To investigate the molecular mechanisms involved in the onset and progression of TDP-43 proteinopathy, we developed an in vitro model of highly homogeneous murine upper motor neurons (UMNs), i.e., glutamatergic neurons of layers V and VI, amenable to lentiviral transduction with human TDP-43, wild-type (wt) or carrying an ALS-mutation (Gitcho et al., 2008). To obtain primary cortical cultures enriched in glutamatergic neurons of layers V and VI, cells were isolated from the cerebral cortex of mouse embryos harvested at embryonic day 14.5 (E14.5), a developmental stage characterized in vivo by a low number of GABAergic neurons and glial cells in the prospective cerebral cortex. Indeed, the immunostaining performed at 3 DIV (Supplementary Figure 1A) revealed that virtually all cells were positive for the neuron-specific β3-tubulin (Ab: TuJ1), and 97% of them stained double-positive for the T-Box Brain Transcription Factor 1 (TBR1), an early marker of deep-layer glutamatergic neurons (Bedogni et al., 2010). Moreover, about 64% of neuronal cells were double-positive for TuJ1 (neuron-specific type III beta tubulin) and COUP-TF-interacting protein 2 (CTIP2), a transient marker of neurons located in cortical layers V and VI (reviewed in Molyneaux et al., 2007). To quantify the GABAergic component of these cultures, we immunostained these cells for glutamic acid decarboxylase-65 (GAD65). As shown in Supplementary Figure 1B (left), GAD65-positive cells were virtually absent, while they were abundant in neurons harvested at a later developmental stage (E17.5; Supplementary Figure 1B, right). Moreover, the results of EdU incorporation revealed that, while a small number of proliferating progenitors were still present at 2 DIV, all cells lost the ability to replicate DNA at 7 DIV (Supplementary Figure 1C).

Finally, we evaluated the presence of glial cells, which are mostly produced at later stages of cortical development. TuJ1 immunostaining, combined with antibodies against specific glial markers (glial fibrillary acidic protein, GFAP, for astrocytes; oligodendrocyte marker 4, O4, for oligodendrocyte precursors; ionized calcium binding adaptor molecule 1, IBA1, for microglia) revealed the virtual absence of all glial cell types in our mouse UMN cultures (Supplementary Figure 1D, positive controls in Supplementary Figures 1E–G).

In summary, E14.5 cortical neuronal cultures are strongly enriched in postmitotic glutamatergic neurons of deep cortical layers, from which the corticospinal tract arises, and contain a negligible contribution of glial cells or GABAergic neurons.

A neuronal model of TDP-43 proteinopathy was produced by transducing the above UMN cultures with lentiviral particles to deliver tRFP alone (ctr neurons) or a human tRFP-TDP-43 N-ter fusion protein, either wt or carrying the A315T mutation (henceforth defined wtTDP-43 and mutTDP-43 neurons, respectively); constructs are sketched in Figure 1A. The transduction efficiency was evaluated using the ArrayScan microscope, as the average percentage of red fluorescent cells over the total cell number (96% for ctr and wtTDP-43 neurons and 97% for mutTDP-43 cells; Figure 1A′). All transduced cells exhibit cytoplasmic aggregates, are viable, and project abundant TuJ1+ neurites (Supplementary Figure 2).

We then assessed the relative expression levels of the exogenous TDP-43 fusion proteins and investigated their subcellular localization, since the cytoplasmic accumulation of TDP-43 is a key feature of ALS neurons (Suk and Rousseaux, 2020). Biochemical analysis, performed on UMN lysates with an antibody recognizing both human and murine TDP-43, detected the expression of total TDP-43 protein (about 4-times higher than the endogenous protein in ctr cells; Figure 1B, solid arrows, quantified in Figure 1B′); notably, the level of endogenous murine TDP-43 is somewhat reduced in wtTDP-43 and mutTDP-43 neurons, probably due to negative post-transcriptional autoregulation (Ayala et al., 2011; Avendano-Vazquez et al., 2012; Figure 1B′′). A ~65 kDa band likely represents a degradation product of the fusion protein. Furthermore, a more prolonged exposure of chemiluminescence-stained membranes revealed, both in wtTDP-43 and mutTDP-43 neurons, the presence of TDP-43 proteolytic fragments, a typical hallmark of ALS neurons (Figure 1B, empty arrows; Neumann et al., 2006).

