Distribution of dipeptide repeat proteins in cellular models and C9orf72 mutation cases suggests link to transcriptional silencing
- First Online:
- Cite this article as:
- Schludi, M.H., May, S., Grässer, F.A. et al. Acta Neuropathol (2015) 130: 537. doi:10.1007/s00401-015-1450-z
- 2k Downloads
A massive expansion of a GGGGCC repeat upstream of the C9orf72 coding region is the most common known cause of amyotrophic lateral sclerosis and frontotemporal dementia. Despite its intronic localization and lack of a canonical start codon, both strands are translated into aggregating dipeptide repeat (DPR) proteins: poly-GA, poly-GP, poly-GR, poly-PR and poly-PA. To address conflicting findings on the predominant toxicity of the different DPR species in model systems, we compared the expression pattern of the DPR proteins in rat primary neurons and postmortem brain and spinal cord of C9orf72 mutation patients. Only poly-GA overexpression closely mimicked the p62-positive neuronal cytoplasmic inclusions commonly observed for all DPR proteins in patients. In contrast, overexpressed poly-GR and poly-PR formed nucleolar p62-negative inclusions. In patients, most of the less common neuronal intranuclear DPR inclusions were para-nucleolar and p62 positive. Neuronal nucleoli in C9orf72 cases showed normal size and morphology regardless of the presence of poly-GR and poly-PR inclusions arguing against widespread nucleolar stress, reported in cellular models. Colocalization of para-nucleolar DPR inclusions with heterochromatin and a marker of transcriptional repression (H3K9me2) indicates a link to gene transcription. In contrast, we detected numerous intranuclear DPR inclusions not associated with nucleolar structures in ependymal and subependymal cells. In patients, neuronal inclusions of poly-GR, poly-GP and the poly-GA interacting protein Unc119 were less abundant than poly-GA inclusions, but showed similar regional and subcellular distribution. Regardless of neurodegeneration, all inclusions were most abundant in neocortex, hippocampus and thalamus, with few inclusions in brain stem and spinal cord. In the granular cell layer of the cerebellum, poly-GA and Unc119 inclusions were significantly more abundant in cases with FTLD than in cases with MND and FTLD/MND. Poly-PR inclusions were rare throughout the brain but significantly more abundant in the CA3/4 region of FTLD cases than in MND cases. Thus, although DPR distribution is not correlated with neurodegeneration spatially, it correlates with neuropathological subtypes.
KeywordsALS FTLD Repeat disorders C9orf72 DPR inclusions Neurotoxicity
About 10 % of all patients with amyotrophic lateral sclerosis (ALS), frontotemporal dementia (FTD) or mixed presentation of both diseases (ALS/FTD) are caused by a massive expansion of a GGGGCC repeat upstream of the C9orf72-coding region [11, 18, 43]. Three main hypotheses have been proposed to explain the pathomechanism of C9orf72 disease. First, reduced expression of the mutant allele suggests a loss of function mechanism [11, 18]. Studies in C. elegans and zebrafish reported motor deficits [7, 51], although loss of C9orf72 has no obvious effect in cultured neurons and mice [25, 55]. Second, the repeat RNA may induce toxicity by sequestering endogenous RNA-binding proteins in nuclear RNA foci . A large number of GGGGCC-interacting proteins have been identified, but their contribution to C9orf72 disease has not been elucidated so far [9, 27, 37]. Additionally, formation of RNA·DNA hybrids of the expanded repeat (so-called R-loops) may contribute to toxicity by interfering with transcription [20, 54]. However, in cultured primary neurons and the fly retina even high-level expression of repeat RNA causes little or no toxicity [35, 55]. Third, although located in an intron and lacking an ATG start codon, sense and antisense transcripts of the expanded repeat are translated by an unconventional mechanism into five dipeptide repeat (DPR) protein species [1, 17, 36, 38, 60]. All DPR species are detected in neuronal inclusions throughout the central nervous system (CNS) of C9orf72 mutation patients, predominantly in the cytoplasm. Inclusions of poly-(glycine–alanine) (poly-GA), poly-(glycine–arginine) (poly-GR) and poly-(glycine–proline) (poly-GP) proteins encoded by the sense strand are far more abundant than poly-(proline–alanine) (poly-PA) and poly-(proline–arginine) (poly-PR) proteins encoded by the antisense strand [17, 36]. None of these mechanisms, however, has so far explained the origin of neuronal and glial TDP-43 inclusions found in almost all cases with C9orf72 mutation, and the variable expression of dementia and motor symptoms even within the same family [16, 33]. Interestingly, the first clinical symptoms and neurodegeneration seem to arise prior to the onset of TDP-43 pathology when DPR inclusion pathology is already widespread [2, 36, 38, 42].
