|Home | About | Journals | Submit | Contact Us | Français|
Ubiquitin-positive inclusion containing Fused in Sarcoma (FUS) defines a new subtype of frontotemporal lobar degeneration (FTLD). FTLD is characterized by progressive alteration in cognitions and it preferentially affects the superficial layers of frontotemporal cortex. Mutation of FUS is linked to amyotrophic lateral sclerosis and to motor neuron disease with FTLD. To examine FUS pathology in FTLD, we developed the first mammalian animal model expressing human FUS with pathogenic mutation and developing progressive loss of memory. In FUS transgenic rats, ubiquitin aggregation and FUS mislocalization were developed primarily in the entorhinal cortex of temporal lobe, particularly in the superficial layers of affected cortex. Overexpression of mutant FUS led to Golgi fragmentation and mitochondrion aggregation. Intriguingly, aggregated ubiquitin was not colocalized with either fragmented Golgi or aggregated mitochondria, and neurons with ubiquitin aggregates were deprived of endogenous TDP-43. Agonists of peroxisome proliferator-activated receptor gamma (PPAR-γ) possess anti-glial inflammation effects and are also shown to preserve the dendrite and dendritic spines of cortical neurons in culture. Here we show that rosiglitazone, a PPAR-γ agonist, rescued the dendrites and dendritic spines of neurons from FUS toxicity and preserved rats' spatial memory. Our FUS transgenic rats would be useful to the mechanistic study of cortical dementia in FTLD. As rosiglitazone is clinically used to treat diabetes, our results would encourage immediate application of PPAR-γ agonists in treating patients with cortical dementia.
Frontotemporal lobar degeneration (FTLD) is a clinical syndrome that is characterized by progressive alteration in behavior, language and personality, with relative preservation of memory at early disease stages (1,2). About half FTLD cases show the pathology of tau-negative, ubiquitin-positive inclusions in the cytoplasm and neurites, particularly in the superficial layers of the frontotemporal cortex and in the dentate gyrus of the hippocampus (1–4). This subtype of FTLD with ubiquitin inclusions is now called FTLD-U. In most FTLD-U cases, ubiquitin inclusions entrap Tar DNA-binding protein 43 (TDP-43) (5), whereas a subtype of FTLD-U shows TDP-43-negative, ubiquitin-positive inclusion that is immunostained of Fused in Sarcoma (FUS) (6). Both TDP-43 and FUS are RNA- and DNA-binding proteins (7), but they define a distinct pathology of FTLD.
FUS can bind to both single- and double-stranded DNA sequences (8), promoting D-loop formation in DNA repair (9). FUS interacts with the p65 subunit of the nuclear factor κB to regulate gene transcription (10). FUS binds RNA molecules, particularly the RNA with GGUG sequence (11), and FUS is associated with the complex of splicing factors, regulating RNA splicing (12). In physiological conditions, FUS is mainly in the nucleus of neurons, but it shuttles between the nucleus and the cytoplasm to perform complex functions (13,14). In hippocampal pyramidal neurons, FUS is translocated to dendrites and accumulates in dendritic spines at excitatory post-synapses upon mGlR5 activation, participating in mRNA sorting to dendritic spines and regulating spine morphology (15,16). Expression of FUS is rapidly downregulated in the peripheral organs during postnatal development, but it remains at significant levels in neurons throughout the lifetime (17), suggesting that FUS may exert important functions in the central nerve system.
