Microtubules Deform the Nuclear Membrane and Disrupt Nucleocytoplasmic Transport in Tau-Mediated Frontotemporal Dementia

Summary The neuronal microtubule-associated protein tau, MAPT, is central to the pathogenesis of many dementias. Autosomal-dominant mutations in MAPT cause inherited frontotemporal dementia (FTD), but the underlying pathogenic mechanisms are unclear. Using human stem cell models of FTD due to MAPT mutations, we find that tau becomes hyperphosphorylated and mislocalizes to cell bodies and dendrites in cortical neurons, recapitulating a key early event in FTD. Mislocalized tau in the cell body leads to abnormal microtubule movements in FTD-MAPT neurons that grossly deform the nuclear membrane. This results in defective nucleocytoplasmic transport, which is corrected by microtubule depolymerization. Neurons in the post-mortem human FTD-MAPT cortex have a high incidence of nuclear invaginations, indicating that tau-mediated nuclear membrane dysfunction is an important pathogenic process in FTD. Defects in nucleocytoplasmic transport in FTD point to important commonalities in the pathogenic mechanisms of tau-mediated dementias and ALS-FTD due to TDP-43 and C9orf72 mutations.


Correspondence
In Brief Paonessa et al. show that mutations in the gene encoding the microtubuleassociated protein tau that cause frontotemporal dementia result in microtubule-mediated deformation of the neuronal nucleus and disrupted nucleocytoplasmic transport. These data indicate that perturbed nucleocytoplasmic transport is a pathogenic mechanism in multiple forms of neurodegenerative disease.

INTRODUCTION
The microtubule-associated protein tau (MAPT; tau) is involved in the pathogenesis of several different forms of dementia, including Alzheimer's disease (AD), progressive supranuclear palsy, Pick's disease, corticobasal degeneration, and frontotemporal dementia (FTD) (Lee et al., 2001;Spillantini and Goedert, 2013). FTD is the third most common cause of dementia, after AD and vascular dementia (Rossor et al., 2010). Autosomaldominant missense and splicing mutations in MAPT are causes of inherited or familial FTD (FTD-MAPT) (D'Souza et al., 1999;Goedert et al., 2012;Hutton et al., 1998). However, although it is well established that these mutations lead to hyperphosphorylation and aggregation of tau protein in vivo (Ballatore et al., 2007;Goedert et al., 2012), the cell biology of neuronal dysfunction and progressive neurodegeneration in this condition are currently not fully understood.
In healthy neurons, tau protein is almost exclusively localized to the axon, and several mechanisms have been suggested for its highly polarized cellular localization, including selective mRNA and protein transport, local translation, and local degradation (Wang and Mandelkow, 2016). Mislocalization and aggregation of tau in neuronal cell bodies are common features of tau-mediated dementias, including FTD and AD (Fu et al., 2016;Thies and Mandelkow, 2007;Zempel and Mandelkow, 2015). Protein aggregation is widely considered as inherently pathogenic in neurodegeneration (Fitzpatrick et al., 2017;Herná ndez-Vega et al., 2017), altering many cellular functions, most notably autophagy and proteostasis (Bence et al., 2001;Caballero et al., 2018;Lim and Yue, 2015). However, how MAPT mutations lead to tau hyperphosphorylation and mislocalization, the effects of this mislocalization on neuronal cell biology, and how this contributes to neuronal dysfunction and neurodegeneration all remain poorly understood.
As a typical microtubule-binding protein, tau has several roles in regulating microtubule function and intracellular transport (Wang and Mandelkow, 2016). Tau binds both alpha and beta tubulin subunits of microtubules and has been demonstrated to both stabilize and promote microtubule growth (Kadavath et al., 2015;Witman et al., 1976). The presence of tau on microtubules can alter directions and rates of axonal transport (Dixit et al., 2008;Trinczek et al., 1999). Tau is a natively disordered protein and has recently been found to undergo fluid phase transitions at higher concentrations, nucleating microtubules when it does so (Herná ndez-Vega et al., 2017). Therefore, it is likely that the changes in tau levels, post-translational modifications, and cellular localization that occur in dementia lead to alterations in microtubule biology, particularly in the neuronal cell body.
