Low-level overexpression of wild type TDP-43 causes late-onset, progressive neurodegeneration and paralysis in mice

Modestly increased expression of transactive response DNA binding protein (TDP-43) gene have been reported in amyotrophic lateral sclerosis (ALS), frontotemporal dementia (FTD), and other neuromuscular diseases. However, whether this modest elevation triggers neurodegeneration is not known. Although high levels of TDP-43 overexpression have been modeled in mice and shown to cause early death, models with low-level overexpression that mimic the human condition have not been established. In this study, transgenic mice overexpressing wild type TDP-43 at less than 60% above the endogenous CNS levels were constructed, and their phenotypes analyzed by a variety of techniques, including biochemical, molecular, histological, behavioral techniques and electromyography. The TDP-43 transgene was expressed in neurons, astrocytes, and oligodendrocytes in the cortex and predominantly in astrocytes and oligodendrocytes in the spinal cord. The mice developed a reproducible progressive weakness ending in paralysis in mid-life. Detailed analysis showed ~30% loss of large pyramidal neurons in the layer V motor cortex; in the spinal cord, severe demyelination was accompanied by oligodendrocyte injury, protein aggregation, astrogliosis and microgliosis, and elevation of neuroinflammation. Surprisingly, there was no loss of lower motor neurons in the lumbar spinal cord despite the complete paralysis of the hindlimbs. However, denervation was detected at the neuromuscular junction. These results demonstrate that low-level TDP-43 overexpression can cause diverse aspects of ALS, including late-onset and progressive motor dysfunction, neuroinflammation, and neurodegeneration. Our findings suggest that persistent modest elevations in TDP-43 expression can lead to ALS and other neurological disorders involving TDP-43 proteinopathy. Because of the predictable and progressive clinical paralytic phenotype, this transgenic mouse model will be useful in preclinical trial of therapeutics targeting neurological disorders associated with elevated levels of TDP-43.

Introduction ALS is a neurodegenerative disease that causes relentless, progressive loss of upper and lower motor neurons leading to paralysis and death. Approximately 90 percent of ALS are sporadic, whereas~10 percent are familial. The mechanism of motor neuron degeneration in ALS is not yet fully understood. Mutations in many genes can cause or enhance the risk of ALS. These genes function in multiple cellular mechanisms, including protein homeostasis, RNA processing, inflammation regulation, cytoskeleton dynamics, and intracellular trafficking, thus suggesting that abnormalities in diverse pathways may lead to motor neuron neurodegeneration in ALS [1,2].
What drives wild type TDP-43 to proteinopathy is not understood. Several studies have suggested that modestly elevated levels of TDP-43 expression are associated with TDP-43 proteinopathy in ALS and other diseases such as FTD and inclusion body myopathy, Paget's disease of the bone and frontotemporal dementia (IBMPFD) [27][28][29][30]. ALS-associated TDP-43 mutations stabilize the TDP-43 protein or mRNA [19,[31][32][33]. TDP-43 levels are increased in cells with TDP-43 mutations from ALS patients and in mutant TDP-43 knock-in transgenic animals [15,34,35]. Remarkably, the mutant stability is inversely correlated with the disease onset age in ALS patients; the more stable the mutant, the earlier the disease onset [19]. These studies raise the question of whether a modestly elevated level of TDP-43 suffices to trigger progressive motor neuron degeneration and ALS.
To answer this question, a mammalian model with modestly elevated levels of TDP-43 should be informative, as such a model will closely mimic the modestly elevated levels in human ALS and create an opportunity to observe chronic TDP-43 toxicity. Previous mammalian models with TDP-43 overexpression have shown a wide range of phenotypes, including both motor and non-motor symptoms [36]. In general, the high TDP-43-expressing lines die rapidly after birth. Some of these mice showed TDP-43 intracellular inclusions and mild motor neuron loss [36]. The low TDP-43 expression lines develop subtle or no motor phenotypes with only mild cellular and molecular changes during the animal's lifespan [35][36][37][38][39][40][41][42]. These models notwithstanding, whether a modestly elevated level of TDP-43 within the range of the increases observed in human ALS can trigger neurodegeneration and ALS phenotypes remain uncertain.
In this study, we have constructed two TDP-43 transgenic lines with modest levels of TDP-43 overexpression. Both lines developed motor weakness after~500 days. The weakness eventually progressed to paralysis in a few months. Conversion of one line into homozygotes accelerated the disease. The homozygous mice developed paralysis during a period of 300 to 430 days. Detailed analysis showed~30% motor neuron loss in layer V cortex, oligodendrocyte degeneration, demyelination, and neuroinflammation, but no motor neuron loss, in the spinal cord. However, denervation at the neuromuscular junction was detected, indicating the presence of distal motor axon degeneration. Thus, this mouse model has replicated several key features of ALS, including progressive motor dysfunction, paralysis, loss of upper motor neurons, oligodendrocyte degeneration, loss of myelin and neuroinflammation. These results demonstrate that a modest elevation of TDP-43 can trigger late-onset neurodegeneration and motor dysfunction and thus may play a causative role in human ALS and other neuromuscular conditions involving TDP-43 proteinopathy.