To further characterize TDP-43 expression, we set up a subcellular fractionation protocol to separate nuclear from soluble and insoluble cytosolic fractions (Supplementary Figure 3A). The results of western blotting performed on subcellular lysates display an enrichment of exogenous wtTDP-43 and mutTDP-43 proteins in the insoluble fraction (upper solid arrow; Supplementary Figure 3B); interestingly, 25 kDa degradation products are restricted to the insoluble fraction (empty arrow; Supplementary Figure 3B). The cytoplasmic accumulation of exogenous TDP-43, a key feature of ALS neurons (Suk and Rousseaux, 2020), was confirmed by immunocytochemistry, revealing a predominantly nuclear localization of the endogenous TDP-43 in ctr neurons (Neumann et al., 2006) in contrast with an enrichment of both wtTDP-43 and mutTDP-43 in cytoplasmic aggregates (Figure 1C). Numerous cytoplasmic aggregates are also found in neurites (Figure 1D). Importantly, cytoplasmic localization is not dependent on overexpression, since an overexpressed TDP-43 fused to a shorter tag (see methods) maintains a nuclear distribution (Supplementary Figure 3C).

Our findings are reminiscent of neuronal cytoplasmic aggregates observed in all cortical layers in motor neuron disease with frontotemporal dementia (Mackenzie et al., 2011). Notably, expression of RFP alone does not lead to aggregates; thus, the observed effects are TDP-43-specific. Overall, our results indicate that neurons expressing RFP-TDP-43 fusion proteins recapitulate several key features of TDP-43 proteinopathy within the brief lifespan of a primary neuron culture. To our knowledge, ours is the first in-vitro neuronal model of TDP-43 proteinopathy.

The analysis of tissue samples obtained from 97% of ALS patients has revealed the presence of TDP-43 pathologic inclusions (Neumann, 2009; Ling et al., 2013). TDP-43 is known to bind other RBPs (e.g., fragile X mental retardation protein, FMRP, and neuron-specific Elav-like Hu RNA-binding protein D, HuC/D), generating granules that transport mRNAs along neurites (Fallini et al., 2012). Therefore, we investigated the nature of exogenous TDP-43 aggregates in our UMN cultures by immunostaining wtTDP-43 and mutTDP-43 neurons with an HuC/D-specific Ab. HuD is involved in mRNA stability, splicing, and positive regulation of translation (Diaz-Garcia et al., 2021). Our results reveal that exogenous TDP-43 granules partially colocalize with HuC/D (Figure 2A). The degree of colocalization, plotted in Figure 2A′, shows that HuD colocalizes slightly more frequently with wtTDP-43 than with mutTDP-43.

To assess whether exogenous TDP-43-positive cytoplasmic granules colocalize with stress granules (reviewed in Dewey et al., 2012), we immunostained cells for Ras-GTPase-activating protein SH3 domain-binding protein 1 (G3BP1)], a known component of stress granules that is involved in their formation (Matsuki et al., 2013). Under basal conditions (37°C), G3BP1 signal displayed a diffuse pattern in ctr, wtTDP-43 and mutTDP-43 neurons alike, while heat shock treatment (43°C, 30 min) generated G3BP1+ granules that contained exogenous TDP-43 (Figure 2B). This result recapitulates findings obtained in cells undergoing a non-lethal injury (Higashi et al., 2013) and suggests that in our conditions TDP-43 does not promote SG formation per se, but retains the ability to be recruited to newly assembled stress granules, once they form (Colombrita et al., 2009), although TDP-43 oligomerization and aggregation takes place in the cytoplasm separate from them (Streit et al., 2022).

In parallel to the functional analysis of our cellular models (see below), we took advantage of a transcriptome and translatome analysis (G. Viero and co-workers, Institute of Biophysics, CNR Unit at Trento; GEO ID: GSE239419) conducted on cell-body and axonal compartments, to explore the gene expression and mRNA translation landscape at a systems biology level. To this end, cortical neurons were plated on microfluidic chambers (Taylor et al., 2005), consisting of two main channels separated by 450 μm-long microgrooves (see Methods); cell bodies are too large to enter them, while dendrites are too short to extend beyond them.

The physical separation of axonal and somatodendritic compartments was demonstrated by immunofluorescence in untransduced neurons, cultured in microfluidic chambers for 9 DIV and immunostained with Abs for Microtubule Associated Protein 2 (MAP2), a dendrite-specific marker, and for TuJ1, which decorates dendrites, cell bodies and axons alike. As shown in Supplementary Figure 4, MAP2 signal is sharply restricted to the whole-cell channel, while axons extend the right channel and are labeled by TuJ1 only.

