Acta Neuropathologica

, Volume 125, Issue 3, pp 413–423

hnRNP A3 binds to GGGGCC repeats and is a constituent of p62-positive/TDP43-negative inclusions in the hippocampus of patients with C9orf72 mutations

Authors

  • Kohji Mori
    • Adolf-Butenandt-Institute, Biochemistry, Ludwig-Maximilians-University
  • Sven Lammich
    • Adolf-Butenandt-Institute, Biochemistry, Ludwig-Maximilians-University
  • Ian R. A. Mackenzie
    • Department of Pathology and Laboratory MedicineUniversity of British Columbia
  • Ignasi Forné
    • Adolf-Butenandt-Institute, Protein Analysis UnitLudwig-Maximilians-University
  • Sonja Zilow
    • Adolf-Butenandt-Institute, Biochemistry, Ludwig-Maximilians-University
  • Hans Kretzschmar
    • Center for Neuropathology and Prion ResearchLudwig-Maximilians-University
  • Dieter Edbauer
    • DZNE, German Center for Neurodegenerative Diseases
  • Jonathan Janssens
    • Neurodegenerative Brain Diseases Group, Department of Molecular GeneticsVIB
    • Laboratory of Neurogenetics, Institute Born-BungeUniversity of Antwerp
  • Gernot Kleinberger
    • Adolf-Butenandt-Institute, Biochemistry, Ludwig-Maximilians-University
    • DZNE, German Center for Neurodegenerative Diseases
    • Neurodegenerative Brain Diseases Group, Department of Molecular GeneticsVIB
    • Laboratory of Neurogenetics, Institute Born-BungeUniversity of Antwerp
  • Marc Cruts
    • Neurodegenerative Brain Diseases Group, Department of Molecular GeneticsVIB
    • Laboratory of Neurogenetics, Institute Born-BungeUniversity of Antwerp
  • Jochen Herms
    • DZNE, German Center for Neurodegenerative Diseases
    • Munich Cluster for Systems Neurology (SyNergy)
  • Manuela Neumann
    • DZNE, German Center for Neurodegenerative Diseases
    • Department of NeuropathologyUniversity of Tübingen
  • Christine Van Broeckhoven
    • Neurodegenerative Brain Diseases Group, Department of Molecular GeneticsVIB
    • Laboratory of Neurogenetics, Institute Born-BungeUniversity of Antwerp
  • Thomas Arzberger
    • Center for Neuropathology and Prion ResearchLudwig-Maximilians-University
    • Adolf-Butenandt-Institute, Biochemistry, Ludwig-Maximilians-University
    • DZNE, German Center for Neurodegenerative Diseases
    • Munich Cluster for Systems Neurology (SyNergy)
Original Paper

DOI: 10.1007/s00401-013-1088-7

Cite this article as:
Mori, K., Lammich, S., Mackenzie, I.R.A. et al. Acta Neuropathol (2013) 125: 413. doi:10.1007/s00401-013-1088-7

Abstract

Genetic analysis revealed the hexanucleotide repeat expansion GGGGCC within the regulatory region of the gene C9orf72 as the most common cause of familial amyotrophic lateral sclerosis and the second most common cause of frontotemporal lobar degeneration. Since repeat expansions might cause RNA toxicity via sequestration of RNA-binding proteins, we searched for proteins capable of binding to GGGGCC repeats. In vitro-transcribed biotinylated RNA containing hexanucleotide GGGGCC or, as control, AAAACC repeats were incubated with nuclear protein extracts. Using stringent filtering protocols 20 RNA-binding proteins with a variety of different functions in RNA metabolism, translation and transport were identified. A subset of these proteins was further investigated by immunohistochemistry in human autopsy brains. This revealed that hnRNP A3 formed neuronal cytoplasmic and intranuclear inclusions in the hippocampus of patients with C9orf72 repeat extensions. Confocal microcopy showed that these inclusions belong to the group of the so far enigmatic p62-positive/TDP-43 negative inclusions characteristically seen in autopsy cases of diseased C9orf72 repeat expansion carriers. Thus, we have identified one protein component of these pathognomonic inclusions.

