Introduction
Amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTLD) are relentlessly progressive neurodegenerative disorders. ALS is characterised by the degeneration of motor neurons in the motor cortex and spinal cord, progressive paralysis and ultimately death due to respiratory failure [
9,
46]. FTLD is the second most common cause of dementia with an onset before 65 years. It is characterised by focal atrophy of the frontal and/or temporal lobes, causing changes in personality, behaviour and language [
52]. ALS and FTLD are increasingly recognised as the phenotypic ends of a clinical spectrum as ~15 % of patients with ALS have cognitive and language deficits akin to FTLD [
47]. Conversely, around 15 % of people with FTLD develop clinical signs of ALS [
32,
40].
ALS and FTLD also have overlapping molecular pathology [
31] as ~95 % of ALS cases and ~50 % of FTLD have neuronal cytoplasmic inclusions (NCIs) containing the TAR DNA binding protein (TDP-43) [
35,
45]. TDP-43 binds to a large number of RNA targets and is involved in the regulation of transcription, splicing and trafficking [
51]. In a minority of ALS and/or FTLD cases, cytoplasmic aggregation of TDP-43 and its loss from the nucleus is associated with pathogenic mutations in the genes encoding TDP-43 (
TARDBP[
50]), progranulin (
PGRN [
33]), valosin containing protein (
VCP[
42]) and Ubiquilin 2 (
UBQLN2 [
10]).
Mutations in another RNA binding protein, fused in sarcoma (FUS), have been identified in ~4 % of autosomal dominant familial ALS [
28,
55] and in rare cases of FTLD [
54]. FUS is ubiquitously expressed and plays an important role in the regulation of RNA transcription, splicing and transport [
4,
15,
58]. Although predominantly nuclear in distribution under physiological conditions, in patients with pathogenic mutations FUS protein accumulates in inclusions in the cytoplasm of lower motor neurons in the spinal cord [
11,
28,
34,
55]. Unlike many other neurodegenerative diseases, mutant FUS aggregates in ALS are not decorated by ubiquitin or p62 [
55]. The majority of FUS mutations occur in the extreme C-terminus of the protein, which contains a nuclear localisation signal [
13], thus it has been proposed that either loss of protein from the nucleus, or toxicity of protein aggregates accumulated in the cytoplasm leads to neurodegeneration [
12,
39].
Cytoplasmic and nuclear FUS inclusions have also been identified as the dominant protein deposited in FTLD cases previously classified as neuronal intermediate filament inclusion disease (NIFID), basophilic inclusion body disease (BIBD) and atypical FTLD with ubiquitinated inclusions [
38,
43,
44,
53]. In contrast to ALS cases, FUS mutations in FTLD are rare and the pathology is distinct. The RNA binding proteins TATA-binding protein-associated factor (TAF15) and Ewings Sarcoma protein (EWS) colocalise with FUS within inclusions in FTLD, which may explain why these inclusions appear to be ubiquitinated.
Knockout of Fus in inbred strains of mice results in chromosomal instability and perinatal death [
20], but only causes male sterility and enhanced sensitivity to radiation in outbred animals [
27]. Neither Fus−/− lines are reported to show motor or cortical neuronal loss. Overexpression of human ALS mutant FUS in adult transgenic rats, however, results in progressive paralysis secondary to the degeneration of motor axons and substantial loss of cortical and hippocampal neurons [
21]. Overexpression of the wild-type protein resulted in cognitive deficits in older animals due to a loss of cortical and hippocampal neurons [
21]. This indicated that mutant FUS is more toxic to lower motor neurons than normal FUS, but that increased expression of wild-type FUS is sufficient to cause neurodegeneration.
Here we report the effect of wild-type FUS overexpression in inbred mice, and demonstrate significant loss of motor neurons, coupled with a major motor and pathological phenotype that recapitulates many aspects of FUS-ALS. This phenotype appears to be crucially dependent on the expression level of the protein, and is associated with a significant shift in FUS localisation to the cytoplasm without concomitant loss of FUS from the nucleus.
Materials and methods
Ethics statement
All experiments were performed under the terms of the UK Animals (Scientific Procedures) Act 1986, and were approved by the Kings College, London ethics review panel.