To address the question of how MAPT mutations lead to neuronal dysfunction and neurodegeneration, we investigated the effects of two different classes of MAPT mutations on the cell biology of human iPSC-derived cortical neurons. We find that both missense and splicing MAPT mutations cause mislocalization of tau to the cell bodies of neurons and marked changes in microtubule dynamics. Microtubules in the cell bodies of FTD-MAPT neurons actively deform the nuclear membrane, disrupting nucleocytoplasmic transport. Defects in nuclear envelope function, including nucleocytoplasmic transport, are an important pathological process in ALS-FTD because of repeat expansions in C9orf72 and TDP-43 mutations (Chou et al., 2018;Zhang et al., 2015Zhang et al., , 2016Zhang et al., , 2018. Our findings demonstrate that dysfunction of the nuclear membrane due to altered microtubule dynamics is a pathogenic process in dementias involving tau, expanding the group of neurodegenerative diseases that involve disrupted nucleocytoplasmic transport and suggesting common mechanisms of neuronal dysfunctional in these heterogeneous conditions.

Increased Phosphorylation and Altered Cellular Localization of Tau in FTD-MAPT Neurons
To study the effects on neuronal cell biology of FTD-MAPT mutations, we generated excitatory cortical neurons (Shi et al., 2012b) from induced pluripotent stem cells (iPSCs) derived from individuals with different autosomal-dominant mutations in MAPT that are causal for early-onset FTD (Figures 1 and S1). We studied two different types of mutations: the MAPT IVS10+16 autosomal-dominant mutation, which increases inclusion of exon 10, encoding the second microtubule-binding repeat and thus altering the ratio of three (3R) and four (4R) tau isoforms (Hutton et al., 1998;Sposito et al., 2015), and the autosomal-dominant MAPT P301L missense mutation that produces an aggregation-prone form of tau (Wang and Mandelkow, 2016) (Figures 1 and S1). The cortical identity of the neurons generated and the reproducibility of the culture compositions among genotypes was confirmed by assessing the expression of a set of classifier genes that define different neuronal cell types ( Figure S1).
Total tau content was similar in neurons of each genotype, collectively referred to here as FTD-MAPT neurons ( Figure 1A). Notably, tau phosphorylation was increased in FTD-MAPT neurons compared with controls ( Figure 1A), including at Ser404 and Ser202/Thr205 (AT8), epitopes typically hyperphosphorylated in tau-mediated dementias (Alonso et al., 2004;Wang et al., 2013). As both mutations are dependent on expression of exon 10 of MAPT, we confirmed translation of exon 10 in neurons generated from all iPSC lines by western blotting and mass spectrometry ( Figures 1B, 1C, and S1). No significant difference in insoluble, aggregated tau was detected by sarkosyl extraction between neurons of each genotype ( Figure S1).
Mislocalization of tau from axons to neuronal cell bodies and dendrites is an early event in FTD in vivo (Gö tz et al., 1995;Hoover et al., 2010;Kowall and Kosik, 1987). As expected, control neurons showed a predominantly axonal distribution of tau, with tau largely absent from MAP2-positive neuronal cell bodies and dendrites ( Figure 1D). In contrast, tau was commonly present in MAP2-positive cell bodies and dendrites in both MAPT IVS10+16 and P301L neurons ( Figures 1D and 1E). Furthermore, tau within cell bodies and dendrites of FTD-MAPT neurons was hyperphosphorylated, as detected by AT8-immunoreactivity (phospho-S202/T205) ( Figure 1D).

