Human TDP-43 and FUS selectively affect motor neuron maturation and survival in a murine cell model of ALS by non-cell-autonomous mechanisms.

TAR DNA-binding protein 43 (TDP-43) and fused in sarcoma (FUS) were recently found to cause familial and sporadic amyotrophic lateral sclerosis (ALS). The mechanisms by which mutations within these genes cause ALS are not understood. We established murine embryonic stem cell (ESC)-based cell models that stably express the human wild-type (WT) and various ALS causing mutations of TDP-43 (A315T) and FUS (R514S, R521C and P525L). We investigated their effect on pan-neuron as well as motor neuron degeneration. Finally, non-cell-autonomous mediated neurodegeneration by muscle cells was investigated. Expression of mutant hTDP-43, but not wild-type TDP-43, as well as wild-type and mutant hFUS proteins induced neuronal degeneration with partial selectivity for motor neurons. Motor neuron loss was accompanied by abnormal neurite morphology and length. In chimeric coculture experiments with control motor neurons and mutant muscle cells (as their major target cells), we detected that mutant hTDP-43 A315T as well as wild-type and hFUS P525L expression only in muscle cells is sufficient to exert degenerative effects on control motor neurons. In conclusion, our data indicate that a selective vulnerability of motor neurons expressing the pathogenic ALS-causing genes TDP-43 and FUS, is, at least in part, mediated through non-cell-autonomous mechanisms.


Introduction
Amyotrophic lateral sclerosis is the most frequent type of motor neuron disease affecting 2 -4 per 100,000 people per annum worldwide. Survival time is approximately 1 -5 years after symptom onset (1 -4). In ALS, motor neurons are predominately affected, with cells undergoing degeneration and death. In 1993, mutations in the human SOD1 gene encoding for the superoxide dismutase-1 were fi rst reported to be causative for familial ALS (FALS) (5). Cell and animal models expressing mutant h SOD1 in glial cells leading to progressive degeneration of motor neurons suggest non-cell-autonomous disease mechanisms (6 -9).
In 2006 and 2009, mutations in the genes TDP-43 and FUS were identifi ed to cause FALS as well as sporadic ALS. The molecular pathophysiology of motor neuron degeneration caused by mutant TDP-43 and FUS genes remains still largely enigmatic, but is suggested to be distinct from those of mutant SOD1 (10,11). Since both TDP-43 and FUS generated from patients (23 -25), and transgenic mice and rat models (26 -30). According to these cell models, murine mESC are uniquely suitable for the analysis of motor neuron survival, function and pathology due to their unlimited proliferation and differentiation potential and their substantially shorter generation and motor neuronal differentiation time compared to human ESC or iPSC models. Moreover, these cell models are particularly suitable to differentiate between cell-autonomous and noncell-autonomous mechanisms by the respective coculture approaches. We thus generated mESCbased in vitro cell models on the basis of mutant h TDP-43 or h FUS genes to dissect the pathophysiological events in mutant TDP-43 and FUS-mediated motor neuron degeneration.

Stable transfection
Murine Hb9::GFP ESCs were transfected with linearized plasmids by electroporation using the Amaxa ™ Mouse ES cell Nucleofector ™ Kit (Lonza). Stable integration of the plasmid into the murine genome was proven by PCR analysis and immunofl uorescence imaging ( Figure 1B, C).

Motor neuron differentiation
For long-term experiments, non-transfected mESCs (non-transgenic control, nt-control) and the stably transfected mESCs were differentiated into the neural lineage by embryoid body (EB) formation (see Figure 1A for a scheme of the experimental design).

Muscle cell differentiation
Induction of muscle differentiation for coculture analysis was achieved by EB formation of mESCs using the hanging drop technique.

Motor neuron/muscle cell cocultures (chimeric cultures)
For a defi ned coculture with motor neurons, nontransfected mESCs were differentiated into motor neurons and sorted for Hb9::GFP fl uorescence as single cell suspension and were plated on maturated muscle cells and cultivated for 24 h in medium.
For details of the experimental set-up please refer to the Supplementary material online to be found online at http://informahealthcare.com/doi/abs/10. 3109/21678421.2015.1055275.