Transgenic animals
For creation of transgenic mice, HA-tagged human FUS cDNA was cloned into a modified mouse prion gene and following removal of vector sequences, C57Bl/6/SJL founder mice were produced by pronuclear injection as described [
30]. Animals were backcrossed onto C57Bl/6 three times prior to further breeding, following which hemizygous animals were interbred to produce homozygous offspring. Direct sequencing of the FUS transgene was performed with Big-Dye
® Terminator v1.1 on an ABI3130 genetic analyser (Applied Biosystems). All sequence traces were analysed using Sequencher
® 4.10 (Gene Codes Corporation). Hemizygous animals were identified using PCR with primers 5′-GCAGGGCTATTCCCAGCAGAGCAG and 3′-CTGGTTCTGCTGTCCATAGCCCTG. Homozygous mice were identified using qPCR with primers 5′-CAGCAAAGCTATGGACAGC and 3′-GCGGTTATGGCAATCAAGAC and a FAM/TAMRA tagged probe-AGCAGAACCAGTACAACAGCAGCA. GAPDH was used as a housekeeping control. Each sample was measured in triplicate. hFUS (+/−) mice will be available through the Jackson Laboratory Repository. They are assigned JAX stock no. 017916.
Evaluation of motor function and health
From 3 weeks of age, NTg (n = 9), hFUS (+/−) (n = 17) and hFUS (+/+) (n = 11) mice were weighed on a weekly basis, and general health status was recorded. Animals showing signs of hind limb paralysis were monitored daily, and disease end-stage and death was defined as the time when animals could no longer obtain food or water, or had lost 25 % of their body weight, at which point they were euthanized.
Motor strength and coordination were longitudinally evaluated on the rotarod (Columbus Instruments), using a 5-min accelerating protocol starting at 2 rpm, and rising to 30 rpm throughout the 5-min testing period. Mice were tested once a week, and latency to fall was recorded. All mice received an initial training session of 2 min at 2 rpm to acclimatise them to the equipment. Data were assessed statistically by two-way analysis of variance (ANOVA) followed by the post hoc Holm-Sidak test. At eight weeks of age, the locomotor activity of animals was assessed in an 80-cm diameter circular open field environment. Mice were allowed to explore the open field for 10 min. Trials were monitored and analyzed using the Ethovision package (Noldus, The Netherlands), and total distance travelled was recorded. Data were assessed statistically by way of one-way ANOVA followed by the post hoc Tukey test.
Histology and immunohistochemistry
Eleven-week-old, end-stage mice, and their age-matched littermates were anaesthetised and transcardially perfused with PBS, followed by 4 % paraformaldehyde (PFA) in phosphate buffer. Brain and spinal cords were postfixed in 4 % PFA in 15 % sucrose for 5 h, cryoprotected in 30 % sucrose for 24 h and cut into 30 μm sections on a cryostat.
For immunohistochemistry, the following antibodies were used: rabbit anti-FUS (1:500, Sigma), rabbit anti-ubiquitin (1:1000, DAKO), rat anti-HA (1:5000, Roche), rabbit anti-GFAP (1:4000, DAKO), mouse anti-CD68 (1:2000, Serotec), goat anti-EWS (C-19; 1:50, Santa Cruz) and TAF15 (TAF II p68; 1:50, Santa Cruz). Sections were washed and incubated with the appropriate biotinylated secondary antibody (1:1000, Vector), and then with an ABC kit (Vector). Sections were imaged using a Zeiss light or confocal microscope and axiovision software.
For motor neuron counts, perfused lumbar spinal cords from four animals per genotype were serially sectioned, and every 6th section (30 μm) was analysed. Sections were mounted, dried, incubated overnight in 1:1 ethanol/chloroform to de-fat the sections, stained for 10 min in warm 0.1 % cresyl violet, dehydrated and coverslipped. To compare the number of motor neurons, large neurons greater than 30 μm in diameter (based on their Fret’s diameter as assessed by Image J software) in the anterior horn of the lumbar spinal cord were counted in 15 sections. Data were assessed statistically by one-way ANOVA, followed by the post hoc Tukey test.
For muscle histology, gastrocnemius or tibialis anterior muscles were dissected fresh, immediately frozen in isopentane cooled in dry ice, and cryostat sections were cut onto slides and stained with haematoxylin and eosin or succinate dehydrogenase (SDH) activity to determine oxidative capacity of the muscle fibres, as described [52].