To ask whether exogenous TDP-43 might bind to components of the mRNA translation machinery, possibly affecting local mRNA translation in the axon, we immunostained wtTDP-43- and mutTDP-43-expressing neurons with antibodies against small and large ribosomal subunit proteins (RPS6 and RPL26, respectively). As shown in Figure 3A and histogram in Figure 3A′, 20% of granules, on average, colocalize with RPS6 and RPL26, in both wtTDP-43 and mutTDP-43 neurons. This finding suggests that TDP-43 aggregates may sequester ribosome components and/or interfere with ribosomal transport, likely resulting in an overall decrease of mRNA translation.

Therefore, we set out to determine whether wtTDP- and mutTDP-expressing neurons showed decreased axonal levels of RPL26, a component of the ribosomal large subunit for which efficient antibodies are available, relative to control axons. To this end, we immunostained wtTDP-43, mutTDP-43 and ctr neurons, cultured in microfluidic chambers for 9 DIV, with antibodies to RPL26 and the axonal marker TuJ1. RPL26 signal was then quantified in the axon and cell body. As shown in Figures 3B,B′, RPL26 protein levels are decreased in both axons and cell bodies of wtTDP-43 and mutTDP-43 neurons relative to ctr neurons. In conclusion, the decrease of axonal RPL26 suggests that the overall amount of 60S ribosomal subunit may be significantly reduced in wtTDP-43 and mutTDP-43 axons and cell bodies, relative to ctr neurons.

Accordingly, we asked if protein synthesis might be globally defective in these neurons. To assess whether mRNA translation is affected in wtTDP-43 and mutTDP-43 neurons, we carried out a puromycylation assay (David et al., 2012). By quantitative immunofluorescence methods, we evaluated puromycin incorporation as a measure of ongoing protein synthesis. As shown in Figures 3C,C′, in both wtTDP-43 and mutTDP-43 neurons puromycin signal is significantly reduced in the cell body, axon and growth cone alike, similar to findings previously observed at the neuromuscular junction (Altman et al., 2021).

To identify transcripts downregulated in the axon of wtTDP-43 and mut-TDP-43 neurons, next-generation sequencing (NGS) of the axon- and cell-body-specific polysome-engaged mRNAs (polysome profiling) from TDP-43-overexpressing neurons was performed by tag-free polysome isolation through a miniaturized sucrose gradient (Supplementary Figure 5; see Methods, Negro et al., 2018).

We found that polysomal mRNA levels are robustly dysregulated in both human TDP-43-expressing neuronal populations compared to control neurons, with 1,598 and 1,601 differentially expressed genes (DEGs) for wtTDP-43 and mutTDP-43, respectively. Compared to the number of DEGs in the cell body alone (467 in human TDP43 and 513 in mouse TDP43), more changes occur in the axon, accounting for 65 and 68% of total DEGs, respectively. Importantly, 1,043/1,131 (92%) axonal DEGs in wtTDP-43 and 877/1,088 (81%) in mutTDP-43 are downregulated, suggesting a widespread loss of mRNAs engaged in local translation (Table 1).

Of all downregulated polysome-engaged transcripts, a large number belonged to gene ontologies related to protein synthesis and displayed lower levels in wtTDP-43 and mutTDP-43 neurons (289 and 313, respectively), compared to control neurons (Figure 3D). Of these, 47 axon-specific polysome-engaged DEGs were shared between the two overexpressing populations, and are listed in Table 2 and Supplementary Table 1. Remarkably, in wtTDP-43 neurons, axon-specific polysomal DEGs make up 97.6% of the total, whereas in mutTDP-43 cells axon-specific DEGs amount to 33.9% of the total.

