TDP-43 transgenic mice develop spastic paralysis and neuronal inclusions characteristic of ALS and frontotemporal lobar degeneration

Dose-Dependent Motor Neuron Disease in TDP43^WT Mice.

We generated several germline transgenic mouse lines overexpressing human wild-type TARDBP (TDP43^WT) under the control of a neuronal murine Thy-1 (mThy-1) promoter that drives transgene expression in virtually all neurons of the central nervous system (Fig. 1A) and becomes active approximately 1 week after birth (16). Transgene levels in total brain lysates were analyzed by semiquantitative real-time PCR (Q-RT-PCR) (Fig. 1B) and immunoblotting with anti-TDP-43 antibodies (Fig. S1). In none of the TDP43^WT lines did TDP-43 expression exceed 1× endogenous TDP-43 expression. To increase transgene dose, we generated homozygous mice from the two highest-expressing founder lines. Homozygous TAR4/4 and TAR6/6 lines showed a double transgene expression (2× and 1.2×, respectively) compared with hemizygous TAR4 and TAR6 lines with 1× and 0.6× hTDP-43 levels over endogenous protein levels (Fig. 1B and Fig. S1). Because Thy-1 expression is limited to neurons, we also analyzed hTDP-43 expression relative to murine TDP-43 by Q-RT-PCR in neurons isolated from 1-month-old TDP43^WT mice (17), and also by immunohistochemical densitometric quantification (18) (Fig. S1). On ≈5-fold-enriched neuronal preparations, TAR4 and TAR6 mice showed 1.4-fold and 0.9-fold hTDP-43 over murine TDP-43. Image densitometry of neuronal TDP-43 staining by a non-species-specific TDP-43 antibody further confirmed that TAR4 and TAR6 had 2.8- and 1.9-fold, and homozygous TAR4/4 and TAR6/6 had 5.1- and 3.8-fold hTDP-43 expression, compared with murine TDP-43 in nontransgenic (Ntg) mice (Fig. S1).

Homozygous TDP43^WT mouse lines were morphologically indistinguishable from their hemizygous or Ntg littermates at birth. At ≈14 days, TAR4/4 mice developed an abnormal hindlimb reflex characterized by retraction of hindlegs toward the trunk upon lifting them by their tail, as opposed to Ntg littermates, which showed normal extension of the legs (Fig. 1C). This reflex was demonstrated as one of the earliest symptoms of loss of motor control in several ALS mouse models (19). A footprint analysis for TAR4/4 mice showed a significant ≈2-fold decrease in the stride of hindlimbs (1.5 ± 0.26 cm vs. 3.4 ± 0.32 cm; P = 0.025) and of forelimbs (1.5 ± 0.19 cm vs. 3.3 ± 0.41 cm; P < 0.01). Moreover, footprints of TAR4/4 mice were characterized by separated forelimb and hindlimb prints, markedly wide-based stance, small stride, and frequent off-line stumbling compared with Ntg littermates, which showed a normal, narrow-based stance with steady, close-proximity forelimb and hindlimb footprints (Fig. 1 E and F). When tested for motor performance on accelerating rotarod, 18-day-old TAR4/4 mice showed a statistically significant ≈2.5-fold reduced performance compared with Ntg littermates (114.2 ± 51.10 s vs. 236.2 ± 31.47 s; P < 0.01) (Fig. 1G). At ≈22 days, fasciculations and spasms of facial muscles were also observed, which was followed by an extremely rapid disease progression, with mice becoming completely paralyzed and dying within 3–4 days (Fig. 1 H and I and Movie S1).

To verify that the hTDP-43 construct did not cause insertional mutagenesis of a mouse endogenous gene leading to the observed phenotype, we characterized the transgene integration site in the TAR4 mouse line. A thermal asymmetric interlaced PCR analysis showed that hTDP-43 had integrated at locus 6qB3 of the mouse genome (nucleotide no. 56,524,796; UCSC Genome Browser, http://www.genome.ucsc.edu/) and did not interrupt any known gene.