Recently, several groups reported toxicity of recombinantly expressed individual DPR species in cell lines, primary neurons and the fly retina. This led to a controversy about the main toxic DPR species. Several groups showed neurotoxicity of poly-GA, the most abundant DPR inclusion protein in C9orf72 mutation patients. Poly-GA toxicity has been attributed to co-aggregation of the transport factor Unc119  and impairment of the proteasome [57, 59]. However, in contrast to TDP-43 inclusions, poly-GA inclusions show no spatial correlation with neurodegeneration in patients [10, 29]. Other reports favor toxicity of the arginine-rich DPR species, poly-GR and poly-PR, by interference with global RNA metabolism and protein synthesis [23, 35, 55]. While poly-GR and poly-PR localization was not analyzed in the fly model , cell culture studies found overexpressed poly-GR and poly-PR (20–400 repeats) predominantly in nucleolar aggregates [23, 34, 55, 57, 59]. This is in strong contrast to the predominantly cytoplasmic localization of poly-GR and poly-PR described in patients so far [17, 36, 38, 60]. Poly-GP also has been reported to induce toxicity in cell lines, although no mechanism was proposed . Only poly-PA was not toxic in any system tested. However, none of the proposed pathomechanisms has been rigorously validated in patient tissue.
Prompted by conflicting reports on the neurotoxicity of DPR proteins in vitro, we carefully compared the expression of recombinant DPR proteins in primary rat neurons of all DPR species with proposed neurotoxicity, including the predominant sense strand-derived DPR inclusions and poly-PR, in patient brain using novel monoclonal antibodies particularly focusing on nuclear and nucleolar pathology. Since toxic overexpressed arginine-rich DPRs mainly aggregate in p62-negative intranuclear inclusions, we tried to identify such inclusions in key areas of neurodegeneration in patient CNS. Additionally, we analyzed the regional distribution pattern of aggregates containing poly-GA, its interacting partner Unc119, poly-GR, poly-GP or poly-PR in brain and spinal cord of autopsy cases with C9orf72 mutation and correlated aggregate frequency with the neuropathological diagnosis.
Materials and methods
Antibodies and reagents
The following antibodies were used: anti-nucleolin (rabbit polyclonal and mouse monoclonal, Abcam, Cambridge England), anti-p62/SQSTM1 (rabbit polyclonal, MBL, Nagoya Japan and mouse monoclonal, BD, Belgium), anti-poly-GA clone 5E9 (mouse monoclonal) , anti-Unc119 (rabbit polyclonal, homemade) , anti-fibrillarin (rabbit polyclonal, Abcam), anti-GST (rabbit polyclonal, Eurogentec, Belgium), anti-H3K9me2 (Cell Signaling Technology, Cambridge, England), anti-HDAC6 (Santa Cruz, Dallas, Texas), anti-CUG-BP1 (Abcam), anti-PML (Abcam), anti-HSF1 (Santa Cruz), anti-CD99/MIC2 (Thermo scientific, Waltham, Massachusetts), anti-PSMC2 and anti-PSMC4 (Bethyl laboratories, Montgomery, Texas), anti-Coilin (Abcam) and anti-p53 (Ventana, Tuscon, Arizona). Poly-GR antibodies 5A2 and 5H9 have been described previously [36, 38]. The novel poly-GR-specific clone 7H1 (rat isotype IgG2c) was identified by rescreening monoclonal antibodies raised against the EBNA2 epitope GQSRGRGRGRGRGRGKGKSRDK with asymmetrically dimethylated arginines  and screened by ELISA against biotinylated (GR)10 peptides (Peps4LifeSciences, Heidelberg, Germany) as described . Like clone 5H9, 7H1 detected (GR)10 with asymmetrically dimethylated arginines and non-methylated arginines, but also weakly cross-reacts with (GR)10 containing symmetrically dimethylated arginines (data not shown). By immunizing rats with synthetic GP10 peptides the poly-GP-specific antibody 7A5 (isotype IgG2c) was raised using previously described protocols . Poly-PR antibody 32B3 (isotype IgG2b) was raised against synthetic PR10 peptides in mouse using the same protocol.
RNA was stained with SYTO12 and SYTO RNAselect (Life Technologies, Darmstadt, Germany) and nuclei were stained with DAPI (Roche Applied Science, Penzberg, Germany).