Immunoblotting
For FUS and HA expression level analysis, whole brains of four end-stage hFUS (+/+) and four age-matched hFUS (+/−) and NTg animals were lysed in low salt buffer (10 mM Tris, 5 mM EDTA, 10 % sucrose) with protease inhibitors (Roche Diagonstics). For cytoplasmic and nuclear fractionation, four brain samples for each genotype were prepared as described earlier [
19]. Briefly, snap-frozen tissue was weighed and homogenised in buffer containing 10 mM Hepes, 10 mM NaCl, 1 mM KH
2PO
4, 5 mM NaHCO
3, 5 mM EDTA, 1 mM CaCl
2, 0.5 mM MgCl
2 and protease inhibitors (10× vol/weight). After 10 min on ice, 2.5 M sucrose (0.5× vol/weight) was added. Tissue was homogenized and centrifuged at 6,300
g for 10 min. The supernatant was collected as the cytoplasmic fraction. The pellet was washed 4 times in TSE buffer [10 mM Tris, 300 mM sucrose, 1 mM EDTA, 0.1 % IGEPAL (Sigma) and protease inhibitors 10× vol/weight], homogenized and centrifuged at 4,000×
g for 5 min. Finally, the pellet was resuspended in RIPA buffer with 2 % SDS (5 × vol/weight) as the nuclear fraction. Protein samples were then separated by SDS/PAGE using 10 % polyacrylamide gels, and transferred to nitrocellulose membranes. Total FUS was recognised by a rabbit polyclonal antibody to FUS (1:2,000, Novus Biologicals), and exogenous HA tagged human FUS was recognised by a mouse monoclonal antibody to the HA tag (1:1,000, Cell Signalling). Fluorescent secondary antibodies conjugated to Dylight 680 or 800 nm (Thermo Scientific) were used to detect protein levels, and results were visualised using the Odyssey Imager (Licor). Data were normalised to GAPDH (1:5,000, Sigma) or Lamin B1 (1:2,000, Abcam). Quantitation of immunoblots was done using Image J software, and data were analysed statistically by way of ANOVA followed by the post hoc Tukey test.
In vivo physiological assessment of neuromuscular function and motor unit survival
Functional analysis of hind limb muscle function was undertaken at 70 days. NTg (
n = 5), hFUS(+/−) (
n = 6) and hFUS(+/+) (
n = 6) mice were anaesthetised (4.5 % chloral hydrate, 1 ml/100 g of bodyweight, i.p) and prepared for in vivo analysis of isometric muscle force as previously described [
24]. The distal tendons of the tibialis anterior (TA) and extensor digitorum longus (EDL) hind limb muscles were dissected free and attached by silk thread to isometric force transducers (Dynamometer UFI Devices, Welwyn Garden City, UK). The sciatic nerve was exposed and sectioned proximally. The length of the muscles was adjusted for maximum twitch tension. The muscles and nerves were kept moist with saline through recordings and all experiments were carried out at room temperature. Isometric contractions were elicited by stimulating the nerve to the TA and EDL using square wave pulses of 0.02 ms duration at supra-maximal intensity, via silver wire electrodes. Contractions were elicited by either a single stimulus for twitch tension or trains of stimuli at frequencies of 40, 80 and 100 Hz for tetanic tension. The maximum twitch and tetanic tension was measured using a computer and Picoscope software (Pico Technology, Cambridgeshire, UK).
Following recording of isometric tension, the contractile and fatigue characteristics of EDL muscles were determined. The time to peak (TTP) was calculated by measuring the time taken (ms) for the muscle to elicit peak twitch tension and the half relaxation time (the time taken for the muscle to reach half relaxation from peak contraction) was also calculated. In addition, the resistance of the EDL muscles to fatigue was assessed by repeated stimulation at 40 Hz for 250 ms every second for 3 min. The tetanic contractions were recorded on a Lectromed Multitrace 2 recorder (Lectromed Ltd, UK). The decrease in tension after 3 min of stimulation was measured and a fatigue index (FI) was calculated, where a FI approaching a value of 1.0 indicating that a muscle is very fatigable [
52].
