HR23B pathology preferentially co-localizes with p62, pTDP-43 and poly-GA in C9ORF72-linked frontotemporal dementia and amyotrophic lateral sclerosis

Human brain sections (6 μm) were deparaffinized in xylene and rehydrated (100%–96%-90%–80%-70%–50% EtOH serie). Antigen retrieval was done in 0.01 M sodium citrate, pH 6.0 using pressure cooker treatment. Endogenous peroxidase activity was blocked with 0.6% H₂O₂ and 1,25% sodium azide in 0.1 M PBS. Immunostaining was performed overnight at 4 °C in PSB block buffer (0.1 M PBS/0.5%protifar/0.15%glycine). Antibodies used in this study are listed in Additional file 1: Table S3, including concentration and brand/catalogue number. Antigen-antibody complexes were visualized by incubation with DAB substrate (DAKO) after incubation with Brightvision poly-HRP-linker (Immunologic) or anti-mouse/rabbit HRP (DAKO). Slides were counterstained with Mayer’s haematoxylin and mounted with Entellan (Merck Millipore International). The slides were then left to dry in the fume hood for an hour and thereafter put in an 37 °C incubater overnight. Pictures were taken by using an Olympus BX40 microscope (Olympus).

Human brain sections were treated as described above. After incubation with the primary antibody, sections were washed with PBS block buffer (1xPBS/0.5%protifar/0.15%glycine) and incubated with secondary anti-mouse/rabbit Cy2/3 linked antibodies (Jackson). To remove background staining, a 10 min incubation with Sudan Black (Sigma, 0.1 g in 100 ml 70% ethanol, filtered) was done. To visualize nuclei, slides were incubated for 10 min with Hoechst 33342(Invitrogen). Slides were mounted with ProLongGold (Invitrogen) and kept at 4 °C until imaging at a Zeiss LSM700 Confocal microscope.

For the neuropathological assessment we used brain sections from 5 C9ORF72 FTD cases (see Table 1). Five different brain regions (frontal cortex, temporal cortex, motor cortex, hippocampus and cerebellum) per patient were semi-quantified on neurodegeneration and pathological score of p62, pTDP-43 and HR23B (Additional file 2: Table S1). Counting was not performed in a blinded fashion. Neurodegeneration was assessed on haematoxylin and eosin (HE) sections and graded as absent (0), mild (1), moderate (2) or severe (3) based on the presence of neuronal loss. The neurodegenerative score from the pathological report was also taken into account. Pathological scores were rated as (0) if completely absent, rare (1) if only a few could be found one brain section, occasional (2) if they not present in every microscopic field, moderate (3) if at least a few examples were present in most microscopic fields, and numerous (4) when many were present in every microscopic field [28]. We reported the overall number of immunoreactive inclusions (total score) as well as the number of neuronal cytoplasmic inclusions (NCI), neuronal intranuclear inclusions (NII) and dystrophic neurites (DN) (Additional file 2: Table S1). Total scores were assessed independent from NCI, NII and DN scores and represent an impression of the global load of pathology and the number of inclusions using the same grading system as described above. Quantification of co-localization of HR23B with DPRs, p62 and pTDP-43 was not performed in a blinded fashion.

Human fibroblast lines from 4 C9ORF72 carriers (13E634, 13E659, 17E0225, 17E0278) and 4 controls (81E253, 86E1375, 06E0717 and 99E0774) were obtained from the cell repository of the department of clinical genetics and the XP25RO fibroblast line was provided by the department of molecular genetics. Fibroblasts were cultured in DMEM medium (Gibco) with 10% fetal calf serum, 1% penicillin/streptavidin and 1% non-essential amino acids. To determine UV-sensitivity, human fibroblasts were seeded in triplicate in 10 cm plates (Greiner Bio-one) in a density of 2000 cells/plate. After 24 h, cells were treated with increasing doses of UV-C (254 nm UV-C lamp, Philips). After 5–7 days, colonies were fixed with 0.1% w/v Coomassie Blue (Bio-Rad) in a 50% Methanol, 10% Acetic Acid solution. Colonies were counted with the integrated colony counter GelCount (Oxford Optronix). Counting was performed automatically with the same settings for each fibroblast cell line, but not in a blinded fashion.