DNA constructs and lentivirus production
Previously described cDNAs of GA175-GFP and GFP-GR149, GP80-V5/His and PR175-GFP with ATG start codon were cloned in a lentiviral packing vector (FhSynW2) containing the human synapsin promoter . Poly-GA, poly-GR and poly-PR were expressed from synthetic genes devoid of GGGGCC repeats, while poly-GP was expressed from a ATG(GGGCCG)80 construct. For poly-GR, the GFP had to be fused to the N-terminus to allow robust expression (for details see ). Lentivirus was produced in HEK293FT cells (Life Technologies) as described previously .
Primary hippocampal and cortical neurons were cultured from embryonic day 19 rats and infected for transduction with lentivirus as described previously [15, 48]. For immunofluorescence, the primary neurons were fixed for 10 min in 4 % paraformaldehyde and 4 % sucrose on ice. Primary and secondary antibodies were diluted in GDB buffer (0.1 % gelatin, 0.3 % Triton X-100, 450 mM NaCl, 16 mM sodium phosphate pH 7.4). Confocal images were taken by a LSM710 confocal laser scanning system (Carl Zeiss, Jena, Germany) with a 63× oil immersion objective.
Patient material, brain slices
Demographic and neuropathological data of patients and control cases
Age at death
Duration of disease
Definition of neuropathological groups
Cases with C9orf72 mutation were stratified into frontotemporal lobar degeneration (FTLD), motoneuron disease (MND) or mixed FTLD/MND according to neuropathological criteria. FTLD was diagnosed when gliosis and/or spongy alterations were seen in the cortex of the superior and/or medial frontal gyrus (Brodman areas 8/9) and/or in the cortex of the parahippocampal and/or fusiform gyrus on hemalum–eosin stainings. MND was diagnosed when either the motor cortex showed gliosis and/or spongy alterations on hemalum–eosin stainings and/or the pyramidal tract showed a microglia activation on immunohistochemical stains using the CR3/43 antibody and/or the hypoglossal nucleus and/or the anterior horn at any spinal cord level showed a loss of motoneurons and/or gliosis and/or p62-positive inclusions in motoneurons.
Immunohistochemistry and immunofluorescence were performed on paraffin sections as previously described . For Unc119 immunohistochemistry, paraffin sections were treated 25 min with 0.1 µg/μl proteinase K in 10 mM Tris/HCl. This pretreatment dramatically increased the number of visible Unc119 aggregates (compare ). Afterwards the slides were incubated with the Unc119 antibody overnight at 4 °C and detected with the DCS SuperVision 2 Kit (DCS innovative diagnostic-system, Hamburg, Germany) according to the manufacturer’s instructions. An additional 0.05 µg/µl proteinase K pretreatment for 1 min before citrate retrieval was used for anti-nucleolin and H3K9me2 immunofluorescence experiments. Anti-poly-GA immunohistochemistry was performed with the Ventana BenchMark XT automated staining system (Ventana) using the UltraView Universal DAB Detection Kit (Roche). Incubation with poly-GR and poly-GP antibodies was done overnight at 4 °C, further steps were an incubation with a rabbit anti-rat antibody (1:2000) for 1 h at room temperature, and a final processing on the Ventana BenchMark XT using the UltraView Universal DAB Detection Kit (Roche). The poly-PR antibody was also incubated overnight at 4 °C and detected the following day on Ventata BenchMark XT. Images of immunohistochemical stainings were taken by CellD, Olympus BX50 Soft Imaging System (Olympus, Tokyo, Japan), confocal images on a LSM710 (Carl Zeiss) with a 40× or 63× oil immersion objective.
RNA in situ hybridization
Paraffin sections were dewaxed in xylene and ethanol followed by microwaving in citrate pH6 buffer for 4 × 5 min. After washing with 2× SSC (0,3 M NaCl, 30 mM sodium citrate, pH7), sections were preincubated (30 min) at 65 °C in 2× SSC containing 40 % formamide and 2.5 % BSA and incubated over night at 50 °C with the Cy3(GGCCCC)4 probe (Integrated DNA Technologies, Coralville, Iowa) in 2× SSC containing 0.8 mg/ml tRNA, 0.8 mg/ml salmon sperm DNA, 0.16 % BSA, 8 % dextran sulfate, 1.6 mM ribonucleoside vanadyl complex and 5 mM EDTA. After washing with 0.5× SSC immunofluorescence was performed as described previously . In all steps, RNase-free Milli-Q ultrapurified water was used.
