Expression of the neuroprotective protein aryl hydrocarbon receptor nuclear translocator 2 correlates with neuronal stress and disability in models of multiple sclerosis

All animal work was in compliance with our protocol approved by the University of British Columbia Animal Care Committee (certificate #A130281) and per the Canadian Council on Animal Care guidelines. Eight-week-old female C57BL/6 (Jackson Laboratories, Bar Harbor, ME) mice were immunized for chronic progressive EAE as previously described [41, 42] with 200 μg of MOG35-55 peptide (MEVGWYRSPFSRVVHLYRNGK; Stanford Pan Facility, Stanford, CA) in 4 mg/mL Mycobacterium tuberculosis (H37Ra) in incomplete Freund’s adjuvant (IFA) (Difco Laboratories, Detroit, MI). Pertussis toxin (200 ng, List Biologicals, Campell, CA) was delivered intraperitoneally that day and 2 days later. Mice were scored daily on a scale from 0 to 5 for severity of clinical symptoms: 0–0.5 indicated no disease/distal limp tail, 1.0 limp tail, 2.0 weakness in one or 2.5 in both hind limbs/slipping on bars, 3.0 paralysis in one or 3.5 both hind limbs, 4.0 hind limb paralysis plus weakness in one or in 4.5 both forelimbs, and 5 for moribund animals.

For qPCR and western blotting analyses, C57BL/6 mice spinal columns and brains were isolated from healthy mice and EAE mice immediately following CO₂ exposure and decapitation at the following intervals: day 7, day 10 (pre-onset), day 14 (onset), day 18 (peak disease), day 25 (recovery), and days 32 and 45 which reflect some degree of worsening. Controls included healthy animals or sham/CFA-immunized (no MOG) animals that also received pertussis toxin harvested at days 7, 14, 25, and 45. For immunohistochemistry, animals were anesthetized with a ketamine/xylazine formulation, with subsequent transcardial perfusion with PBS followed by buffered formalin. Tissues were fixed for 3–5 days, and spinal cords removed before cryoprotection in 30% sucrose solution for at least 5 days. Spinal cords were cut into 5 mm pieces representing five to seven different levels and embedded in optimal cutting temperature compound (OCT) (VWR, Radnor, PA) and frozen until sectioning at 5 μM on a cryostat.

Sectioned tissues were permeabilized with 0.05% triton and blocked with 10% normal goat or donkey serum. Tissues were incubated with antibodies to NeuN, microtubule-associated protein 2 (MAP2), a neuronal marker) and glial fibrillary acidic protein (GFAP, a marker of astrocytes, all from EMD Millipore, Etobicoke, ON), myelin basic protein (MBP, Santa Cruz Biotechnology, Dallas, TX), SMI-32 (Biolegend, San Diego, CA) or ARNT2 (Santa Cruz) overnight. After incubation, tissues were washed with TBST followed by incubation with secondary antibody at room temperature. Normal goat, chicken, rat, mouse, and rabbit IgG (Santa Cruz) served as isotype controls.

Tissue sections and cells were visualized on a Zeiss Axio Vert 200 Inverted Fluorescence Microscope (Oberkochen, DE), and images acquired with an Axiocam 506 monochrome camera. Analysis was performed with Zeiss software Zen (Version 2.3). To quantify immunostaining results, sections from five to seven spinal levels spanning the entire cord (C4/5, T1/2, T5/6, T12/13, L5/6, S3/4) were examined from seven healthy/CFA control-immunized mice and five EAE mice at day 18 for a total of 26–40 sections per treatment group. Disease scores ranged from 2.5 to 4.0 with an average score of 3.25 ± 0.67, and all animals showed evidence of inflammatory infiltration of the spinal cord. Images were captured at × 20 with identical light intensity and exposure times applied to all images from each experimental set. ARNT2 expression in neurons was quantified by measuring the average intensity per pixel for ARNT2 over the area of NeuN-positive nuclei. First, a binary mask was created to select only the gray matter (based on regions of NeuN staining positivity) and then another one to select all nuclei of the section using the DAPI staining. Nuclei with a high fluorescence intensity (set at greater than 1500 mean intensity per pixel for NeuN) were selected, and the average fluorescence intensity per pixel for ARNT2 over the area of the DAPI-positive nucleus was measured. Neuronal cells were quantified by counting the NeuN/DAPI⁺ cells per square millimeter in the whole gray matter region, and ARNT2 intensity was quantified in cells in this larger region as well as those in the dorsal horn at each level based on anatomical outlines in the © 2008 Allen Institute for Brain Science Allen Spinal Cord Atlas available from http://​mousespinal.​brain-map.​org. The values of all of the cells within one side of the spinal cord were first normalized to the average value for ARNT2 mean intensity measured in central canal ependymal cells (negative for ARNT2 expression) to account for any background staining between sections. The values for each spinal cord section within this region from each mouse were then pooled and averaged for analysis. The number of astrocytes in the gray matter was assessed using the same gray matter mask to select all the nuclei in the gray matter with DAPI staining. The nuclei with a high fluorescence intensity (greater than 1600 mean intensity per pixel) for GFAP were selected as GFAP⁺ cells and normalized to the surface area of the gray matter mask to obtain cell number per square millimeter.

