Strawberry notch homolog 2 regulates the response to interleukin-6 in the central nervous system

All mice were on the C57BL/6 background, were maintained in the Molecular Bioscience (building G08) animal house facility at the University of Sydney under specific-pathogen-free conditions and received food and water ad libitum. The generation of Sbno2 “floxed” mutant mice was performed by Ozgene Pty Ltd, Australia. Briefly, a conditional allele of Sbno2 was created by flanking exons 8 to 10 (corresponding to Genbank accession NM_183426.1) with loxP sites (Fig. 1A). Gene targeting was performed in C57BL/6 embryonic stem cells and the neomycin cassette introduced during this process was removed. The Sbno2-floxed mice were then crossed with male B6.Cg-Tg(Gfap-cre)73.12Mvs/J mice which express the enzyme Cre recombinase in gametes. Cre-mediated recombination of the floxed exons was designed to cause a translational frame shift, introducing an early stop codon in exon 11, resulting in nullizygous (Sbno2^−/−) mutant mice. Furthermore, to generate GFAP-IL6 × Sbno2^−/− mice, the Sbno2^−/− mice were crossed with GFAP-IL6 mice (originally obtained from the Scripps Research Institute, La Jolla, CA, USA, where they were developed by I. L. Campbell), which chronically produce IL-6 from astrocytes in the CNS at pathophysiological levels [16]. Genotyping primer pairs used to detect Sbno2^+/+ (wildtype, WT) or floxed Sbno2 sequence were 5ʹ-GACTGCCAGCTGAATCCAAAC-3ʹ (forward) and 5ʹ CCCACAAGCCTCCATTTTCC-3ʹ (reverse); to detect recombined sequence, 5ʹ-GGGAAAACTGAGACCCCACC-3ʹ (forward) and the same reverse as for the WT/floxed sequence. Clinical scoring was based on the Table 1 of Metten et al. 2004 [34] with the following modifications: individual mice for observation were removed from home cage and placed into a separate cage with no housing; mice were immediately observed by eye for 30 s and “splay” and “wobble” scores recorded; scoring was always performed by the same person; a wobble score of “1” was only recorded when the mouse wobbled while moving; “onset” was determined by a score above 0. For euthanasia and tissue collection, mice were subjected to deep anaesthesia by intraperitoneal injection of ketamine at 75 mg/kg in combination with xylazine (Troy Laboratories) at 10 mg/kg (10 µl/g body weight) before perfusing with approximately 20 mL of ice-cold phosphate buffered saline (PBS) (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1.8 mM KH2PO4) using a gravity-fed set-up (approximate pressure 120 mmHg, approximate flow rate 3 mL/min). Unless otherwise stated, sex was not controlled for. Tissue collected for histochemistry and immunohistochemistry was placed into PBS-buffered 4% (w/v) formaldehyde overnight. The tissue was processed to extract RNA and protein as previously described [13].

