Interferon-stimulated genes—essential antiviral effectors implicated in resistance to Theiler’s virus-induced demyelinating disease

Twenty-four 5-week-old female SJL/J HanHsd mice and 11 C57BL/6 mice (Harlan Winkelmann, Borchen, Germany) were kept in a microisolator cage system (Tecniplast, Hohenpeißenberg, Germany) with standard bedding material (ssniff Spezialdiäten GmbH, Soest, Germany) under a 12:12 h light/dark cycle. Food (R/M-H V1534; ssniff Spezialdiäten GmbH) and water were available ad libitum. They were inoculated into the right cerebral hemisphere with 1.63 × 10⁶ plaque-forming units (PFU) per mouse of the BeAn strain of TMEV in 20 μl Dulbecco’s modified Eagle medium (DMEM; PAA Laboratories, Cölbe, Germany) with 2 % fetal calf serum and 50 μg/kg gentamycin under general anesthesia using medetomidine (0.5 mg/kg, Domitor; Pfizer, Karlsruhe, Germany) and ketamine (100 mg/kg, Ketamin 10 %; WDT eG, Garbsen, Germany). Twenty-four 5-week-old female SJL/J HanHsd and 12 C57BL/6 mice were mock-infected using 20 μl DMEM only. TMEV- and mock-infected mice were sacrificed in groups of five to six animals at 14, 42, 98, and 196 days post infection (dpi), respectively. Euthanasia was performed with a fivefold overdose of the described anesthetics. Segments of the spinal cord of mice were removed immediately after death and either formalin-fixed and paraffin-embedded or snap-frozen in OCT embedding compound (Sakura Finetek Europe, Zoeterwoude, Netherlands) using liquid nitrogen [37]. Microarray analysis was performed on 23 TMEV- (5 mice at 98 dpi; 6 mice at 14, 42, and 196 dpi) and 24 mock-infected SJL/J mice (6 mice at all time points). Spinal cord tissue obtained from these 47 SJL/J mice was also studied using immunohistochemistry at all four time points. ISG immunoreactivity and transcript numbers of selected genes were compared between SJL/J and C57BL/6 mice at 14 dpi (6 mock- and 6 TMEV-infected mice, respectively) and 98 dpi (6 mock- and 5 TMEV-infected mice, respectively) using immunohistochemistry and reverse transcription quantitative polymerase chain reaction (RT-qPCR), respectively. Moreover, immunofluorescence was performed on one additional TMEV-infected SJL/J mouse due to the lack of adequate tissue from the other SJL/J mice. Animal experiments were authorized by local authorities (Niedersächsisches Landesamt für Verbraucherschutz- und Lebensmittelsicherheit, Oldenburg, Germany, permission numbers: 509c-42502-02/589, 509.6-42502-04/860, 33-42502-05/963, 33-42502-04-07/1292).

For microarray analysis, RNA was isolated from snap-frozen spinal cord specimens using the RNeasy Mini Kit (Qiagen, Hilden, Germany), amplified and labeled employing the MessageAmp II Biotin Enhanced Kit (Ambion, Austin, TX, USA), and hybridized to GeneChip mouse genome 430 2.0 arrays (Affymetrix, Santa Clara, CA, USA) as previously described [38, 39]. Background adjustment and quantile normalization was performed using RMAexpress [40]. MIAME compliant data sets are deposited in the ArrayExpress database (E-MEXP-1717; http://​www.​ebi.​ac.​uk/​arrayexpress). Bottom-up analysis of manually selected candidate genes was done using classical statistics as previously described [41‐43]. Accordingly, 111 candidate genes comprising 10 pattern recognition receptors (PRRs), 10 IFN regulatory factors, 9 type I/II IFNs, 4 type I/II IFN receptors, 8 signal transducers, 60 IFN-dependent antiviral effectors, and 10 MHC class I/II genes were manually extracted from peer-reviewed published literature (Additional file 1: Table S1) [19, 22, 23, 44, 45]. The fold change was calculated as the ratio of the inverse-transformed arithmetic means of the log2-transformed expression values of TMEV-infected vs. mock-infected mice. Downregulations are shown as negative reciprocal values [39, 46]. The normalized expression values of these genes were evaluated for significant differences between TMEV- and mock-infected mice employing independent pairwise Mann-Whitney U tests (Prism 6, GraphPad Software, La Jolla, CA, USA). Statistical significance was designated as P ≤ 0.05.