Microtubules Invade the Nucleus in FTD-MAPT Neurons
Given the mislocalization of tau to the cell bodies of FTD-MAPT neurons, we studied neuronal microtubule dynamics in control and FTD-MAPT neurons (Figures 2 and S2). Actively extending microtubules were live-imaged in iPSC-derived neurons of each genotype by expression of GFP-tagged EB3 (Figure 2; Videos S1 and S2), the microtubule plus-end binding protein (+TIP) (Akhmanova and Steinmetz, 2008). Total microtubule movements were not different between non-demented control and FTD-MAPT neurons, with similar rates of extensions and retractions measured among the various genotypes (Table S1).
However, microtubule dynamics were qualitatively different in the cell bodies of control neurons compared with FTD-MAPT neurons. Non-demented control neurons typically had many actively growing microtubules within the cell body that extended around a smooth, oval nucleus (Figures 2A and 2B; Video S1). In contrast, many FTD-MAPT neurons had microtubules with plus ends projecting into the nucleus (15 of 26 FTD-MAPT neurons) ( Figure 2B), an event that was infrequently detected in both groups of control neurons (2 of 19) ( Figure 2B). Notably, those microtubules that abnormally projected into the nucleus in FTD-MAPT neurons frequently originated from a pronounced focus that resembled a microtubule organizing center (Figure 2A; Video S1). These pronounced foci were not detected in either of the control neuronal lines. We confirmed that the EB3+ microtubules that project into the neuronal nucleus are dynamically growing microtubules, as the mobility of EB3+ comets was greatly reduced following acute microtubule depolymerization (with nocodazole) (Figures S2Band S2C; Video S2).

Microtubules Deform the Nuclear Envelope in FTD-MAPT Neurons
Microtubules couple to the nuclear membrane through the LINC complex (Crisp et al., 2006). This physical association results in transmission of mechanical forces that influence nuclear shape and integrity (Chang et al., 2015), affecting the function of the nuclear envelope (Webster et al., 2009). Given the abnormal projection of microtubules into the nucleus in FTD-MAPT neurons, we studied the shape of the nuclear envelope in iPSC-derived neurons. Marked differences were present in nuclear shape between non-demented controls and FTD-MAPT neurons, as demonstrated by large folds, or invaginations, of the laminB1-positive inner nuclear lamina within the nucleus (Figure 3). The neuronal identity of invaginated cells was confirmed by the co-staining with the cortical deep layer transcription factor CTIP2 and the pan-neuronal protein b3-tubulin ( Figure 3).
Quantification of the proportions of neurons with deformation of the nuclear membrane, as defined by the presence of laminB1-positive regions within the nucleus, (Figures 3B and S3), demonstrated that deep nuclear invaginations were present in approximately 25% of MAPT P301L and 40% of MAPT IVS10+16 neurons, compared with fewer than 10% of control neurons ( Figure 3B). To confirm that microtubules actively deformed the nucleus in FTD-MAPT neurons, we acutely depolymerized microtubule with the small molecule nocodazole. This significantly reduced the proportion of neurons with nuclear invaginations and restored round nuclear morphology ( Figures  3C and 3D). We conclude that the pronounced deformations of the neuronal nuclear membrane in FTD-MAPT neurons are actively mediated by microtubules.