To determine DNA damage recruitment of NER proteins, human fibroblasts were seeded on coverslips and after 1 week in culture irradiated with 60 J/m2 (254 nm UVC lamp, Philips) through an 8 μm microporous filter (Millipore) to induce sub-nuclear local DNA damage. Cells were fixed after 30 min with 2% paraformaldehyde and permeabilized with 0.1% Triton X-100 for 20 min. Next, cells were incubated with fresh 0.07 M NaOH in PBS for 5 min to denature DNA and enable CPD staining. Cells were then washed with PBS containing 0.15% glycine and 0.5% BSA and incubated with primary antibodies (see Additional file 1: Table S3) overnight at 4 °C. The next day, cells were washed with 0.1% Triton X-100 and incubated with Alexa Fluor conjugated secondary antibodies (488, 555 and 633; Invitrogen) for 1 h at RT. Coverslips were mounted using ProLongGold with DAPI (Invitrogen) and imaged using a Zeiss LSM700 microscope.

To measure NER capacity, human fibroblasts were grown on coverslips for 1 week in culture and irradiated with 16 J/m² UV-C. After irradiation, cells were incubated for 1 h in UDS medium (F10, 1% PS, 10% dialyzed serum) with 2% HEPES and 1% 5-ethynyl-2′-deoxyuridine (EdU, Invitrogen). After 1 h, medium was changed to normal DMEM medium (10% FCS, 1% NEAA, 1% PS) for 10 min. Next, cells were fixed in 4% paraformaldehyde and permeabilized with 0.1% Triton X-100. Blocking was done using 1.5% BSA in PBS for 30 min. EdU incorporation was visualized by incubating cells for 1 h at room temperature with Click-it reaction cocktail containing Atto 594 Azide (60 μM, Atto Tec.), Tris-HCl (50 mM, pH 7.6), CuSO₄*5H₂O (4 mM, Sigma) and ascorbic acid (10 mM, Sigma). After the Click-it reaction, cells were washed with 0.1% Triton X-100, incubated with Hoechst 33342 (Invitrogen) for 10 min and mounted with ProLongGold (Invitrogen). Images were acquired using a Zeiss LSM700 microscope. UDS levels were quantified by measuring the total nuclear fluorescence intensities (in at least 50 cells per experiment) with FIJI image analysis software. Intensity levels were averaged and normalized to the fluorescence levels in unirradiated cells. Quantification of intensity levels was done automatically with FIJI but was not performed in a blinded fashion.

We started with assessing the localization of ras-related nuclear protein GTPase activating protein (Ran-GAP) in C9FTD patient brain sections. We studied five FTLD-TDP cases with the C9ORF72 repeat expansion and three non-demented cases (for information about our cohort, see Table 1). We predominantly found a diffuse nuclear staining or nuclear envelop staining of Ran-GAP. Most nuclei were round-shaped but occasionally we observed misfolded nuclei or the invagination of the nuclear membrane (Additional file 3: Figure S1). The number of oddly-shaped nuclei did not clearly differ between C9FTD cases and non-demented controls. Next, we investigated Adenosine Deaminase, RNA Specific B2 (ADARB2) localization. ADARB2 immunostaining showed both nuclear and cytoplasmic localization in both C9FTD and non-demented controls. We could detect some intranuclear inclusions in the hippocampus dentate gyrus of C9FTD patients (Additional file 4: Figure S2A). However, IF double staining of ADARB2 with p62 showed similar results between hippocampal nuclei of C9FTD patients and non-demented cases (Additional file 4: Figure S2B).