Immunostained sections from five levels of each EAE mouse spinal cord (n = 5, total of 25 levels) were assessed for the degree of inflammatory infiltration according to the criteria in Additional file 1: Table S1. MBP-negative regions of the white matter were outlined in Zen, and demyelinated areas (in μm²) were summed from each level and divided by the area of the entire white matter with the same binary mask approach used for gray matter to get a % area of demyelination for each level. SMI32-positive regions of interest, or spheroids, were selected as those regions with a staining intensity above the threshold set by averaging regions of white matter in healthy animals; the number of SMI32+ spheroids was evaluated per square millimeter of white matter at each of five levels from five EAE mice as for infiltrates. The relative sum of inflammation, % demyelination, and SMI32⁺ staining per level of the spinal cord was compared to the relative intensity of ARNT2 in neurons at that level.

Embryonic day 18 (E18) rat cortices were graciously provided by Dr. Shernaz Bamji (University of British Columbia) and were cultured as previously described [43]. Cortices were washed with 37 °C Hank’s Balanced Salt Solution (HBSS, Gibco, Carlsbad, CA), incubated in 0.25% Trypsin-EDTA (Sigma-Aldrich®, St. Louis, MO), DNase (Sigma-Aldrich) and HBSS, and dissociated and seeded at 50,000 cells/cm² for western blotting/qPCR and 35,000 cells/cm² for immunocytochemistry on poly-l-lysine-coated plates in Neurobasal® media (Thermo Fisher Scientific, Waltham, US) supplemented with B-27 (Gibco, Thermo Fisher Scientific). Cells were maintained at 5% CO₂ and 37 °C until experiments at 7–16 days in vitro (DIV). Hydrogen peroxide (H₂O₂, Fisher, Waltham, MA) was used to mimic oxidative damage, and staurosporine (Sigma-Aldrich), a non-selective protein kinase inhibitor, was used to induce apoptosis, modeling pathological processes associated with degeneration in MS [44, 45].

Cells were lysed in radioimmunoprecipitation assay buffer (RIPA, tris-hydrochloric acid (MP Biomedicals, Santa Ana, CA), sodium chloride (Fisher), sodium deoxycholate (Fisher), NP-40 (Thermo Fisher Scientific, sodium dodecyl sulfate (SDS, MP Biomedicals), and EDTA (MP Biomedicals)) with 5% protease inhibitor (Roche, Basel, CH) and lysate stored at − 20 °C. Samples run on 12% tris/glycine gels (Bio-Rad, Hercules, USA) were transferred to nitrocellulose membranes, blocked with 2% skim milk, and incubated with primary antibodies overnight at 4 °C. Membranes were washed and incubated with secondary antibody at room temperature with Clarity™ Western ECL Blotting Substrate (Bio-Rad) added before imaging on a Bio-Rad ChemiDoc™ MP System. Blots were analyzed using quantitative densitometry on Image Lab™ (Bio-Rad 5.2) and normalized to tubulin as a loading control following normalization of data on different blots to a positive control spinal cord lysate that was loaded on each blot run; this enabled blot to blot comparisons when run in an identical manner and using the same exposure times.