Semi-quantitative analysis of inclusion pathology
Frequency of poly-GA, poly-GR, poly-GP and Unc119 inclusion pathology was analyzed separately for neuronal cytoplasmic inclusions (NCI), neuronal intranuclear inclusions (NII) and dystrophic neurites (DN) in a semi-quantitative manner for 36 different CNS regions of five representative cases (C9-1 to 5) with C9orf72 mutation with a Zeiss Axioplan microscope. In neocortical regions, in the granular and molecular cell layers of the cerebellum and in spinal cord, each type of inclusion pathology was considered as “few” if less than half of 12 representative visual fields (using a 20× objective) showed at least one inclusion, as “some” if more than half but not all visual fields showed at least one inclusion, as “many” if in every visual field at least 4 inclusions were detectable and as “abundant” if each visual field showed more than 20 aggregates. This method was also used for counting dystrophic neurites in all regions. In structures of hippocampus, subcortical nuclei, brain stem and the Purkinje cell layer of the cerebellum, NCIs and NIIs were considered as “few” if less than 2 % of the neurons contained aggregates, “some” if 3–25 % of the neurons contained aggregates, “many” if 25–50 % of the neurons contained aggregates and “abundant” if more than 50 % of the neurons contained aggregates.
Quantitative analysis of inclusion pathology
The following areas with high loads of DPR protein aggregates but diverging neurodegenerative vulnerability were selected for quantification of NCIs and NIIs: cortex of the superior frontal gyrus, motor cortex, striate area of the occipital cortex, granular cell layer of the dentate gyrus, cornu ammonis regions 3/4, granular cell layer of the cerebellum, molecular cell layer of the cerebellum (superior part).
In all cases with C9orf72 mutation, 3–12 pictures adjacent to each other were taken from a representative area of each region of interest with a digital camera (Olympus Cam SC30) at an Olympus BX41 microscope using a 40× objective for cerebellar granular cell layer and a 20× objective for all other regions. Three to four pictures were taken from each cerebellar and hippocampal region. In neocortex, pictures were taken in a columnar orientation covering all six cell layers. The inclusions of one such column represented by 6–12 adjacent pictures were counted. All NCIs and NIIs were manually counted on each digital picture separately using the CellCounter plugin in Fiji ImageJ. For each region in each case, the total number of inclusions was divided by the number of pictures taken, and the average value was determined. Finally, the average of the values for each region was determined in each neuropathological group (FTLD, MND, FTLD/MND) separately.
Statistical analysis was performed with GraphPad Prism software (version 6.01). The groups with neuropathological diagnosis MND, FTLD and FTLD/MND were compared and analyzed by two-way ANOVA followed by Tukey’s post hoc test. Nucleolus size (Feret diameter) was quantified from confocal images, taken on a LSM710 with a 40× oil immersion objective, using Fiji ImageJ particle analyzer and statistically evaluated by an unpaired t test followed by an F-test to compare variances. Multiple comparison of the size of the nucleoli in the frontal cortex was done by one-way ANOVA followed by Tukey’s post hoc test. Significance level was set at p < 0.05 (two sided).
Intranuclear poly-GR and poly-PR inclusions are nucleolar in cell models, but para-nucleolar in patients
To compare DPRs expressed from synthetic genes and DPR inclusions in C9orf72 mutation patients under optimal conditions, we raised novel monoclonal antibodies. Rat poly-GP antibody 7A5, rat poly-GR antibody 7H1 and mouse poly-PR antibody 32B3 specifically detected the respective 15-mer DPRs fused to GST (Fig. S1a). 7A5 and 7H1 robustly detected SDS-insoluble aggregates in frontal cortex of patients but not of controls cases (Fig. S1b). In patients, poly-GR antibody 7H1 detected more neuronal cytoplasmic inclusions than the previously used clone 5H9 (Fig. S1c). The monoclonal poly-GP and poly-PR antibodies also allowed a more sensitive detection of poly-GP and poly-PR inclusions than our previous polyclonal antibodies [36, 38]. With the new antibodies, poly-GR and poly-GP aggregates were found in various brain areas and in spinal cord motoneurons of C9orf72 mutation patients, but not of control cases (Fig. S2a, b). Poly-PR inclusions were much less common in all brain regions (Fig. S2c). Despite a recent report of preferential aggregation of poly-PR in spinal cord motoneurons , we found no such inclusions with both the mouse poly-PR antibody 32B3 and our rabbit polyclonal antibody .