Keywords

ALSC9orf72FTLDhnRNP A3NeurodegenerationTDP-43

Introduction

Neurodegenerative disorders are generally characterized by disease-signifying protein deposits. Moreover, in a number of neurodegenerative diseases mutations causing genetically inherited variants of the disease were associated with the genes encoding the protein deposits, their precursors or their modulating enzymes. Functional analysis of these genetic variants fundamentally helped to understand disease-associated mechanisms of Alzheimer’s disease (AD) and Parkinson’s disease [13, 16]. Similarly, research into amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTLD) was dramatically accelerated by the identification of the RNA/DNA-binding protein TDP-43 (Tar DNA-binding protein of 43 kDa) as an abundant deposited protein [2, 25] and by the discovery that mutations in TARDBP cause familial variants of both diseases [4, 36]. These findings also helped to develop the concept that ALS and FTLD are multisystem disorders with overlapping clinical and pathological characteristics and similar functional and genetic causes [27, 33]. Besides TDP-43 and the long known SOD1 (super oxide dismutase 1) gene [30], a number of other ALS and/or FTLD-related genes/risk factors were discovered including FUS (fused in sarcoma) [20, 40], OPTN (optineurin) [23], Ataxin 2 [11], Chmp2B [35], VCP (valosin containing protein) [41], TMEM106B [38], GRN (progranulin) [3, 7], PFN (profilin) [43] and the C9orf72 gene [8, 14, 29]. Pathological repeat expansions in C9orf72 have been found in about 40 % of familial ALS patients and 20 % of familial FTLD [27], demonstrating that C9orf72 is the most common genetic cause for these incurable disorders. Thus, C9orf72 may provide a key not only for the understanding of ALS/FTLD causing mechanisms, but also for the identification of drug targets and the design of therapeutic strategies. Independent evidence originally obtained by three different research groups [8, 14, 29] and subsequently confirmed by numerous other groups demonstrated that a large expansion of a hexanucleotide repeat (GGGGCC) in the first intron of C9orf72 segregates with the disease. In healthy humans the repeat length is apparently below 26, whereas for ALS/FTLD patients a repeat length from 60 to 1,600 was determined [8, 14, 29, 39]. Repeat expansions in non-coding regulatory regions of genes can cause a disease principally by two different mechanisms. Due to the immense length of the repeat expansion transcription and/or splicing may be affected leading to haploinsufficiency [39]. On the other hand, RNA toxicity caused by sequestration of RNA-binding proteins may also be causative [28]. Currently, evidence for both possibilities exists. The observation of nuclear RNA foci in patients with GGGGCC hexanucleotide repeat expansions [8], a finding which however is still controversially discussed, suggests that trapping of essential RNA-binding proteins may be involved in the disease. On the other hand, the finding of decreased expression of C9orf72 mRNA and decreased transcriptional activity of the C9orf72 promoter on intermediate (7–24 repeats) alleles [8, 14, 39] implies a loss of function as a disease-causing mechanism. These scenarios are not mutually exclusive and may even occur in parallel.

We aimed to identify proteins capable to bind to the GGGGCC repeat and to verify their pathological relevance by investigating their distribution in cases with C9orf72 hexanucleotide expansion mutations.

Materials and experimental procedures

DNA synthesis and plasmid construction

120 base single-stranded DNA containing GGGGCC or AAAACC hexanucleotide repeats with restriction enzyme sites (NheI or HindIII) were synthesized. 100 μM of complementary DNA strands were annealed in the presence of 10 % GC-RICH solution (Roche) and GC-RICH PCR Reaction buffer (Roche) and cloned into pcDNA3.1 (+) vector (invitrogen). Plasmids containing 7, 17 and 23 repeats of GGGGCC and 17 repeat of AAAACC were obtained. The sequence of all constructs was verified.

In vitro transcription of RNA probes

pcDNA3.1-GGGGCC×23 and pcDNA3.1-AAAACC×17 constructs were linearized with HindIII and used as templates for RNA synthesis. In vitro RNA transcription was performed with T7 Ribomax Express Large Scale RNA Production System (Promega) as described by the manufacturer supplemented with 40 U of RNase inhibitor (RiboLock, Thermo Scientific). To achieve equal levels of biotinylation between these probes, different concentrations of biotin-14-CTP (2 mM for GGGGCC probe and 0.29 mM for AAAACC probe) were added in each reaction. Following DNase treatment, biotinylated RNA products were purified with phenol/chloroform. Biotinylation efficacy of RNA probes was evaluated using the BrightStar BioDetect kit (Ambion). For competition experiments non-biotinylated GGGGCC repeat RNA was in vitro RNA transcribed using the MEGA script kit (Ambion) as described by the manufacturer. Expected lengths of repeat RNA probes and competitor were confirmed with formaldehyde gel electrophoresis.

Cell culture and preparation of nuclear extracts

HEK293 cells were grown with DMEM supplemented with 10 % FCS and penicillin/streptomycin. Nuclear extract was prepared as previously described [9]. Protein concentration was determined by the BCA method.

siRNA-mediated knockdown of hnRNP A3

Cells were plated without antibiotics at a density of 600,000 cells per 6 wells. Transfection of ON-TARGET plus SMARTpool Human HNRNPA3 or ON-TARGETplus Non-targeting pool siRNA (Thermo SCIENTIFIC) was performed with Lipofectamine RNAiMax. Three days after transfection cell lysates were harvested.