Paraffin sections were prepared and routine histochemistry (hematoxylin and eosin (H&E); luxol fast blue and cresyl violet (LFB)) was performed as previously described [35]. For immunohistochemistry (IHC), tissue was deparaffinised in xylene and rehydrated in a graded ethanol series. Antigen retrieval was performed at sub-boiling temperature as follows: for GFAP, in 25 mM Tris (pH 9.0) for 25 min; for IBA1, in 25 mM Tris–HCl (pH 8.0), 5 mM ethylenediaminetetraacetic acid, 0.05% (w/v) sodium dodecyl sulfate for 45 min; for CD3, in 10 mM citric acid for 45 min. After incubation, slides were cooled in retrieval buffer to approximately room temperature on ice then briefly washed in PBS. Endogenous peroxidases were blocked with 0.3% (w/v) hydrogen peroxide in PBS for 10 min then briefly washed. Tissue sections were encircled using a ‘pap’ pen (Sigma-Aldrich) to form a hydrophobic barrier, blocked with IHC blocking buffer (1% (v/v) normal goat serum in PBS with 0.1% (v/v) Triton-X100) at room temperature for 30 min in a humidified chamber and then incubated with primary antibody diluted in IHC blocking buffer at 4 °C overnight in a humidified chamber (GFAP, 1:2,000, Dako #Z0334; IBA1, 1:2,000, Wako #019-19741; CD3, 1:200, Abcam #16669). The next day, slides were incubated with biotinylated secondary antibody (Rabbit IgG-Biotin, 1:200, Vector Labs #BA1000) for 30 min in a humidified chamber, followed by incubation with avidin-conjugated horse radish peroxidase (Vector Labs) for 20 min in a darkened humidified chamber. Immunocomplexed protein signal was detected using 3,3'-diaminobenzidine HRP substrate (Vector Labs) then counterstained with hematoxylin. Sections were dehydrated in a graded ethanol series, then cleared with xylene before coverslips were mounted using dibutylphthalate polystyrene xylene mountant (Sigma-Aldrich). Bright field images of sections were acquired using a ZEISS Axio Scan.Z1 Slide Scanner and ZEISS Zen Blue software. For quantification, images were deconvoluted using ImageJ to remove the signal from hematoxylin counterstaining. Densitometric analysis of 3,3'-diaminobenzidine (DAB) intensity from GFAP and IBA1 staining in the cerebellar white matter, cerebellar molecular layer and cerebellar granular layer was performed using ImageJ. Per section, the average signal intensity for each region was quantified from 5 locations of a uniform area. Cell counting of CD3-positive cells in the cerebellum was performed manually and was normalised to the measured area (arbitrary units).

Levels of IL-6 in cerebellar and forebrain lysates were detected and quantified using the IL-6 Mouse Uncoated ELISA Kit according to the manufacturer’s instructions (ThermoFisher Scientific #88-7064-88). The plate was scanned with a FLUOstar Omega microplate reader.

Primary astrocyte cultures were derived, cultured and treated as described previously [36]. Briefly, brains from 1- to 3-day-old mice were enzymatically dissociated into mixed glial cell cultures. CD11b-positive cells were removed using the Miltenyi Magnetically-Activated Cell Sorting procedure (Miltenyi) to yield astrocyte cultures. For treatment, hyper-IL-6 [37] was used at 50 ng/mL. To determine the stability of transcripts, cells were treated with 50 ng/mL hyper-IL-6 for 2 h then with actinomycin D (10 µg/mL) (Sigma-Aldrich) or the equivalent volume of DMSO as a vehicle control.

RNase protection assay (RPA) and generation of probes was performed as previously described [38]. Probes used in this study, their NCBI accession number and target regions were: Sbno2, NM_183426.1, 800-1149; Fos, NM_010234.2, 466-677; Gbp2, NM_010260.1, 1100-1255; Icam1, X16624, 98-398; Igtp, NM_018738.4, 1168-1291; Irgm1, NM_008326.1, 953-1176; Irf1, M21065, 211-326; Nlrc5, NM_001033207.3, 5374-5739; Serpina3n, M64086, 1914-2016; Socs3, U88328, 361-555; Timp1, X04684, 118-366; L32, K02061, 61-139.

Immunoblotting was performed as described previously [27] using the following antibodies and dilutions: SBNO2 (C-terminal #2, custom-made, validated in [27]), 1:640, Biomatik; GAPDH, 1:100,000, Sigma-Aldrich G8795; ERK1/2, 1:10,000, Sigma-Aldrich M5670; pT202-ERK1/pY204-ERK2, 1:5,000, Cell Signaling Technology (CST) 9101; STAT1, 1:1,000, CST 9172; pY701-STAT1, 1:1,000, CST 9167; STAT3, 1:2,000, CST 4904; pY705-STAT3, 1:2,000, CST 9131; pS536-NF-kB, 1:1,000, CST 3033; NF-kB, 1:3,000, CST 8242; Rabbit IgG-peroxidase, 1:30,000, Santa Cruz SC2004; Mouse IgG-peroxidase, 1:10,000, Sigma-Aldrich A0168.