RT-qPCR was performed for IFN-α (using consensus primers annealing with all IFN-α subtypes), IFN-β, IRF-7, ISG15, PKR, and three housekeeping genes (GAPDH, β-Actin, HPRT) using standard protocols, the Mx3005P qPCR System (Agilent Technologies Deutschland GmbH, Böblingen, Germany), and Brilliant III Ultra-Fast SYBR® QPCR Master Mixes [47]. The following primers were used: IFN-α forward ATGGCTAGRCTCTGTGCTTTCCT; IFN-α reverse AGGGCTCTCCAGAYTTCTGCTCTG; IFN-β forward TGAATGGAAAGATCAACCTCACCTA; IFN-β reverse CTCTTCTGCATCTTCTCCGTCA; IRF7 forward CGAGTGCTGTTTGGAGACTG; IRF7 reverse GGCCTTGAAGATCTGTGCAT; ISG15 forward AACTGCAGCGAGCCTCTGA; ISG15 reverse CACCTTCTTCTTAAGCGTGTCTACAG; PKR forward GTACAAGCGCTGGCAGAACTCAAT; and PKR reverse AAGAGGCACCGGGTTTTGTAT. Tenfold serial dilution standards ranging from 10⁸ to 10² copies/μL were used to quantify the results. A normalization factor achieved from the housekeeping genes was calculated to correct for experimental variations [48]. Specificity of each reaction was controlled by melting curve analysis. Similar to microarray analysis, fold changes were calculated to compare the respective results.

Two- to three-micrometer paraffin spinal cord sections of SJL/J mice at 14, 42, 98, and 196 dpi and C57BL/6 mice at 14 and 98 dpi were routinely stained with hematoxylin and eosin (HE) or used for immunohistochemistry as described [47]. Briefly, dewaxed and rehydrated sections were incubated in ethanol with 0.05 % hydrogen peroxide for 30 min to block endogenous peroxidase. Antigen retrieval was performed for the detection of PKR (boiled in 10 mM citrate buffer, 15 min, pH 6.0) and ISG15 and OAS1 (trypsin solution with CaCl₂ × 2 H₂O, 20 min, pH 7.8). Subsequently, sections were blocked with phosphate-buffered saline (PBS) containing 20 % horse serum (Mx1/2/3) or 20 % goat serum (PKR, ISG15, OAS1) for 30 min at room temperature. Sections were incubated with an affinity-purified goat polyclonal anti-Mx1/2/3 antibody (sc-34130, Santa Cruz Biotechnology, Santa Cruz, CA, USA; 1:100), a rabbit monoclonal anti-PKR antibody (ab-32036, Abcam, Cambridge, MA, USA; 1:400), or rabbit polyclonal antibodies directed against ISG15 (sc-50366, Santa Cruz Biotechnology; 1:200) or OAS1 (sc-98424, Santa Cruz Biotechnology,; 1:400). Negative control sections were incubated with goat serum (G9023, Sigma-Aldrich, Taufkirchen, Germany; 1:5650) and rabbit serum (R4505, Sigma-Aldrich) diluted at the same gamma globulin concentration as the primary antibodies. Biotinylated horse-anti-goat antiserum (BA-9500, Vector Laboratories, Burlingame, CA, USA; 1:200) and goat-anti-rabbit antiserum (BA-1000, Vector Laboratories; 1:200) were used as secondary antibodies, and immunolabeling was visualized by the avidin-biotin-peroxidase complex (ABC) method (PK-6100, Vector Laboratories) with 3,3-diaminobenzidine (DAB) as substrate. Finally, sections were slightly counterstained with Mayer’s hematoxylin.

Photos of ISG15, PKR, and OAS1 immunohistochemically stained sections were generated by a digital microscope (HS All-in-one Fluorescence Microscope BZ-9000 Generation II; HS All-in-one Fluorescence Microscope BZ-II Analyzer, BIOREVO, KEYENCE Deutschland GmbH, Neu-Isenburg, Germany) and evaluated using analysis software (analySIS 3.1 software package; Soft Imaging system, Münster, Germany). Light microscopy was used to count Mx-positive cells. In addition, ISG positive and negative cells present in the Virchow-Robin space of meningeal and intraparenchymal blood vessels were counted to calculate the percentage of ISG15-, PKR-, OAS1-, and Mx-positive perivascular cells. Immunohistochemical results are expressed as median with minimum and maximum.