Next, we stained for HR23B and found different types of inclusions in cortical areas, hippocampus and cerebellum of C9FTD cases (Fig. 1 and Additional file 5: Figure S5). Most cytoplasmic inclusions showed a round and perinuclear appearance (Fig. 1a and c), while some inclusions had a negative central core surrounded by a positive halo labeling (Fig. 1e). Neuropils were mostly found in layer 2 of cortical areas (Fig. 1a). Intranuclear inclusions were also observed, including cat-eye inclusions (Fig. 1b). Overall, layer 2/3 and layer 5/6 of frontal and temporal cortices showed the highest HR23B pathological burden in C9FTD cases, followed by motor cortex, hippocampus and cerebellum (Fig. 1g and Additional file 2: Table S1). Hippocampus dentate gyrus (DG) harbored some perinuclear inclusions and hippocampus cornu ammonis (CA) showed some cells with strong nuclear staining (Additional file 5: Figure S5). Cerebellum showed low HR23B staining with some nuclear and perinuclear inclusions in the granular layer and nearly absent in the molecular layer (Additional file 5: Figure S5). Non-demented controls showed normal nuclear localization of HR23B (Additional file 5: Figure S5).

Furthermore, we assessed fragile X mental retardation protein (FMRP) localization in C9FTD cases. Punctuated FMRP staining in the cytoplasm indicative of the formation of stress granules was absent in our C9FTD cases (Additional file 6: Figure S3). FMRP was evenly distributed in the cytoplasm in both C9FTD cases and controls. Very occasionally, we could detect inclusions in the hippocampus dentate gyrus in both C9FTD cases and controls (Additional file 6: Figure S3). Finally, we set out to test Pur-alpha staining, which showed to be located in stress granules in both C9FTD cases and controls (Additional file 7: Figure S4). We did not find Pur-alpha intranuclear inclusions in the cerebellum or other brain areas of C9FTD cases nor controls (Additional file 7: Figure S4).

In order to validate the observed HR23B pathology, we used a second independent HR23B antibody (for details see methods) that revealed similar results (Additional file 8: Figure S6). Furthermore, we expanded our cohort with three ALS cases with the C9ORF72 repeat expansion, three sporadic ALS cases with unknown genetic cause, two FTD with GRN mutation and two FTD-MAPT cases. In non-demented controls, we observed immunoreactivity for HR23B in nuclei (Additional file 5: Figure S5 and Additional file 8: Figure S6). C9ALS cases showed very strong HR23B staining in nuclei and some cytoplasmic inclusions in motor cortex (Fig. 2a) and spinal cord sections (Fig. 2b). Sporadic ALS cases also showed very strong nuclear HR23B staining but no pathology in motor cortex (Fig. 2c) nor spinal cord (Fig. 2d). GRN FTD cases showed the same extend of HR23B pathology as C9FTD in frontal cortex, consisting of some intranuclear inclusions, cytoplasmic inclusions and neuropils (Fig. 2e). HR23B pathology was absent in MAPT FTD cases (Fig. 2f).

To evaluate HR23B’s aggregation process in C9FTD, we performed double labeling with known pathological hallmarks, including DPRs, pTDP-43 and p62 in five C9FTD frontal cortices (Fig. 3). HR23B was mostly found to be co-localized with p62 (Fig. 3), as 66% of all HR23B inclusions were also positive for p62 (Additional file 9: Table S2). Next is pTDP-43 with 22.6% of HR23B inclusions being positive for pTDP-43 (Fig. 3 and Additional file 9: Table S2). From all DPRs, HR23B showed partial co-localization with poly-GA (Fig. 3), for 6.6% (Additional file 9: Table S2). The other DPRs only co-stained with 0–3% of all HR23B inclusions. We did not evaluate poly-PA because we found too few inclusions. Interestingly, hippocampus dentate gyrus showed a much higher co-localization between DPRs and HR23B than frontal cortex (Fig. 4 and Additional file 9: Table S2). We found 60.6% of HR23B inclusions being positive for poly-GA in the dentate gyrus, which is nearly a 10-fold increase compared to the frontal cortex (Fig. 4 and Additional file 9: Table S2). Poly-GP increased from 0.8% in frontal cortex to 4.7% in hippocampus and poly-GR also showed a 10-fold increase from 1% in frontal cortex to 10.5% in hippocampus (Fig. 4 and Additional file 9: Table S2). Only poly-PR stayed fairly undetectable with a slight increase from 0.65% in frontal cortex to almost 1% in hippocampus. HR23B co-localization with p62 in dentate gyrus was also higher in hippocampus than in frontal cortex (87.4% vs 66%) (Fig. 4 and Additional file 9: Table S2). We performed a 2-way ANOVA to compare co-localization percentages of HR23B with pathological hallmarks between different brain areas (p < 0.0001). Post Bonferroni test indicated that percentages of poly-GA and p62 were significantly different between hippocampus DG and frontal cortex (Fig. 4 and Additional file 9: Table S2). For the other DPRs and pTDP-43, differences were not significant (Fig. 4 and Additional file 9: Table S2). These data suggest a difference in aggregation formation and co-localization patterns between different brain areas or cell types.