Mutation of FUS is linked to amyotrophic lateral sclerosis (ALS) (18–20), MND with dementia (21) and FTLD (22). Mutation in FUS changes the profile of its RNA binding (23), suggesting that pathogenic mutation in FUS gains toxic properties. Indeed, overexpression of mutant, but not the normal, FUS causes rapid paralysis in Caenorhabditis elegans and in rats (24–26). FUS deficiency causes early death in homogeneous C57/BL6 background (27), but causes only male infertility in heterogeneous genomic background (28). Studies in yeast and Drosophila suggest that cytoplasmic localization of mutant FUS is required for cytotoxicity (29,30). FUS and TDP-43 are two similar ribonucleoproteins and may interact with each other in a common pathway leading to neurodegeneration (31). In Drosophila and zebrafish (32,33), overexpression of FUS rescues the phenotypes of TDP-43 deficiency, but does not vice versa, suggesting that FUS may function downstream of TDP-43 in a pathway. Till now, the only mammalian animal model expressing human FUS develops rapid paralysis as the predominant phenotype, with limited loss of cortical neurons (25). How FUS pathology correlates with cortical dementia in FTLD remains to be determined.
To examine FUS in the pathogenesis of FTLD, we developed novel transgenic rats expressing mutant FUS restrictedly in neurons of the forebrain. This novel rat model reproduced the core phenotypes of FTLD: severe loss of neurites and dendritic spines; progressive loss of neurons particularly in the superficial layers of the cortex; and gradual development of FUS mislocalization and ubiquitin aggregates primarily in the entorhinal cortex of the temporal lobe. In this novel rat model, the spatial memory was well preserved when the neurites and dendritic spines were rescued by treatment with rosiglitazone.
Although mutation in the fus is causative of some familial ALS (18,19), FUS pathology defines a distinct subtype of FTLD that displays ubiquitin- and FUS-positive, but TDP-43 negative, inclusions (34,35). Ubiquitous expression of mutant human FUS in rats causes early-onset paralysis without a significant loss of motor neurons (25). To examine FUS pathobiology in FTLD, we attempted to overexpress mutant human FUS restrictedly in the forebrain and thus to induce cortical dementia in transgenic rats (Figs 1–3).
To express a transgene in the forebrain, we isolated the promoter (9 kb) of the mouse camk2a from a BAC clone and used it to drive tetracycline-responsive transactivator (tTA) in rats (Fig. 1A). We established one transgenic line that carries two copies of the tTA transgene as determined by quantitative PCR. To determine the profile of transgene expression, we crossed Camk2a-tTA transgenic rats with a LacZ reporter line (36). X-gal staining shows that expression of the reporter gene was restricted in the forebrain (Fig. 1B). X-gal-stained cells were observed in the hippocampus and cortex (Fig. 1C and D), but not in the corpus callosum (Fig. 1D). The corticospinal tracts in the spinal cord were also stained (Fig. 1E). Expression of the tTA in the rat forebrain did not affect neuron development and maturation and thus did not affect rats' spatial learning (Supplementary Material, Fig. S1).
To induce expression of mutant human FUS in the rat forebrain, we crossed Camk2a-tTA transgenic line with mutant FUS transgenic rats (line 16) (25). In the double-transgenic rats, we observed expression of human FUS only in the neurons (Fig. 2A–F), but not in the glia (Fig. 2G–O). Profiles of transgene expression in Camk2a-tTA/TRE-FUS-R521C rats were consistent with that in Camk2a-tTA/TRE-LacZ rats (Figs 1 and and2),2), validating that the Camk2a promoter drives transgene expression restrictedly in the neurons of the rat forebrain.
We attempted to induce expression of mutant FUS in postnatal rats and thus to induce disease phenotypes in adult rats (Fig. 3). Expression of mutant FUS is subject to the regulation by doxycycline (Dox) (Fig. 1A). Dox was withdrawn from postnatal rats at the age of 30 days and expression of the FUS transgene was initiated gradually (Fig. 3A–I). After Dox withdrawal, the mutant FUS transgene was activated first in the hippocampus and then in the cortex (Fig. 3B–I). Overexpression of mutant FUS in the forebrain impaired rats' spatial memory, which was detected by Barnes maze assay (Fig. 3J). The rats began losing spatial memory (disease onset) about 5 weeks after Dox withdrawal, and by the age of 18 weeks most of the rats had difficulty in locating the escaping hole in the Barnes maze and reached disease end-stages (Fig. 3J). In the FUS transgenic rats, neurons were progressively lost in the hippocampus and cortex (Fig. 3K–R). Neuronal loss was quantified with stereological cell counting (Fig. 3S and T). The brain regions of neuronal loss were correlated with the initiation of transgene expression: expression of the mutant FUS transgene was detected earlier in the hippocampus than in the cortex (Fig. 3B–I) and thus the neuronal loss was first detected in the dentate gyrus (Fig. 3K–T). At the onset of the disease, neurons were lost severely in the dentate gyrus, but were only lost moderately in the cortex (Fig. 3). At disease end-stages, most of the neurons were lost in the frontal cortex (Fig. 3S and T). Our findings suggest that both the dentate gyrus and the frontal cortex contribute to spatial learning and spatial memory in rats.