Generation of hTDP-43 and hFUS transgenic mESC lines
Murine embryonic stem cells (Hb9::GFP mESCs) were used to generate stable cell lines expressing the human variants of wild-type and mutant TDP-43 and FUS . TDP-43 and FUS cell lines were separately generated and stable colonies were selected by G418 treatment. Stable insertion of the constructs was confi rmed by PCR analysis with specifi c primers detecting the fusion site of the transgenic fusion protein. All of the generated mESC lines showed a specifi c DNA band representing either a part of the neomycine resistance gene or a part of the appropriate transgenic fusion protein ( Figure 1B). Transgenepositive lines were differentiated and analysed for hTDP-43 and hFUS fusion protein expression by fl uorescence imaging. To distinguish plasmid-derived exogenous proteins from endogenous wild-type TDP-43 and FUS the transgenes were tagged with DsRed-Monomer C1fl uorescent protein (named ' DsRed ' ). As expected, hTDP-43 and hFUS transgenic proteins were mainly localized in the nucleus while the DsRed fl uorescence protein alone (vector control) was present throughout the whole cell ( Figure 1C), thereby proving that the physiological protein distribution was not impaired by the fusion to the DsRed. Quantifi cation of the total amount of transgenic cells in differentiated cultures resulted in 12.2 Ϯ 1.9% transgene-positive cells of the total cell number without variation between cell lines (one-way ANOVA, F ϭ 2.021, p ϭ 0.0715).

Expression of ALS causing genes affects neuron survival
We used ESCs from mice expressing green fl uorescent protein (GFP) downstream from the motor neuron specifi c Hb9 promoter to label motor neurons. Differentiation of these murine ESCs into motor neurons was induced by embryoid body formation for fi ve days and followed by neuron maturation in monolayer cultures that were analysed after an additional 5 -14 days. The number of motor neurons and other neuron types were determined by immunofl uorescence imaging. ESC lines transfected with the empty vector showed similar results compared to the non-transgenic (nt)-control ESC line with no signifi cant difference in the number of Tuj1 ϩ The presence of differentiated motor neurons and other neuronal subtypes was directly affected by the expression of mutant hTDP-43, but not wild-type hTDP-43 protein: after fi ve days of maturation, hTDP-43 A315T expressing neural cultures lost all GFP-expressing motor neurons, whereas wild-type hTDP-43 expressing cultures behaved similarly to non-transgenic control cultures (nt-control) ( Analysing the amount of GFP ϩ motor neurons after 24 h of neuron maturation, we found approximately 31% GFP ϩ motor neurons under control conditions, whereas wild-type and mutant hTDP-43 expressing cultures harboured 65 -70% fewer GFP ϩ motor neurons compared to control conditions (vector control, 29%; wild-type hTDP-43, 9%; hTDP-43 A315T, 11.5%).
After 14 days of maturation/aging, we observed a dramatic loss of 1.4% Tuj1 ϩ and 1.1% Map2 ϩ neurons in hTDP-43 A315T neuronal monolayer cultures compared to nt-control conditions ( Figure  2A, Supplementary Tables 2/3 to be found online at http://informahealthcare.com/doi/abs/10.3109/2167 8421.2015.1055275). The survival of Map2 ϩ neurons was signifi cantly higher in both nt-control and wildtype hTDP-43 conditions compared to hTDP-43 A315T cultures. Already by day 5 of maturation, analyses of GFP ϩ to Map2 ϩ neuron numbers revealed a selective loss of GFP ϩ motor neurons in the cultures expressing the hTDP-43 A315T system ( In contrast, analysis of cultures expressing hFUS revealed no reduction of GFP ϩ motor neurons after fi ve days in wild-type and mutant hFUS culture conditions, whereas after 14 days of maturation/aging, GFP ϩ motor neuron numbers were signifi cantly lost in cultures expressing wild-type and mutant hFUS ( Figure 2B, Supplementary Table 5 to be found online at http://informahealthcare.com/doi/abs/ 10.3109/21678421.2015.1055275). Tuj1 ϩ and Map2 ϩ neurons were in significantly reduced amounts in wild-type and mutant hFUS cultures as well ( Figure 2B, Supplementary Tables 6/7 to be found online at http://informahealthcare.com/doi/abs/10. 3109/21678421.2015.1055275). The most striking difference between TDP-43 and FUS cultures was that wild-type hFUS expression led to complete loss of motor neurons when analysed at day 14 of maturation/aging, suggesting that expression of wildtype hFUS has the same effect as expression of mutant hTDP-43 or mutant hFUS proteins (Figure 2A, B). The reduced ratio between the numbers of GFP ϩ and Map2 ϩ neurons in cells expressing wild-type and mutant hFUS showed a selective vulnerability of GFP ϩ motor neurons only at day 14 ( Figure 2B, right panel; Supplementary Quantifi cation of the total amount of transgenic cells in differentiated cultures showed no differences between cell lines over time (two-way ANOVA, F ϭ 0.1297, p ϭ 0.8787), indicating an additional effect of neuronal maturation/aging on neurodegeneration seen in our cultures. These data suggest that the expression of hTDP-43 A315T and wild-type and mutant hFUS proteins induce neuronal degeneration with partial selectivity for GFP ϩ motor neurons in differentiated and maturated/aged transgenic murine ESCs cultures.