Transgenic mice
A cDNA encoding mouse wild-type TDP-43 and EGFP linked by the internal ribosome entry site (IRES) was generated by standard PCR using the full-length mouse TDP-43 cDNA and EGFP as templates. The resulting cDNA (mTDP43-IRES-GFP) was cloned into the XhoI site of the MoPrp.Xho plasmid (ATCC#JHU-2) [43]. After the sequence verification and tests in cultured cells, the prp-mTDP43-IRES-EGFP construct was linearized by Not1 and injected into the pronuclei derived from FVB/NJ mice. The founder mice were screened by PCR using the primers complementary to the GFP and vector DNA sequence: forward TGCTGCTGCC CGACAACCA and reverse ATAACCCCTCCCCCAGCCTAGA. The positive founders were bred with wild type FVB/NJ mice, and the offspring were sacrificed and characterized for their expression of GFP in the CNS. Their brain and spinal cord were examined under a fluorescence microscope. The lines were terminated if 3 to 5 animals from the line showed no detectable GFP fluorescence. The lines surviving this screen were further analyzed for their GFP expression levels by immunoblot. Two lines, 19 and 42, were selected for a further screen for motor phenotypes. Both lines developed incompletely penetrant, late-onset motor dysfunction and paralysis. Line 19 was successfully converted to a homozygous line and further analyzed in detail. This line has been deposited at The Jackson Laboratory and will be available as Stock No. 031609.

Behavioral analysis
Mice were housed in the University of Massachusetts Medical School animal facility managed by the Department of Animal Medicine. This facility is a specific pathogen-free (SPF) facility. Each cage housed 1-5 animals ad libitum. The rooms were maintained at 20-22˚C and with a 12-12 light-dark cycle. All the behavioral experiments were approved by IACUC and conducted according to University of Massachusetts Medical School policies and procedures regulating the use of animals in research and the provisions of the PHS/NIH Guide for the Care and Use of Laboratory Animals.
Home cage observation. Mice were observed daily on weekdays for their general health and motor behavior, and their body weights monitored biweekly. The disease stages were assigned as follows: pre-symptomatic (pre-sym), slightly weak (swk), weak (wk), and paralysis (par). At the pre-sym stage (usually <10 months), the motor behavior was indistinguishable from the nTg mice. At the swk stage (at~10 months), the mice showed slight foot-dragging in their gait. At the wk stage (~11-14 months), the foot-dragging became readily observable, and the movement became noticeably slowed. At the paralysis stage, two or more limbs became paralyzed, and the mouse was incapable of locomotion. The paralysis stage was the endpoint of the experiment, and the mouse was sacrificed for tissue harvesting.
HomeCageScan. To monitor the mouse behavior in their home cage continuously, we used the HomeCageScan system as previously described [44,45]. Briefly, mice were housed individually in polycarbonate cages with minimal bedding (about 200 ml). A digital video camera was mounted on one side of the wall. Each mouse was recorded for 24h per week, with 12h daylight and 12h dim red light, and then returned to its cage with its littermates. Video data were analyzed by HomeCageScan software (Clever Systems, Reston, VA, USA) to quantify travel. Travel measures the overall motor/muscle functions by recording the distance traveled in meters.
Accelerating rotarod. Transgenic animals and age-matched controls were tested for time on accelerating rotarod from 12 to 72 RPM over three trials with a maximum time of 300s per trial at different age points. The longest time of three trials on the rotarod was recorded in seconds once the mice fell from the bar.
Grip strength. Transgenic animals and age-matched controls were tested for loss of fourlimb grip strength using a grip strength meter at different time points. Mice were allowed to grip on a horizontal metal wire grid with four limbs. They were gently pulled back by their tails with steady force until they release their grip from the grid. Peak tension was recorded from five consecutive trials.

Sedimentation assay
Mouse lumbar spinal cords were homogenized using a handheld polytron for 20 sec in lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 0.5% deoxycholic acid, 1% Triton X-100, 20 mM NaF, 1 mM Na 3 VO 4 , 5 mM EDTA) with protease inhibitor (1:100 dilution, P8340, Sigma, St Louis, MO, USA) and phosphatase inhibitor cocktails (Thermo Fisher). The homogenates (100 μL/sample) were centrifuged at 12,000g at 4˚C for 5 min. The supernatants were moved to new tubes and measured for protein concentration as described above. The pellets were rinsed 3 times with the lysis buffer and resuspended in 20 μL 1X Laemmli buffer. Ten micrograms of protein from the supernatant were mixed with 2X Laemmli buffer. The supernatant sample and an equivalent volume of pellet sample were heated at 95˚C for 5 min, cleared by centrifugation, and then resolved by SDS-PAGE. The gel was then immunoblotted, as described above.