The number of motor units innervating the EDL muscles was determined by stimulating the nerve with stimuli of increasing intensity, resulting in stepwise increments in twitch tension due to successive recruitment of motor axons with increasing stimulus thresholds. The number of stepwise increments was counted to give an estimate of the number of functional motor units in the EDL muscles [
52].
Discussion
ALS and FTLD are two neurodegenerative disorders that fall within an overlapping clinical and pathological disease spectrum that includes the TDP-43 and FUS proteinopathies [
31,
32,
40]. FUS and TDP-43 are structurally related RNA-binding proteins that may have overlapping functions and form complexes with other RNA-binding proteins [
25]. Both proteins reside predominantly in the nucleus, and shuttle between the nucleus and the cytoplasm to perform various functions [
3,
14,
18,
36]. It has been proposed that both FUS and TDP-43 may induce disease via either a gain of function within the cytoplasm, or a loss of function from the nucleus. The loss of function hypothesis is supported by evidence of nuclear depletion of TDP-43 in neurons containing cytoplasmic inclusions [
2,
12,
45], a finding that has not yet been conclusively reported for FUS.
FUS accumulates within cytoplasmic inclusions in ALS patients (and rarely FTLD cases) when it is associated with
FUS gene mutations [
28,
54,
55]. FUS containing inclusions have also been identified as the dominant pathology in a subset of FTLD patients accounting for most cases that have TDP-43 and tau negative pathology [
38,
43,
48,
53]. We generated hemizygous and subsequently homozygous FUS transgenic mouse lines expressing HA-tagged human wild-type FUS at 1.4- and 1.9-fold above endogenous levels in non-transgenic littermates (respectively). Overexpression of human FUS resulted in significantly decreased expression of mouse Fus. The overall pattern of Fus positive staining in all animals, regardless of genotype, was consistent with that previously reported [
1], with intense neuropil staining in the cord and to a lesser extent the brainstem, together with nuclear staining throughout the CNS, although neuropil staining throughout the brain appears slightly increased in hFUS (+/+) mice compared to their NTg and (+/−) littermates, which is consistent with the increase in cytoplasmic Fus observed in these mice.
Although additional attempts were made, we were unable to generate further viable transgenic lines over-expressing human wild-type FUS under the control of the Prion promoter for comparison to the line reported here. This is similar to problems described in TDP-43 rodent models of ALS [
56,
59], and is likely to be due to selective pressure against the expression of FUS above endogenous levels during early development, given its apparent dose-dependent toxicity in cellular models [
7,
16,
23].
Homozygous transgenic mice developed a rapidly progressive motor deficit as measured by rotarod, locomotor activity and neurophysiological testing. Tremor began around 4 weeks and progressed to paralysis at 10–12 weeks, necessitating euthanasia. End-stage pathology revealed approximately 60 % motor neuron loss from the lumbar spinal cord. Many surviving neurons contained FUS-positive, ubiquitin-negative inclusions in the spinal cord, and there was evidence of microglial and astrocytic activation in the anterior horn and white matter of the dorsal columns. These changes are the pathological hallmark of human mutant FUS-mediated ALS cases [
55]. Most ALS-related FUS mutations reside in the C-terminal nuclear localising signal (NLS). They disrupt binding to the nuclear transport factor, transportin, which impairs their nuclear import, leading to cytoplasmic accumulation within motor neurons and toxicty [
13]. The degree of cytoplasmic mislocalisation in cellular studies for each ALS mutant appears to correlate with the age of disease onset in ALS patients implying that cytoplasmic mislocalisation is directly toxic [
13]. Conversely, deletion of the nuclear export signal strongly suppressed mutant FUS toxicity in
Drosophila, implying that cytoplasmic localisation is required for neurodegeneration [
29]. This finding is supported by a recent study in
C. elegans demonstrating that cytoplasmic mislocalisation of FUS induced by mutant FUS is sufficient to cause motor dysfunction, even in the presence of functional levels of wild-type FUS in the nucleus [
39]. Here, we show that overexpression of wild-type FUS is toxic to motor neurons when it accumulates in the cytoplasm. Thus, the toxicity of FUS mutations may be solely due to their impact on nuclear import [
6,
13,
17,
22,