In contrast to transduced hippocampal neurons, poly-GR and poly-PR antibodies labeled mainly cytoplasmic inclusions in C9orf72 mutation patients (Fig. S2a, c), an observation consistent with previous reports [17, 36, 38, 60]. However, a fraction of neurons also contained small poly-GR and poly-PR inclusions in the nucleus (Fig. 1b, first and second row). Quantitative analysis revealed that 78 % of the poly-GR NIIs were attached to the nucleoli, whereas the remaining NIIs were randomly distributed (Fig. 1b, first row, Fig. S4a). In contrast to GFP-GR149 and PR175-GFP expressing neurons, we never saw a colocalization of poly-GR or poly-PR and nucleolin in three C9orf72 cases investigated. Immunofluorescence with two other monoclonal poly-GR antibodies (5H9 and 5A2) [36, 38] confirmed these results (Fig. S4b). Moreover, poly-GR did not colocalize with fibrillarin, another nucleolar marker (Fig. S4c). Intranuclear poly-GA and poly-GP showed a very similar pattern of para-nucleolar inclusions in C9orf72 mutation patients (Fig. 1b, rows three and four; Fig. S4a). Thus, current cellular DPR models cannot fully replicate the pattern of intranuclear aggregates found in patient tissue.
Para-nucleolar DPR aggregates colocalize with silent DNA
Since arginine-rich DPR proteins and transcription of the expanded repeat have been shown to induce nucleolar stress in cellular models [20, 50], we also investigated nucleolar size and morphology. Nucleolin stainings of the CA3/4 layer of the hippocampus, a region with abundant DPR pathology, revealed no differences in nucleolus shape and size between C9orf72 patients and controls (Fig. S5a, b). In the frontal cortex of C9orf72 FTLD cases, the size of the nucleoli did not differ from nucleoli of healthy controls regardless, whether the cells contained cytoplasmic or para-nucleolar or no DPR inclusions (Fig. S5c).
Nucleolar stress typically results in nucleolar p53 accumulation , which we did not observe in C9orf72 cases (Fig. S5d). Thus, the expanded hexanucleotide repeat DNA and/or RNA may interfere with transcriptional processes without inducing overt nucleolar stress in the hippocampus and cortical areas.
Overexpressed and patient poly-GR, poly-PR and poly-GP show different p62 labeling
We wondered whether such p62-negative poly-GR inclusions occur in patients, particularly in the nucleolus. In frontal cortex, double immunostaining revealed a strong colocalization of poly-GR and p62 in the cytosol and the nucleus, similar to poly-GA (Fig. 3b, first and second row, Fig. S7a, first row). Only very few poly-GR inclusions in the cytosol (Fig. 3b, second row) as well as in the nucleus (Fig. S7a, second row) were not labeled with p62. Similarly, the vast majority of poly-GP and poly-PR inclusions co-stained with p62 (Fig. 3b, third row and Fig. S7b).
Moreover, double immunostaining of p62 and nucleolin revealed no colocalization of ubiquitinated inclusions and the nucleolus (Fig. 3c). However, occasionally p62 labeling was observed next to the nucleolus, which was consistent with the findings for specific DPR antibodies (Fig. 1b). Together, these findings indicate that in patients with C9orf72 mutation most intranuclear DPRs aggregate in a p62-positive para-nucleolar compartment and not directly within the nucleolus.
Poly-GR and poly-GP inclusion types resemble poly-GA pathology and also occur in glia
Although DPR proteins had previously been described exclusively in neurons, we noticed intranuclear inclusions in ependymal cells of the spinal cord central canal in C9orf72 cases with MND most prominently with poly-GA antibodies (Fig. 4m, n), but also with poly-GR and poly-GP antibodies (Fig. 4o, p). Such glial inclusions were not detected in an FTLD–MND–FUS case confirming antibody specificity (Fig. S7c). Strikingly, the vast majority of these inclusions were intranuclear, while most neuronal DPR inclusions were cytoplasmic. In contrast to neuronal intranuclear DPR inclusions, the ependymal inclusions were not associated to the nucleolus (Fig. S7d). We observed further glial intranuclear poly-GA inclusions in ependymal and subependymal cells lining the lateral ventricle (Fig. 4q). Thus, not only TDP-43 pathology but also DPR pathology extends to glial cells in C9orf72 mutation patients.
Taken together, the poly-GR and poly-GP inclusion pattern resembled that of poly-GA in C9orf72 mutation patients [10, 29, 38]. Poly-PR inclusions were very rare and were not found in DNs. The identification of different types of inclusions in neuronal and glial cells suggests cell type-dependent differences in DPR aggregation or degradation.
