Overexpression of the ubiquitin‐editing enzyme A20 in the brain lesions of Multiple Sclerosis patients: moving from systemic to central nervous system inflammation

Human post‐mortem snap‐frozen brain tissues

For this study, 38 brain tissue blocks obtained from five control cases (CC) with no neurological diseases and 11 secondary progressive (SP) and six primary progressive (PP) MS cases were used. Demographic and clinical features are reported in Table 1. Post‐mortem brains were cut into blocks (2x2x1cm) and unfixed blocks were immediately frozen by immersing in isopentane, pre‐cooled on a bed of dry ice and stored at −75°C. Human post‐mortem brain tissues were obtained from the United Kingdom (UK) MS Tissue Bank at Imperial College via a UK prospective donor scheme with full ethical approval (08/MRE09/31). Neuropathological confirmation of the diagnosis of MS was carried out according to the International Classification of Diseases of the Nervous System criteria (www.ICDNS.org), and confounding pathologies were excluded, including Alzheimer‐like changes (56).

The project has been approved by the University Bioethics Committee of the University of Turin on 28 February 2015.

Neuropathological characterization of human post‐mortem snap‐frozen brain tissue by immunohistochemistry

Air‐dried 10‐µm‐thick cryosections cut from snap‐frozen tissue blocks were characterized by immunohistochemistry (IHC) with specific antibodies. The complete list of used antibody, the working dilution and the postfixation method are reported in Table 2. Immunostaining for myelin‐oligodendrocyte glycoprotein (MOG) was performed to discriminate presence of white matter (WM) and grey matter (GM) and for the occurrence and extension of demyelinated lesions in WM (WML) and GM (GML). Immunostaining for major histocompatibility complex class II (MHC II) molecules was performed to assess the stage of lesion activity. The polyclonal anti‐A20 antibody was used to detect A20.

Primary antibodies indicated in Table 2 were diluted in PBS containing 0.1% of Triton X‐100 and 10% of normal serum and incubated overnight at 4°C. Next, sections were incubated with the appropriate biotinylated secondary antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA), followed by avidin‐biotin peroxidase complex (ABC; Vector Laboratories Burlingame, CA). The immune reactions were visualized with 3′3‐diaminobenzidine (DAB, Sigma, St. Louis, MO). Hematoxylin was used to counter‐stain nuclei in the slices. All the sections were sealed with DPX (Sigma, St. Louis, MO).

Control blocks contained normal WM and GM areas, while MS blocks included normal‐appearing WM (NAWM), normal‐appearing GM (NAGM), and lesions. Pre‐active (pre‐AL), active (AL), and chronic active (CAL) WM lesions (WML) and active (aGML) or inactive (iGML) GM lesions (GML) were identified based on the anti‐MHC II immunostaining according to Reynolds and colleagues (56).

The quantification of the A20 immunoreactivity in the lesions and in the normal‐appearing areas was performed on IHC images obtained by frozen brain sections deriving from 10 SPMS and 3 PPMS cases. A picture containing lesions and neighboring normal‐appearing matter was obtained for each tissue block by combining several IHC images acquired at 4X magnification with Nikon Eclipse microscope. Total coverage area, number of sections, mean and standard deviation (SD) of the section area were the following: 2.6 × 10⁸ µm² (8 sections, mean 3.7*10⁷ µm², SD 4.6 × 10⁷) for the AL; 1.2 × 10⁸ µm² (12 sections, mean 1.1 × 10⁷ µm², SD 9.6 × 10⁶) for the CAL; 2.8 × 10⁸ µm² (20 sections, mean 1.4 × 10⁷ µm², SD 1.6 × 10⁷) for the matched NAWM; 1.6 × 10⁷ µm² (27 sections, mean 5.8 × 10⁵ µm², SD 8.5 × 10⁵) for the aGML; 2.7 × 10⁶ µm² (9 sections, mean 3.0 × 10⁵ µm², SD 2.0 × 10⁵) for the iGML; 1.3 × 10⁷ µm² (36 sections, mean 3.5 × 10⁵ µm², SD 3.2 × 10⁵) for the NAGM.

Images were acquired and analyzed with Nikon Eclipse microscope. A20 densitometric analysis was manually performed on IHC images using ImageJ. Signal intensity, corresponded to the mean pixel intensity of the area, ranged from 0 (=white or total absent) to 255 (=black or total present). The A20 optic density (OD) in lesions area was normalized for the background by subtracting the OD value of the normal‐appearing neighboring area.