Cerebella were homogenised in TRI Reagent (Sigma-Aldrich) and RNA was purified using the Direct-zol RNA MiniPrep kit according to the manufacturer’s instructions (Zymo Research). Purified RNA was dissolved in sterile, RNase-free water and purity and concentration of RNA were assessed. Samples were processed by the Ramaciotti Centre for Genomics (University of New South Wales, Sydney, Australia) for cRNA preparation, hybridisation and scanning of Affymetrix murine Clariom S microarrays according to the manufacturer’s instructions (ThermoFisher Scientific). Transcriptome Analysis Console software (version 4.0.1.36, ThermoFisher Scientific) was used to perform gene-level normalisation and summarisation via the in-built Signal Space Transformation-Robust Multi-Chip Analysis algorithm [39, 40]. The Mouse Genome Informatics Gene Expression Database [41] was used to select a background signal value based on the expression of 195 genes deemed not to be expressed in the murine cerebellum, as well as the expression of 219 Y chromosome genes in female samples. A signal intensity threshold of 6 (log2) was found to be a robust cut off and so genes with a signal below 6 were considered to be either not expressed or below the detection limit of the assay. As sex was not controlled for in this study, × and Y chromosome genes were excluded from the analysis. Genes were considered to be differentially regulated between two genotypes if they had a statistically significant (false discovery rate (FDR)-adjusted p value < 0.05) fold change of ± 2. Gene ontology (GO) overrepresentation was performed using the PANTHER Overrepresentation Test (Released 20190711; annotation version: GO Ontology database (Biological Process) Released 2019-12-09; Test Type: Fisher's Exact; Correction: FDR) [42]. The background population was defined as genes from the microarray which was expressed (i.e. ≥ 6) in any one genotype. Probes that could not distinguish multiple gene products (denoted by a semicolon in the gene lists) were excluded from the analysis.

Statistical analyses were performed as follows unless otherwise stated. Analyses and graphing were performed using GraphPad Prism software (version 8). The data were compared by two-way ANOVA with Tukey's post test and presented as the mean ± SEM. Differences were considered significant at p < 0.05. The number of samples analysed is given in the figure legends. Detailed statistics are presented in Additional file 11.

To study the role of SBNO2 in the CNS, we first generated mice in which Sbno2 was disrupted (Sbno2^−/−) (Fig. 1A). The recombination of Sbno2 exons 8 to 10 was confirmed in both DNA and RNA samples (Fig. 1B, C). Although Sbno2 transcript upstream of the recombined region was present, no SBNO2 protein was detected using an antibody that targets a region downstream of the recombination site (Additional file 1). Breeding male Sbno2^−/− with female Sbno2^± mice resulted in offspring that were viable and suckled to weaning age, however, we found an underrepresentation of female offspring (40% female versus 60% male) while obtaining genotypes in the expected Mendelian ratios. Adult Sbno2^−/− mice showed a slight decrease in body weight compared with Sbno2^+/+ (wild type, WT) mice (Fig. 1D), but were otherwise physically normal and showed no signs of disease (Fig. 1F). Histological analysis of peripheral organs (kidney, liver, spleen, small intestine) and the CNS (brain and spinal cord) revealed no overt abnormalities in adult Sbno2^−/− mice compared with WT mice (Fig. 2 and data not shown).