Spectrum and distribution of DPR inclusions
To further elucidate the spectrum of DPR pathology in C9orf72 mutation patients, we analyzed the load of NCI, NII and DN pathology in 36 CNS regions using monoclonal antibodies for poly-GA (clone 5E9), poly-GR (clone 7H1) and poly-GP (clone 7A5) in five representative cases with comprehensive tissue collection, including two MND cases and three FTLD/MND cases (C9-1 to C9-5, see Table 1). Overall poly-PR distribution pattern appeared similar (not shown), but the number of inclusions was too low for a reliable semi-quantitative analysis.
Poly-PR but not poly-GR inclusions show different distribution in FTLD and MND cases
Poly-PR inclusions were scarce throughout the CNS with the highest frequency in the hippocampus. In three cases (MND and FTLD/MND, 6 sections each), we found no poly-PR in spinal cord motoneurons. Poly-PR was significantly more abundant in the CA3/4 region of FTLD cases compared to MND cases (Fig. 6b). Thus, poly-PR, but not poly-GR, distribution differs between C9orf72 disease subtypes, although it is not spatially correlated with neurodegeneration.
Spectrum of Unc119 inclusion pathology
Regional distribution of poly-GA and Unc119 inclusions differs between MND and FTLD cases
A quantitative analysis of the complete patient cohort revealed no difference in the poly-GA and Unc119 frequency in cortical regions and hippocampus between the MND, FTLD or FTLD/MND patients (Fig. 8b, c; Table S2). As for poly-GR (Fig. 6a), the poly-GA load was similar in the non-degenerating occipital cortex and the degenerating frontal cortex of FTLD and FTLD/MND patients (Fig. 8b). Unexpectedly, poly-GA and Unc119 inclusions were significantly more common in the cerebellar granular cell layer of FTLD patients compared to MND or FTLD/MND patients (Fig. 8b, c). Interestingly, these patients showed a trend towards lower levels of poly-GR inclusions (Fig. 6a), suggesting differential translation or aggregation of these DPR species in the cerebellum. These findings are consistent with an emerging role of the cerebellum in the pathophysiology of C9orf72 disease [13, 14, 30, 31, 32, 52, 56].
With this study, we provide the first quantitative analysis of the three major DPR species poly-GA, poly-GR and poly-GP as well as poly-PR in a neuropathologically characterized cohort of C9orf72 mutation patients using monoclonal antibodies. Despite ample in vitro evidence especially for poly-GA, poly-GR and poly-PR toxicity [23, 34, 35, 50, 55, 57, 59, 60], we could not identify a spatial correlation between DPR inclusion pathology and neurodegeneration in patients, although poly-GA and poly-PR showed different distribution in MND and FTLD cases. Different localization and aggregation behavior especially of poly-GR and poly-PR proteins in cellular models and patients may explain the poor translatability of the in vitro results. The newly identified para-nucleolar aggregation of DPR proteins in heterochromatin structures in patient neurons hints for repeat-associated alterations in transcription.
Subcellular localization of DPR proteins
In patients, poly-GA, poly-GR, poly-GP and poly-PR showed remarkably similar regional and subcellular expression patterns, suggesting that these proteins are co-translated in most cells and then co-aggregate in p62-positive inclusions . In transduced primary neurons, only poly-GA expression gives rise to p62-positive compact cytoplasmic inclusions. Consistent with previous reports overexpressed poly-GR and poly-PR predominantly localized to the nucleolus and was p62 negative in primary neuron culture [34, 50, 57]. However, in patients with C9orf72 mutation, poly-GR and poly-PR inclusions were predominantly cytoplasmic, and we did not find a single nucleolar inclusion. Overexpression of poly-GP in neurons resulted either in diffuse cytoplasmic or more often diffuse pan-nuclear accumulation similar to previous reports . We found both expression patterns in patients, although compact NCIs were much more common.
We noticed abundant poly-GA pathology in ependymal cells of the spinal cord central canal and the lateral ventricle. Poly-GP and poly-GR were detected at a lesser extent. While TDP-43 pathology and RNA foci have been detected in neurons and glia , DPR proteins had been described only in neurons and in Sertoli cells of testis so far [1, 38]. In contrast to neurons, ependymal cells harbor almost exclusively intranuclear inclusions. The pathogenic role of glial DPR inclusions remains unclear, since it does not extend to astrocytes and oligodendrocytes . However, trophic support from ependymal cells has been linked to ALS either directly or via altering neurogenesis [6, 12]. Moreover, poly-GP has been detected in the CSF, which may reflect neuronal death or active secretion . Additionally, ependymal cells may release DPR proteins into the CSF more efficiently than neurons.