Immunofluorescence

For indirect double immunofluorescence (IF), air‐dried 10‐µm‐thick cryosections cut from snap‐frozen tissue blocks were stained with the A20‐specific polyclonal antibody and the following mouse monoclonal antibodies reported in Table 2: anti‐CD4, anti‐CD8, anti‐CD68, anti‐CD20, anti MHC II, anti‐GFAP, and anti‐NEFH. For antibodies directed against CNS markers (i.e., anti‐MHC II, anti‐GFAP, and anti‐NEFH antibodies) a final incubation with 0.05% Sudan Black B (SBB) (Sigma Aldrich) for 5 minutes a room temperature (RT) was performed to efficiently reduce the auto‐fluorescence signal of the snap‐frozen brain tissues. Staining was visualized using secondary antibodies Alexa Fluor 488 conjugated donkey anti‐mouse IgG and Alexa Fluor 555 conjugated donkey anti‐rabbit IgG (Invitrogen, Eugene, OR). Sections were sealed in ProLong Gold anti‐fade reagent with 4′,6′‐diamidino‐2‐phenylindole (DAPI, Invitrogen, Carlsbad, CA). Images were acquired using Leica TCS SP5 confocal microscope.

Total RNA extraction, retrotranscription, and gene expression analysis from human brain tissues

For RNA extraction, 50‐μm‐thick cryosections were cut from selected snap‐frozen tissue blocks from CC and MS cases. The area of interest (WM, GM, NAWM, NAGM, and relative lesions) were identified based on immunohistochemistry characterization with anti‐MOG and anti‐MHC II antibodies and manually dissected. The dissection was performed by gently incising the frozen tissue blocks with a needle, before performing the cut of the tissue block. Two or three cryosections were obtained from each tissue block, depending of the size of the area of interest.

Total RNA was extracted from the cryosections using the RNeasy Lipid Tissue Mini Kit (#74804, Qiagen, Hilden, Germany), according to the manufacturer’s instruction. Total RNA was reverse‐transcribed at final concentration of 20 ng/μL using random hexamer primers (High‐Capacity cDNA Reverse Transcription Kit, #4368814, Thermo Fisher Scientific). Gene expression analysis was performed by real‐time PCR using Applied Biosystems’ TaqMan gene expression products (Thermo Fisher Scientific) for the following genes: A20, TNFR1, TNFR2; TNF, IL1B, CXCL10, TGFb, IL10, IL6, NF‐kB, STAT3, TRAFF6, CASP3, CASP8, BCL2, and XIAP. Transcriptional expression was normalized using GA3PDH as a housekeeping gene. Expression levels of the analyzed genes were calculated by the normalized comparative cycle threshold (Ct) method (2^−ΔΔCt), using the Universal Human Reference RNA (Stratagene, Santa Clara, California) for calibration.

Statistical analysis

Normality of distributions was assessed by graphical inspection and by the Shapiro–Wilk test. A20 OD was compared between lesioned and neighboring normal‐appearing areas of MS cases by means of the paired Mann–Whitney U test. Normalized OD signals of AL and CAL and of iGML and aGML were compared by the Mann–Whitney U test. The association between A20 levels and clinical and demographical variables was evaluated by regression models. Kruskal–Wallis with Dunn post hoc test was applied to compare gene expression data between CC samples and normal‐appearing or lesioned areas of MS cases. Lesioned and neighboring normal‐appearing areas of MS cases were compared by the paired Mann–Whitney U test. p‐values were adjusted for multiple comparisons using the false discovery rate (fdr) correction. Spearman correlation was evaluated between gene expression data. Statistical analyses were performed using R software 3.5.3 version (www.r‐project.org).

Characterization of A20 protein expression in human post‐mortem control brain tissues

To investigate the A20 protein expression in brain, we performed IHC and double IF using A20‐specific antibodies on brain slices from 5 CC. In both WM (Figure 1A‐D) and GM (Figure 1E‐J) of CC, the IHC revealed the A20 expression in cells morphologically resembling astrocytes (Figure 1C‐C′ and I‐I′) and the double confocal IF confirmed that A20 is basally expressed by scattered GFAP‐positive (GFAP+) astrocytes (Figure 1D, J and Figure S1). In GM, the IHC also showed the A20 expression in the perinuclear area of cells morphologically resembling neurons (Figure 1G‐G′), as confirmed by double confocal IF showing a population of NEFH+ neuronal cells expressing A20 (Figure 1H and Figure S1). We did not detect A20 in MHC II+ microglial cells in either WM or GM (data not shown).