Two other gene ontologies affected by TDP-43 overexpression were related to synaptic function and to the response to oxidative stress. The presence of high levels of ROS and the consequent oxidative stress-dependent neuronal damage are among the main hallmarks of ALS (Barber and Shaw, 2010). Therefore, we asked whether our neuronal model of TDP-43 proteinopathy exhibits this characteristic aspect of neurotoxicity. We performed this analysis in neurons loaded with H₂DCFDA (DCF), both under untreated conditions (Figure 4A) and upon mild iron overload (48 h in the presence of 20 μM ferric iron); the latter provides information about the ability of neurons to counteract and detoxify an oxidative condition, promoted here by the iron-catalyzed Fenton reaction. Not only did our results show a significant increase in basal ROS levels in both overexpressing populations compared to ctr neurons (Figure 4B, red dot bars), but they also revealed a sharply increased oxidative effect of iron overload in TDP-43 overexpressing neurons vs. ctr cells (Figure 4B, green dot bars). As mentioned, numerous polysome-engaged somatic and axonal mRNAs involved in the response to oxidative stress are downregulated in neurons expressing wtTDP-43 (78) and mutTDP-43 (74), compared to ctr neurons. Notably, a majority of differences observed in wtTDP-43 neurons were relative to axonal mRNAs, suggesting that locally translated axonal mRNAs may play a major and as yet uncharacterized role in the detoxification of oxidative stress. Out of the axonal DEGs, 9 were shared between wtTDP-43 and mutTDP-43 neurons (Figure 4C; Table 2; Supplementary Table 1), and include transcripts encoding nucleotide exchange factors for chaperone proteins, kinases, ER membrane sensors involved in stress response and Prdx2, encoding an abundant cytosolic peroxidase.

As mentioned, gene ontologies related to synaptic function are affected in wtTDP-43 and mutTDP-43 neurons, compared to ctr neurons. Synaptic abnormalities have been reported in the neurons of both ALS patients and animal models (primarily SOD1 transgenics), ranging from morphological alterations to functional impairment, thereby supporting the hypothesis that ALS might either result from, or be exacerbated by synaptic dysfunction (Fogarty, 2019).

Since synaptic alterations seem to involve primarily glutamatergic neurons, as suggested by recent genetic studies (van Rheenen et al., 2021), we investigated whether wtTDP-43 and mutTDP-43 overexpression could induce synaptic impairment in our cultures. Whole-cell patch clamp experiments revealed spontaneous, AMPA-dependent mini excitatory postsynaptic currents (mEPSCs), recorded at a holding potential of −70 mV and in the presence of a Na⁺ channel blocker (TTX, 1 μM) and GABA_A receptor antagonist (gabazine, 10 μM). No significant differences were detected in the average amplitude, decay time constant, or frequency of mEPSCs (Supplementary Figure 6). However, spontaneous synaptic activity in TTX- and gabazine-free ACSF revealed recurrent bursts of postsynaptic currents (Figure 5A). While burst duration was not affected by TDP-43 overexpression (Figure 5A′′; either wild-type or mutant), the frequency of bursts was significantly reduced in both overexpressing populations compared to ctr neurons (Figure 5A′).

The decrease of spontaneous bursting activity observed in TDP-43 overexpressing neurons prompted us to investigate the role of calcium (Ca²⁺) homeostasis in UMNs. To this end, neurons were loaded with the fura-2 calcium dye which, thanks to its ratiometric properties, makes it possible to compare intracellular Ca²⁺ levels both at rest and at maximal glutamate (100 μM) stimulation. In keeping with the decreased burst frequency, both wtTDP-43 and mutTDP-43 UMNs showed decreased basal levels of intracellular Ca²⁺ concentration [Ca²⁺]_i compared to ctr neurons, therefore potentially affecting spontaneous electrical activity. Notably, the amplitude of [Ca²⁺]_i response to glutamate stimulation was also significantly reduced in wtTDP-43 but not in mutTDP-43 neurons (Figure 5B).

The impairment of spontaneous synaptic activity correlates with a depletion of vesicles in presynaptic active zones in TDP-43-overexpressing neurons, as revealed by ultrastructural analysis (Figure 5C) and by the quantification of presynaptic vesicle numbers (graph in Figure 5C′). To confirm in live neurons the results obtained by electron microscopy, we exploited the properties of the FM1-43 dye (Amaral et al., 2011), since the loading and loss of the fluorescent dye provide a reliable measurement of synaptic vesicle retrieval and release in cultured neurons (Sambri et al., 2020). Exo-endocytosis was evoked by a first pulse of depolarizing stimulus (30 mM KCl added to the bath together with FM1-43 dye; see Methods); after washing the non-vesicle bound FM1-43, we quantified the fluorescence signal within synaptic boutons, correlating with their content of endocytic FM1-43-labeled vesicles (see image in Figure 5D, left image). The graph in Figure 5D′ shows a trend of signal decrease in wtTDP-43 neurons and a sharp decrease in mutTDP-43 cells, indicating, as in graph Figure 5C’, a lower number of vesicles in TDP-43-overexpressing neurons compared to ctr ones. The FM1-43-based experiments also allowed us to analyze the rate of synaptic exocytosis; indeed, after inducing fluorescent dye endocytosis, a second depolarizing KCl pulse caused vesicle exocytosis, evaluated by FM1-43 fluorescence decay over 20 s. As shown in Figure 5D (right) and Figure 5D′′mutTDP-43 and, even more so, wtTDP-43 neurons, where calcium alterations are more pronounced, show a major impairment in their rate of exocytosis compared to ctr cells.