After the mutant FUS transgene was turned on in postnatal rats, the rats began losing spatial memory by the fifth week of transgene expression (Fig. 3J). At disease onset, no significant loss of neurons was detected in the cortex, though it was detected in the dentate gyrus (Fig. 3K–T). As neurites and spines play a critical role in neurotransmission, these structures might be destroyed and the neuronal functions might be disturbed before the neurons die. Using Golgi staining, we detected that the neurons in the frontal and entorhinal cortexes were severely deprived of neurites before memory loss (Fig. 4A–L). In the frontal cortex, neurons of layer II were first affected (Fig. 4A–F), displaying short and retracted neurites. Similarly, neurons in the entorhinal cortex lost neurites severely (Fig. 4G–L). Quantitative measures show that neurite branches and the spine density on both the apical and basal dendrites were markedly reduced in the transgenic rats (Fig. 4M and N). Neurite and dendritic spines are the primary targets of degeneration caused by mutation in FUS.
Mutation of FUS is linked to ALS and FTLD, and Golgi fragmentation is frequently observed in the motor neurons of ALS patients (18,19,37,38). We examined Golgi apparatus in mutant FUS transgenic rats and observed that both cis- and trans-Golgi complexes were fragmented in the neurons expressing mutant human FUS (Fig. 5, Supplementary Material, Figs S2 and S3). After the mutant FUS transgene was turned on, Golgi complexes were fragmented progressively in the rats at pre-symptomatic stages (Fig. 5A–I), suggesting that Golgi fragmentation is an early event in FUS pathology. Golgi fragmentation was observed in all of the brain regions examined, but appeared to be severe in the entorhinal cortex of the temporal lobe (Fig. 5). Fragmented Golgi complexes were restricted to the neurons expressing mutant human FUS (Fig. 5 and Supplementary Material, Fig. S3), suggesting that Golgi fragmentation is a result of the overexpression of mutant human FUS.
FUS shuttles between the nucleus and the cytoplasm in physiological conditions and its normal localization might be interrupted by pathogenic mutations (39,40). We examined FUS localization in mutant FUS transgenic rats at presymptomatic and symptomatic stages and we observed that mutant FUS accumulated in the cytoplasm and in the neurites (Fig. 6A–G and Supplementary Material, Fig. S4A–F). Compared with the other brain regions expressing mutant FUS, the entorhinal cortex was predisposed to FUS mislocalization (Fig. 6G). As the disease was progressing, mutant FUS progressively accumulated in the neurites and in the dendritic spines (Supplementary Material, Fig. S4). The entorhinal cortex is part of the temporal lobe that is primarily affected in FTLD. Our finding suggests that entorhinal cortical neurons are predisposed to mutation of FUS.