Characterization of A20 protein expression in NAWM and WML in human post‐mortem MS brain tissues

We further evaluated A20 expression in human post‐mortem brain tissues from progressive MS patients, both in normal‐appearing and demyelinated areas of WM and GM. In NAWM (Figure 2A‐E′) of MS cases, we detected A20 expression on ramified cells resembling astrocytes (Figure 2C‐C′). Double confocal microscopy analysis demonstrated that the A20+ cells were mostly GFAP+ astrocytes, although not all the astrocytes expressed A20 (Figure 2D‐D′ and Figure S2). However, differently from the CC WM, we observed A20 expressed also by few MHC II+ microglia cells (Figure 2E‐E′ and Figure S2). In the demyelinated area (Figure 2F‐M′), including pre‐AL, AL, and CAL, we detected A20 expressed by a huge number of highly ramified resident cells (Figure 2H‐H′ and L‐L′). Particularly in CAL we observed a greater number of ramified cells expressing A20 in the active plaque rim compared to few A20+ cells in the lesion core (Figure 2L‐L′). Double confocal microscopy clarified that in WML A20 is expressed by all the GFAP+ astrocytes (Figure 2I‐I′ and Figure S2) and by numerous MHC II+ microglia cells (Figure 2M‐M′ and Figure S2). In addition, we detected A20 in perivascular infiltrates in or near the active lesions (Figure 3C‐C′). Double confocal microscopy showed that A20 was not expressed by lymphocyte populations including CD4+ T helper (Th), CD8+ cytotoxic T cells, and CD20+ B cells (Figure 3D‐O′). However, it was expressed by almost all the CD68+‐activated microglia/macrophages (Figure 3P‐S′).

Characterization of A20 protein expression in NAGM and GML in human post‐mortem MS brain tissues

In the NAGM (Figure 4A‐H), the A20 expression was observed in the perinuclear area of groups of cells morphologically resembling neurons (Figure 4C‐C′) and in rare scattered ramified cells (Figure 4G‐G′). Confocal IF analysis revealed that, similar to CC GM, in NAGM A20 is expressed in the soma of some NEFH+ neurons (Figure 4D and Figure S3) and in sporadic GFAP+ astrocytes (Figure 4H and Figure S3). On the contrary, in aGML (Figure 4I‐L) A20 expression was detected in a network of ramified cells (Figure 4K‐K′), demonstrated to be GFAP+ astrocytes by double confocal IF (Figure 4L and Figure S3). We did not observe a co‐localization of A20 and MHC II+ staining (data not shown), but this could be due to the intense A20 signal of the astrocytes network, thus, we could not exclude that A20 is also expressed by some microglia cells.

Quantification of A20 protein expression in human post‐mortem MS brain tissues

To quantify the A20 protein expression in both WML and GML, we performed a densitometric analysis on IHC images from 13 MS cases (10 SPMS and 3 PPMS). We excluded major differences in the A20 expression pattern between PPMS and SPMS by graphical inspection. When we examined all the MS cases together, for WML, we measured A20 optic density (OD) in 8 AL, 12 CAL, and 20 neighboring NAWM areas, classified based on MOG and MHC II immunostaining. For AL, we considered the entire area of the lesion, while for CAL only the active rim. A20 OD was significantly higher in both AL and CAL compared to their neighboring NAWM (paired Mann–Whitney U test, P = 0.0078 and P = 0.0005, respectively) (Figure 5A,B). Furthermore, we compared normalized OD signals of AL and CAL, obtained by subtracting the OD value of the neighboring NAWM area to that of the lesion. As a result, we did not observe significant differences (Mann–Whitney U test, P = 0.34) (Figure 5C).

For GML, we measured A20 OD in 27 aGML, 9 iGML, and 36 neighboring NAGM areas, classified through MOG and MHC II immunostaining. We found higher A20 immunoreactivity in the aGML as compared to the closest NAGM areas (paired Mann–Whitney U test, P < 0.0001) (Figure 5D). In contrast, we did not observed significant differences between A20 immunoreactivity in iGML and neighboring NAGM (paired Mann–Whitney U test, P = 0.10) (Figure 5E). As expected, A20 normalized expression was higher in the aGML compared to iGML (Mann–Whitney U test, P < 0.0001) (Figure 5F).