Purification of hexanucleotide-repeat-binding proteins

A total of 0.6 mg of HEK293 cell nuclear extract was diluted in 4 ml of protein-binding buffer [10 % glycerol, 10 mM HEPES, 50 mM KCl, 1 U/ml RNase inhibitor, 0.15 μg/ml yeast tRNA (10109495001, Roche), 1 mM EDTA, 1 mM DTT and 0.5 % Triton X100 in DEPC water]. The diluted extract was then precleared with heparin-agarose (H6508, Sigma) and streptavidin-agarose (15942-050, Invitrogen). The precleared nuclear extracts were incubated with streptavidin μMACS-microbeads (MACS molecular) and 150 pmol of biotinylated repeat RNA in the presence of 50 mM KCl for 1 h. For competition experiments, nuclear extracts were incubated for 1 h with 7.5 nmol (50-fold excess) of non-biotinylated GGGGCC competitor RNA before addition of the biotinylated probe. The reaction mixture was loaded on a μMACS column and subsequently washed three times with protein-binding buffer. GGGGCC-repeat binding proteins were sequentially eluted with increasing concentrations of NaCl. Each eluate was TCA precipitated and subjected to SDS-PAGE. Gels were stained with Silver Quest Silver staining kit (Invitrogen) according to manufacturer’s protocol and prepared for mass spectrometry.

Protein identification and quantification by LC–MS

Gel lanes of interest were excised into 25 pieces per lane and transferred into a 96-well plate and digested as described before with minor modifications [32, 42]. For the LC–MS analysis, the peptides were injected in an Ultimate 3000 HPLC system (LC Packings) and the effluent was directly electrosprayed into a LTQ-Orbitrap mass spectrometer (Thermo) as described elsewhere [31]. Proteins were identified using Mascot (Matrix Science, London, UK; version Mascot) against SwissProt_2011.02 database for human proteins (Fragment Tolerance: 0.80 Da, Fixed Modification for carbamidomethyl cysteine, Variable Modification for methionine oxidation, Max Missed Cleavage:2). Proteins were quantified using the “Quantitative Value” from Scaffold (version Scaffold_3.4.9, Proteome Software Inc., Portland, OR; Peptide Thresholds: 90.0 % minimum, Protein Threshold: 99.0 % minimum and 2 peptides minimum).

Western blotting

Samples were separated on 10 % Tris–glycine gel and transferred on PVDF membranes. After blocking for 1 h with 0.2 % I-Block (Applied Biosystems) in PBS, the membrane was incubated with indicated antibody overnight. The antibody signal was detected with HRP-conjugated secondary antibodies (Promega) using the ECL reagents (GE healthcare) and exposed to X-ray films (SuperRX, Fujifilm).

Immunohistochemistry

Immunohistochemistry was performed on 3-μm-thick paraffin sections. Following antigen retrieval by microwaving in 0.1 M citrate buffer pH 6 and blocking of endogenous peroxidase with 5 % H2O2 in methanol, sections were settled in PBS with 0.02 % Brij35 (Sigma-Aldrich). After blocking with 2 % FBS in PBS for 5 min, the respective primary antibody was applied for 1.5 h at RT or overnight at 4 °C. When the first antibody was derived from goat or rat, rabbit-anti-goat antibody (DAKO) or rabbit anti-rat IgG (PARIS-biotech) was used as secondary antibody for 1 h. After rinsing with 0.02 % Brij35 in PBS, antibody binding was detected and enhanced by DCS Super Vision 2 HRP-Polymer-Kit (DCS Innovative Diagnostik-Systeme, Hamburg, Germany) or NovoLink DS Polymer Detection Systems (Leica) using DAB as chromogen. Counterstaining for cellular structures was performed with haemalum. Microscopic images were obtained using a BX50 microscope and Cell-D software (Olympus).

Immunofluorescence

For double immunofluorescence, the secondary antibodies conjugated to Alexa Fluor 555 or Alexa Fluor 488-conjugated antibodies (anti-rabbit, anti-mouse, and anti-rat) were used with TO-PRO-3 (Invitrogen) for nuclear counterstaining. To reduce autofluorescence, slides were incubated with 0.3 % Sudan Black in 70 % ethanol. Confocal images were obtained with an inverted laser scanning confocal microscope (Zeiss LSM510 or LSM710) with a 40× or 63×/1.4 oil immersion lens, using a pinhole diameter of 1 Airy unit. If necessary for printing, brightness and contrast of each image were linearly enhanced using LSM software (Zeiss). Co-localization of inclusions on these images was manually determined using ImageJ software with “Cell Counter” plug-in. Six merged images per section per case were analyzed.

Human tissue

All cases provided by the Neurobiobank Munich, Ludwig-Maximilians-University Munich and the University of British Columbia were collected and distributed according to the guidelines of the local ethical committee. Autopsy material of a control individual was provided by the Antwerp Bio Bank at the Institute Born-Bunge, University of Antwerp, Antwerp, Belgium. Brain autopsy was performed on the basis of informed consent.