Subsequently, motor problems also developed in TAR6/6 and TAR4 mice (Table S1). TAR6/6 mice developed abnormal limb reflex at 2 months, showed up to 6-fold reduced motor performance at ≈4 months (P < 0.01; Fig. S2), and an average survival of 6.7 months (Fig. 1I). Similarly, TAR4 mice showed abnormal limb reflex at ≈14 months and 40% reduced motor performance at ≈15 months of age (P < 0.01; Fig. S2). Thus, correlation of hTDP-43 expression in TAR4/4, TAR6/6, and TAR4 mice with age of disease onset was best explained by a log-log regression model (correlation coefficient R ² = 0.98; P < 0.001; Fig. 1D). No clinical phenotype has been observed in the TAR6 hemizygous mice analyzed until 19 months or in any other founder lines analyzed until 7 months, all expressing hTDP-43 ≤ 0.6× the endogenous TDP-43 level on total brain lysates. Thus, TDP43^WT mice have a dose-dependent ALS-like phenotype, influencing both onset and progression of the disease.

TDP43^WT Mice Display Neuropathology Reminiscent of FTLD/ALS.

One of the chief pathological findings in ALS and FTLD-TDP brains are neuronal nuclear and cytoplasmic inclusions composed of ubiquitinated and phosphorylated TDP-43 (4, 5, 20). We therefore first examined brains of TAR4/4, TAR6/6, and TAR4 mice by ubiquitin immunohistochemistry. Despite mThy-1 promoter driving transgene expression in virtually all neurons of the brain, dense, large NCIs and NIIs characteristic of FTLD and ALS patients were present solely in cortical layer V of the anterior cortex, including primary motor cortex and somatosensory areas of the hind- and forelimbs, and also to some extent in hippocampal/subicular neurons (Fig. 2 A–C and Fig. S3). A dose-dependent, diffuse ubiquitin immunostaining was also present in brainstem, cranial motor nuclei, and Purkinje cells, but was absent from other brain regions, including neostriatum, substantia nigra, and thalamus. Using a panel of TDP-43-specific antibodies, including phosphospecific TDP-43 antibodies (21), we further showed that though NIIs always costained for TDP-43, a proportion of NCIs did not stain with TDP-43 antibodies (Fig. 2 B and C). The presence of NCIs, and to some extent NIIs, also coincided with an overall decreased nuclear TDP-43 reactivity, or “clearing,” as described for ALS and FTLD patients (4, 5).

Interestingly, dose-dependent astrogliosis and microgliosis were also observed in TAR4/4, TAR6/6, and TAR4 mice in the same regions where NCIs, NIIs, and diffuse cytoplasmic ubiquitin staining was observed (Fig. S4). For TAR4/4, glial fibrillary acidic protein (GFAP) selectively stained cortical layer V of the anterior cortex, including the motor and somatosensory cortex (Fig. 2D). Neuronal loss was also present in affected brain regions of all mouse lines in both superficial and deep cortical layers of the anterior cortex, especially in higher-expressing TAR4/4 and TAR6/6 mice. Quantitative neuronal loss was shown in motor cortex of TAR4/4 mice compared with Ntg littermates at 24 days (867 per mm² ± 195 vs. 1,231 ± 181; P < 0.001) and of TAR6/6 mice at 6 months (969 ± 187 per mm² vs. 1,149 ± 185; P < 0.01; Fig. 2E and Fig. S3). Vacuolar degeneration of several cranial motor nuclei was also observed, especially for the highest-expressing TAR4/4 line. In addition, TAR4/4 mice showed loss of CA3 hippocampal neurons and degeneration of Purkinje cells. These data are consistent with mutant superoxide dismutase 1 (SOD1) mice, where in low doses, the spinal cord is primarily affected, and in higher doses, other regions are also affected (22, 23). Moreover, a preferentially higher expression of hTDP-43 is shown to occur in the hippocampus (24), and degeneration of susceptible hippocampal cellular population is commonly observed in FTLD-TDP patients (25). Cerebellum is also shown to be severely affected in ALS patients (26). Thus, TDP43^WT mice recapitulate FTLD/ALS pathology, including the characteristic NCIs and NIIs. Interestingly, these pathological changes were accompanied by increased levels of activated caspase-3 and cleaved poly(ADP-ribose) polymerase (PARP) (Fig. S5), additionally suggesting that neuronal apoptosis is a hallmark of TDP-43 proteinopathies.