RT-PCR and qRT-PCR
For total RNA extraction, frozen tissues or sorted cells were homogenized in cold TRIzol reagent (Invitrogen) following the manufacturer's protocol. RNA was then reverse transcribed to cDNA using qScript cDNA SuperMix (Quanta BioSciences). For testing candidate splicing targets, RT-PCR amplification using between 33 and 37 cycles were performed from at least three nTg mice and three Tg mice. Products were separated on 2% agarose gels and visualized by staining with ethidium bromide and photographed. For qRT-PCR measurements of candidate gene targets, real-time PCR was performed on the cDNA using the primers for the targets. The PCR cycles were carried out in a Bio-Rad Real-Time PCR system (C1000 Thermal Cycler, Biorad), and the PCR product was detected using Sybr Green. The levels of target genes were standardized to the housekeeping gene GAPDH in individual animals and then further normalized to the mean ΔCT of the wild type mice.
For quantification of TDP-43 signal intensity in the nucleus and cytoplasm, sections were double-stained for TDP-43 and cellular markers. After staining, the cells were visualized and photographed using confocal microscopy. The cells in the ventral horn of the spinal cord were measured for their fluorescence intensity using the Nikon NIS Elements software. For each cell, the average fluorescence intensity was calculated. Cells on at least 5 different sections from each of the three or more mice per genotype were measured.
For immunohistochemistry, sections were washed 3 times in PBS containing 0.25% Tween 20 and then stained following the manufacturer's instructions for Vectastain ABC kit, Elite PK-6100 standard ImmPact tm DAB peroxidase Substrate kit SK-4105 (Vector Lab). The sections were then mounted on slides and dried overnight at 55˚C. After soaking in Xylenen 2 times for 2 minutes each, the slides were sealed with Permount (Vector Lab).

Detection of demyelination
To detection of demyelination, mice were fixed by transcardial perfusion using 4% paraformaldehyde and 2.5% glutaraldehyde in 0.1M sodium phosphate (pH 7.6). Tissues were further fixed by soaking in the same fixative at 4˚C for 24 hours. Luxol fast blue staining was performed on 10-μm spinal cord or brain paraffin-embedded sections for demyelination.

Visualization and quantification of neurodegeneration
Visualization and quantification of the cortical neurons were carried out as described previously {Aliaga, 2013 #1266}. Briefly, whole mouse brains were placed in 30% sucrose solution for 2 days, frozen, and sectioned sagittally at 50 μm thickness. Layer V pyramidal neurons were counted using every ninth section, with a total of nine sections per half brain. Sections were mounted onto gelatin coated slides and stained with CTIP2 or Cresyl violet. Stereological counting was performed using Stereo Investigator software (MBF Bioscience, Williston VT). Counting was performed within the motor area of cortical layer V. Only pyramidal neurons with a soma greater than 15 μm in diameter were included. A single experimenter who was blinded to the genotype performed all counts.
For visualization of ventral root axons, mice were fixed by transcardial perfusion using 4% paraformaldehyde and 2.5% glutaraldehyde in 0.1M cacodylate (pH 7.6). Tissues were further fixed by soaking in the same fixative at 4˚C for 24 hours. L4 and L5 roots attached to dorsal root ganglia were dissected and postfixed with 2% osmium tetroxide in 0.1 M phosphate (pH 7.6), dehydrated in a graded ethanol series, and embedded in Epon-Araldite resin. Onemicron sections were stained with toluidine blue and examined and photographed by light microscopy.
For quantification of ventral horn motor neurons, lumbar enlargement of the fixed spinal cords was sectioned on a cryostat at 20μm thickness. Every other section was collected until a total of ten sections were collected from each spinal cord. The sections were stained with goat ChAT antibody at 4˚C overnight. A secondary donkey anti-goat biotinylated antibody and a Vectastain ABC and DAB peroxidase Substrate kit (Vector Lab) were used to reveal motor neurons. Images of the spinal cord sections were taken using a Nikon microscope, and motor neuron numbers in the ventral horn region were counted manually from each section.
For muscle histology, isopentane in a container was pre-chilled with liquid nitrogen until the isopentane started to solidify at the bottom of the container. A fresh specimen isolated from gastrocnemius muscle was placed on a cork disc with a drop of OCT, which kept the muscle in the desired orientation. The specimen was frozen by immersion into the isopentane for about 5 seconds and then stored at -20˚C. The frozen tissue was sectioned using a cryostat and stained with Hematoxylin & Eosin (H&E).