Immunofluorescence was used to identify cells expressing ISG15 and PKR as described [49, 50]. Briefly, after citrate buffer heat-mediated antigen retrieval and incubation with PBS containing 20 % horse serum or 20 % goat serum to block non-specific binding, paraffin sections were co-incubated with antibodies directed against ISG15 (sc-50366; 1:100) or PKR (ab-32036; 1:200) and antibodies directed against GFAP (ab-3554, Abcam; 1:200), Olig2 (ab-85900, Abcam; 1:500), or CD107b (MCA2293, AbD Serotec, Puchheim, Germany; 1:200) overnight at 4 °C. Negative controls included sections incubated with rabbit serum (R4505, Sigma-Aldrich; 1:3000/1:6000), goat serum (I9140, Sigma-Aldrich; 1:4500), sheep serum (1:5000), and rat serum (R9759, Sigma-Aldrich; 1:16000), respectively. After washing, sections were incubated Cy2- and Cy3-conjugated secondary antibodies (goat anti-rabbit, A-11034, Invitrogen; 1:200; goat anti-rat, 112-165-003; donkey anti-rabbit, 711-545-152; donkey anti-sheep, 713-765-147; donkey anti-goat, 705-765-147, Jackson ImmunoResearch, Suffolk, UK; 1:200) in the dark for 1 h at room temperature. Nuclei were counterstained with 0.01 % bisbenzimide (H33258, Sigma-Aldrich), and sections were mounted in Dako Fluorescence Mounting Medium (S3023, DakoCytomation GmbH, Hamburg, Germany).

RT-qPCR data and immunohistochemical data of SJL/J and C57BL/6 mice were analyzed using Kruskal-Wallis and non-parametric Mann-Whitney U tests (IBM SPSS, version 19.0, IBM Corporation, Armonk, NY, USA). Ratios of the geometric means and 95 % confidence intervals were calculated with independent T tests to compare ISG protein expression between C57BL/6 and SJL/J mice (IBM SPSS, version 19.0). Spearman’s rank correlation coefficients were calculated to analyze the correlation between ISG protein and gene expression as well as TMEV RNA copy numbers, inflammation, and demyelination in the spinal cord of SJL/J mice (IBM SPSS, version 19.0). Data concerning TMEV, inflammation (determined on hematoxylin and eosin and immunohistochemically stained sections), and demyelination (determined on luxol fast blue stained sections) have been taken from previous publications [39, 51]. Statistical significance was designated as P ≤ 0.05.

Recently, Ulrich et al. (2010) performed a microarray analysis of mock- and TMEV-infected SJL/J mice at 14, 42, 98, and 196 dpi and detected transcriptional changes of 1001 genes in the spinal cord. In the present study, the log2-transformed normalized data were independently re-evaluated focusing on critical genes of the type I IFN pathway employing conventional statistical methods as described [41, 43, 52]. Most genes were significantly up-regulated by TMEV infection and peaked at 98 dpi (Fig. 1; Table 1; Additional file 1: Table S1). Only a mild increase in Toll-like receptor (Tlr)3 gene expression (fold change <2) was found, whereas other receptors of viral RNA including Pkr, melanoma differentiation-associated gene (Mda5), Tlr7, and Tlr13 were moderately up-regulated (fold change up to 5.77). Nfkb1, lrf3, and Irf5 gene expression was barely influenced by TMEV infection. In contrast, Irf1 and Irf7 were strongly up-regulated (fold change up to 4.40 and 11.42, respectively). Interestingly, no changes were found in type I IFN gene expression, and Ifnar genes were only mildly increased in the late phase of TMEV-IDD. However, gene expression of several factors critically involved in IFN signal transduction, especially of the ISGF3 components IRF9, STAT1, and STAT2, was moderately to strongly up-regulated (fold change up to 15.39). In addition, gene expression of various IFN-dependent antiviral effectors was strongly increased in TMEV-IDD (fold change up to 31.22). Similarly, several MHC class I and II genes were strongly induced by TMEV infection (fold change up to 37.55).

RT-qPCR of selected genes was performed to validate microarray results of SJL/J mice and to compare gene expression of SJL/J and C57BL/6 mice (Additional file 2: Figure S1). In general, there was a good correlation between RT-qPCR and microarray data in SJL/J mice revealing highly similar fold changes of IRF7, ISG15, and PKR at both 14 and 98 dpi. Nevertheless, minor changes in type I IFN mRNA transcription were detected in SJL/J mice using RT-qPCR, which were not found in the microarray analysis.