Tissues were processed with the AllPrep kit (Qiagen, Hilden, DE) per manufacturer guidelines and RNA stored until use. In vitro cell culture RNA was isolated using the Qiagen RNeasy Mini Kit, and protein was precipitated using buffer APP (Qiagen) per manufacturer’s guidelines). RNA was measured using a NanoDrop™ spectrophotometer (Wilmington, USA). Reverse transcription was performed using the QuantiTect® Reverse Transcription Kit (Qiagen) with Fast Start Essential DNA Green Master (Roche). cDNA, SYBR Green and ultrapure dH₂O were combined and loaded on to LightCycler® 480 Multiwell plates (Roche) and run on a LightCycler® 480 Instrument II (Roche); qPCR primers are included in Additional file 1: Table S2. Data was subsequently analyzed using the LightCycler® 480 software (Version 1.5, Roche). Standard curves (Arnt2, Npas4, Bdnf, and β-actin cDNA generated from mouse brain, diluted in 10-fold successions) were included on each run for absolute quantification. Precipitated proteins from spinal cord were quantified and analyzed by western blotting as outlined above for cell cultures.

Cell viability measurements were performed using a lactate dehydrogenase (LDH) cytotoxicity assay kit (Thermo Fisher Scientific™ Pierce™) per protocol guidelines. Controls included samples from dead cells (killed with lysis buffer provided for maximum LDH release) against which % cytotoxicity of treatments could be measured. Absorbance was measured using a Spectra Max plate reader (Molecular Devices, Sunnyvale, USA).

Cells in culture were fixed with 10% phosphate-buffered formalin (Fisher), permeabilized with 0.1% Triton-X 100 (Fisher), and blocked with 10% normal goat serum (Gibco). Cells were washed with PBS, and incubated with primary antibodies for ARNT2, NeuN, MAP-2, GFAP, or cleaved caspase 3 (R&D Systems, Minneapolis, MN) in 2% normal goat serum overnight at 4 °C and secondary antibodies at RT (Additional file 1: Table S2). After counterstaining with 4′,6-diamidino-2-phenylindole (DAPI), images were acquired and analysis was performed with Zeiss software Zen (Version 2.3). ARNT2 staining was examined in NeuN-, GFAP-, or MAP-2-positive cells and cleaved caspase 3 in neuronal cells.

We began our investigation by characterizing ARNT2 levels in the spinal cord where inflammation and degeneration are most prevalent in the MOG-induced model of MS. This is also the region where tissue changes are thought to contribute to disability in this model. qPCR analysis of chronic EAE spinal cord samples was performed with tissues from pre-onset, onset, peak, recovery/stabilization, and worsening stages of the disease compared to sham-immunized controls or healthy non-immunized littermates (Fig. 1a). Compared to CFA-immunized littermates, Arnt2 levels in the spinal cord are decreased by onset (day 14) and are lowest at peak disease (Fig. 1b). Notably, Arnt2 message levels have begun to increase again by day 25 where the animals have shown some recovery. By day 45 however, the RNA levels are still significantly below levels observed in healthy or sham-immunized animals. We also examined expression of the primary CNS partner for ARNT2, Npas4, and one of their downstream targets, Bdnf, over the course of EAE (Additional file 2). While Arnt2 message tended to show increases prior to disease onset, both Npas4 message and Bdnf message were elevated compared to sham-immunized animals prior to disease onset (at days 7 and 10 post EAE induction respectively). Like ARNT2, expression of Npas4 and Bdnf decreased significantly over the course of EAE, with levels at their lowest at peak disability and recovering as animals recovered. Notably, ARNT2 protein was also lowest at peak clinical disease on day 18 (Fig. 1c) and was the only time point where ARNT2 protein levels were significantly different than healthy or sham CFA-immunized control mice.

We next examined the expression of ARNT2 in the spinal cord of EAE animals at peak disease compared to controls, localizing ARNT2 to distinct cell populations and regions. In the spinal cords of sham-immunized animals, ARNT2 was localized to regions of both the gray as well as the myelin-rich white matter. In gray matter, ARNT2 staining was frequently associated with the nuclei of NeuN⁺ cells, either larger motor neurons in the ventral horns (arrows) or smaller cells of the dorsal horns or intermediate regions (arrowheads, Fig. 2a). While ependymal cells lining the central canal were negative for ARNT2 expression (Fig. 2b), ARNT2 was commonly associated with the nuclei of GFAP⁺ cells lining the midline and surrounding the central canal (insets, Fig. 2b, c respectively). ARNT2 positivity was also detected in several GFAP⁺ astrocytes lining the meninges.