Electromyography
Mice were anesthetized by inhalation of isoflurane. Animals were placed immediately on a heating pad to maintain their core temperature at 37˚C. Measurements were performed using a Cardinal Synergy electromyography (EMG) machine. A ground self-adhesive gelled surface electrode was placed over the tail. Potentials were recorded from several sites of the muscles of all four limbs with a concentric needle electrode (30G) using a gain of 50 μV/division and a bandpass filter with low and high cut-off frequency settings of 20 and 10,000 Hz, respectively. The entire recording process took 15-20 minutes per mouse, after which the mice were euthanized by isoflurane overdose or used for tissue collection.

Visualization and quantification of neuromuscular junctions
Animals are euthanized via an overdose of isoflurane, then transcardially perfused with PBS for 2 minutes, followed by 4% paraformaldehyde for 5 minutes. Gastrocnemius muscles were dissected out and placed in 1.5% paraformaldehyde for 24 hours at 4˚C. The muscles were then washed with PBS for 30min at 4˚C and placed in 25% sucrose overnight at 4˚C. Muscles were embedded in OCT medium, frozen rapidly, and stored at -80˚C. Sections were cut at 35nm thickness using a Leica Cryostat, placed on Superfrost Plus slides, and stored at -80˚C. Slides were allowed to defrost for 30 minutes before use. Slides were washed 3 times with PBS for 5 minutes and then 3 times with 4% Triton X-100 for 5 minutes. Sections were blocked using 10% donkey serum in PBS for 3.5 hours at room temperature. A primary antibody solution of Rabbit anti-synaptophysin (ThermoFisher, 1:1000) and rabbit anti-Neuronal class III Beta-Tubulin (Biolegend, 1:1000) diluted in blocking solution was applied for 24 hours at 4˚C. Slides were washed again with PBS. A secondary antibody solution of Alexa-488nm-labelled Donkey anti-Rabbit (ThermoFisher, 1:500) and Alexa-555nm α-Bungarotoxin (ThermoFisher, 1:500) diluted in PBS was applied overnight at 4˚C in the dark. Slides were imaged on a Nikon microscope. Neuromuscular junctions were then counted and quantified as innervated or denervated Based on nerve occupancy of the endplates. Those with>50% occupancy was counted as innervated and those with <50% were counted as denervated.

Prion promoter drives TDP-43 expression primarily in the CNS and causes age-dependent, progressive weakness, and paralysis
To determine the effects of an elevated level of TDP-43, we constructed a transgene in which the mouse prion promoter (Prp) [43] drives the expression of wild type murine TDP-43 cDNA. We employed the authentic TDP-43 sequence to avoid any possible effects of a tag. GFP was co-expressed from an internal ribosomal entry site (IRES) to monitor the transgene expression (S1A Fig). The founder lines were screened for their expression of GFP in the CNS. Two lines (lines 19 and 42) were obtained. Both expressed similar levels of GFP and TDP-43 (S1B Fig). Western blots of tissues from various organs showed that the transgene was predominantly expressed in the CNS (S1C and S1D Fig). Despite the readily detectable GFP expression, changes in TDP-43 levels were hardly noticeable in these lines compared with non-transgenic (nTg) mice (S1B Fig), possibly as a result of TDP-43 autoregulation, as previously reported [46]. A small cohort from each of the two lines was monitored for up to 750 days. Five of the ten mice from line 19 and four of the five mice from line 42 developed motor deficits ending in full paralysis (S1E and S1F Fig). The symptomatic mice also showed progressive weight loss towards the paralysis stage (S1G Fig), similar to other animal models for ALS [36].
To facilitate the analysis of these mice, we sought to elevate the gene dosage by making homozygous lines. We were unable to generate homozygous mice from line 42 but succeeded from line 19. As anticipated, the homozygous mice developed motor symptoms earlier than the original hemizygous line and became paralyzed before 400 days (Fig 1A and 1B; S1 Video). This phenotype was completely penetrant. Males developed paralysis at younger ages (~356 ±19 days) than the females (~385±26 days) (Fig 1B). The motor symptoms progressed over approximately three to five months before ending in paralysis based on several different