IFN-α and IFN-β transcript numbers were not significantly changed at 14 dpi in SJL/J mice (fold changes: IFN-α: 1.33, IFN-β: −1.07). At 98 dpi mRNA levels of both type I IFNs were downregulated (fold changes: IFN-α: −2.12, IFN-β: −2.78), but only IFN-β differed significantly (IFN-α: P = 0.052, IFN-β: P = 0.030). IRF7, ISG15, and PKR were significantly up-regulated at 14 dpi (fold changes: IRF7: 4.96 (P = 0.009), ISG15: 4.13 (P = 0.009), PKR: 1.64 (P = 0.026)) and 98 dpi (fold changes: IRF7: 11.53 (P = 0.017), ISG15: 21.12 (P = 0.004), PKR: 3.19 (P = 0.004)) in SJL/J mice.

C57BL/6 mice showed a significant increase in mRNA levels of IFN-β at 14 dpi (fold change 3.73, P = 0.041) and of IFN-α at 98 dpi (fold change 2.49, P = 0.017). TMEV infection did not significantly influence the transcript numbers of IFN-β at 98 dpi (fold change 1.42, P = 0.247) and of IFN-α at 14 dpi (fold change −1.01, P = 1.000). Similar to IFN-β, ISG15 was up-regulated at 14 dpi (fold change 2.11, P = 0.026) but not at 98 dpi (fold change 1.59, P = 0.177) in C57BL/6 mice. Statistical analysis detected a tendency of increased IRF7 mRNA levels at both time points (fold changes: 14 dpi: 2.24, P = 0.093; 98 dpi: 2.02, P = 0.052). PKR transcript numbers were not significantly influenced by TMEV infection in this mouse strain (fold changes: 14 dpi: 1.26, 98 dpi: 1.06).

Significant differences between the two mouse strains at the mRNA levels of the investigated genes were predominantly found at 98 dpi. In mock-infected mice, there were higher transcript numbers of IFN-α (P = 0.041) and IFN-β (P = 0.009) and lower mRNA levels of ISG15 (P = 0.009) in SJL/J compared to C57BL/6 mice at this time point. In TMEV-infected mice, there were higher transcript numbers of IRF7 (P = 0.008), ISG15 (P = 0.008), and PKR (P = 0.008) and lower mRNA levels of IFN-α (P = 0.016) in SJL/J compared to C57BL/6 mice at 98 dpi. At 14 dpi, a significant difference between the two mouse strains was only detected for IFN-β, which was higher in mock- (P = 0.002) and TMEV-infected (P = 0.002) SJL/J compared to C57BL/6 mice.

Double-staining with cellular markers revealed PKR expression in Olig2-positive oligodendrocytes (Fig. 6d) and MAC-3-positive microglia/macrophages (Fig. 6f), whereas astrocytes did not express PKR (Fig. 6b). In contrast, ISG15 co-localized with GFAP confirming its astrocytic expression (Fig. 6a). In addition, few oligodendrocytes (Fig. 6c) and microglia/macrophages (Fig. 6e) expressed ISG15.

ISG gene and protein expression was correlated to several parameters characterizing the progression of TMEV-IDD in SJL/J using Spearman’s rank correlation coefficients in order to quantify the relationship of ISGs with viral load, inflammation, and demyelination. The gene expression of ISG15, PKR, OAS1, and Mx strongly correlated with TMEV RNA, inflammatory cells including CD3⁺ T cells, CD45R⁺ B cells, and MAC3⁺ microglia/macrophages, and demyelination (r _s  ≥ 0.6; P ≤ 0.05; Table 2). In addition, low but significant correlations were found between the protein expression of ISG15, PKR, and Mx and their respective gene expression, TMEV RNA, and all investigated inflammatory cells (r_s ≥ 0.4; P ≤ 0.05; Table 3). ISG15 (r _s  = 0.3) and PKR (r _s  = 0.4) protein expression was also significantly correlated to the extent of demyelination in the spinal cord. OAS protein expression was only significantly correlated to the overall number of inflammatory cells (r _s  = 0.3) and the number of intralesional MAC3⁺ microglia/macrophages (r _s  = 0.4).