Super-resolution Imaging Demonstrates Close Apposition of Tau and Tubulin within Nuclear Lamina Invaginations
To further study the spatial relationships between tau, microtubules, and the nuclear envelope, we conducted a detailed (B) Tau isoforms with three (3R; red) or four (4R; inclusion of region 2; green) microtubule-binding regions were detected by western blot analysis of dephosphorylated protein extracts from iPSC-derived control and FTD-MAPT cortical neurons (120 DIV) and from post-mortem human cerebral cortex (non-demented individual). Tau isoforms were identified relative to a commercial tau ladder (Sigma). Molecular weight (kDa) is indicated. (C) Peptide sequences identified by tau immunoprecipitation (IP)/mass spectrometry from iPSC-derived cortical neurons, confirming the inclusion of repeat 2 (corresponding to exon 10) of 4R tau. In MAPT P301L neurons, both proline and leucine were identified at position 301 (highlighted red). See also Figure S1. (D) Tau protein is mislocalized to MAP2-positive cell bodies and dendrites in iPSC-derived FTD-MAPT neurons. Confocal images of iPSC-derived control and FTD-MAPT neurons (120 DIV; tau, green; MAP2, red; DAPI, blue). Hyperphosphorylated, AT8-positive tau (AT8; green) is found in cell bodies of FTD-MAPT neurons (arrows) but not in controls (b3-tubulin, red; DAPI, blue). Scale bars, 20 mm. (E) Increased co-localization of tau and MAP2 protein in FTD-MAPT neurons, compared with non-demented control neurons, analyzed by Pearson's R correlation (control lines, gray bars; FTD-MAPT lines, black). Significance was determined for three-sample comparison of non-demented control and two MAPT IVS10+16 lines using one-way ANOVA followed by Tukey's test (*p < 0.05 and ***p < 0.001). Pairwise comparison of the MAPT P301L line and its isogenic control was carried out using Student's t test (*p < 0.05); error bar represents SEM; n = 3 independent experiments. See also Figure S1. analysis of the neuronal nucleus in iPSC-derived neurons using three-dimensional (3D) stimulated emission depletion (STED) super-resolution imaging. Three-dimensional STED imaging demonstrated that invaginations of the nuclear lamina present in FTD-MAPT neurons commonly extended deeply into the nucleus, in some cases traversing the entire length of the nucleus, forming pronounced folds ( Figure 4A). In comparison, nuclei from non-demented controls had a regular, smooth morphology, with few examples of nuclear lamina invaginations ( Figure 4A).
As observed by confocal microscopy, tau protein was found to be abundant in the cell bodies of FTD-MAPT neurons by STED imaging and in those neurons was closely apposed to the outer nuclear membrane ( Figure 4B). Both tau protein and neuronal tubulin were found within nuclear lamina invaginations in FTD-MAPT neurons ( Figure 4B). STED imaging demonstrated that tau within nuclear membrane invaginations is within hundreds of nanometers of the nuclear lamina ( Figure 4C). Given that laminB1 filaments line the inner surface of the nuclear envelope, which is typically of the order of 15-60 nm in width (Burke and Stewart, 2013; Gerace and Huber, 2012), we conclude that tau is in close proximity to proteins in the outer membrane of the nuclear envelope in FTD-MAPT neurons.

Tau-Containing Nuclear Lamina Invaginations in Neurons of the Post-mortem FTD-MAPT Cerebral Cortex
Having identified nuclear lamina defects in iPSC-derived FTD-MAPT neurons, we asked whether alterations of the nuclear lamina are also a feature of FTD-MAPT in vivo. To do so, we studied the incidence of invaginations of the nuclear lamina in the frontal and temporal cortex from two separate cohorts from independent brain banks, both containing individuals diagnosed with FTD due to MAPT IVS10+16 mutations and compared with age-matched non-demented controls. These cohorts were analyzed separately using different methods for detecting laminB1. We quantified the fraction of all nuclei with invaginations within each brain region ( Figure 5). Similar results were obtained in both cohorts: the frequency of nuclear lamina invaginations was higher in the deep cortical layers in post-mortem cerebral cortices from individuals with FTD due to the MAPT See also Video S1, Table S1, and Figure S2.
Data from iPSC-derived FTD-MAPT neurons suggested that laminB1-positive nuclear invaginations would be associated with the presence of phosphorylated tau within the neuronal cell body. Consistent with this, we found that the nuclear lamina was grossly disrupted in neurons that had high levels of hyperphosphorylated (AT8+) tau and neurofibrillary tangles in the post-mortem FTD-MAPT IVS10+16 cerebral cortex (Figures 5C and S4B), and those neurons frequently contained pronounced nuclear lamina invaginations ( Figures 5C and S4B). Furthermore, nuclear lamina invaginations in such neurons also commonly contained AT8-positive hyperphosphorylated tau ( Figure 5C).