PLOS ONE
Low-level TDP-43 overexpression causes late-onset neurodegeneration in mice behavioral measures, including body weight, travel velocity, grip strength, and rotarod performance (Fig 1C-1F, S1 Video). Given the accelerated and completely penetrant phenotype, we focused on these mice for further analysis and referred to the TDP-43 line 19 homozygous transgenic mice as TDP-43 transgenic (Tg) mice in the following text.
As expected, the increased gene dosage in the homozygous mice increased the expression of the transgene compared with the hemizygous mice, as shown by the Western blots for TDP-43 (Fig 2A). The hemizygous mice expressed TDP-43 at~10% above the nTg level in the spinal cord and~20% above in the frontal cortex ( Fig 2B). By comparison, the homozygous mice expressed~30% above the nTg level in the spinal cord and~40% above in the frontal cortex ( Fig 2B). Higher levels of mRNA were observed. The hemizygous mice expressed TDP-43 at 3 and~4 fold of the nTg levels in the spinal cord and the frontal cortex, respectively. By comparison, the homozygous mice expressed~5 and~8 fold of the nTg levels in the spinal cord and the frontal cortex, respectively ( Fig 2C).
TDP-43 is a gene-splicing modulator. To determine whether the modest elevation of TDP-43 level in the Tg mice impacted TDP-43 function, we measured the splicing patterns of some TDP-43-regulated mRNAs [47,48]. For comparison, we also examined splicing patterns of the same target mRNAs in the TDP-43 knockdown (KD) mice [45]. Of the fifteen alternatively spliced genes that we measured, seven changed in the opposite direction in the TDP-43 Tg mice compared to the TDP-43 KD mice, with exon exclusion increased the Tg mice but decreased in the KD mice ( Fig 2D, red genes); four genes changed in the same direction, with exon exclusion enhanced in both the Tg and KD mice (Fig 2D, blue genes); the final four  [45]. The PCR primers and the exon numbers are listed in S1 Table. (E) A sagittal section from a Tg mouse brain and a cross-section from either Tg or nTg mouse spinal cord were stained for GFP. Notice that the broad GFP staining throughout the Tg brain and spinal cord but not in the nTg.

PLOS ONE
Low-level TDP-43 overexpression causes late-onset neurodegeneration in mice genes did not change significantly in the Tg mice, but showed increased exon exclusion in the KD mice ( Fig 2D, black genes). Overall, the splicing modulation in our Tg mice was biased toward exon exclusion, consistent with the gain of TDP-43 function as reported in the literature [39,47].
To determine where the transgene is expressed in the CNS, we stained brain and spinal cord sections for GFP. We observed a broad expression pattern in all regions of the CNS ( Fig  2E). The elevated level of expression was also sustained throughout the lifespan of the mice (Fig 2F). To determine the cell types that express the transgene, we conducted double immunofluorescence staining. In spinal cord, we observed high levels of GFP in oligodendrocytes and astrocytes, and low levels in microglia and neurons, including motor neurons (Fig 3). In brain, we observed relatively high levels of GFP in oligodendrocytes, astrocytes and neurons, and low levels in microglia (Fig 3).
The increased TDP-43 expression may perturb the nuclear and cytoplasmic distribution of TDP-43. Because such perturbation may impact neurodegeneration [49], we quantified the nuclear and cytoplasmic TDP-43 staining intensities. TDP-43 was increased in both the nucleus and cytoplasm in all cell types that we measured (Fig 4B). The overall increase was larger in glial cells than in neurons. In the nucleus, TDP-43 was up by~150-220% in glial cells, compared with up by~30-50% in neurons (Fig 4B). Likewise, in the cytoplasm, TDP-43 was increased by~90-145% in glial cells, compared with an increase of~80-90% in neurons. Interestingly, the cytoplasmic increase was larger than the nuclear increase in neurons but was In the spinal cord, the expression was relatively strong in oligodendrocytes (ODC) and astrocytes but weak in microglia and neurons, including motor neurons. In the frontal cortex (top four rows under the brain label), the transgenes are expressed strongly in ODC, astrocytes, and neurons (arrows) but weakly in microglia. Sections were double immunostained for GFP and various cellular marks as indicated. The transgenes were weakly expressed in Purkinje cells compared with the surrounding neuropils in the cerebellum. All panels are in the same magnification. All animals used are pre-symptomatic and less than 100 days old.
https://doi.org/10.1371/journal.pone.0255710.g003 smaller in glial cells (Fig 4B), resulting in a decreased cytoplasmic-to-nuclear ratio of TDP-43 in glial cells but an increased ratio in neurons compared with nTg mice (Table 1). Despite the increase, we did not observe TDP-43 aggregates in any disease stages, even though sedimentation experiments showed a~25% increase in the detergent-insoluble TDP-43 and ubiquitinated proteins in the Tg mice at the end disease stage (S2 Fig), suggesting a modest increase in protein aggregation. To determine the pathological basis of the clinical phenotype, we examined the spinal cord throughout the various disease progression stages in the Tg mice. The most conspicuous pathological feature was white matter degeneration. Luxol fast blue staining revealed normal myelination at age five months, but the staining became slightly pale at eight months. At ten months, the staining became paler (Fig 5A). At the end-stage (12 months), the staining was completely lost in the ventral spinal cord (Figs 5A, 5D and S3A), indicating loss of myelin,  The myelin loss was accompanied by increasing astrogliosis (Fig 5B and 5E) and microgliosis (Fig 5C and 5F). Myelin loss could be associated with injury in oligodendrocytes as a result of TDP-43 overexpression. Therefore, we doubly stained the spinal cord sections for oligodendrocyte marker APC and cell death marker activated caspase 3. We found an increasing association of activated caspase 3 and oligodendrocytes during the progression of the clinical phenotypes ( Fig  6A), and a reduction in oligodendrocyte-associated proteins, including CNPase, MBP, and MCT1 (Fig 6B). Staining of CNPase in the white matter was also broadly lower than the nTg mice ( Fig 6C).