Since the localization of DPR aggregates is already differing between neurons and glia in patients, cell type-specific effects may contribute to the aberrant expression pattern of overexpressed poly-GR and poly-GP in cellular models. Further explanations may be the faster expression kinetics and the lack of expression of the other DPR species and hexanucleotide repeat RNA in most current model systems. Since aberrantly localized DPR proteins may invoke different toxic pathways, future studies of cellular and animal models of DPR toxicity will benefit from the careful analysis of the subcellular localization of the aggregates.
Para-nucleolar DPR aggregates and nucleolar stress
While intranuclear DPR inclusions appear randomly distributed throughout the nucleus in glia, we noticed that intranuclear inclusions in neurons are predominantly para-nucleolar. To elucidate the function of para-nucleolar DPR aggregates, we tested several markers for known nucleolus-associated compartments. Robust co-staining with p62 is reminiscent of the ubiquitinated Marinesco bodies, found in the aging brain particularly in neuromelanin containing neurons of the substantia nigra [3, 40]. However, the para-nucleolar DPR inclusions lack the characteristic eosinophilic staining and we could not detect colocalization with HDAC6, which had previously been identified in Marinesco bodies . The “perinucleolar compartment” has been implicated in RNA polymerase III-dependent transcription , but the marker proteins CUG-BP1, PML, HSF1 and CD99 did not colocalize with DPR NIIs. In cells with elevated proteasomal activity, proteasomes congregate in “clastosomes” close to the nucleolus , but the para-nucleolar DPRs were negative for proteasomal subunits PSMC2 and PSMC4. Block of transcription leads to a segregation of nucleolar subcompartments and formation of the so-called “nucleolar caps”, but the marker proteins fibrillarin, PML and coilin were not detected in DPR inclusions .
Colocalization of para-nucleolar DPR proteins with heterochromatin in DAPI staining and with H3K9me2, a prominent marker of transcriptional repression, suggests a link between DPRs and transcriptional regulation. This is most consistent with transcriptional stalling and nucleolar stress due to formation of RNA·DNA hybrids (so-called R-loops) from hexanucleotide repeats . Importantly, H3K9 dimethylation has been linked to R-loop-induced transcriptional silencing . This potential link of DPR proteins with DNA/RNA-based disease mechanisms may also explain why para-nucleolar DPR aggregates were not found in transduced neurons expressing DPR proteins from synthetic genes. We found no colocalization of para-nucleolar DPR inclusions with GGGGCC repeat RNA foci. Consistent with previous reports, there was rather an inverse correlation of foci and (cytoplasmic) DPR inclusions . Nucleolar stress is typically associated with nucleolar enlargement and nucleolar accumulation of p53 particularly when it is caused by proteasomal inhibition [21, 26]. Interestingly, two groups reported proteasomal impairment by poly-GA in vitro [57, 59]. However, we detected no nucleolar accumulation of p53 and no change in nucleolar size and morphology in C9orf72 patients. Thus, neither the C9orf72 mutant allele nor cytoplasmic or para-nucleolar DPR inclusions affected nucleolar size in the brain.
Correlation of DPR and Unc119 inclusion pathology with neuropathological subtypes
Our cohort of 14 C9orf72 mutation patients represents the whole spectrum of clinical and neuropathological subtypes, including three cases each with either MND or FTLD and eight cases with a mixed disease. We chose five representative cases with comprehensive tissue collection for the semi-quantitative analysis of 36 CNS regions of the sense strand-derived DPR species and Unc119. We had previously shown that poly-GA sequesters Unc119, a protein that regulates trafficking of lipidated cargo proteins, such as transducin α in the retina [34, 58]. Loss of Unc119 is neurotoxic and Unc119 overexpression rescues poly-GA toxicity in vitro. Using improved antigen retrieval with proteinase K, we could detect Unc119 in about 40 % of poly-GA inclusions in all analyzed brain regions. Although these data corroborate Unc119 as a specific component of poly-GA inclusions, selective co-aggregation of Unc119 cannot easily explain selective vulnerability in certain brain regions. However, proteinase K pretreatment precludes analyzing the residual soluble Unc119 in affected cells. Identification of Unc119 cargos essential for neuronal survival and analysis of their localization in C9orf72 patients will be necessary to determine functional Unc119 inactivation and its correlation to neurodegeneration.