As the disease was progressing, ubiquitin was accumulated and aggregated in affected neurons particularly in the entorhinal cortex (Fig. 6H–N and Supplementary Material, Fig. S5). Similar to mislocalized FUS, aggregated ubiquitin was detected primarily in the entorhinal cortex (Fig. 6N). Aggregated ubiquitin and mislocalized human FUS were simultaneously detected in the neurites of the same affected neurons, but double-labeling fluorescence staining revealed that these two proteins were not physically colocalized in the neurites or the cytoplasm (Fig. 7A–I). Intriguingly, some neurons with ubiquitin aggregates were deprived of endogenous TDP-43 protein (Fig. 7J–L). Aggregated ubiquitin was not colocalized with fragmented Golgi complex (Fig. 7M–R). As mutant FUS was expressed, some neurons displayed an enhanced staining of Cox-IV indicative of mitochondrial clustering and damage (Supplementary Material, Fig. S5A–I). Aggregated ubiquitin also was not colocalized with damaged mitochondria (Fig. 7S–X). In mutant FUS transgenic rats, ubiquitin-positive inclusions entrap something other than fragmented Golgi complex and damaged mitochondria.
Glial reaction is a common feature of neurodegenerative diseases, including FTLD. We examined astrocytic and microglial reaction in mutant FUS transgenic rats and observed that both microglia and astrocytes were activated in all of the brain regions examined (Fig. 8). Glial reaction was closely associated with neuron death. As neuronal loss occurred earlier in the hippocampus than in the cortex, reactive astrocytes and microglia were detected first in the hippocampus (Fig. 8E and F) and then in the cortex (Fig. 8A–D). In the frontal cortex, neurons of layers II and III were first affected (Fig. 4). Accordingly, reactive microglia and astrocytes were first detected in the layers II and III of frontal cortex (Fig. 8A3 and C3). Glial reaction is related to neuronal death.
In mutant FUS transgenic rats, neurites and dendritic spines were the primary targets of neurodegeneration, and glial cells were remarkably activated in response to the neuronal death (Figs 4 and and8).8). Rosiglitazone is a potent agonist of peroxisome proliferator-activated receptor gamma (PPAR-γ) and possesses anti-inflammation effects on glial reaction (41,42). Recent studies showed that rosiglitazone rescues dendritic spine loss in cultured primary neurons (43). We examined the neuroprotective effects of rosiglitazone in mutant FUS transgenic rats.
Mutant rats were treated with rosiglitazone or vehicle at the onset of memory loss (Fig. 9). By the time of disease onset, neurons in the dentate gyrus were lost severely and glial cells were overtly reactive (Figs 3 and and8).8). Treatment with rosiglitazone slowed down the progression of cortical dementia in FUS transgenic rats (Fig. 9A). We examined neuronal loss in rosiglitazone-treated rats by stereological cell counting and found that neuronal loss was not rescued by rosiglitazone (Fig. 9B and C). We further examined dendritic spines by Golgi staining and observed that rosiglitazone rescued the loss of dendritic spines caused by mutation of FUS (Fig. 9D–I). Rosiglitazone is known for its effects on glial reaction. Indeed, rosiglitazone mitigated astroglial and microglial reaction (Fig. 10A–F), with a dramatic effect on microglial activation. Moreover, treatment with rosiglitazone mitigated Golgi fragmentation and ubiquitin aggregation in the neurons expressing mutant FUS (Fig. 10G–R). Consistent with previous findings in culture (43), our data show that rosiglitazone prevented cortical dementia from progression by preserving dendritic spines.
Overexpression of mutated human FUS in the rat forebrain initiated a dying-back process of neurodegeneration that was accompanied by gradual development of ubiquitin aggregates and FUS mislocalization. Similar to the pathology of FTLD, the pathological changes observed in FUS transgenic rats were predominant in the temporal lobe, and the neuronal death was preceded by remarkable Golgi fragmentation. In FUS transgenic rats, anti-glial inflammation treatment prevented the loss of neurites and dendritic spines and preserved spatial memory.