Results

Identification of GGGGCC hexanucleotide-repeat-binding proteins

To identify proteins, which could be bound by GGGGCC hexanucleotide repeats, we performed pull down assays using in vitro-transcribed biotinylated RNAs containing either 23 GGGGCC or 17 AAAACC repeats as control. We choose this repeat length to circumvent the still unsolved problems of cloning and transcribing the extremely long hexanucleotide repeats observed in patients with C9orf72 mutations. We expect such shorter repeats to have a similar ability and specificity for selective protein binding like expanded repeats, because the repetitive sequence will likely form a similar secondary structure and RNA-binding proteins typically bind to rather short sequences motifs or structures. In fact a four-repeat GGGGCC repeat RNA is sufficient to form a complex G-quadruplex structure [12]. Moreover, repeats of intermediate length, such as the 23 repeats used here, may be a risk for developing the disease [15, 39]. Biotinylated in vitro-transcribed RNA probes were incubated with nuclear extracts from HEK 293 cells in the absence or presence of a 50-fold excess of non-biotinylated competitor RNA containing the GGGGCC repeat (Fig. 1a). The competitor RNA prevented binding as indicated by strongly enhanced flow through (Fig. 1a, lane 4). Using increasing salt concentrations for the elution of bound proteins, we observed differential protein-binding affinity to the GGGGCC/AAAACC repeats upon elution with 500 mM NaCl (Fig. 1a, lanes 17 and 18). Proteins eluting with 500 mM NaCl from the GGGGCC or the AAAACC repeats were subjected to LC–MS/MS. This allowed the identification of 235 proteins in three replicates (Suppl. Tab. 1). 188 proteins were identified at least twice in the three pull down experiments. Of these 188 proteins, binding of 127 proteins could be competed with an excess of non-biotinylated GGGGCC probes. 72 proteins showed at least a twofold stronger binding to GGGGCC than to AAAACC. All proteins with an abundance higher than 20 in the GGGGCC fraction were finally selected. These stringent selection criteria resulted in 20 top candidate proteins, of which most were known RNA interacting factors such as heterogenous ribonucleoproteins (hnRNPs), splicing factors and mRNA-binding proteins (Table 1). A selection of these 20 proteins for which specific antibodies were commercially available (Suppl. Tab. 2) was then confirmed by Western blotting of the elution fractions. All antibodies tested including antibodies to hnRNP A3, hnRNP A2B1, SFPQ, ILF3, NONO, hnRNP L, IL2BP1, ILF-2, and FUS revealed strong and selective binding to the GGGGCC repeat in the 500 mM NaCl fraction. These signals were completely blocked by 50-fold excess of the non-labeled GGGGCC probe (Fig. 1b, lanes 8–10).
https://static-content.springer.com/image/art%3A10.1007%2Fs00401-013-1088-7/MediaObjects/401_2013_1088_Fig1_HTML.gif
Fig. 1

Identification of GGGGCC hexanucleotide repeat-specific binding proteins. a Representative silver-stained gels showing proteins pulled down by the respective repeat containing RNAs. HEK293 nuclear extracts were incubated with indicated RNA probes with (+) or without (−) 50-fold excess of non-biotinylated RNA competitor. RNA–protein interaction was weakened with increasing concentrations of NaCl. In the presence of GGGGCC competitor (G+), RNA–protein binding was inhibited (lanes 10, 13, 16, 19, 22, 25), and proteins in the flow through fraction (lane 4) were increased. Boxed lanes in 500 mM NaCl elution fractions were excised for protein identification by LC–MS/MS. b Western blot analysis confirming GGGGCC-repeat-specific binding of selected proteins. Aliquots of proteins eluted at different salt concentrations were subjected to electrophoresis, and Western blotting was performed using the indicated antibodies. All proteins show GGGGCC-repeat-specific binding at high NaCl concentrations (compare lane 8 and 9). Note that binding was completely blocked by a 50-fold excess of non-labeled GGGGCC (+) (lanes 7, 10). hnRNP F, which is not one of the 20 GGGGCC-repeat-specific binding proteins, is used as negative control. A AAAACC repeats, FT flow through, G GGGGCC repeats

Table 1

List of 20 selected proteins specifically binding to the GGGGCC repeat

 

Identified proteins

Accession

Average quantitative value

GC/AC ratio

AC

GC

Competition

1

Heterogeneous nuclear ribonucleoproteins A2/B1

ROA2_HUMAN

110.3

369.8

1.7

3.4

2

Splicing factor, proline- and glutamine-rich

SFPQ_HUMAN

51.7

130.5

0.1

2.5

3

Splicing factor 3B subunit 3

SF3B3_HUMAN

51.1

118.8

6.3

2.3

4

ELAV-like protein 1; (Hu-antigen R)

ELAV1_HUMAN

9.6

117.1

0.0

12.1

5

Interleukin enhancer-binding factor 3

ILF3_HUMAN

24.7

87.4

0.0

3.5

6

Non-POU domain-containing octamer-binding protein

NONO_HUMAN

13.4

80.4

0.0

6.0

7

Heterogeneous nuclear ribonucleoprotein R

HNRPR_HUMAN

28.8

74.1

0.0

2.6

8

Heterogeneous nuclear ribonucleoprotein A3

ROA3_HUMAN

23.8

70.6

0.0

3.0

9

Heterogeneous nuclear ribonucleoprotein L

HNRPL_HUMAN

21.7

57.2

0.0

2.6

10

Scaffold attachment factor B1

SAFB1_HUMAN

27.5

55.0

0.0

2.0

11

Insulin-like growth factor 2 mRNA-binding protein 1 (IMP1)