Because TAR4/4, TAR6/6, and TAR4 mice also showed an ALS phenotype, we next focused our attention on the spinal cord of these TDP43^WT mice (Fig. 3 and Fig. S3). Atrophy and increased number of pyknotic neurons in the ventral horn region of the lumbosacral and cervical spinal cord occurred in all three transgenic mouse lines with a severity that was highly dose-dependent. For TAR4/4 and TAR6/6 mice, although all spinal segments were affected, loss of the characteristic cervical and lumbosacral enlargements that innervate the extremities was remarkable. The number of neurons in the lumbosacral region in TAR4/4 mice was significantly lower than in Ntg littermates (608 ± 112 vs. 803 ± 125 per mm²; P < 0.001) as was also observed in TAR6/6 mice (689 ± 107 vs. 768 ± 108 per mm²; P < 0.05; Fig. 3H and Fig. S3). Further using ubiquitin immunohistochemistry, a dose-dependent accumulation of ubiquitinated proteins in spinal motor neurons was also observed for all lines, shown here for TAR4/4 mice (Fig. 3A). Employing phosphospecific TDP-43 antibodies that can recognize pathological TDP-43, we showed a punctate staining pattern of the nucleus and cytoplasm in ≈10%, ≈2%, and less than 1% of spinal neurons in TAR4/4, TAR6/6, and TAR4 mice, respectively. TAR4/4 mice also showed larger cytoplasmic aggregates as granulovacuolar bodies. More interestingly, NIIs of 2–2.5 μm in diameter were observed in ≈1% of spinal neurons in 22- to 24-day-old TAR4/4 mice, but not in 18-day-old or younger TAR4/4 or in other mouse lines (Fig. 3 B–D). The presence of NIIs was also accompanied by nuclear TDP-43 clearing. Transmission electron microscopy confirmed light microscopy data by showing presence of non-membrane-bound NIIs of 2–3 μm in diameter, composed of loosely packed granular material and 10- to 20-μm fibrils, and was also accompanied by neuritic changes (Fig. 3 E and F) as shown in ALS and FTLD patients (27, 28). Similar to GFAP staining in the motor cortex, neuronal inclusions were also marked by astrogliosis (Fig. 3G). In none of the mouse models at any disease stage did the NIIs or NCIs in brain or spinal cord stain for tau or α-synuclein, which are important exclusion criteria for TDP-43 proteinopathies.

Accumulation of Detergent-Soluble ≈25-kDa CTFs in Nuclear Fraction.