TDP-43 mice show no motor neuron loss but a mild muscle denervation
The progressive clinical motor phenotype suggests the presence of motor neuron injury. By Nissl staining and choline acetyltransferase (ChAT) staining, we did not observe gross morphological deviations from the controls (Figs 5D and 7A). Quantification of ChAT-positive neurons in the ventral horn also did not reveal a significant difference between the Tg mice and the nTg mice ( Fig 7B). To verify this result, we examined and quantified ventral root axon numbers. We did not observe a significant change in the number of ventral root axons in the Tg mice from the nTg mice (Fig 7C-7E). Although we could not detect any motor neuron number changes, the Tg mice might have been at an early stage of motor neuron degeneration, which had not yet been reflected in the motor neuron numbers. To examine this possibility, we stained for activated caspase 3. We did not detect an association between the activated caspase 3 with motor neurons, although caspase staining of surrounding cells, possibly glial cells, were increased ( Fig 7F).
Despite the preservation of motor neuron cell bodies and their proximal axons, the distal axons and neuromuscular junction could be affected. To test this possibility, we examined muscle morphology and physiology. We did not observe muscle fiber atrophy (Fig 8A and  8B). By needle electromyography (EMG), we observed that~half (3 of 5 homozygotes and 1 of 3 hemizygotes) of the paralyzed mice had a completely normal EMG pattern that was indistinguishable from the control nTg mice (Fig 8C1). The other half of the paralyzed mice showed various degrees of positive sharp waves (PSWs) in different muscles ranging from normal ( Fig  8C1) to single and to multiple PSWs (Fig 8C2 and 8C3). Quantification of endplate nerve occupancy in the gastrocnemius muscle unveiled an average of~30% denervation in the paralyzed homozygous Tg animals compared with~10% denervation in the age-comparable control nTg mice (Fig 8D).

TDP-43 mice show elevated neuroinflammation in their spinal cord
Demyelination and gliosis are commonly associated with neuroinflammation [50]. Therefore, we examined the state of neuroinflammation in the TDP-43 transgenic mice. We found that the mRNAs of neuroinflammation-associated genes such as TNF-α and NF-κB p65 subunit were increased along with the disease progression ( Fig 9A). Furthermore, the p65 protein and its phosphorylation were also increased in parallel with clinical symptom development in the Tg mice compared with the nTg mice (Fig 9B). By immunostaining, the increased p65 was associated with oligodendrocytes ( Fig 9C, arrows). These changes suggested the presence of neuroinflammation in the TDP-43 transgenic mice.
To further confirm this observation, we examined additional molecular markers for neuroinflammation. The expression of NADPH oxidase (NOX), which is involved in inflammation and oxidative stress, was dramatically increased. The three mRNAs of its subunits, gp91 phox , p22 phox , and p67 phox , were increased by~600,~80, and~9 fold, respectively ( Fig  10A-10C). Two genes involved in the production of prostaglandins, the inducible cyclooxygenase 2 (COX-2) and hemopoietic prostaglandin D synthase (HPDGS), were increased ( Fig  10D and 10E). GLT-1, an astrocytic glutamate transporter, was decreased at the end-stage ( Fig  10F). LCN2 and IL-6, two proinflammatory and neurotoxic cytokines, were also highly upregulated (Fig 10G-10I). Although not all the markers that we measured showed an increase (S4

TDP-43 mice show a mild white matter degeneration in CNS areas beyond the lumbar spinal cord
Because the TDP-43 transgenic mice expressed an elevated level of TDP-43 in the forebrain (Figs 2A, 2E and 3), it is possible that the upper motor neuron degeneration could contribute

PLOS ONE
Low-level TDP-43 overexpression causes late-onset neurodegeneration in mice

PLOS ONE
Low-level TDP-43 overexpression causes late-onset neurodegeneration in mice

PLOS ONE
Low-level TDP-43 overexpression causes late-onset neurodegeneration in mice to the paralysis phenotype. To examine this possibility, we investigated whether there was neurodegeneration in the forebrain. The brain weight in the TDP-43 mice did not differ from those of the nTg mice (S5 Fig), suggesting no widespread neuronal loss in the brain. Staining for myelin at different CNS levels showed that demyelination was most severe in the lumbar spinal cord and became progressively less severe towards the rostral CNS (Fig 11A). The myelin staining was almost absent in the lumbar spinal cord, and paler compared to the nTg mice in the cervical spinal cord and lower brainstem (Fig 11A). At the level of pons and above, no apparent difference in the staining was noticeable between the nTg and Tg mice (Fig 11A). At the cervical spinal cord level, the myelin staining intensity throughout the different ages showed a late-onset, progressive demyelination (Fig 11B), a pattern similar to the lumbar