Also, we did not highlight relevant correlations between A20 immunoreactivity in both WML and aGML and plaque dimension, demographical and clinical features of subjects, such as gender, MS type, disease length, age and cause of death, and post‐mortem delay (data not shown).

Gene expression analysis of A20 and related molecules in human post‐mortem CC and MS brain tissues

We further investigated the snap‐frozen brain tissues of CC and MS cases by quantifying the gene expression level of A20 (Figure 6) and other 13 genes belonging to the A20 pathways and involved in the regulation of inflammation and apoptosis (Figure S4). In particular, we analyzed 4 WM and 4 GM samples from 4 CC and 12 NAWM, 9 WML, 12 NAGM, and 12 GML samples from 10 progressive MS patients. Given the high variability between‐subjects in the gene expression data, we first compared CC samples (WM and GM) with normal‐appearing and lesioned areas of MS cases (Figure 6A,C), then, we restricted the comparison of normal‐appearing and lesioned areas of MS cases to matched samples from the same slice of brain tissue blocks (Figure 6B,D).

The A20 expression level was significantly higher in WML compared to both WM of CC and matched NAWM (WM vs. WML: Kruskal–Wallis with Dunn post hoc test, P = 0.048; NAWM vs. WML: paired Mann–Whitney U test, P = 0.008) (Figure 6A,B). A20 was more expressed also by GML compared to matched NAGM, although the difference was less pronounced than in the WM (paired Mann–Whitney U test, P = 0.02) (Figure 6D). On the contrary, we did not observe significant differences between NAGM or aGML and GM of CC (Kruskal–Wallis test, P = 0.27) (Figure 6C).

None of the genes belonging to the A20 pathways and involved in the regulation of inflammation and apoptosis showed significant differences among GM tissues (Figure S4). However, most of them showed differences between WM tissues. TNF receptors 1 (TNFR1) and 2 (TNFR2), whose stimulation by TNFα induces an A20 upregulation, were higher expressed in the WML compared to the matched NAWM (paired Mann–Whitney U test, P = 0.02 and P = 0.02, respectively) (Figure S4A‐B). The pro‐inflammatory cytokines TNFα and IL‐1β, both able to upregulate A20 expression through the binding to their receptors, showed higher levels in the WML compared to the neighboring matched NAWM (paired Mann–Whitney U test, P = 0.02 and P = 0.02, respectively) (Figure S4C‐D). However, we did not observe any difference in the expression of the pro‐inflammatory cytokine IL6 (Figure S4E). The anti‐inflammatory/regulatory cytokines IL10 and TGFβ were highly expressed in the WML compared to the matched NAWM (paired Mann–Whitney U test, P = 0.04 and P = 0.008, respectively) and to WM of CC, although for TGFβ this difference was not significant after adjusting for multiple comparisons (Kruskal–Wallis with Dunn post hoc test, P = 0.03 and P = 0.11, respectively) (Figure S4F‐G). Regarding the apoptotic pathways, we observed higher expression levels of the caspases CASP3 and CASP8 in the WML compared to the neighboring NAWM (paired Mann–Whitney U test, P = 0.02 and P = 0.008, respectively) and, for CASP8 only, also compared to WM of CC (Kruskal–Wallis with Dunn post hoc test, P = 0.04) (Figure S4H‐I). The anti‐apoptotic molecule XIAP, but not BCL2, showed an increased expression in WML compared to matched NAWM (paired Mann–Whitney U test, P = 0.02) (Figure S4J‐K). Finally, the pro‐inflammatory transcription factors NF‐kB and STAT3 were highly expressed in WML compared to the matched NAWM (paired Mann–Whitney U test, P = 0.02 and P = 0.03, respectively) (Figure S4L‐M). As shown in the correlation plots, A20 expression was significantly correlated with the expression level of most of the genes in its pathway, in both WML and GML (Spearman correlation coefficient, P < 0.05) (Figure S4N‐O). Notably, in WML the strongest correlation was observed between A20 and TNF‐related molecules including TNF, TNFR1, and TNFR2, the caspases CASP3 and CASP8, the chemokines IL10 and TGFβ, the transcription factors STAT3 and NF‐kB, which induce the expression of A20 itself (Spearman correlation coefficient ≥ 0.7, P < 0.001). TNFR1, IL10, and CASP8 showed the higher correlation with A20 also in the GML (Spearman correlation coefficient ≥ 0.7, P < 0.0001).