Compared to the gray matter of healthy animals which typically showed numerous ARNT2⁺/NeuN⁺ cells, ARNT2 expression was detected at lower levels in the NeuN⁺ cells in EAE mice (Fig. 3). In the gray matter of an EAE spinal cord with moderate disability but at a higher level of the spinal cord (C4/5) with few infiltrates (Fig. 3a, b), there were far fewer double-staining NeuN⁺ cells in the EAE mouse than in the healthy mouse; this pattern was prominent in the ventral horn and intermediate zones of the gray matter though it seemed most pronounced in many levels within the dorsal horns (Fig. 3b, c insets). Notably, there was also a significant reduction in the number of NeuN⁺ cells/mm² of gray matter in EAE animals compared to healthy controls (Fig. 4a). The reduction observed when samples were pooled from all levels of the cord (cervical, thoracic, and lumbar) was 535.9 ± 23.3 in healthy vs 281.7 ± 26.3 in EAE animals (p < 0.001); the greatest differences were observed in the thoracic and lumbar levels while differences in the cervical regions were not significant (430.7 ± 34.8 vs 389.9 ± 36.3, data not shown). Using a threshold for negative staining set by ARNT2 staining of the ependymal cells in each section, we measured the mean fluorescence intensity per pixel of each NeuN⁺ nucleus to compare healthy animals to EAE animals. We analyzed the entire gray matter and also performed a separate analysis of cells in the dorsal horn only. Separate from the determination of positive cells, we also distinguished the percent of moderate (> 1.5× negligible intensity threshold) versus highly staining cells (> 2× negligible intensity threshold) as illustrated in Fig. 4b. We did not detect a significant difference in the % of NeuN⁺ cells which were positive for ARNT2 between the two groups, 65.8 ± 3.1 versus 66.6 ± 3.4% in healthy vs EAE animals respectively over the entire gray matter region; numbers were also similar when only the dorsal horns were compared in the healthy (66.6 ± 3.4%) vs. EAE animals (63.0 ± 4.1%). However, the relative intensities of the stained cells were significantly lower in EAE animals than healthy mice (Fig. 4c). Over the entire gray matter region, the percentage of NeuN⁺ cells staining moderate to high was significantly lower in EAE animals (33.7 ± 2.6% of cells vs 22.6 ± 2.7%, p = 0.0065). This decrease was particularly marked in highly staining ARNT2 cells which dropped from 11.0 ± 1.4% of NeuN⁺ cells in healthy animals to 5.1 ± .9% of NeuN⁺ cells in EAE animals (p = 0.0024). Decreases in ARNT2 expression were even more pronounced within the dorsal horns; the proportion of moderate to highly stained NeuN⁺ cells dropped by 50% (36.0 ± 3.0% in healthy vs 18.2 ± 3.0% in EAE mice) whereas the proportion of highly expressing ARNT2 cells dropped by more than 80% (10.1 ± 1.7% in healthy vs 1.9 ± .9% in EAE mice). The number of GFAP⁺ cells in the gray matter of EAE mice was also significantly increased compared to healthy sham-immunized mice (Fig. 4d, e; 419.2 ± 40.4 vs 161.7 ± 14.1 cells/mm², p < 0.0001). In Fig. 4f, notably, the intensity of ARNT2 expression in NeuN⁺ cells correlated inversely with the degree of immune cell infiltration. Declines in neuronal ARNT2 intensity were correlated with increased infiltrates (r = − 0.508, R² = 0.259, p = 0.0047). In this EAE model, regions of immune cell infiltration often correlated with areas of demyelination (reduced staining for myelin basic protein, MBP) and also axonal damage which is apparent as axonal spheroids detectable by staining for non-phosphorylated neurofilaments with the SMI32 antibody [26, 46]. As expected, demyelination was not detected in our sham immunized animals (Fig. 4g). In contrast, all levels of EAE mice examined exhibited some degree of demyelination, typically associated with areas of inflammatory infiltration; the degree of demyelination ranged from 1.4 to 21.5% of the white matter area in a single level with an average of 4.7 ± 0.9% demyelination over all levels examined from EAE mice. Similarly, few SMI32⁺ axons were detected in sham-immunized animals at any level. In contrast, SMI32⁺ spheroids were detected in all levels of EAE animals examined, with an average of 62.7 ± 13.3 (range 6 to 300) spheroids per square millimeter. In EAE animals, we found a direct correlation between the frequency of SMI32⁺ spheroids and the degree of demyelination (r = 0.878, R² = 0.772, p < 0.0001 (data not shown). Notably, the frequency of SMI32⁺ spheroids or axon “rings” stained with SMI32 (as in insets) per square millimeter correlated inversely with the ARNT2 intensity of NeuN⁺ cells at each level of the EAE mice examined (Fig. 4h; r = − 0.376, R² = 0.141, p = 0.032). The degree of demyelination did not correlate with the ARNT2 intensity as closely throughout all levels examined (Fig. 4i; r = − 0.229, R² = 0.052, p = 0.136). Examination of different levels of the spinal cord showed that SMI32⁺ were frequent in areas of extensive demyelination yet were also detected in areas where demyelination was less apparent, such as those in higher/thoracic levels of the cord (Fig. 4f, insets).