IF2B1_HUMAN

19.5

47.9

0.6

2.5

12

Scaffold attachment factor B2

SAFB2_HUMAN

20.0

45.0

0.0

2.3

13

Heterogeneous nuclear ribonucleoprotein A1

ROA1_HUMAN

20.8

44.9

0.1

2.2

14

Double-stranded RNA-specific adenosine deaminase (ADAR1)

DSRAD_HUMAN

16.5

42.3

0.3

2.6

15

Putative pre-mRNA-splicing factor ATP-dependent RNA

DHX15_HUMAN

13.0

31.0

1.0

2.4

16

Interleukin enhancer-binding factor 2

ILF2_HUMAN

8.3

25.3

0.0

3.0

17

Putative ATP-dependent RNA helicase DHX30

DHX30_HUMAN

6.3

23.5

0.0

3.8

18

Heterogeneous nuclear ribonucleoprotein K

HNRPK_HUMAN

10.7

23.2

0.0

2.2

19

Nucleolar RNA helicase 2

DDX21_HUMAN

7.5

23.0

0.7

3.1

20

RNA-binding protein FUS

FUS_HUMAN

7.0

20.0

0.0

2.9

20 proteins were selected as specific GGGGCC-repeat binding proteins based on the criteria described in the result section and in the legend of supplementary Table 1. The quantitative value reflects the relative protein amount estimated from the intensity of LC–MS/MS signal derived from the 500 mM elution fraction using Scaffold software. In the presence of non-biotinylated GGGGCC repeat competitor (competition) binding of all proteins is efficiently suppressed. Furthermore, these proteins show at least 2 times more binding to the GGGGCC RNA repeat (GC) compared to AAAACC RNA repeat (AC). Quantitative values listed here are the averages of three independent experiments. Standard errors of these average quantitative values are shown in supplementary Table 1

Positive results are described in bold

Identification of hnRNP A3-positive inclusions in C9orf72 mutation cases

To investigate a potential pathological involvement of these proteins, we performed immunohistochemical screenings using antibodies against the identified GGGGCC RNA repeat binding proteins on hippocampal paraffin sections derived from an autopsy case (case EM1) with genetically confirmed C9orf72 repeat expansion. Hippocampus was selected as a representative region, in which the characteristic inclusion body pathology of cases with C9orf72 mutations is seen consisting of TDP-43-positive neuronal cytoplasmic inclusions (NCI) (Fig. 2a), and p62-positive/TDP-43-negative dot-like NCIs, dot-like neuronal intranuclear inclusions (NII) (Fig. 2b) and star-like NCIs (Fig. 2c) [1]. Among the antibodies tested (Suppl. Tab. 2; Suppl. Fig. 1) only the antibody against hnRNP A3 [Suppl. Tab. 2 (ab1)] visualized specific inclusion pathology in the C9orf72 cases (Fig. 2d). This antibody recognized a 40 kDa protein in cerebellar lysates of C9orf72 carriers and controls (Suppl. Fig. 2a). siRNA-mediated knockdown of hnRNP A3 messenger RNA results in a significant reduction of hnRNP A3 protein in cultured cells (Suppl. Fig. 2b) again confirming the specificity of the hnRNP A3 antibody. To investigate if these hnRNP A3-positive inclusions are characteristic for C9orf72 GGGGCC repeat expansion carriers, we stained a series of hippocampal sections from 13 cases with C9orf72 mutations, 7 FTLD-TDP cases without C9orf72 mutation, 2 ALS-TDP cases without C9orf72 mutation, 1 case with Lewy body disease (LBD), 1 case with AD, 1 case with a combination of AD and LBD, 1 case with Pick’s disease, and 2 cases with Huntington’s disease (HD). To determine the physiological distribution pattern of hnRNP A3, five control cases without neurodegenerative alterations were added. All cases investigated are listed in Table 2. In most (4 out of 5) control cases hippocampal neurons and some glial cells showed a moderate to intense nuclear and a weak cytoplasmic hnRNP A3 expression (Fig. 2e; Suppl. Fig 3a). In contrast to that, a significant reduction of intranuclear hnRNP A3 staining expression was observed in 10 out of 13 C9orf72 mutation cases (Fig. 2d, f–i). In all C9orf72 mutation cases dot-like hnRNP A3-positive NCIs consistently occurred in the granular layer of the dentate gyrus (Fig. 2d, f–h; Table 2). However, hnRNP A3-positive dot-like inclusions were less frequently observed than p62-positive dot-like inclusions (compare Fig. 2b and 2d). In addition, a subset of star-like p62-positive/TDP-43-negative NCIs known to be pathognomonic for cases with C9orf72 hexanucleotide repeat expansions [1] were immunopositive for hnRNP A3 (arrowheads in Fig. 2c, i). 10 out of 13 C9orf72 mutation cases also showed hnRNP A3-positive NIIs (Fig. 2d, f–g; Table 2). Additionally, immunopositive dystrophic neurites (DN) were found in 9 out of 13 C9orf72 mutation cases (Fig. 2j; Table 2). No hnRNP A3-positive NCIs or NIIs were detectable in FTLD-TDP and ALS-TDP cases without GGGGCC repeat expansions (Fig. 2k, l; Suppl. Fig. 3b), AD, LBD, Pick or HD cases (Suppl. Fig. 3c–g; Table 2). However, few hnRNP A3-positive DNs were also found in 3 out of 7 FTLD-TDP cases without C9orf72 hexanucleotide expansions, but not in cases with other diseases (Table 2). No hnRNP A3-positive NCIs, NIIs or DNs were seen in any control case (Fig. 2e; Suppl. Fig. 3a; Table 2). These findings were confirmed by another anti-hnRNP A3 antibody (A3 ab2), which also visualized neuronal inclusions in a case with C9orf72 hexanucleotide expansions (Suppl. Fig. 3h) but not in a control case (Suppl. Fig. 3i). In the granular layer of cerebellum dot-like neuronal hnRNP A3 aggregates were found in 6 out of the 13 C9orf72 cases (positive and negative examples are shown in Fig. 3a–g; Table 2), while p62-positive inclusions were consistently observed in all cases (e.g., case EM1 in Fig. 3h). No hnRNP A3 positive NCIs or NIIs were detectable in the cerebella of 3 control, 2 FTLD-TDP and 2 ALS-TDP cases (Fig. 3i–l; Table 2). In C9orf72 mutation cases co-localization experiments revealed that 16.5–27.6 % of the p62 inclusions in the granular layer of the dentate gyrus were also immunopositive for hnRNP A3 (Fig 4a, d), but only 3.2 % of phosphorylated TDP-43 inclusions were also immunopositive for hnRNP A3 (Fig 4b, data not shown). In cerebellum of three C9orf72 cases with the most frequent cerebellar hnRNP A3 inclusions, 2–17 % of p62 inclusions also contained aggregated hnRNP A3 proteins (Fig. 4c, d).
https://static-content.springer.com/image/art%3A10.1007%2Fs00401-013-1088-7/MediaObjects/401_2013_1088_Fig2_HTML.jpg
Fig. 2