Disrupted Nucleocytoplasmic Transport in FTD-MAPT Neurons
Abnormalities of the nuclear lamina are also found in aging diseases, such as Hutchinson-Gilford progeria syndrome (Broers et al., 2006). Nuclear membrane distortion in response to mechanical forces leads to deleterious effects on many aspects of nuclear function, disrupting nucleocytoplasmic transport (Kelley et al., 2011). We confirmed that the nuclear lamina/membrane invaginations present in iPSC-derived FTD-MAPT neurons also (B) FTD-MAPT neurons have increased numbers of cells with deformed nuclear membranes, as defined by the shape of the inner nuclear lamina. Schematic of image analysis method used to quantify nuclear invaginations, as a measure of distortion of the nuclear membrane: nuclear area was established using DAPI (blue), and nuclear lamina (laminB1, green) was assigned as either nuclear boundary or invaginated (i.e., within the nucleus). The fraction of total laminB1 that was invaginated was used empirically to define a threshold for defining neurons as having nuclear membrane invaginations (see Figure S3 for details). Between 25% (MAPT P301L) and 40% (MAPT IVS10+16) of FTD-MAPT neurons have nuclear invaginations, compared with fewer than 10% of control neurons. Significance was determined for non-demented control and two MAPT IVS10+16 lines using one-way ANOVA followed by Tukey's test (*p < 0.05); pairwise comparison of the MAPT P301L line and its isogenic control was carried out using Student's t test (*p < 0.05); error bar represents SEM; n = 3 independent experiments. (C) Acute depolymerization of microtubules reverses nuclear lamina invaginations and restores rounded nuclear shapes. Confocal images of control and FTD-MAPT neurons (using genotypes described in A; 120 DIV), treated with DMSO (vehicle) or 10 mM nocodazole for 3 h (laminB1, green; b3-tubulin, red; DAPI, blue). Scale bars, 10 mm. (D) The proportion of FTD-MAPT neurons with nuclear lamina invaginations is significantly reduced by nocodazole treatment. Quantification of neurons with abnormalities of the nuclear lamina was carried out as in B. n = 3 independent experiments; error bars represent SEM. Significance was determined using one-way ANOVA followed by Tukey's test (*p < 0.05); error bar represents SEM; n = 3 independent experiments. See also Figure S3. contained nuclear pores within these membrane folds, with nuclear pores (labeled by NUP98) co-localizing with laminB1positive invaginations ( Figure S5A).
To assess whether alterations in the nuclear membrane in FTD-MAPT neurons result in defects in nucleocytoplasmic transport, we expressed NES:GFP and NLS:RFP from a single construct in iPSC-derived neurons . This assay enables measurement of the integrity of both nuclear localization and accumulation and cytoplasmic retention and nuclear exclusion within individual neurons (Figures 6A and S5B). Control iPSCderived neurons had discrete cellular distributions of each protein, with prominent nuclear RFP and cytosolic GFP ( Figure 6B). In contrast, localization of NLS:RFP was altered in FTD-MAPT neurons such that there was a marked decrease in the nuclear/ cytoplasmic RFP ratio ( Figure 6C). Conversely, nuclear exclusion of NES:GFP was reduced in FTD-MAPT neurons, with an increase of GFP within the nucleus ( Figure 6C). Together, these data demonstrate defects in the selective permeability of the nuclear envelope in FTD-MAPT neurons, indicating a general failure of nucleocytoplasmic transport within FTD-MAPT neurons.
Defective nucleocytoplasmic transport may be a cumulative phenomenon due to protracted nuclear membrane damage, or an acute process due to microtubule-mediated nuclear membrane deformation. To distinguish between these mechanisms, we acutely depolymerized microtubules with nocodazole in control and FTD-MAPT neurons and quantified nucleocytoplasmic transport using the NES:GFP/NLS:RFP assay. This treatment corrected the distribution of both NES:GFP and NLS:RFP in FTD-MAPT neurons, restoring the nucleocytoplasmic ratios to the level of healthy control neuron ( Figure 6D). We conclude that defective nucleocytoplasmic transport in FTD-MAPT neurons is an ongoing process due to microtubule-mediated deformation of the nuclear membrane.