Protein aggregation and cytoplasmic mislocalization by expression of hTDP-43 A315T
The formation of aggregates or inclusions and protein mislocalization of pathogenic TDP-43 and FUS are hallmarks of ALS (31 -33). We therefore investigated the presence of aggregate formation and mislocalization of the transgenic proteins. Notably, only hTDP-43 A315T, but not cultures expressing wild-type hTDP-43 or wild-type or mutant hFUS, exhibited aggregation formation or abundance of the respective protein in the cytoplasm (Figure 3). Interestingly, cytoplasmic mislocalization of hTDP-43A315T was only found in cells of mesodermal origin ( Figure 3A). We could detect only rarely cytoplasmic localization of hTDP-43 by wild-type hTDP-43 expression (Supplementary Figure 1

Non-cell-autonomous effects cause motor neuron degeneration in vitro
We observed initial hints for non-cell-autonomous effects in cultures expressing hTDP-43 but not hFUS. Even though the transgene was not expressed by neurons themselves, there was obvious neuronal  To prove whether non-cell-autonomous effects are suffi cient to evoke motor neuron degeneration, we generated transgenic muscle cells as the major connective cell type of motor neurons by overexpressing wild-type and mutant hTDP-43 or hFUS proteins and cocultured them with healthy (ntcontrol) GFP ϩ motor neurons ( Figure 4A, representative images). Quantitative analysis of murine ESC muscle cell differentiation resulted in a total amount of 48.7% (nt-control) and 74.2% (transgenic cell lines) of myosin ϩ , SMA ϩ and desmin ϩ muscle cells of which 18.8 Ϯ 5.6% of the cells showed positive transgene expression by a DsRed fl uorescence signal (data not shown). The amount of GFP ϩ motor neurons collected via FACS immediately before plating yielded approximately 20%. Within the remaining A315T-, wild-type hFUS-and hFUS P525L expressing muscle cells were analysed. All data are depicted relative to the non-transgenic motor neurons growing on the control muscle cells. Total neurite lengths of motor neurons were signifi cantly reduced if cocultured on hTDP-43 A315T expressing muscle cells ( p ϭ 0.0002; two-sided unpaired t -test). The expression of both wild-type hFUS and hFUS P525L proteins in muscle cells signifi cantly affected the total neurite length (wild-type hFUS: p ϭ 0.025, FUS P525L: p ϭ 0.001; two-sided unpaired t -test). * p Յ 0.05, * * p Յ 0.01, * * * p Յ 0.001. motor neurons, the neurite lengths and the branching points of GFP ϩ motor neurons were analysed by immunofl uorescence imaging after 24 h in coculture. The number of motor neurons was not altered by transgene expression ( Figure 4B). Cocultures of nt-control and vector control muscle cells with healthy motor neurons revealed no differences in neurite lengths and number of branching points. In contrast, motor neurons growing on top of transgenic hTDP-43 A315T and hFUS muscle cells showed a signifi cantly reduced total neurite length compared to the nt-control. Strikingly, wild-type hFUS expression also led to a signifi cant reduction in the total neurite length.
In summary, we showed that hTDP-43 A315T as well as both wild-type and hFUS P525L expressed in muscle cells clearly exert a non-cell-autonomous degenerative effect on nt-control motor neurons ( Figure 4B).