Immunohistochemical detection of phosphorylated TDP-43 (pTDP), p62, and hnRNP A3 in hippocampi of cases with C9orf72 hexanucleotide expansions, a FTLD-TDP case without C9orf72 mutation, and a control case. acC9orf72 mutation case EM1 with characteristic large pTDP-43-positive (red arrowheads in a) [24], small dot-like p62-positive NCIs (red arrowheads in b), small dot-like p62-positive NIIs (red arrows in b) in GL and large star-like p62-positive NCIs in CA3 (green arrowheads in c). d In GL of C9orf72 mutation case EM1 small dot-like hnRNP A3-positive NCIs (red arrowheads) and NIIs (red arrow) are seen whose shapes are similar to those of p62-positive inclusions shown in b. e In GL of non-diseased control case EM16, there is a strong nuclear and a weak cytoplasmic immunoreactivity for hnRNP A3, but there are no hnRNP A3-positive NCIs or NIIs. fh In GL of other cases with expanded C9orf72 repeats (EM3, EM10, EM8) hnRNP A3-positive dot-like NCIs (red arrowheads) and NIIs (red arrows) are detectable with varying frequency confirming the findings in case EM1 (d). i In CA4 of a case with expanded C9orf72 repeats (case EM10), a star-like hnRNP A3-positive NCI is seen (green arrowhead), whose size and shape is similar to the p62-positive NCIs characteristically found in CA regions of cases with expanded C9orf72 repeats (green arrowheads in c). j In CA4 of a C9orf72 mutation case (case EM7) a hnRNP A3-positive dystrophic neurite can be detected. k, l In a FTLD-TDP case without C9orf72 repeat expansion (case EM20), there are numerous often ring-shaped pTDP-positive NCIs in GL (k), but no hnRNP A3-positive NCIs or NIIs (l). Counterstains for visualizing cellular structures were done with haemalum. CA3 cornu ammonis region 3, CA4 cornu ammonis region 4, GL granular layer of the dentate gyrus, NCI neuronal cytoplasmic inclusion, NII neuronal intranuclear inclusion. Scale bar 20 μm

Table 2

Cases investigated with source, neuropathological diagnosis, C9orf72 genotyping and results on immunhistochemical hnRNPA3 stains

Case

Source

Neuropathological diagnosis

FTLD-TDP type

GGGGCC repeat expansion

hnRNP A3 expression

DG

CA

 