DISCUSSION
The cellular and molecular biology of the pathogenesis of FTD due to MAPT mutations is not well understood. Currently, it is thought that MAPT mutations all lead to tau protein aggregation and that protein aggregation is the primary driver of neurodegeneration (Ballatore et al., 2007;Spillantini and Goedert, 2013). However, protein aggregation may represent only the late stage of the disease, and the processes preceding and leading to neurofibrillary tangle formation and cellular dysfunction remain to be elucidated. Here we report the use of human stem cell systems to study the effects of those mutations on neuronal cell biology, finding that tau-mediated dementias are accompanied by defective neuronal nucleocytoplasmic transport.
Focusing on two different types of MAPT mutations causal for FTD, we have found that both IVS10+16 and P301L mutations lead to marked defects in nucleocytoplasmic transport in human  (C) Quantification of the nuclear/cytoplasmic ratio for both NES:GFP and NLS:RFP demonstrates altered nuclear transport in FTD-MAPT genotypes relative to controls: NES:GFP is increased in the nuclei of FTD-MAPT neurons, whereas NLS:RFP is decreased (points indicate quantifications from individual fields, different shapes indicate measurements from different experiments). Significance was determined for non-demented control and two MAPT IVS10+16 lines using one-way ANOVA followed by Dunnett's test (*p < 0.05, **p < 0.01, and ***p < 0.001). Pairwise comparison of the MAPT P301L line and its isogenic control was performed using Student's t test (*p < 0.05); error bar represents SEM; n = 4 independent experiments.
(legend continued on next page) neurons. We find that both missense and splicing mutations in MAPT alter tau protein localization and phosphorylation within iPSC-derived neurons within 4 months in cell culture, recapitulating a well-described aspect of early FTD pathology in vivo (Gö tz et al., 1995;Hoover et al., 2010;Kowall and Kosik, 1987), without detectable tau aggregation. Mislocalization of tau in the cell bodies of FTD-MAPT neurons in culture leads to marked changes in microtubule dynamics, causing deformation of the nuclear membrane both in cell culture and in the human FTD-MAPT cortex in vivo. Disruption of the nuclear lamina is commonly associated with dysfunction of the nuclear envelope, and we find marked disruption of nucleocytoplasmic transport in FTD-MAPT neurons. Disrupted nucleocytoplasmic transport is due to ongoing microtubule-mediated nuclear membrane deformation, as it is corrected by acute microtubule depolymerization. Together, these data indicate that perturbation of the function of the nuclear membrane and disruption of nucleocytoplasmic transport is an important pathological process in FTD due to MAPT mutations.
An early event in FTD is the mislocalization of tau from axons to cell bodies and dendrites, and this key stage in disease progression is also an early event in iPSC-derived models of FTD-MAPT. In vivo, mislocalization of tau is typically associated with tau hyperphosphorylation (Gö tz et al., 1995;Spillantini and Goedert, 2013). We find this also occurs in iPSC-derived FTD-MAPT neurons, where we detected increased tau phosphorylation at serine 202/threonine 205 (the AT8 epitope) and also at serine 404. The sequence in which mislocalization and hyperphosphorylation take place in FTD in vivo, and in iPSC-derived FTD-MAPT neurons in culture, is not currently clear, nor are the mechanisms by which these processes occur. The appearance of tau within cell bodies and dendrites indicates a breakdown of the cellular polarity mechanisms that maintain the axonal enrichment of tau protein and its exclusion from the somatodendritic compartment, mechanisms that are poorly understood.
The two heterozygous, dominant MAPT mutations studied here have different effects on tau protein in neurons. The MAPT P301L missense mutation, like many missense mutations in the microtubule-binding region domain of MAPT, increases the tendency of tau to aggregate in cell-free assays and in transgenic mouse models (von Bergen et al., 2005;Lewis et al., 2000;Shammas et al., 2015). In contrast, the IVS10+16 mutation is not a coding mutation but rather is an intronic single base change that favors the inclusion of exon 10 in the MAPT mRNA, increasing the amount of tau containing four microtubule-binding repeats (4R), relative to the three-repeat (3R) form (Hutton et al., 1998). However, despite these differences, the changes in the forms of tau in both MAPT P301L and IVS10+16 neurons both lead to mislocalization and increased phosphorylation of tau. This finding suggests that either the presence of a pool of P301L tau, or a shift in the 3R/4R tau ratio, alter a common pathway that regulates tau distribution within neurons, tau phosphorylation, or both.