Again, a high number of DEGs were observed in the axonal translatome, underlying the key contribution of axonal protein synthesis to the regulation of synaptic activity and maintenance (Figure 5E). The list of significantly downregulated synaptic mRNAs shared by wtTDP-43- and mutTDP-43 axons includes transcripts encoding synaptic adhesion molecules, regulators of cytoskeletal dynamics, regulatory and structural proteins involved in synaptic vesicle trafficking/recycling (Table 2; Supplementary Table 1).

Animal handling and experimental procedures were performed in accordance with the EC guidelines (EC Council Directive 86/6091987) and with the Italian legislation on animal experimentation (Decreto L.vo 116/92) and approved by our Institutional Animal Care and Use Committee.

Mouse embryonic cerebral cortices (strain C57BL/6 N) were harvested at gestational day 14.5 (E14.5) and digested with Trypsin (Gibco, 15090-046). To obtain a single cell suspension the cells were mechanically dissociated with a glass Pasteur in neuronal culture medium (Neurobasal™ Plus Medium, Gibco A3582901, 1X B-27™ Plus Supplement, Gibco, A35828-01); the cell concentration was calculated with the T20 automated Cell Counter (Biorad). Cortical neurons were plated at the density of 42,000/cm² on plastic or glass slides treated with Poly-D-Lysine (100 μg/mL, Sigma, P6407), and cultured in an incubator at 37°C, 5% CO2. For specific experiments, cortical neurons were plated on microfluidic chambers (Xona Microfluidics, SND-450). In the cell body compartment, more neuron culture medium volume was added relative to the axonal compartment to generate hydrostatic pressure. At 2 days in vitro (DIV) 20 ng/mL BDNF (Proteintech, 450-02-B) was added to the axonal compartment; fresh BDNF (20 ng/mL) was added every 3 days.

For overexpression experiments in cortical neurons, we generated constructs encoding the human wt TDP-43 (wtTDP-43) and the human mutant TDP-43 (mutTDP-43), fused N-terminally with the turbo-Red Fluorescent Protein or the FLAG tag (DYKDDDDK, Einhauer and Jungbauer, 2001); the turbo-Red Fluorescent Protein (tRFP) was used as a control. For calcium assays control neurons were transduced with empty vector p277 instead of tRFP due to technical issues with the imaging instrument. The vectors were engineered in the laboratory, starting from the lentiviral vector pLenti277-GFP kindly provided by L. Naldini (San Raffaele-Telethon Institute of Gene Therapy, Milan, Italy). Lentiviral particles were prepared as described previously (Amendola et al., 2005). Briefly HEK293T cells were transiently co-transfected using the calcium-phosphate precipitation method with the transfer vectors, the MDLg/pRRE plasmid, the RSV-Rev plasmid and the MDLg plasmid encoding the G glycoprotein of the vesicular stomatitis virus. Cell supernatants containing lentiviral particles were collected 72 h after transfection, filtered and subjected to ultracentrifugation. The pellets were resuspended, divided into aliquots and stored at −80°C. To calculate the titer of lentiviral particles, cortical neurons were seeded in a 24-well plate and infected at 5 DIV with serial dilutions (10⁻³ to 10⁻⁷) of lentiviral particles overexpressing tRFP-TDP-43(wt), tRFP-TDP-43(A315T) and tRFP. Uninfected cortical neurons were used as controls. The experiment was repeated in triplicate for each condition. Cortical neurons were fixed after 9 days in vitro (DIV) and processed for immunofluorescence with tRFP and TuJ1 primary antibodies. ArrayScan microscope (Thermo Fisher ArrayScan XTI HCA Reader) was used to acquire 20 random fields/well. For every field we counted tRFP-positive infected cells (red) and the total number of cells, stained with a nuclear marker (DAPI). A ratio of total tRFP-positive cells to DAPI-positive cells was calculated for each well. To calculate the titer a given ratio of tRFP-positive cells to total DAPI positive cells between 0.1% and 10% (dynamic range) was chosen, and this formula was used: (Infected cells/Total cells) *Dilution factor *Number of seeded cells. In our experiments a multiplicity of infection (MOI) = 4 was used to achieve the transduction of most cells (average > 96%, see Figure 1A′), without detectable cellular toxicity. Cortical neurons were infected at 5 DIV and treated, fixed or lysed at different DIV based on the experimental needs.