To assess the contribution of SBNO2 to IL-6-mediated neuroinflammation in vivo, we crossed the Sbno2^−/− mouse with the GFAP-IL6 mouse model. At 1 month of age, both the GFAP-IL6 mice and the GFAP-IL6 × Sbno2^−/− mice were physically indistinguishable from WT mice. However, GFAP-IL6 and GFAP-IL6 × Sbno2^−/− mice had progressively diminished body weight compared with WT and Sbno2^−/− mice and body weight was significantly lower in GFAP-IL6 × Sbno2^−/− mice compared with GFAP-IL6 mice at 3 and 7 months of age (Fig. 1D). However, IL-6 levels were relatively unchanged, being slightly, but not significantly decreased in GFAP-IL6 × Sbno2^−/− mice compared with GFAP-IL6 mice at 1 and 3 months of age (Fig. 1G). Transgene expression in the GFAP-IL6 brain is highest in the cerebellum [20]. Accordingly, GFAP-IL6 mice progressively develop mild ataxia from around 2 months of age, which at 7 months presents as mild splaying of the back feet with an otherwise normal gait (Fig. 1E). In contrast, GFAP-IL6 × Sbno2^−/− mice quickly developed overt ataxia and by 7 months nearly all mice were affected, exhibiting a wobbly gait with occasional falls. The age at onset of the disease was comparable between both genotypes.

Histological and immunohistochemical analysis (Fig. 2) of the cerebellum of 7 month-old GFAP-IL6 mice showed a vacuolated appearance in the cerebellar white matter apparent in both H&E- (Fig. 2C) and LFB-stained sections (Fig. 2G), in line with previous reports [16, 18, 19, 43]. This was accompanied by a loss of myelin, which was most apparent at the outer branches of the cerebellar arbor vitae, gliosis of the Bergmann glia in the molecular layer with thickened processes (Fig. 2K), hypertrophy of microglia with stunted processes (Fig. 2O), meningeal leukocyte infiltrates and a diffuse infiltration of the brain’s parenchyma by CD3-positive cells (Fig. 2S). The pathology in the GFAP-IL6 × Sbno2^−/− cerebellum was more severe when compared with the GFAP-IL6 cerebellum. H&E staining revealed a near complete loss of architecture at the centre of the cerebellum (Fig. 2D), with the granule cell layer having degraded to the point of being indistinguishable and the molecular layer being thinned. LFB staining showed a near complete loss of myelin in the white matter in the cerebellum of GFAP-IL6 × Sbno2^−/− mice (Fig. 2H). GFAP immunohistochemistry appeared diffuse and disorganised, with individual astrocytes nearly impossible to distinguish and the GFAP-stained radial processes of the Bergman glia were not distinguishable in areas where the granular layer had degraded (Fig. 2L), although GFAP DAB intensity was not significantly different between the two genotypes (Additional file 2). IBA1 immunohistochemistry, a marker for microglia and monocytes/macrophages, was significantly increased throughout the GFAP-IL6 × Sbno2^−/− cerebellar white matter only (Fig. 2P; Additional file 2). Sulci contained a large number of infiltrating cells, many of which were IBA1- or CD3-positive indicating the presence of monocytes/macrophages and T cells (Fig. 2P, T). The infiltration of CD3-positive T-cells throughout the cerebellum was significantly increased in GFAP-IL6 × Sbno2^−/− compared with that in the GFAP-IL6 cerebellum (Fig. 2T; Additional file 2).

Strawberry notch family members have been implicated in the regulation of gene expression and we have demonstrated that Sbno2 is markedly upregulated in response to IL-6 [27]. Therefore, we hypothesised that SBNO2 is a putative feedback transcriptional repressor and that the exacerbated pathology in the GFAP-IL6 × Sbno2^−/− cerebellum was due to an enhanced transcriptional response to IL-6 in the absence of SBNO2. To investigate this, differential gene expression profiling was performed on RNA isolated from the cerebellum of WT, Sbno2^−/−, GFAP-IL6 and GFAP-IL6 × Sbno2^−/− mice (Fig. 3). We used 1-month old mice as at this age neuropathology in both GFAP-IL6 and GFAP-IL6 × Sbno2^−/− mice was modest (Additional file 3), allowing us to assess the disruption of Sbno2 more directly.