PLOS ONE
Low-level TDP-43 overexpression causes late-onset neurodegeneration in mice region ( Fig 5A) though less severe. These results indicate that the overt demyelination in the Tg mice is mostly confined to below the brainstem.
We also surveyed gliosis in various regions of the brain in the paralyzed mice. We found elevated levels of astrogliosis and microgliosis in all regions, including the cerebral cortex, hippocampus, corpus callosum, internal capsule, cerebellum, and brainstem (Fig 12), suggesting that elevated expression of TDP-43 provokes an inflammatory response by glial cells.

TDP-43 mice develop neuronal loss in the motor cortex
Because TDP-43 was expressed in both neurons and glia in the cortex (Fig 3), we investigated whether there was a neuronal loss in the motor cortex. Ctip2 staining revealed dramatically the loss of Ctip2 expression might have caused the lowered neuronal counts. Therefore, we applied Nissl staining (Fig 13A and 13B) and repeated the quantification of neurons in layer V cortex. This approach revealed~30% neuronal loss (Fig 13C). These results indicate that there is motor neuron loss in layer V of the motor cortex.

Discussion
This study has established a line of TDP-43 transgenic mice that express exogenous TDP-43 protein at modestly elevated levels. Respectively for hemizygous and homozygous mice, the levels are~10% and~30% above the nTg level in the spinal cord and~20 and~40% increase frontal cortex (Fig 2). These levels are comparable with the increase reported in human ALS and FTD [15,28,30,34]. These TDP-43 mice show several key features of ALS, including lateonset progressive motor dysfunction ending in paralysis in mid-life, oligodendrocyte injury, demyelination, neuroinflammation, distal motor axon degeneration in the periphery, and cortical motor neuron loss. Thus, this transgenic mouse model demonstrates the multiple longterm adverse consequences of low-level TDP-43 overexpression in vivo and demonstrates that such low-level elevation of TDP-43, as observed in ALS and other neurodegenerative conditions, can trigger neurodegenerative clinical phenotypes.