Cells were fixed with 4% Paraformaldehyde (PFA)/4% Sucrose in 1X PBS for 15 min, then incubated with primary antibodies in 10% Goat Serum (GS), 0.1% Triton in 1X PBS overnight at 4°C. For antibodies directed against nuclear epitopes a permeabilization step of 10 min with 0.5% Triton in 1X PBS was added. Subsequently, cells were incubated with secondary antibodies, washed three times with 1X PBS and counterstained with DAPI (D9542, Sigma). Images were acquired using Leica Confocal SP8 (Leica TCS SP8 SMD FLIM Laser canning Confocal), Nikon Spinning Disk (Nikon CSU-X1 Spinning Disk, Nikon TE2 inverted microscope), Axio Observer (Zeiss Axio Observer.Z1 with Hamamatsu EM9100). The quantification of all experiments was performed with the NIS-element Software (Nikon). When IF analyses were combined with heat shock treatments, cortical neurons were seeded in 24-well plates and infected as usual. At 14 DIV the plates were incubated in a water bath at 43°C for 60 min, washed in 1X D-PBS and processed for immunofluorescence.

Cortical neurons were scraped at room temperature using 50 μL of Lysis Buffer (5% SDS, 10 mM EDTA, 50 mM Hepes pH 7.4) with the addition of protease inhibitors (Leupeptin, Sigma, L2884; Aprotinin, Sigma, A1153; PMSF, Sigma, P7626; Pepstatin A, Sigma, P5318; Sodium Orthovanadate, Sigma, S6508; Sodium Pyrophosphate, Sigma, 221,368; Sodium Fluoride Sigma, S6776; β-Glycerophosphate Sigma, G6251). All protein extracts were sonicated and quantified with Pierce ™ BCA Protein Assay kit (ThermoFisher, 23225). 25 μg protein lysate were resuspended in Laemmli buffer 1X (Sigma, S3401-1VL), denatured at 95°C for 5 min and loaded onto a polyacrylamide gel (Mini-PROTEAN(R) TGX ™ Precast Gels 4%–20%, Biorad, 456-1094). Proteins were transferred to a nitrocellulose membrane (Biorad, 1704,159) using Trans-Blot(R) Turbo™ (Biorad). Membranes were incubated in blocking buffer (5% milk in TTBS) for 1 h at RT, then overnight at 4°C with primary antibodies diluted in blocking buffer. The membranes were incubated with secondary antibodies in blocking buffer. Bands were revealed using the Clarity ™ Western ECL Substrate (Biorad, 170-5061). The images were acquired using the ChemiDoc™MP (Biorad) and the quantification was performed with Image Lab ™Software (Biorad).

See Supplemental Table 2.

All described procedures were performed at 4°C. Cortical neurons were washed in 1xPBS, scraped and lysed for 10 min in Lysis Buffer (50 mM Tris–HCl pH 8.0, 10 mM NaCl, 5 mM MgCl2, 0.1% Nonidet P-40) with the addition of protease inhibitors. The lysate was spun at 1,000 × g for 15 min to pellet the nuclear fraction (P1). The supernatant (S1) was centrifuged at 50.000 rpm for 60 min in a TLA55 rotor (Beckman) to yield crude cytosol (S2) and crude membrane pellet (P2). The pellet (P2) was resuspended in Lysis buffer without Nonidet P-40 and centrifuged again at 50.000 rpm for 60 min to yield washed crude membrane pellet (P2’). The nuclear fraction (P1) was washed three times with Lysis buffer without Nonidet P-40 at 1,000 × g for 15 min. Pellet was resuspended in a high-salt buffer (20 mM HEPES pH 7.5, 0.5 M NaCl, 1 mM EDTA, 1 mM dithiothreitol and protease inhibitors) and rotated for 20 min. The pellet was centrifuged at 17.700 g for 30 min and the supernatant, containing the nuclear extract, was collected.

DIV 14 cortical neurons plated on microfluidic chambers were treated with 2 μM puromycin (Sigma, P8833) for 5 min. Pulse-labeling with a low concentration of puromycin, a structural analog of aminoacyl-tRNAs that is incorporated by nascent polypeptide chains, causing peptide release and disassembly of the two ribosomal subunits, was used as a readout of new protein synthesis (David et al., 2012). After treatment, the cultures were fixed and processed for immunofluorescence using an anti-puromycin primary antibody. Puromycin mean pixel intensity was measured in the soma, in 30 μm axon shaft and in the growth cone of at least 10 neurons per condition for each experiment.