Approximately 11,000 genes were expressed in each genotype (Additional file 4). No genes were differentially regulated in Sbno2^−/− when compared with WT mice, which was reflected in the principal component analysis (Fig. 3A). However, 385 genes were upregulated and 37 downregulated in GFAP-IL6 compared with WT mice (Fig. 3B, Additional file 5) and 696 genes were upregulated and 117 downregulated in GFAP-IL6 × Sbno2^−/− compared with WT mice (Fig. 3B, Additional file 6). Further, when comparing GFAP-IL6 and GFAP-IL6 × Sbno2^−/− mice, 217 genes were upregulated and 36 downregulated in GFAP-IL6 × Sbno2^−/− mice (Additional file 7). These findings further support the notion that SBNO2 acts predominantly as a repressor in the context of the transcriptional response to IL-6 in the murine CNS. The findings also suggest that the role of SBNO2 in regulating gene expression in adolescent mice under physiological conditions is, at the best, minor.

To better understand the biological outcomes brought about by differential gene regulation, we used the PANTHER overrepresentation test [44, 45] to identify the ‘biological process’ gene ontology (GO) terms overrepresented in each of the gene lists. There were no overrepresented GO terms amongst genes downregulated in any genotype comparison. Amongst genes upregulated in GFAP-IL6 mice compared with WT mice, 616 GO terms were enriched (Additional file 8), most of which related to the immune system as reflected in the most significantly enriched GO term being ‘immune system process’ (GO:0002376). Further, many GO terms related to phenotypes of the GFAP-IL6 model which have previously been experimentally verified, such as microglial cell activation [18, 46], endothelial involvement and angiogenesis [16, 19], cell adhesion [47], chemotaxis [20], complement [43] and the acute-phase response [16, 48]. Amongst genes upregulated in GFAP-IL6 × Sbno2^−/− mice compared with WT mice, 817 GO terms were enriched (Additional file 9), 68% of which were also enriched in GFAP-IL6. Terms highly enriched in both analyses included those relating to hypersensitivity (largely due to Fc receptor genes), complement and ‘microglial cell activation involved in immune response’. Notably, terms relating to MHC molecules were highly enriched in GFAP-IL6 × Sbno2^−/−, specifically. These results were re-enforced by examining the 275 GO terms enriched amongst genes upregulated in GFAP-IL6 × Sbno2^−/− compared with GFAP-IL6, in which the MHC-related terms were the most highly enriched (Fig. 3C, Additional file 10). These findings suggest that many of the biological processes which underlie the pathology of the GFAP-IL6 brain also underlie that in the GFAP-IL6 × Sbno2^−/− brain, but furthermore, that some biological processes may be unique to the GFAP-IL6 × Sbno2^−/− model.

Noting that the phenotype of the GFAP-IL6 × Sbno2^−/− brain appeared to be largely an exacerbation of the response to IL-6, yet IL-6 itself was not increased (Fig. 1G), we sought to clarify whether there were changes in the activation of the canonical IL-6/gp130 signalling pathways that might account for this. For this, we examined the key IL-6-responsive gp130 signalling components, STAT1, STAT3 and ERK, as well as the potential SBNO2-interactor, NF-κB [28] in cerebellar lysates from 1 month-old mice (Fig. 4). In lysates from WT and Sbno2^−/− mice, phospho-(p)STAT1 and pSTAT3 were undetectable, while pERK was present and comparable between both genotypes. Lysates from GFAP-IL6 mice showed pSTAT1 and pSTAT3 signals, while pERK was comparable to WT and Sbno2^−/− mice. In GFAP-IL6 × Sbno2^−/− mice, levels of pSTAT3 were significantly reduced when compared with GFAP-IL6 mice, while pSTAT1 and pERK levels were lower, albeit not significantly. Phosphorylation of NF-κB was undetectable in all genotypes. These findings show there is diminished IL-6-stimulated gp130 pathway activation in the GFAP-IL6 × Sbno2^−/− cerebellum.