FUS pathology in our rat model preferentially affected the temporal lobe that is primarily affected in FTLD. Although mutant FUS was expressed in the whole forebrain, including the frontal, parietal and temporal cortexes and the hippocampus, FUS mislocalization and ubiquitin aggregates were primarily developed in the entorhinal cortex that is part of the temporal lobe. The pathological changes preceded the neuronal loss and possibly are related to the neurodegeneration in FUS transgenic rats. Consistent with the findings in patients with FTLD (1–4), neurodegeneration in our FUS transgenic rats was a dying-back process and preferentially affected the superficial layers of the cortex at early disease stages. Although aggregated ubiquitin and mislocalized mutant FUS were simultaneously found in the same neurons, ubiquitin immunostaining was not physically colocalized with the staining of mutant FUS, differing from clinical observation that reveals the colocalization of ubiquitin with FUS in the same inclusions (6). This difference in FUS pathology might be caused by a relatively faster progression of disease phenotypes in rats than that in patients or might result from the distinction of species. FUS and TDP-43 are two similar ribonucleoproteins and are both involved in ALS and FTLD. Viral delivery of TDP-43 induces cytoplasmic mislocalization of TDP-43 in monkey, but expresses TDP-43 only in the nucleus in rats (44). Mitochondrion and Golgi apparatus are involved in neurodegeneration (45–47). Expression of human FUS with pathogenic mutation induced Golgi fragmentation in the cortical neurons, which preceded neuronal death (Fig. 5). Damaged mitochondria form aggregates that can be labeled of Cox-IV (48). Expression of mutant FUS in rats induced aggregation of mitochondria (Fig. 7). Intriguingly, ubiquitin-positive inclusions did not contain damaged Golgi or mitochondria (Fig. 7). What ubiquitin-positive inclusions entrap in FUS transgenic rats remains to be determined. FUS may function in the downstream of TDP-43 in a common pathway as FUS rescues TDP-43-deficient phenotypes in invertebrate animal models (32,33). Unexpectedly, few cells with ubiquitin-positive inclusion were deprived of TDP-43 in FUS transgenic rats (Fig. 7). It remains to determine how TDP-43 depletion is related to FUS pathology. Our FUS transgenic rat is the first mammalian animal model recapitulating the core phenotypes of FTLD and would be useful to the mechanistic study of FUS pathogenesis.
Astroglia and microglia are commonly reactive to neurodegeneration in patients and in the animal models of neurodegenerative diseases (49–53). In response to neurodegeneration in FUS transgenic rats, astrocytes and microglia were overtly activated and their activation was closely related to neuronal death (Fig. 8). Reactive glial cells produce neurotoxic cytokines and the other unidentified mediators that induce neuroinflammation and neurotoxicity (51,54–57). Activation of PPAR-γ suppresses the secretion of cytokines from reactive glia and neuronal response to glial activation (58,59). Rosiglitazone is a potent PPAR-γ agonist (60), preserves memory in transgenic mouse models of Alzheimer's disease (41,61) and improves the cognition in patients with Alzheimer's disease (62,63). Recent study showed that rosiglitazone preserves the neurites and dendritic spines of cortical neurons in culture (43). As the mouse models tested of PPAR-γ agonists show cognitive deficits without significant loss of neurons (41,61), it is not known whether rosiglitazone can rescue both neurons and neurites in an animal model of cortical dementia. In our FUS transgenic rats, treatment with rosiglitazone preserved the spatial memory by rescuing neurites and dendritic spines rather than neurons (Fig. 9). Activation of PPAR-γ receptor exerts neuroprotective effects selectively on neuronal terminals rather than on the cell body. Rosiglitazone has a long history in clinical use and is worth testing of its therapeutic effects in FTLD patients.
Animal use followed NIH guidelines and was approved by the Institutional Animal Care and Use Committee (IACUC) at Thomas Jefferson University.
Creation of FUS and LacZ transgenic rats was reported (25,36). Transgenic rats expressing tTA were created using a similar procedure described previously (36). The promoter of the mouse calcium/calmodulin-dependent protein kinase type II subunit alpha (Camk2a) gene was isolated from a BAC clone (CHORI: RP24-243J21) and was used to drive the tTA transgene. Linearized transgenic DNA was purified from agar gel and was injected into the pronuclei of fertilized eggs of Sprague–Dawley rats to produce transgenic founder rats (36). Transgenes were maintained on the SD genomic background and the Camk2a-tTA transgene was identified by PCR analysis of the rat tail DNA with the following primers: 5′-TGAAAGGCAGGCAGGTGTTG-3′ (forward) and 5′-TCCAAGGCAGAGTTGATGAC-3′ (reverse).