Discussion
Here we report the generation of a murine ESCbased cell model of ALS by expression of wild-type and mutant h TDP-43 and h FUS genes. Investigating heterogeneous neuro-ectodermal cultures, we found increased neurodegeneration with partial selectivity for motor neurons by expression of hTDP-43 A315T as well as wild-type and mutant hFUS. Furthermore, we identifi ed degenerative non-cell-autonomous effects of motor neurons caused by muscle cells expressing hTDP-43 A315T or wild-type and hFUSP525L. Interestingly, wild-type TDP-43 did not induce a comparable pathology.
Analysing murine hTDP-43 animal models revealed contrasting results, including the hypothesis of both loss-of-function and gain-of-function pathology (34). However, a common feature is that mutant hTDP-43 (A315T or M337V) leads to premature death and motor neuron degeneration. It is, however, still unclear whether overexpression of wild-type hTDP-43 also causes motor neuron disease. Using both Thy-1 promoter-and prion protein promoter (PrP)-driven overexpression of wild-type hTDP-43 was reported to cause a phenotype with many features of motor neuron disease (28,29,35). Another study using expression of wild-type hTDP-43 and Q331K or M337V mutants, however, reported only tremor in wild-type animals while the variant hTDP-43 strains caused signifi cant age-dependent motor defi cits. Interestingly, these mice did not die prematurely, but showed rather a stable phenotype after the initial motor impairment. Finally, in a rat model of hTPD43 overexpression, only rats expressing mutant hTPD43 developed symptoms of motor neuron disease (36). In line with that, we detected degeneration of motor neurons only in the mutant TDP-43 A315T line. In older mice representing a more advanced form of the disease, neurons other than motor neurons also were affected by degeneration (36), perfectly fi tting our data showing decreased Map2 ϩ neurons next to the complete loss of motor neurons in aged cultures. Importantly, and in contrast to the human neuropathology, many of the murine TDP-43 models showed neurodegeneration in the absence of TDP-43 aggregation and with normal nuclear 30,37,38). This might explain why we could hardly detect cytoplasmic TDP-43 mislocalization and aggregation in our murine cell model.
Since FUS was discovered to cause ALS more recently than was TDP-43, there are currently few murine TDP-43 models. Mitchell et al. reported a progressive motor neuron degeneration in mice when overexpressing wild-type hFUS (39), while Huang et al. reported that only mutated FUS caused motor neuron disease in a rat model. The different reports showed the appearance of aggregated cytoplasmic FUS in different amounts depending upon the severity of the underlying mutation (39,40). This is different from our cell model in which we did not fi nd FUS aggregation. One reason might be that within a monolayer cell model a toxic gain of function could become relevant much earlier and, thus, the respective motor neurons already degenerated before FUS accumulation become obvious. Similar to the mouse models, we detected a complete loss of motor neurons in the wild-type and mutant hFUS lines, suggesting that FUS effects are distinct from those of TDP-43.
For both mutations we show a stable expression of the transgenes over time; thus, there is no direct correlation of transgene accumulation and neurodegeneration. The progressive decline in motor and other neurons in our aging cultures may be due to different reasons, including cumulative toxicity in such an artifi cial system or cellular aging. Stem cell-derived motor neurons are very young neurons compared to neurons from adult animals/humans. This means that a stepwise acquisition of neuronal function implicates an increase in their vulnerability to many different cues. Importantly, the appearance of pathophysiological phenotypes in long-term cultures supports the hypothesis of a degenerative disease rather than a developmental disorder.