The present study investigated the expression of genes critically involved in the type I IFN pathway during TMEV-IDD. PRRs recognize pathogen-associated molecular patterns (PAMPs) such as viral and bacterial components and induce intrinsic cellular defenses. Double-stranded RNA formed during TMEV replication is typically identified by PKR, MDA5, and TLR3 [31, 53, 54]. PKR, MDA5, and TLR2 gene expression was moderately increased in the present study, whereas TLR3 mRNA levels were hardly affected by TMEV infection. Nevertheless, previous studies already demonstrated that TMEV induces an up-regulation of TLR2 via TLR3 signal and subsequently an activation of nuclear factor of kappa light polypeptide gene enhancer in B cells (NF-κB) causing the expression of several pro-inflammatory cytokines and chemokines including IL-1β, IL-6, CCL2 (MCP-1), and CCL5 (RANTES) [32, 33, 55, 56]. In contrast to IRF3 and IRF5, IRF7 was strongly up-regulated in the present study. Interestingly, the IRF7 protein must be continuously produced due to its short half-life [57]. In addition, IRF3 function can be blocked by the TMEV L protein [2, 15, 16]. Consequently, several PRRs might be induced by TMEV and implicated in the activation of the innate and adaptive immune response via NF-κB and IRF7. A high NF-κB expression was already described in microglia/macrophages in demyelinating TMEV lesions [55]. The activation of the type I IFN pathway by TMEV was also demonstrated by increased transcription of various ISGs including MHC class I and II genes, especially in the late phase of TMEV-IDD. Correspondingly, high correlation coefficients were found between ISG gene expression and TMEV RNA as well as inflammation and demyelination.

However, type I IFN production was blocked in the spinal cord of TMEV-infected SJL/J mice, and up-regulation of ISG transcription might depend on increased expression of ISGF3 or even IRF1, which can directly induce ISG transcription and thereby exert antiviral functions in the absence of IFN [23]. The lack of type I IFN might be related to the strong OASL1 expression found in the present study, which inhibits the translation of the potent type I IFN inducer IRF7 [58]. In addition, the increase in ISG protein expression, which only weakly correlated with its gene expression, was mainly restricted to late white matter lesions. Low ISG protein levels in the early phase of TMEV-IDD in this susceptible mouse strain indicate an intrinsic and/or virus-dependent impairment of these IFN-dependent antiviral effectors. This most likely favors virus spread within the CNS and infection of white matter glial cells. Furthermore, an impaired immune response gives the virus enough time to cause initial damage to myelin sheaths resulting in the release of myelin breakdown products. As soon as a DTH response directed against myelin sheath components has established, pro-inflammatory effects of high ISG expression within white matter lesions can even promote demyelination by enhancing antigen presentation and activating additional autoreactive immune cells. Consequently, a strong antiviral immune response in the early phase of TMEV-IDD is critical to prevent virus persistence and initiation of the demyelination process, a time point which might even define an antiviral vs. pro-inflammatory outcome of ISG actions.

In contrast to SJL/J mice, TMEV infection induced type I IFN transcription in C57BL/6 mice. Nevertheless, absolute IFN-β mRNA levels were higher in SJL/J compared to C57BL/6 mice at 14 dpi. RT-qPCR also revealed higher IRF7, ISG15, and PKR transcript number in TMEV-infected SJL/J at 98 dpi. However, immunohistochemistry demonstrated similar or higher protein levels of ISG15, PKR, and OAS1 in the spinal cord white and gray matter of resistant C57BL/6 mice. Moreover, ISG15 and PKR protein levels were further increased by TMEV infection. ISG15 has been reported to inhibit IRF3 degradation [59] and enhance NF-κB signaling via suppression of protein phosphatase 1B (PPM1B) [60]. Downstream targets of the transcription factors IRF3 and NF-κB including IFN-β and pro-inflammatory cytokines play an important role in the innate immune response to TMEV infection. Interestingly, the main ISG15 expressing cells were astrocytes, endothelial cells, and ependymal cells, which line the ventricular system and the central canal of the spinal cord. This expression pattern puts ISG15 in the first line of defense to restrict virus spread in the CNS via the liquorogenic route [3]. PKR is required for type I IFN production in response to TMEV by acting as cytoplasmic RNA sensor and stabilizing IFN mRNA integrity [32]. Moreover, PKR is itself up-regulated by type I IFNs and exerts direct antiviral effects [25, 61]. Similarly, OAS1 protein might act as antiviral effector in TMEV-infected C57BL/6 mice restraining virus spread in the spinal cord before demyelination starts. Nevertheless, due to the inhibition of the OAS/RNase L pathway by the TMEV L* protein [35], the antiviral activity of RNase L might be similar in TMEV-infected SJL/J and C57BL/6 mice despite different OAS1 protein levels.