To assess any changes in the normal cellular function of HR23B, we first focused on its role in DNA damage repair. HR23B interacts with and stabilizes Xeroderma pigmentosum, complementation group C (XPC) protein [32], which is involved in the recognition of bulky DNA adducts in global genome nucleotide excision repair (GG-NER). Staining with XPC antibody did not reveal any gross differences between C9FTD and non-demented control post-mortem brain sections (Fig. 5). To assess nucleotide excision repair capacity in living cells, we performed UV-sensitivity assays with four C9ORF72 patient fibroblast lines and compared these to four healthy control fibroblast lines. Intriguingly, C9ORF72 fibroblasts were more sensitive to UV-C damage than healthy control fibroblasts (Additional file 10: Figure S7A), but not as sensitive as a fibroblast line that is fully deficient in NER (XP25RO, homozygous for 619C > T causing an ARG207X change in exon 5 of the XPA gene). The GG-NER pathway is initiated by recognition of DNA damage by the HR23B/XPC/CETN2 complex, which is then followed by multiple downstream steps to verify and excise the damage [29]. We therefore assessed whether factors involved in each of the steps of the NER pathway were recruited normally to DNA damage in C9ORF72 fibroblasts. To do so, we evoked local DNA damage by UV-C irradiation through a microporous filter in our fibroblast lines and performed immunofluorescence to visualize DNA damage (by cyclobutane pyrimidine dimers (CPD) antibody) and DNA damage recruitment of NER factors XPC, XPB, XPA, XPF and XPG. All tested NER proteins clearly co-localized with DNA damage in C9ORF72 patient fibroblasts, indicating that recognition and processing of CPDs is still functional in these cells (Additional file 10: Figure S7B). To finally verify that NER is fully operational, we measured incorporation of the thymidine analogue 5-ethynyl-2-deoxiuridine (EdU) after UV-C irradiation to quantify the efficiency of DNA repair in our fibroblast lines. The NER-deficient XP25RO cell line, which we used as negative control, did not show any EdU incorporation. In contrast, we still observed efficient EdU incorporation in our four C9ORF72 patients cell lines to similar levels as in four healthy control lines (Additional file 10: Figure S7C), indicating that NER is not deficient in C9ORF72 patient fibroblasts.

Besides its role in DNA damage repair, HR23B is also known for its function in the ubiquitin-proteasome system. Various HR23B binding partners have been identified which are involved in the unfolded protein response (UPR), transcriptional regulation, cell cycle control and ER-associated degradation (ERAD) [51]. To assess if HR23B aggregation also evokes changes in the localization of the proteasome, we stained our sections for proteasome subunit 20S. However, we did not observe any obvious changes in 20S normal localization (Fig. 5). Another binding partner of HR23B is ataxin-3, a deubiquitinase enzyme in which a poly-glutamine expansion is linked to SCA3 [46]. HR23B-positive inclusions have been found in post-mortem brain material of SCA3 patients [6] however we could not detect ataxin-3 pathology in C9FTD cases (Fig. 5). Besides the proteasome, we wondered if we could find any changes in ERAD. HR23B did not sequester NGly1, one of its known binding partners involved in ERAD, into protein inclusions. Strikingly, however, we did observe clearance of NGly1 staining in C9FTD frontal cortex (Fig. 5). Although some neurons showed the same strong peri-nuclear staining as almost all non-demented control cells, in the majority of patient neurons no NGly1 staining was observed, indicating reduced expression of NGly1 in C9FTD brains. This may suggest a partial loss of function of ERAD in a subset of neurons in the brain of C9FTD patients.