due to high basal expression of antiviral genes, fast response via IFNAR signaling, or both. In order to distinguish between these possibilities, monoclonal antibodies were used to block IFNAR signaling in WT astrocytes [63] (Fig. 8a‐c). Interestingly, the numbers of TBEV-infected cells and TBEV replication were increased to levels similar to those measured in IFNAR^−/− cells. This indicated that response to type I IFN is more important for restricting viral growth than the basal expression level of antiviral ISGs. Previous studies have shown that astrocytes are resistant to TBEV-induced cytopathic effects [34, 35]. We can now show that the viability of WT astrocytes after TBEV infection is dependent on the type 1 IFN response, since IFNAR^−/−, as well as WT astrocytes treated with an IFNAR-specific antibody, rendered them susceptible to TBEV-induced cytopathic effect 48 hpi (Fig. 8d).

Neurons are the main target of neurotropic flavivirus infection; however, it is not clear if and how astrocytes influence viral growth in neurons. Therefore, supernatants from infected astrocytes were collected at different time points post infection and inactivated. Primary neurons were treated with supernatants before TBEV infection. The number of infected neurons decreased already when pretreated with supernatants from astrocytes infected for 3 h (Fig. 8e). These data indicate that astrocytes can mediate a very fast antiviral effect on neurons already 3 h post infection.

Here, we show that astrocytes are in an antiviral state and respond quickly to flavivirus infection by upregulating type 1 IFNs which limits neurotropic flavivirus spread. Secreted IFNs control viral replication by upregulating several innate immune pathways and induce cell survival. Our data suggest that although neurons seem to represent the main target for neurotropic flavivirus infection in vivo [64], astrocytes are likely to play an important role in responding to infection, amplifying the type I IFN response and limiting viral spread in both astrocytes and neurons.