After establishing the relationship between inflammation and ARNT2 expression in vivo, we set out to examine factors which alter ARNT2 expression as well as the outcome of decreasing ARNT2 expression in vitro. Cultures routinely consist of 92.6 ± 6.9% MAP2⁺ neuronal cells (n = 3) by DIV10-14 (Additional file 3: Table S3) with some GFAP⁺ astrocytes (7.4 ± 6.9%). Healthy neurons, characterized by large cell bodies with complex axonal and dendritic processes, ranged in nuclear area from 50 to 200 μm². Occasional astrocytes exhibited a reactive morphology with extensive ramifications in comparison to non-reactive astrocytes that are more flat and polygonal [47].

Staurosporine potently induces apoptosis by preventing ATP binding to kinases, preventing protein kinase activity and resulting in growth factor withdrawal [48]. Reduced growth factor activity is linked to caspase-3 and cytochrome c activity, resulting in activation of the apoptotic pathway [49]. We tested staurosporine at a range of doses from 5 to 1000 nM for 24 h, where previous studies showed neuronal apoptosis in vitro with 500–1000 nM [48, 49]. Ten nanomolars of staurosporine increased ARNT2 levels by greater than 50% compared to vehicle-treated cells at 24 h (Fig. 5a, n = 3, p ≤ 0.05). At 500 to 1000 nM, cell death became evident by phase microscopy showing numerous cell bodies and cellular debris (Fig. 5b), which was not observed with ARNT2 levels nearer to vehicle controls. We next examined ARNT2 and cleaved caspase 3 intensities in enriched neuronal cultures and found that 6 or 12 h of exposure to 1 μM staurosporine significantly decreased ARNT2 expression and these decreases in ARNT2 were accompanied by increases in cleaved caspase 3 (Fig. 5c–e). In staurosporine-treated wells, cells staining brightly for cleaved caspase 3 typically showed negligible levels of ARNT2 compared to vehicle-treated controls.

To model mild to severe oxidative stress in primary cortical-enriched neuronal cultures, we applied 25, 50, 100, and 300 μM H₂O₂ (Fig. 6). Twenty-five micromolars of H₂O₂ was associated with a significant increase in ARNT2 tested over 0.5 to 3 h compared to untreated controls (Fig. 6a, n = 6, p ≤ 0.01). Notably, phase microscopy showed no changes in cell morphology or signs of blebbing and LDH release was also negligible over the 24 h treatment period (Fig. 6a). There were no changes in ARNT2 from 10 to 24 h in longer term experiments (data not shown).

Similar to 25 μM, 50 μM H₂O₂ increased ARNT2 protein levels with peak levels measured at 2 h after addition (Fig. 6b, n = 6, p ≤ 0.05); ARNT2 levels then returned to baseline by 3–4 h and ultimately dropped below baseline. The dropping of ARNT2 levels below baseline was not associated with morphological alterations within the first 5 to 10 h (Fig. 6e). Cell cytotoxicity measured by LDH release, while not apparent at 6 h, was significantly increased to 15–17% cytotoxicity by 12 and 18 h of cell exposure.