Another characteristic feature of FTLD and ALS is accumulation of ≈25-kDa CTFs in affected brain and spinal cord regions (4, 5, 29). We thus characterized TDP-43 CTFs from brain of different TDP43^WT mice and different disease stages of TAR4/4 mice by immunoblotting with a panel of TDP-43 antibodies. Also, because TDP-43 has important nuclear functions, we analyzed the cytoplasmic and nuclear fractions separately. We first showed hTDP-43 full-length (FL) protein in both nuclear and cytoplasmic fractions; however, compared with Ntg mice, TAR4/4 mice showed increased levels of TDP-43 in the cytoplasmic fraction, whereas in the nuclear fraction, FL TDP-43 levels did not differ between TAR4/4 and Ntg mice (Fig. 4A and Fig. S6). These data suggest that levels of FL TDP-43 in the nucleus are tightly regulated. Similar observations were also made on TAR6/6 mice. Reprobing these blots revealed lamin A/C detection only in the nuclear fraction and GAPDH only in the cytoplasmic compartment, confirming a clear separation of nuclear and cytoplasmic fractions (Fig. 4A Lower). The ≈35-kDa CTFs were also recovered from both cytoplasmic and nuclear fractions from TDP43^WT mice but not from Ntg littermates. When analyzing different disease stages of TAR4/4 mice, the ≈35-kDa CTF levels decreased with increasing disease severity (Fig. 4A and Fig. S6). The identity of these fractions was also confirmed by immunoblotting with a C-t antibody (30) (Fig. 4B). Finally, the ≈25-kDa fragment was also specifically observed in TDP43^WT mice but only in the nuclear fraction, and increased with increasing disease severity in TAR4/4 mice (P < 0.001; Fig. 4B). These data suggest that ≈25-kDa TDP-43 fragments play a pathogenic role in these mouse models.

Because decreased solubility of TDP-43 or its cleaved fragments is also proposed as a disease mechanism (4, 5), we further determined the solubility of TDP-43 by sequential brain extraction with buffers of increasing detergent strengths (31). Though the majority of ≈35-kDa CTFs were recovered in low-salt (LS) fraction, and to some extent in Triton X-100 (Tx) fraction, FL TDP-43 was recovered from LS, Tx, sarkosyl-soluble (SARK), and, in a very small proportion, from the detergent-insoluble (urea) fraction. However, the ≈25-kDa CTFs were recovered solely in the Tx fraction that also solubilized nuclear lamin A/C protein (Fig. 4C). Interestingly, though presence of ≈25-kDa CTFs in the detergent-insoluble urea fraction is a hallmark feature of TDP-43 proteinopathy, these CTFs are also consistently recovered in the Tx-soluble nuclear fractions from brains of FTLD/ALS patients (Fig. 4D and Fig. S7). Because ≈25-kDa CTFs lack the predicted NLS sequence (6 –9), we tested whether this hinders their transport to nucleus. Transiently transfecting two human cell lines (SH-SY5Y neuron-like cells and H4 neuroglioma cells) with a caspase-cleaved, GFP-tagged, TDP-43 CTF construct (TDP-43-Δ1–219-GFP) (32), we showed a predominant nuclear localization of this fragment (Fig. 4E). These data all indicate that ≈25-kDa CTFs can localize to the nucleus and progressively deposit in affected brain and spinal cord regions and associate with the phenotype we describe here, and is consistent with the idea that ≈25-kDa CTFs play a direct role in TDP-43-led neurodegeneration (4, 5, 29).

Construction of Transgenic Mice Expressing Wild-Type Human TDP-43.

TARDBP (NM_007375) was amplified from a human cDNA library (hTDP-43) and cloned into an mTUB expression vector containing a murine Thy-1 promoter (JSW Research). The expression vector was microinjected into pronuclear oocytes of Bl6/SJL mice by the Yale Transgenic Mouse Service Facility (New Haven, CT). Offspring were genotyped, and from 108 potential transgenic pups, 23 were identified to carry the transgene. Founders were bred with Ntg C57Bl6/J mice to establish stable transgenic lines. Additionally, homozygous hTDP-43-overexpressing mice were generated for two selected lines by crossbreeding hemizygous hTDP-43-overexpressing mice. Homozygosity was determined by semiquantitative PCR, and homozygous expression was confirmed by Q-RT-PCR and immunoblotting on brain tissue of 3-week-old mice. Although homozygous hTDP-43 mice were generated each time by crossbreeding of hemizygous mice, for sake of convenience, homozygous mice are also referred to as mouse lines. All animal experiments were approved by the University of Antwerp Ethics Committee and conducted according to the guidelines of the Federation of European Laboratory Animal Science Associations (FELASA).