26]. Although nuclear Fus levels also increase in these animals, it is unlikely that this is responsible for the main pathological phenotype, as hemizygous animals show a significant increase in nuclear Fus, and yet show no signs of overt phenotype after two years of age. In the brain, there was no evidence of microglial and astrocytic activation or neuronal loss despite abundant granular and skein-like FUS-positive inclusions in the cytoplasm of multiple neuronal populations. Given the early onset severe motor dysfunction in these animals, it was not possible to assess them for cognitive impairment. In contrast, hemizygous animals have no gross pathological changes in either the brain or spinal cord at 12 weeks and show no evidence of motor dysfunction out to 104 weeks. They did however, show mild astroglial and microglial activation in the spinal cord at 12 weeks. There were also subtle changes within muscle architecture in hemizygous mice, without signs of neuronal loss or neurophysiological evidence of neuromuscular dysfunction. A cohort of hemizygous mice is currently being aged and will be screened for signs of cognitive dysfunction. Recent findings in FUS transgenic rats overexpressing wild-type FUS revealed cognitive defects in aged animals [
21]. Fus overexpression within the CNS of our mice is variable, with some cell populations displaying more expression than others. This suggests that the PrP-driven expression may display some mosaic expression properties in this model. However, the lack of obvious cell loss in several brain cell populations showing high levels of Fus overexpression supports the idea that lower motor neurons are selectively vulnerable to Fus overexpression. We therefore conclude that the overexpression of human wild-type FUS is particularly toxic to motor neurons.
These findings are consistent with recent reports of wild-type FUS toxicity in other species. Overexpression of wild-type Fus results in punctate cytoplasmic aggregates and dose-dependent toxicity in yeast [
16,
23] and apoptotic cell death in human prostate cancer cells [
7]. Overexpression of human wild-type FUS restricted to neurons in
Drosophila results in a dose-dependent decrease in life span [
37], and an impaired locomotor phenotype accompanied by morphological abnormalities in motor neurons and neuromuscular junctions [
8].
Overexpression of the R521C ALS FUS mutant in the rat [
21] resulted in an aggressive motor phenotype leading to death of all animals by 10 weeks. This was accompanied by disruption of the neuromuscular junction and a ~10 % motor neuron loss. They report some FUS mislocalisation to the cytoplasm without discrete inclusion formation and although ubiquitinated inclusions were observed within spinal neurons they do not colocalise with FUS. Rats transgenically overexpressing wild-type human FUS have no motor phenotype or spinal cord pathology but develop cognitive dysfunction and neuron loss in the frontal cortex and dentate gyrus by 12 months. Here again, FUS localised to the nucleus with little cytoplasmic mislocalisation and no inclusions. Ubiquitinated inclusions were detected in FUS expressing cells but did not colocalise with FUS. Our transgenic mice display a progressive and lethal motor phenotype similar to that seen in the rats expressing mutant FUS. In our mice, neuronal loss was only seen in homozygous animals where significant cytoplasmic mislocalisation of FUS was observed without loss of nuclear FUS. Both mouse and rat transgenic models support a toxic gain of function due to the accumulation of FUS rather than a loss of nuclear function.
Our mice constitutively overexpress wild-type human FUS under the control of the prion promoter. In comparison, the transgenic rat lines, wild-type and mutant, employed a cDNA construct driven by the Tet-off TRE and tTA promoter system whereby expression is induced only on weaning [
21]. Thus, it is possible that the different promoter systems and onset of expression, combined with species and strain variations, may account for some of the different specific effects that overexpression of the wild-type FUS has in these two models. Nevertheless, neurodegeneration was observed in both our wild-type FUS mice, and the wild-type FUS transgenic rats, demonstrating that FUS overexpression is pathogenic.
The phenotype observed in our hFus (+/+) mice bears some similarities to mice overepxressing wild-type TDP-43 [
57], such as a failure to gain weight, tremor, and motor dysfunction. Since both Fus and TDP-43 are RNA-binding proteins, the similarities in dysfunction that occur in response to overexpression of either protein may suggest that dysfunctions in RNA processing are important in the development of disease.
We conclude that overexpression of wild-type human FUS will induce motor neuron degeneration in mice when protein levels are sufficient to cause significant cytoplasmic accumulation. Our mice reproduce many aspects of the clinical phenotype and pathological features of human mutant FUS-mediated ALS, and as such, this model will facilitate the exploration of disease mechanisms and opportunities for therapy.