Exposure to a higher dose of H₂O₂ (100 μM) increased detectable ARNT2 protein at 0.5 and 1 h compared to controls (Fig. 6c, n = 7, p ≤ 0.05). By 3–4 h exposure, ARNT2 levels dropped below baseline in most experiments, but cells maintained their complex morphology through 5 h of treatment. Increases in LDH release became significant by 6 h (p ≤ 0.05) and were followed by a relative loss of dendrites and increased neuronal blebbing, nuclear condensation, and cell death apparent by 10 h (Fig. 6e). Notably, exposure to the highest tested dose H₂O₂ at 300 μM did not significantly increase ARNT2 levels and instead was associated with significant declines in ARNT2 expression beginning at 3–4 h which preceded blebbing of many cells by 5 h and loss of all dendritic morphology by 10 h (Fig. 6e).

We analyzed mRNA for Arnt2, Npas4, and Bdnf in H₂O₂-treated neuronal-enriched cultures compared to PBS-treated cells. Notably, Arnt2 mRNA was not influenced by oxidative stress in primary cortical neuron-enriched cultures, indicating no relative change in transcript levels (Additional file 4). Over the time intervals where ARNT2 protein changed (0.5 to 2 h in response to H₂O₂), we saw no change in ARNT2 message. However, Npas4 mRNA increased rapidly and specifically in response to oxidative stress and was followed closely by increases in Bdnf mRNA.

We next examined the relationship between declines in ARNT2 intensity and expression of cleaved caspase 3. Using early DIV7 neuronal cultures, we minimized the likelihood of glial contamination with the potential for MAP2 losses by neurons following exposure to stressors or undergoing cell death or apoptosis. Exposure to 25 μm H₂O₂ for 1 h had no effect on ARNT2 expression but 50 μM significantly reduced the mean ARNT2 intensity in neurons (Fig. 7a, p < 0.05). In this 1 h of exposure, there were no increases in cleaved caspase 3 expression in either group (Fig. 7b). In contrast, 3 h exposure to both 25 and 50 μM of H₂O₂ both significantly reduced ARNT2 intensity in neurons (Fig. 7c). Only exposure to the greater dose of H₂O₂ was associated with increases in cleaved caspase 3, although the majority of these cells also showed condensed nuclei (Fig. 7b, c) indicative of reduced cell viability that was also apparent in our earlier phase micrographs (Fig. 6e).

We also examined ARNT2 expression at the individual cell level in our enriched neuronal cultures; 15–20% of the cells in the culture stained negligibly for ARNT2 (Fig. 8a). The majority of cells (60–70%) were low to moderate in their expression of ARNT2 while 10–15% had more intense ARNT2 staining. Notably, the few contaminating astrocytes were also positive for ARNT2. We tested the same doses of H₂O₂ (25 μM and 100 μM) previously used, focusing on earlier time points (0.5, 1, 2, 4, and 6 h). Total DAPI⁺ cells, MAP2⁺ cells, and GFAP⁺ cells were analyzed (Fig. 8b). With 100 μM H₂O₂, there was a significant increase in ARNT2 protein detected at 1 h exposure compared to untreated controls (n = 3, p ≤ 0.05 for DAPI⁺, 0.001 for MAP2⁺, and 0.01 for GFAP⁺ cells); this matched the increases in ARNT2 observed by western blotting (Fig. 6c). The staining intensity of ARNT2 was at or below baseline in neurons and astrocytes treated for 2–4 h and reached levels significantly lower than baseline (i.e., all three data groups, n = 3 for each; at 2 h: p ≤ 0.01 for GFAP⁺; at 4 h: p ≤ 0.05 for DAPI⁺, p ≤ 0.01 for GFAP⁺; at 6 h: p ≤ 0.001 for all three) which preceded changes in cell viability. Using this cell-specific analysis, we report a significant early upregulation of ARNT2 by 100 μM H₂O₂ in both neurons and astrocytes, which after extended exposure results in similar decreases in ARNT2 expression that precede cytotoxicity as in Fig. 6a–d.