In both MAPT IVS 10+16 and P301L mutant neurons, the appearance of tau in cell bodies is accompanied by marked qualitative changes in neuronal microtubule dynamics. Microtubules in FTD-MAPT neurons actively deform the nuclear envelope, which we find can be reversed by depolymerization of microtubules. Tau has multiple roles in stabilizing microtubules (Wang and Mandelkow, 2016), and microtubules are coupled to the nuclear membrane through the LINC complex (Chang et al., 2015;Crisp et al., 2006;Luo et al., 2016). Therefore, it is likely that the overall effect of the presence of tau in the cell body is to promote microtubule stability, leading to increased pushing forces on the nuclear membrane and the formation of invaginations in the nuclear membrane. As tau has recently also been found to promote microtubule nucleation when undergoing phase transitions at high concentration (Herná ndez-Vega et al., 2017), accumulation of tau in the cell body may also lead to increased pushing forces on the nucleus by facilitating microtubule nucleation.
Alterations in nuclear shape and nuclear membrane function are a common feature of cellular aging, including in the nervous system, and are associated with multiple deleterious changes in nuclear biology, including chromatin changes and disrupted nucleocytoplasmic transport (Frost, 2016;Oberdoerffer and Sinclair, 2007). Drosophila models of FTD, with neuronal expression of human MAPT R406W, have nuclear shape abnormalities and chromatin changes . Recently, nuclear membrane disruption has been reported in a MAPT P301L transgenic mouse model of FTD (Eftekharzadeh et al., 2018). Perturbations of the nuclear lamina have been described in the post-mortem AD brain , including the juxtaposition of neurofibrillary tangles of tau with the nuclear membrane (Sheffield et al., 2006). We also find here an increase in nuclear invaginations in neurons of the human post-mortem MAPT IVS10+16 cortex, and the presence of hyperphosphorylated tau within nuclear invaginations in tangle-bearing neurons. Together, these different studies are consistent with a pathological effect of tau within the neuronal cell body in FTD and AD, whereby the presence of tau alters microtubule biology, resulting in pronounced abnormalities of the neuronal nucleus and defective nucleocytoplasmic transport. Recent findings have shown a direct interaction of aggregated tau with the nuclear pore complex (Eftekharzadeh et al., 2018), suggesting that the microtubule disruption of nuclear pore function reported here may be mediated by tau protein at the nuclear membrane.
Microtubule deformation of the nucleus is a phenotype also seen in classic laminopathies such as the accelerated aging disorder Hutchinson-Gilford progeria syndrome (HGPS), in which the primary defect is due to mutant lamin A/C protein (Capell (D) Effect of nocodazole (Noc) on the nuclear/cytoplasmic ratio for both NES:GFP and NLS:RFP. Nuclear/cytoplasmic ratio is restored in the presence of nocodazole (+) in FTD-MATP neurons compared with vehicle (À) treated neurons (points indicate quantifications from individual fields, different shapes indicate measurements from different experiments). Significance was determined using one-way ANOVA followed by Dunnett's test (*p < 0.05, **p < 0.01, and ***p < 0.001); error bar represents SEM; n = 3 independent experiments. See also Figure S5. and Collins, 2006). In that case, microtubules also contribute to nuclear deformations, leading to defects in nucleocytoplasmic transport (Kelley et al., 2011;Larrieu et al., 2014;Snow et al., 2013). Changes in nuclear envelope function in other neurodegenerative diseases, including ALS-FTD due to repeat expansions in C9orf72, Huntington's disease, and AD, have recently been reported (Freibaum et al., 2015;Grima et al., 2017;Jovi ci c et al., 2015;Zhang et al., 2015).