Our study has two major limitations. First, our differentiation protocol led to less than 20% motor neurons. Even though we had approximately 20% motor neurons in young cultures with progressive degeneration in aged cultures, we cannot completely rule out that this is partly due to a maturation defect by the transgenes. However, even though the yield of motor neurons was limited, we were able to show a robust motor neuron disease phenotype. Secondly, we only rarely observed transgene expression in motor neurons themselves, whereas other neuron types expressed it more obviously. However, we clearly saw a selective vulnerability of motor neurons due to transgene expression and we could detect a progressive decline in motor neuron numbers in long-term cultures (mimicking cellular aging) that was much higher in cells expressing the transgene. This suggests that motor neurons expressing the transgenes died early in motor neuron development and, thus, were not detectable at later stages. This would be in line with data from a TDP-43 rat model claiming that embryonic expression of mutated hTPD-43 is lethal (36). Additionally, Sephton et al. published that TDP-43 is developmentally regulated and highly expressed during embryonic development and decreases in levels in postnatal development (14). Expression of TDP-43 and FUS showed a remarkable reduction in adulthood in peripheral tissues but not in motor neurons (41). Thus, future studies should use inducible expression systems to overcome this early toxicity.
Even though the transgenes were barely observable in motor neurons themselves, the motor neurons underwent degeneration in cultures expressing hTDP-43 A315T as well as in wild-type and mutant hFUS proteins, leading us to the assumption of noncell-autonomous effects by transgenic cells growing in the immediate microenvironment of motor neurons. The majority of these cells were of muscle origin. Coculture experiments confi rmed a non-cellautonomous effect on healthy motor neurons by transgene-expressing muscle cells, the major connective cells of motor neurons. Consistently, a recent report showed that the expression of pathogenic hSOD1 in muscle cells alone causes motor neuron degeneration (42). For SOD1-ALS, similar results were additionally found to be mediated by glial cells in human ESC-based and murine ALS models (6 -9). Recently, Grad et al. demonstrated that non-cellautonomous SOD1 can be taken up by neighbouring cells and, thereby, propagate SOD1 misfolding (43). Re et al. found that non-cell-autonomous toxicity by human primary SALS and FALS astrocytes selectively contributes to the death of human motor neurons. The process is named necroptosis (44).
Whereas non-cell-autonomous effects of SOD1 are well studied, there exist only preliminary data on TDP-43 and no data on FUS for such effects. Pathological TDP-43 A315T expression in astrocytes did not result in a pathological effect on motor neurons (38). Correspondingly, the study of Serio et al. could not fi nd any non-cell-autonomous effects in a human iPSC model using cocultures with mutant hTDP-43 astrocytes (16). These differences among SOD1 and TDP-43 non-cell-autonomous effects generated by different cell types that interact with motor neurons (glial cells and muscles) evoke mechanistic differences in the pathologies of SOD1 and TDP-43.
Recently, there have been discussions regarding the role of muscle cells in motor neuron disease (MND), and it has been suggested that muscle in MND is not simply the victim of, but rather a major contributor to, the disease. Selective expression of mutated genes was shown to be causal for the MND phenotype and for survival in an SOD1 ALS mouse model (45) as well as in a Kennedy ' s disease mouse model (46). Furthermore, peripheral depletion of polyQ-AR in a mouse model of Kennedy ' s disease signifi cantly improved survival and motor neuron degeneration (47). Signifi cant neuropathological changes are seen in skeletal muscle of sporadic ALS patients (48). Currently there are no data regarding any pathological non-cell-autonomous effects of TDP-43 or FUS expression in muscle cells. We thus present novel initial data on the existence of noncell-autonomous causative factors for motor neuron degeneration of both wild-type and hFUS P525L and hTDP-43 A315T induced by cocultured muscle cells.

Conclusions
Here we show selective motor neuron degeneration in hTDP-43 A315T as well as in wild-type and mutant hFUS using in vitro models of ALS. In longterm cultures we also observed non-specifi c neuronal loss, not just in motor neurons, resembling spreading of cortical neuron pathology also seen in ALS (49) and frontotemporal lobar degeneration. Finally, using a coculture system we observed non-cell-autonomous effects mediating motor neuron degeneration by hTDP-43 A315T and both wild-type and mutant hFUS protein expression in muscle cells, as the major connective cell type of motor neurons.