Colocalization analysis between tRFP-tagged TDP43 and endogenous HuC/D, RPS6 and RPL26 was performed in the soma of DIV 9 cortical neurons, at least 10 neurons per condition for each experiment. Pearson’s correlation analysis was carried out using the ImageJ plugin JACoP; thresholds were set as described (Bolte and Cordelieres, 2006).

Cortical neurons, plated in a 96-well plate (Greiner Bio-One, 655090) and cultured up to DIV 13, were loaded with 5 μM dichlorofluorescein diacetate (CM-H2DCFDA, Thermofisher, C6827), a probe that turns into a fluorescent molecule (2′,7′-dichlorofluorescein, DCF) upon oxidation by intracellular ROS. The CM-H2DCFDA loading was performed in KRH pH 7.4 buffer (125 mM NaCl, 25 mM Hepes/NaOH pH 7.4, 5 mM KCl, 1.2 mM MgCl2, 2 mM CaCl2, 6 mM Glucose) for 30 min at 37°C, followed by 10 min of incubation with the nuclear dye 1X Hoechst (Sigma, 33258; Vannocci et al., 2018). For iron overload, neurons were treated with 20 μM ferric ammonium citrate for 2 days before the analysis (Codazzi et al., 2016). The images were acquired using ArrayScan (ThermoFisher ArrayScan XTI HCA Reader) and the fluorescence intensity analysis was carried out by the ALEMBIC Facility of the San Raffaele Hospital.

Individual slides with cortical neuronal cultures were placed in a recording chamber mounted on the stage of an upright BX51WI microscope (Olympus) equipped with differential interference contrast optics (DIC). Cultures were perfused with artificial cerebrospinal fluid (ACSF) containing (in mM): 125 NaCl, 2.5 KCl, 1.25 NaH₂PO₄, 2 CaCl₂, 25 NaHCO₃, 1 MgCl₂, and 11 D-glucose, saturated with 95% O₂ and 5% CO₂ (pH 7.3). ACSF was continuously flowing at a rate of 2–3 mL/min at 32°C. For mEPSC recordings, ACSF was added with the Na + channel specific blocker tetrodotoxin (TTX, 1 μM). Whole-cell patch-clamp recordings were performed in cortical neurons using pipettes filled with a solution containing the following (in mM): 124 KH₂PO₄, 10 NaCl, 10 HEPES, 0.5 EGTA, 2 MgCl₂, 2 Na2-ATP, 0.02 Na-GTP (pH 7.2, adjusted with KOH; tip resistance: 4–6 MΩ). All recordings were performed using a MultiClamp 700B amplifier interfaced with a computer through a Digidata 1440A (Molecular Devices, Sunnyvale, CA, United States). Traces were sampled at a frequency of 10 kHz and low pass filtered at 2 kHz. Data were acquired using Clampex software (Molecular Devices) and analyzed with Clampfit and GraphPad Prism applications. Statistical comparisons were obtained using SigmaStat 3.5 (Systat, San Jose, CA).

Cortical neurons were incubated with 4 μM fura-2 acetoxymethyl ester (AM, Calbiochem, CAS 108964-32-5) 40 min at 37°C; the ratiometric properties of fura-2 (excitation at 340 nm and 380 nm and emission at 510 nm) permit the analysis of intracellular Ca2+ levels both at basal conditions and upon stimulation with 100 μM glutamate. Imaging setup and analysis are the same of FM1-43 assay. Single-cell video imaging setup consists of an Axioskope 2 microscope (Zeiss, Oberkochen, Germany) and a Polychrome IV (Till Photonics, GmbH, Martinsried, Germany) light source. Fluorescence images were collected by a cooled CCD video camera (PCO Computer Optics GmbH, Kelheim, Germany), with a rate of 1 ratio image every 2 s. The “Vision” software (Till Photonics) was used to control the acquisition protocol and to perform data analysis (Codazzi et al., 2006; Rosato et al., 2022).