Our finding here of disruption of the neuronal nuclear membrane as a consequence of MAPT mutations in FTD extends this pathogenic mechanism to dementias in which protein aggregation has been thought to be the primary driver of neurodegeneration. These data suggest that dysfunction of the nuclear membrane may be a common pathogenic process in diverse neurodegenerative diseases, which could be targeted therapeutically with agents that regulate microtubule functions, nucleocytoplasmic transport, and/or associated processes.

CONTACT FOR REAGENT AND RESOURCE SHARING
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Rick Livesey (r.livesey@ucl.ac.uk).

EXPERIMENTAL MODEL AND SUBJECT DETAILS
Human iPSC lines MAPT IVS10+16-A and MAPT IVS10+16-B mutant iPSCs were as reported in (Sposito et al., 2015). MAPT P301L was generated from Janssen Pharmaceutica by TALEN editing the line MAPT P301-isogenic, under the IMI STEMBANCC project agreement ICD 483960. The non-demented control line was previously reported (Israel et al., 2012). iPSC cells were growth and expanded in feeder-free conditions using Essential 8 Medium (Thermo Fisher Scientific), at 37 C with 5% CO 2 . Essential 8 Medium was replaced daily.
Human post-mortem brain sections Human brain sections were obtained from the Queen's Square Brain Bank, Institute of Neurology, University College London (cohort one) and from the London Neurodegenerative Diseases Brain Bank and Brains for Dementia Research (cohort two). For cohort one, control brains included one male (age 71) and one female (age 56). FTD-MAPT IVS10+16 brains were from two males (age 52 and 66). For cohort two, control brains were from five males (age 63-77) and one female (age 43). FTD-MAPT IVS10+16 brains were from four males (age 48-71) and two females (age 58 and 63). The use of human post-mortem tissues for this study has been approved with Research Ethics Committee reference ID 08/H0718/54+5 for cohort one and reference ID 08/MRE09/38+5 for cohort two.

METHOD DETAILS
Generation of iPSC-derived cortical neurons and drug treatments Differentiation of iPSCs to cortical neurons was carried out as described, with minor modifications (Shi et al., 2012b;2012a). Briefly, dissociated iPSCs were plated on Geltrex (Thermo Fisher Scientific)-coated plates to reach full confluence. Neural induction was initiated the next day (Day 0) by changing the culture medium to a 1:1 mixture of DMEM/N-2 (DMEM/F-12 GlutaMAX; 1 3 N-2; 5 mg ml À1 insulin; 1 mM L-glutamine; 100 mm non-essential amino acids; 100 mM b-mercaptoethanol; 50 U ml À1 penicillin and 50 mg ml À1 streptomycin) and Neurobasal/B-27 (Neurobasal; 1 3 B-27; 200 mM L-glutamine; 50 U ml À1 penicillin and 50 mg ml À1 streptomycin) media (hereafter referred as N2B27) supplemented with 1 mM dorsomorphin and 10 mM SB431542 to inhibit TGFb signaling and support neuronal differentiation and neurogenesis, media was replaced every 24 hours. At day 12 neuroepithelial sheet was harvested and dissociated using the enzyme Dispase and replated on laminin-coated plates. The day after, media was replaced with N2B27 containing 20 ng/mL FGF2. N2B27+FGF2 was added freshly daily for 4 days to promote the maturation of neural rosettes. After 4 days FGF2 was withdrawn and neural rosettes were maintained in N2B27 refreshing medium every other day. At day 30 neural rosettes were dissociated using Accutase and neural progenitor cells were plated on laminin-coated plates at 150,000 cells/mm 2 . Plated neurons were maintained for up to 120 days with a medium change every other day. To establish identity and quality of cortical neuronal inductions, gene expression profiling was performed on a custom gene expression panel. RNA was isolated from induced cortical neurons using TRIzol (Thermo Fisher Scientific), according to the manufacturer's instructions.