Analysis of Integration Site.

Thermal asymmetric interlaced (TAIL) PCR was performed to define the hTDP-43 chromosomal integration site essentially as described previously (45). Following TAIL-PCR, products were directly sequenced, and the specific chromosomal integration sites were identified using the UCSC BLAST database for Mus musculus genomic DNA (http://www.genome.ucsc.edu/, assembly: July 2007).

Tissue Preparation.

Brains were harvested, weighed, and cut midsagittally. Left hemispheres were fixed in 2% paraformaldehyde (18-20 h) or 4% glutaraldehyde (4 h) and further prepared for histological or electron microscopy analysis, respectively. Right hemispheres were snap frozen in liquid nitrogen and stored at −80 °C for subsequent mRNA and protein analysis. Similarly, spinal cords subdivided into cervical, thoracic, lumbar, and sacrococcygeal regions were also processed as described for brain.

Immunoblotting.

Cytoplasmic and nuclear fractions from brain were prepared as described earlier (46). Briefly, frozen tissue was weighed and homogenized in buffer containing 10 mM Hepes, 10 mM NaCl, 1 mM KH₂PO₄, 5 mM NaHCO₃, 5 mM EDTA, 1 mM CaCl₂, 0.5 mM MgCl₂ (10× vol/wt). After 10 min on ice, 2.5 M sucrose (0.5× vol/wt) was added. Next, tissue was homogenized and centrifuged at 6,300 × g for 10 min. The supernatant was collected as cytoplasmic fraction. The pellet was washed three times in TSE buffer (10 mM Tris, 300 mM sucrose, 1 mM EDTA, 0.1% Nonidet P-40, 10× vol/wt), homogenized, and centrifuged at 4,000 × g for 5 min. Finally the pellet was resuspended in RIPA with 2% SDS (5× vol/wt) as the nuclear fraction.

Also, proteins were sequentially extracted with buffers of increasing detergent strength (hypotonic, Triton X-100, myelin floatation, sarkosyl, and urea buffers) as described previously (31). All buffers were substituted with protease inhibitors (Complete Mixture; Roche) and phosphatase inhibitors (PhosphoStop; Roche) before use. The following antibodies were used: rabbit anti-TDP-43 (ProteinTech Group), human specific mouse anti-TDP-43 (against residues 205–222; Abnova), rabbit anti-C-terminal TDP-43 (C-t TDP-43) (29), rabbit anti-TDP-43 phosphorylation sites 409/410 (TDP-43 p409/410) (21), rabbit anti-cleaved caspase-3, and rabbit anti-cleaved PARP (Cell Signaling Technology). Detection was performed using HRP-conjugated secondary antibodies and ECL Plus Chemiluminescent Detection System (Amersham Biosciences). Data were normalized to β-actin (Sigma), GAPDH (Meridian Life Science), or Lamin A/C (N-18; Santa Cruz Biotechnology).

Histology.

Classical histochemistry was performed according to standard protocols (47). For immunohistochemistry, the following antibodies were used in addition to the ones mentioned for immunoblotting: mouse anti-ubiquitin (Zymed), rabbit anti-GFAP (Dako), and rabbit anti-Iba1 (Wako Chemicals). Antigen retrieval was performed with boiling in citrate buffer. All immunohistochemical analyses were performed on an Axioskop 50 light microscope (Zeiss) equipped with a CCD Olympus DP camera, an Axiovert 200M fluorescent microscope (Zeiss), or an LSM 700 confocal laser scanning microscope (Zeiss) as detailed elsewhere (27, 47).

Electron Microscopy.

Brain and spinal cord were fixed in 4% buffered glutaraldehyde (4 h), transferred to glucose buffer, and finally to 2% buffered osmium tetraoxide before embedding and processing with standard protocols (27, 47). Grids were viewed and photographed on a Philips CM10 electron microscope.