CBL

NCI

NII

NCI

NII

DN

NCI

NII

EM1

LMU

FTLD/ALS-TDP

B

+

+

+

+

+

EM2

LMU

FTLD/ALS-TDP

B

+

+

+

+

+

EM3

LMU

FTLD/ALS-TDP

B

+

+

+

+

+

+

+

EM4

VA

FTLD-TDP

B

+

+

+

+

+

+

EM5

VA

FTLD/ALS-TDP

B

+

+

+

EM6

VA

FTLD-TDP

A + B

+

+

+

+

+

+

EM7

VA

FTLD-TDP

A + B

+

+

+

+

+

+

EM8

VA

FTLD/ALS-TDP

B

+

+

EM9

VA

FTLD-TDP

A + B

+

+

+

+

+

+

+

EM10

VA

FTLD-TDP

B

+

+

+

+

+

+

+

EM11

VA

FTLD-TDP

B

+

+

+

EM12

VA

FTLD-TDP

B

+

+

+

+

EM13

VA

FTLD/ALS-TDP

B

+

+

+

EM14

LMU

Control

EM15

LMU

Control

EM16

VA

Control

nd

nd

EM17

VA

Control

nd

nd

EM18

VIB

Control

EM19

LMU

FTLD-TDP

B

+

EM20

LMU

FTLD-TDP

B

+

EM21

VA

FTLD-TDP

B

nd

nd

EM22

VA

FTLD-TDP

A

nd

nd

EM23

VA

FTLD-TDP

C

+

nd

nd

EM24

VA

FTLD-TDP

B

nd

nd

EM25

VA

FTLD-TDP

A

nd

nd

EM26

LMU

ALS-TDP

B

EM27

LMU

ALS-TDP

EM28

VA

LBD

nd

nd

EM29

VA

AD/LBD

nd

nd

EM30

VA

AD

nd

nd

EM31

VA

Pick

nd

nd

EM32

LMU

HD

nd

nd

nd

EM33

LMU

HD

nd

nd

nd

FTLD-TDP subtyping is based on Mackenzie et al. [22]

AD Alzheimer’s disease, ALS amyotrophic lateral sclerosis, CA cornu ammonis, CBL cerebellum, DG dentate gyrus, DN dystrophic neuritis, FTLD frontotemporal lobar degeneration, HD Huntington’s disease, LBD Lewy body disease, LMU Ludwig-Maximilians-University, Munich; NCI neuronal cytoplasmic inclusion, NII neuronal intranuclear inclusion, nd not determined, VA Vancouver General Hospital, Vancouver; VIB Antwerp VIB Department of Molecular Genetics, Antwerp

https://static-content.springer.com/image/art%3A10.1007%2Fs00401-013-1088-7/MediaObjects/401_2013_1088_Fig3_HTML.jpg
Fig. 3

Immunohistochemical detection of hnRNP A3 in the cerebellar granular layer of cases with C9orf72 hexanucleotide expansions, a control case, a FTLD-TDP, and a ALS-TDP case without C9orf72 mutation. ae In C9orf72 mutation cases EM3, EM10, EM6 and EM9 hnRNP A3-positive neuronal cytoplasmic inclusions (NCIs red arrowheads), neuronal intranuclear inclusions (red arrows), and a reduced nuclear staining are observed in varying frequency. In C9orf72 mutation case EM8, there is a strong nuclear staining of hnRNP A3 but no hnRNP A3-positive NCIs (f). In C9orf72 mutation case EM1 no hnRNP A3-positive NCIs are seen (g), while p62-positive dot-like NCIs (red arrowheads) are present (h). Strong nuclear staining of hnRNP A3 in control case EM18 (i). No hnRNP A3-positive NCIs and a variable degree of nuclear staining is observed in the cerebellar granular layer of control case EM15 (j), FTLD-TDP case EM20 (k), and ALS-TDP case EM27 (l). Counterstains for visualizing cellular structures were done with haemalum. CBL cerebellum. Scale bar 20 μm

https://static-content.springer.com/image/art%3A10.1007%2Fs00401-013-1088-7/MediaObjects/401_2013_1088_Fig4_HTML.gif
Fig. 4

Double immunofluorescence analysis of p62, phosphorylated TDP-43 (pTDP), and hnRNP A3 in hippocampus and cerebellum of cases with C9orf72 hexanucleotide expansions. a In DG-GL of a C9orf72 mutation case EM1 double immunofluorescence for p62 (green signal) and hnRNP A3 (red signal) reveals dot-like aggregates for both proteins. Merging both signals demonstrates that hnRNP A3 and p62 aggregates are co-localized in the cytoplasm of granular cells (arrow). Note that not all p62 aggregates contain hnRNP A3 (arrowheads). Double immunofluorescence for pTDP (green signal) and hnRNP A3 (red signal, arrow) in b does not show co-localization of the cytoplasmic aggregates. c Partial co-localization of p62 and hnRNP A3 the in granular layer of the cerebellum (arrows). Again, not all p62 aggregates contain hnRNP A3 (arrowheads). d Percentage of p62 inclusions that co-localize hnRNP A3 in DG-GL and CBL of indicated C9orf72 cases. Nuclei were marked with TO-PRO-3. DG-GL granular layer of the dentate gyrus, CBL cerebellum. Scale bar 10 μm