Neurons were loaded with fura-2 AM as previously described, before being subjected to the FM1-43 assay. FM1-43 is a styryl dye whose loading and release from synaptic vesicles provides a reliable measurement of vesicle release. The neurons were treated with 20 μM FM1-43 (Sigma, cat. SCT126; Betz and Bewick, 1992), dissolved in high K+ (60 mM)-containing KRH (HK-KRH; Na+ concentration was adjusted to maintain isotonicity), diluted 1:1 in normal KRH to obtain the required final KCl and FM1-43 concentration (30 mM and 10 μM respectively). FM1-43 was kept in the extracellular solution for 2 min, to complete endocytosis; this step was followed by several washes with dye-free KRH buffer, to eliminate the excess of FM1-43; after an additional stimulation with HK-KRH, fluorescence decay of FM1-43 over time was measured. The imaging setup is the same described for fura-2 calcium assay. FM1-43 fluorescence analyses were performed as described (Lazarenko et al., 2018).

Neuronal cultures were fixed with 4% paraformaldehyde (PFA) and 2% (wt/vol.) glutaraldehyde in cacodylate buffer 0.12 moL/L pH 7.4 overnight at 4°C, followed by incubation at room temperature for 2 h in 1% (wt/vol.) OsO4, 1.5% potassium ferrocyanide in 100 mM cacodylate buffer pH 7.4 for 1 h on ice. After dehydration in a graded series of ethanol preparations, tissue samples were cleared in propylene oxide, embedded in epoxy medium (Epoxy Embedding Medium kit; Sigma-Aldrich, St. Louis, MO 63103 USA), and polymerized at 60°C for 72 h. From each sample, one semi-thin (1 μm) section was cut with a Leica EM UC6 ultramicrotome (Leica Microsystems, Vienna, Austria). Ultra-thin (60 nm thick) sections of areas of interest were then obtained, counterstained with uranyl acetate and lead citrate, and examined with a transmission electron microscope (Talos 120C Fei), image were acquired with a 4kx4k Ceta CMOS camera (Thermo Fisher Scientific).

Miniaturized sucrose gradients were used to isolate free or polysomal RNA from cell bodies and axons of CNs cultured in microfluidic chambers, as described above. The procedure for polysome profiling by miniaturized sucrose gradient fractionation was adapted from published protocols (Negro et al., 2018). The procedure consists of the following steps: (i) wtTDP-43, mutTDP-43 and ctr neurons are plated in microfluidic chambers for 9 DIV; (ii) axonal and cell body compartments of these neurons are lysed and loaded onto a miniaturized sucrose gradient to separate polysome-engaged mRNAs from sub-polysomal mRNAs; (iii) polysomal mRNAs and sub-polysomal mRNA fractions are isolated; (iv) mRNAs are extracted from these fractions and sequenced (see also Supplementary Figure 5).

Sequencing data were retrieved from Gene Expression Omnibus (GEO ID: GSE239419) and processed as described by the authors. Briefly, reads were first preprocessed for adapter removal and trimming, then mapped to the mouse genome (GRCm38.p6, ENSEMBL release 92 and Gencode M17 gene annotation) and finally deduplicated. After sample size normalization based on the trimmed mean of M-values method, only genes with FPKM > 10 in all replicates of at least 1 sample were kept for subsequent analysis. Differentially expressed genes were identified by the edgeR applying multiple thresholds (CPM > 0.05, absolute log2 fold change > 0.75, value of p < 0.05). Annotation enrichment analysis with Gene Ontology terms, REACTOME and KEGG pathways were performed with the clusterProfiler Bioconductor package. GO analysis was performed as described (Bernabo et al., 2017).

Gene ontology tables were searched using the following query words: translation, ribosom* and polysom* for mRNA translation; oxidative, peroxid* and reactive oxygen species for oxidative stress; synap*; vesicle, exocyt*, endocyt* and neurotransmitter for synaptogenesis and synaptic function.

All values are expressed as mean ± standard error of the mean (SEM) of at least 3 independent experiments. Statistical analysis was performed using the GraphPad Prism software for the choice of optimal statistical tests based on data distribution. The statistical tests for each experiment are reported in the figure legends. Differences yielding a p value ≥ 0.05 were regarded as non-significant. With respect to decay curves (Figure 5), we fitted the model fluorescence=2αctr+∑IiΔitεwhere α_ctr is the coefficient for the controls; I_i is the indicatrix functions for experimental condition and ∆i represents the difference between control and the experimental condition i. ε is supposed to have lognormal distribution.This model allows an easy interpretation of fluorescence decay because α_ctr, (α_ctr+∆_hTDP-43) and (α_ctr+∆_mutTDP-43) are related to the fluorescence half-life in the three experimental conditions. Furthermore, multiplicative log-normal distributed errors address for the higher variability observed in correspondence of higher mean values.