Decreased phosphosignalling in the GFAP-IL6 × Sbno2^−/− cerebellum suggested that SBNO2 may alter IL-6/gp130 signal pathway activation. In addition, as SBNO2 production plateaued after 6 h of hyper-IL-6 treatment in WT astrocytes [27], we asked whether sustained gene expression in hyper-IL-6-treated Sbno2^−/− astrocytes may reflect a need for SBNO2 to repress excessive signal pathway activation. Therefore, we treated Sbno2^−/− astrocytes with hyper-IL-6 for 6 h and examined IL-6/gp130 signalling pathway components STAT1, STAT3, ERK as well as NF-κB (Fig. 5A). A two-way ANOVA was performed to analyse the effect of Sbno2 disruption and of hyper-IL-6 treatment on protein levels (Additional file 11). The main effects analysis showed that the treatment had a statistically significant effect on pSTAT1, pSTAT3, pERK, ERK and pNF-κB. Sbno2 disruption had a statistically significant effect on pSTAT1, pSTAT3, pERK, pNF-κB and NF-κB. There was a statistically significant interaction between these effects on pSTAT1, pSTAT3 and pNF-κB. In both WT and Sbno2^−/− astrocytes, hyper-IL-6 treatment increased pSTAT1, pSTAT3 and pERK levels, but to a significantly lesser degree in Sbno2^−/− cells. The pNF-κB level increased significantly with treatment in WT, but not Sbno2^−/− cells. These results reveal that disruption to Sbno2 causes decreased levels of hyper-IL6-stimulated gp130 pathway activation in primary cultured astrocytes.

The information gleaned from the GFAP-IL6 × Sbno2^−/− model provided a chronic, in vivo view of SBNO2-modulated gene expression. Next, we wished to delineate the cellular function of SBNO2 and how it affected the temporal response to an acute exposure to IL-6. We had previously shown that Sbno2 is predominantly, but not exclusively, expressed by astrocytes in the CNS [27]. Therefore, we treated primary cultured astrocytes from WT and Sbno2^−/− mice with hyper-IL-6 and examined the expression of several genes associated with CNS inflammation over 24 h (Fig. 5B). A two-way ANOVA was performed to analyse the effect of Sbno2 disruption and of hyper-IL-6 treatment time on gene expression (Additional file 11). Main effects analysis showed that treatment time had a statistically significant effect on the expression of all genes examined, and that Sbno2 disruption had a statistically significant effect on Fos, Gbp2, Icam1, Igtp, Irf1, Irgm1, Nlrc5 and Socs3. There was a statistically significant interaction between the effects on all genes besides Socs3. The expression of Gbp2, Icam1, Irf1, Nlrc5 and Serpina3n was significantly higher in untreated Sbno2^−/− cells than in untreated WT cells, suggesting some requirement for SBNO2 to maintain basal expression of these genes in vitro. The expression of Fos, Gbp2, Icam1, Igtp, Irf1, Irgm1 and Nlrc5 was significantly higher in Sbno2^−/− cells than in WT cells following hyper-IL-6 treatment (at 2 h, 6 h, 12 h and 24 h), while expression of Serpina3n, Socs3 and Timp1 was significantly higher only at certain times. Expression of Igtp, Irf1, Irgm1 and Nlrc5 returned to their basal level of expression in WT cells, but levels were sustained in Sbno2^−/− cells. These results suggest that SBNO2 is required to actively dampen expression of a subset of genes as part of the negative regulation of the IL-6 transcriptional response.

Having determined that increased signal pathway activation was not responsible for sustained gene expression in Sbno2^−/− astrocytes, we next sought to address whether changes in gene transcript stability played a role. To examine this, Sbno2^−/− primary cultured astrocytes were treated with hyper-IL-6 for 2 h to stimulate gene expression, then actinomycin D (ActD) was added, which inhibited further transcription, revealing transcript stability over the following 2 h. No differences between Sbno2^−/− and WT cells were found for Fos, Gbp2, Icam1, Igtp, Irf1, Irgm1, Nlrc5, Serpina3n and Socs3 transcripts, while the Icam1 transcript was found to be significantly different (Fig. 6). Overall, these results suggest that sustained gene expression in Sbno2^−/− astrocytes is not elicited through SBNO2 modulating transcript stability.