PLOS ONE
Low-level TDP-43 overexpression causes late-onset neurodegeneration in mice How low-level overexpression of TDP-43 causes the late-onset neurotoxicity remains unknown. Pathological studies have consistently detected various forms of TDP-43 aggregates in ALS and FTD CNS tissues [51], suggesting an association between TDP-43 aggregation and the pathogenesis of these diseases. In vitro studies have shown that overexpression of TDP-43 can induce cytoplasmic liquid-liquid phase separation (LLPS), which, upon stress, transform into permanent aggregates. These aggregates can then draw TDP-43 out of the nucleus and kill the cell [52,53]. However, in vivo models have shown divergent results. Some models show cytoplasmic TDP-43 aggregates either with or without nuclear TDP-43 depletion [37, 54,55]. Other studies suggest that neither TDP-43 aggregation nor its nuclear depletion is required for neurodegeneration [39,56]. Our results present a nuanced picture. Although we did not observe cytoplasmic aggregates and nuclear depletion in cells, we detected an increase in the detergent-insoluble TDP-43 and high molecular-weight ubiquitinated protein species (S2 Fig), suggesting there is TDP-43 protein aggregation, albeit modest, in the TDP-43 mice. These results suggest that a modest elevation of TDP-43 levels can lead to TDP-43 protein aggregation, which could contribute to neurodegeneration in the TDP-43 mice.
Previous studies suggest that TDP-43 aggregation and nuclear depletion may lead to gain of toxicity as well as loss of function that ultimately cause cellular degeneration, including both neurons and oligodendrocytes [22,23,26,[57][58][59][60][61]. Analysis of mRNA splicing in mouse models has shown that a loss of TDP-43 function increases cryptic exon inclusion, whereas a gain of TDP-43 function leads to an increase in skiptic exons [9,39]. Both the increases in the cryptic exon inclusion and the skiptic exons have been observed in human ALS samples, supporting the dual gain-and loss-of-function hypothesis [9,39]. Consistent this hypothesis, both loss-and gain-of-function mouse models, including the model described in this report, produce motor neuron degeneration and ALS-like phenotypes [36,39,45,62,63], suggesting that each type of these models represent one aspect of the mechanism.
An intriguing observation in this study is that different cell types may be responding to TDP-43 overexpression differently. In oligodendrocytes, TDP-43 overexpression elicited similar percentage of increases in nucleus and cytoplasm (Fig 4B), thereby maintaining the same cytoplasmic-to-nuclear TDP-43 ratio as in the nTg mice (Table 1). In astrocytes and microglia, the percentages of nuclear increase were approximately twice the cytoplasmic increases ( Fig 4B). As a result, the cytoplasmic-to-nuclear TDP-43 ratios were reduced, though not statistically significant (Table 1). By contrast, in motor neurons and non-motor neurons, the percentage of cytoplasmic increases was approximate twice the nuclear increases ( Fig 4B). Consequently, the ratio of cytoplasmic-to-nuclear TDP-43 was increased (Table 1). These observations suggest that neurons, particularly motor neurons, may handle the TDP-43 overexpression differently. It will be interesting to determine whether this observation can be replicated in other TDP-43 overexpression models in future experiments because of the potentially detrimental effects of an increased TDP-43 level in the cytoplasm, which may define motor neuron vulnerability [49].
However, whether this is the case for TDP-43 is less clear. Astrocyte-specific expression of mutant TDP-43 in rats induces motor neuron degeneration as the mutant expression in motor neurons [73,74]. However, mutant TDP-43-expressing astrocytes fail to show toxicity in motor neuron-astrocyte coculture or after being transplanted in rat spinal cord [75,76]. Our TDP-43 transgenic mice showed two neurodegeneration patterns, one in the spinal cord and the other in the frontal cortex. In the spinal cord, the TDP-43 transgene was expressed highly in glial cells but lowly in neurons (Figs 3 and 4). The primary pathology was oligodendrocyte injury, demyelination, gliosis, and neuroinflammation (Figs 5,6,(9)(10)(11)and S3). However, no motor neuron loss was detected (Fig 7). In the motor cortex, the TDP-43 transgene was expressed in both neurons and glial cells (Fig 3). Approximately 30% of large pyramidal neurons were lost in layer V (Figs 13 and S6). These results suggest that motor neurons can tolerate the detrimental effects invoked by modestly elevated levels of TDP-43 in their neighboring cells so long as the TDP-43 levels in themselves are maintained at relatively low levels. Even so, the tolerance of motor neurons to the injuries to their neighboring cells and neuroinflammation is probably limited, as illustrated by the presence of distal motor axon degeneration and neuromuscular denervation (Fig 8). Because we have to sacrifice the mice at their paralysis stage, we do not know whether letting the disease progress further will eventually lead to the loss of the lower motor neurons. In any case, our results are consistent with contributions to motor neuron toxicity from both cell-autonomous and non-cell-autonomous sources.
A unique feature of our TDP-43 mice, compared with other established TDP-43 mouse lines, is the development of a fully penetrant progressive disease course (Fig 1, S1 Video). The phenotype of late-onset, slowly progressing motor dysfunction to complete paralysis is reproducible for over >10 generations of homozygous breeding. This feature contrasts with other TDP-43 mouse models reported thus far. We note that there is a dose-dependent response to increased levels of TDP-43. In some reports, excessive expression of TDP-43 (e.g.,>3 fold of nTg level) causes acute toxicity and rapid demise of the animals. At the other end of the spectrum, models that express too little TDP-43 provoke no phenotype or at most mild phenotypes that require careful measurements to unveil [35][36][37][38][39][40][41][42]. Few transgenic lines developed motor dysfunction and paralysis but symptoms from the digestive system, highly variant phenotypes and lifespan (e.g.,~1 to >15 months), and low penetrance complicate the phenotypes (e. g. 5% transgenic animals) [57,[77][78][79]. Thus, our TDP-43 mice provide a useful model for in vivo study of chronic TDP-43 toxicity derived from the modest elevation of TDP-43 levels and in vivo preclinical tests of experimental drugs targeting chronic TDP-43 toxicity in the CNS.

Conclusions
We established a transgenic mouse model with a modestly elevated TDP-43 level. This model displays several ALS characteristics, including late-onset and progressive paralytic motor dysfunction ending in paralysis, neuroinflammation, and neurodegeneration. This study demonstrates that modest elevations in TDP-43 expression can trigger neurodegeneration and clinical phenotypes of ALS, suggesting that modestly elevated TDP-43 levels in humans could cause ALS and other neuromuscular disorders involving TDP-43 proteinopathy. Because of the easily observable, predictable, and progressive clinical paralytic phenotypes, this transgenic mouse model may be useful in preclinical trials of therapeutics targeting neurological disorders associated with elevated levels of TDP-43.  Table. PCR primers for data shown in Fig 2D. (JPG) S1 Video. Tg mice developed progressive motor dysfunction and paralysis. The transgenic mouse was marked with red ink on its tail. The other mouse in this video was a control nTg mouse. Notice the Tg mouse goes through stages from hyperactive to final paralysis. (MP4) S1 Raw images.