Rat's spatial learning and memory were examined with a Barnes maze (Med Associates) as described previously (25). Mutant FUS transgenic rats were trained to locate the escaping hole on a Barnes maze three times per day, for 3 consecutive days. After the training sections, the rats were examined of spatial memory once a week for determining the progression of disease phenotypes: disease onset was defined as unrecoverable increase in the time of locating the escaping hole on a Barnes maze, and disease end-stage was defined as the inability to locate the escaping hole on a Barnes maze within 60 s.
Rosiglitazone was dissolved in sterilized water and was administered to rats by gavage. From the disease onset, rats were given rosiglitazone at 10 mg/kg body weight, once per day until they reached the disease end-stages.
As described previously (25,36), rats were deeply anesthetized and were transcardially perfused with 4% paraformaldehyde dissolved in 1× PBS buffer. After perfusion, rat tissues were dissected and were dehydrated as described previously (25,36). Tissue sections of 12 μm were immunostained with the following primary antibodies described previously (25,36): rabbit polyclonal antibody to human FUS (made in-house) (25), chicken antibody to ubiquitin (Sigma), mouse monoclonal antibodies against Iba-1 (Wako Chemical) or GFAP (Millipore), mouse monoclonal anti-APC (Calbiochem), rabbit polyclonal antibody to TDP-43 (Proteintech), mouse monoclonal antibody to GM130 (BD Bioscience), rabbit polyclonal antibody to GLG1 (Abjent), rabbit polyclonal antibody to Cox-IV (Cell Signaling) and mouse monoclonal antibody to NeuN (Millipore). For histochemistry, immunostained sections were visualized with an ABC kit in combination with diaminobenzidine (Vector) and counterstained with hematoxylin to display nuclei. For immunofluorescent staining, tissue sections were incubated with specific primary antibodies and then with secondary antibodies labeled with fluorescent dyes (Jackson ImmunoResearch). The primary antibodies were incubated overnight at 4°C and the secondary antibodies were incubated for 2 h at room temperature. In determining the colocalization of two proteins, fluorescent staining was documented by confocal microscopy (Imaging Facility of Kimmel Cancer Center at Jefferson) and the single-layer image was scanned with a Zeiss LSM510 META confocal system. To reveal the integrity of Golgi trans and cis complexes, Z-stacks of confocal images (at 1 µm of intervals) were projected to reconstruct Golgi structure.
The total number of neurons was estimated with stereological cell counting for the following brain regions in one hemisphere: the frontal cortex (from the apical forebrain to the first occurrence of corpus callosum), the parietal cortex (from the first occurrence of corpus callosum to the first occurrence of hippocampus) and dentate gyrus. Rat forebrains were cut into serial coronal sections (20 μm) and every 12th section (a total of 15 to 18 sections) was counted for neurons in the defined brain regions. Tissue sections were stained with Cresyl violet and mounted in sequential order (rostral–caudal). The number of targeted neurons was estimated using a procedure described previously (25,36).
Dendrites and dendritic spines were visualized by Golgi impregnation method following the manufacturer's instruction (FD NeuroTechnologies). Five neurons in each selected brain region of individual rats were examined for dendrite branches and dendritic spine density (64). Dendritic spine density was calculated by dividing the number of spines with the length of dendrites. Both apical and basal dendrites were examined for spine density.
The number of neurons in the defined region was statistically compared between groups of transgenic rats, and comparison among experimental groups was performed by one-way ANOVA followed by Tukey's post hoc test. The null hypothesis was rejected at the level of 0.05.
We thank Ms Xiao-Tao Wei for technical assistance.
Conflict of Interest statement. None declared.
This work was supported by the National Institutes of Health (NS072696 and NS072113 to X.-G.X and NS073829 to H.Z.).