In all patients, DPR and Unc119 pathology showed a stereotypic expression pattern with highest abundance in cortex, hippocampus, thalamus and cerebellum. In contrast to previous semi-quantitative studies restricted to poly-GA pathology [10, 29], we performed a quantitative analysis of poly-GA, poly-GR, poly-PR and Unc119 pathology in seven critical brain regions in all 14 patients. The amount of poly-GA, poly-GR, poly-PR and Unc119 aggregates was similar in frontal cortex, motor cortex and occipital cortex, although the latter is not affected by neurodegeneration in C9orf72 mutation patients. Moreover, the extent of DPR pathology in frontal cortex and motor cortex did not correlate with neurodegeneration in FTLD or MND cases. Interestingly, poly-GA and poly-PR, the DPR species with the strongest toxic effects in cell culture, showed distinct depositions in FTLD vs. MND cases with C9orf72 mutation cases [34, 35, 55, 57, 59]. Poly-PR aggregates were significantly more common in the CA3/4 region of FTLD than of MND cases. Due to the very low frequency of poly-PR inclusions, the pathophysiological relevance remains unclear. Interestingly, nuclear foci of antisense repeat RNA have recently been linked to motor neuron degeneration . Poly-GA and Unc119 pathology was significantly higher in the cerebellar granular cell layer of FTLD patients compared to MND and FTLD/MND patients. At the same time, there was a trend for lower poly-GR pathology in FTLD patients, which suggests that the composition of the DPR inclusions in these patients is significantly altered, although it is unclear if and how this is related to pathogenesis. In our previous study, focusing on poly-GA pathology no similar correlation was observed , but both studies differ in staging of the cases (clinically vs. neuropathologically) and in analyzing the extent of DPR pathology (semi-quantitative vs. quantitative approaches). Interestingly, there is considerable somatic heterogeneity in the length of the expanded C9orf72 repeat and only the repeat length in the cerebellum but not in the frontal cortex is inversely correlated with disease duration, arguing for an underappreciated role of the cerebellum in the pathogenesis of FTLD [49, 52].
DPR inclusions are not actually involved in the C9orf72 pathomechanism but only TDP-43 inclusions. The strongest counterarguments are rare C9orf72 cases without TDP-43 pathology and abundant DPR pathology [2, 36, 38, 42]. In addition, DPR pathology seems to precede TDP-43 pathology, although it is not spatially correlated [2, 33]. Moreover, introducing stop codons into the GGGGCC repeat expansion prevented toxicity in the fly model, strongly arguing for a critical role of DPR proteins . Methylation in the C9orf72 promoter region is associated with reduced RNA foci and DPR pathology and prolonged disease duration presumably by inhibition of repeat transcription, which supports a toxic gain of function pathomechanism [4, 28, 44].
Soluble DPR proteins, rather than inclusions, may cause neurodegeneration. Although diffuse poly-GA coalesces into inclusions in cell culture systems , it remains unclear whether DPR proteins in cells with diffuse staining patterns cause enhanced toxicity. Soluble poly-GR/PR may interfere with the overall cellular RNA metabolism . Intercellular spreading of DPR proteins may trigger pathogenic mechanisms leading to TDP-43 phosphorylation or seed TDP-43 aggregation in a non-cell autonomous manner. Spreading and seeding have been reported for other intracellular aggregating proteins in neurodegenerative diseases, but have not been claimed to be the main source of toxicity [22, 53].
Finally and most likely, a combination of DNA•RNA hybrids, RNA foci and protein toxicity, together with a potential C9orf72 haploinsufficiency and unknown cell type-specific susceptibility factors are responsible for the selective neurodegeneration in certain brain regions in C9orf72 mutation carriers. This is supported by a very recent mouse model showing TDP-43 pathology, neurodegeneration, RNA foci and DPR proteins upon high-level viral expression of the GGGGCC repeat .
This interaction of DNA/RNA toxicity and DPR toxicity may be represented by the newly described para-nucleolar DPR aggregates. Thus, models expressing both repeat RNA and DPR proteins and constant comparison with pathological analysis of patient samples are needed to elucidate the cause of neurodegeneration in C9orf72 repeat expansion carriers, and how this can lead to either FTLD or MND.
We thank Mrs. B. Kraft, Mrs. I. Pigur and Mr. M. Schmidt for excellent technical assistance. We thank C. Lehmer, K. Mori, Y. Ohki, D. Orozco and B. Schmid for critical comments and technical advice. DE was supported by the Helmholtz Young Investigator program HZ-NG-607 and the Hans und Ilse Breuer Foundation. DE, MD and TK received funding from the Munich Cluster of Systems Neurology (SyNergy).The research leading to these results has received funding from the European Research Council under the European Union’s Seventh Framework Programme FP7/2014-2019 under Grant agreement No. 617198 [DPR-MODELS] to DE.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.