Discussion

Taken together, we identified hnRNP A3 as a first component of the so far enigmatic TDP-43-negative/p62-positive NCIs and NIIs in hippocampus, which are pathognomonic for cases with C9orf72 repeat expansions [1]. hnRNP A3 is a member of hnRNP A/B-type family of proteins (A1, A2/B1, A3, and A0) that contain two N-terminal RNA recognition motifs followed by a C-terminal glycine-rich auxiliary domain [17]. The members of the hnRNP A/B family are known to shuttle between nucleus and cytoplasm. The hnRNP A/B-type family of proteins performs multiple functions in alternative pre-mRNA splicing, nuclear import and cytoplasmic trafficking of mRNA, mRNA stability and turnover, and translation [17]. Most studies have been performed with the homologs hnRNP A1 and A2/B1. However, evidence exists that hnRNP A3 and A2/B1, but not A1 act as mRNA trafficking trans-acting factors in neurons [21]. Consistent with our findings in human non-diseased control cases, hnRNP A3 is mainly expressed in cell nuclei of mouse brain and human cell lines and to a lesser extent in the cytosol [26]. However, in C9orf72 cases an apparent redistribution from the nucleus to the cytosol is observed, which may cause a loss of an essential nuclear function. Of note, we also observed a similar redistribution in some cases with other neurodegenerative disorders. Therefore, further studies are required to prove the selectivity of this observation. Interestingly, nuclear clearance accompanied by a potential loss of function seems to be a more general phenomenon, as it is also observed for TDP-43 and to some extend for FUS as well [2, 20, 25, 40]. Furthermore, the major cause of ALS-FUS are mutations affecting a PY-nuclear localization signaling, further supporting a loss of nuclear function as one reason for the disease [10].

We found that hnRNP A3 binds to the GGGGCC repeats. Moreover, since hnRNP A3 is known to act in mRNA export [21], we speculate that enhanced binding to the C9orf72 repeat expansions could initiate aberrant export of C9orf72 pre-mRNA to the cytosol for subsequent degradation as it is observed for other splicing defective pre-mRNAs [18, 37] or even aberrant ATG-independent translation of the repeats as it has been described for the CAG-repeats in ataxin 8 [44]. hnRNP A3 is therefore a novel RNA-binding protein, which together with TDP-43 and FUS may contribute to the pathogenesis of familial ALS and FTLD. Moreover, hnRNP A3 is one protein constituent of some of the p62 positive TDP-43 negative C9orf72 specific inclusions. However, their relation to the described RNA foci [8, 34], whose existence is still controversial, remains to be determined. Furthermore, consistent with the lack of cerebellar hnRNP A3 deposits in some C9orf72 cases and with the staining of only a subset of the disease-signifying hippocampal inclusions additional protein constituents of the p62-positive/TDP-43-negative inclusions are to be expected. Other proteins may include RBM45 [6], ubiquilin [5], Rho-guanine nucleotide exchange factor (RGNEF) [19], or even peptides derived from aberrant ATG-independent translation of the repeats as it has been described for the CAG-repeats in ataxin 8 [44].

Acknowledgments

We thank Iryna Pigur for expert technical assistance, Axel Imhof, Harald Steiner, Akio Fukumori, Richard Page, and Eva Bentmann for providing tools and technologies and Dorothee Dormann for critically reading the manuscript. This work was supported by the Deutsche Forschungsgemeinschaft (SFB-596), the Competence Network for Neurodegenerative Diseases (KNDD) of the Bundesministerium für Bildung und Forschung (BMBF) to C.H. and the Consortium of Centers of Excellence in Neurodegenerative Brain Diseases (CoEN) to C.H., M.C., D.E., and C.V.B. K.M. was supported by a postdoctoral fellowship from the Alexander von Humboldt Foundation. D.E. was supported by the Helmholtz Young Investigator Program HZ-NG-607. The Agency for Innovation by Science and Technology provides a PhD fellowship to J.J. The authors acknowledge the Antwerp biobank of the Institute Born-Bunge for the brain samples as well as the neurologists S. Engelborghs and P.P. De Deyn and neuropathologist J.J. Martin for the clinical and pathological diagnoses. The Antwerp site is supported for the genetic research of neurodegenerative brain diseases by the Belgian Science Policy Office Interuniversity Attraction Poles program, the Foundation for Alzheimer Research (SAO/FRA), the Medical Foundation Queen Elisabeth, the Flemish Government Methusalem excellence program, the research Foundation Flanders (FWO) and the Special Research Fund of the University of Antwerp, Belgium.

Conflict of interest

The authors declare that they have no conflict of interest.

Supplementary material

401_2013_1088_MOESM1_ESM.xlsx (62 kb)
Supplementary material 1 (XLSX 62 kb)
401_2013_1088_MOESM2_ESM.pdf (536 kb)
Supplementary material 2 (PDF 536 kb)
401_2013_1088_MOESM3_ESM.docx (33 kb)
Supplementary material 3 (DOCX 33 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2013