Abnormal microtubule dynamics disrupt nucleocytoplasmic transport in tau-mediated frontotemporal dementia

The neuronal microtubule-associated protein tau (MAPT) is central to the pathogenesis of many dementias, including Alzheimer’s disease. 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 mislocalises to neuronal cell bodies and dendrites in cortical neurons, recapitulating a key early event in FTD. Mislocalised tau in the cell body leads to abnormal microtubule dynamics in FTD-MAPT neurons that grossly deform the nuclear membrane, resulting in defective nucleocytoplasmic transport. Neurons in the post-mortem human FTD-MAPT cortex have a high incidence of nuclear deformation, 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 both tau-mediated dementias and ALS-FTD due to TDP-43 and C9orf72 mutations.


INTRODUCTION
The microtubule-associated protein tau (MAPT, tau) is central to 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). Frontotemporal dementia is the third most common cause of dementia, after Alzheimer's disease and vascular dementia (Rossor et al., 2010). Autosomal dominant missense and splicing mutations in MAPT are causes of inherited or familial frontotemporal dementia (FTD-MAPT) (D'Souza et al., 1999;Goedert et al., 2012;Hutton et al., 1998). However, while 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 localised to the axon, and several mechanisms have been suggested for its highly polarised cellular localisation, including selective mRNA and protein transport, local translation and local degradation (Y. Wang and Mandelkow, 2015). Mislocalisation 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;Hernández-Vega et al., 2017), altering many cellular functions, most notably autophagy and proteostasis (Bence et al., 2001;Caballero et al., 2018;Lim and Yue, -4 -2015). However, how MAPT mutations lead to tau hyperphosphorylation and mislocalisation, the effects of this mislocalisation 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 (Y. Wang and Mandelkow, 2015). Tau binds both alpha and beta tubulin subunits of microtubules, and has been demonstrated to both stabilise 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 (Hernández-Vega et al., 2017). Therefore, it is likely that the changes in tau levels, post-translational modifications and cellular localisation 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 mislocalisation 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 -5 -nucleocytoplasmic transport. Deficits in nuclear envelope function, including nucleocytoplasmic transport, are an important pathological process in ALS-FTD due to repeat expansions in C9orf72 and TDP-43 mutations (Chou et al., 2018;K. Zhang et al., 2018;Y.-J. Zhang et al., 2016). 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 localisation 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 (iPSC) derived from individuals with different autosomal dominant mutations in MAPT that are causal for early-onset FTD (Figs. 1 & 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 aggregationprone form of tau (Y. Wang and Mandelkow, 2015) (Fig. S1).
Total tau content was similar in neurons of each genotype, collectively referred to here as FTD-MAPT neurons (Fig. 1A). However, tau phosphorylation was increased in FTD-MAPT neurons compared to controls (Fig. 1A), notably at Ser404 and Ser202/Thr205 (AT8), epitopes typically hyper-phosphorylated in tau-mediated dementias (Alonso et al., 2004;J.-Z. 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 (Fig. 1B, C & S1).
-7 -Mislocalisation of tau from axons to neuronal cell bodies and dendrites is an early event in FTD in vivo (Götz 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 (Fig. 1D). In contrast, tau was commonly present in MAP2-positive cell bodies and dendrites in both MAPT IVS10+16 and P301L neurons (Fig. 1D, E). Furthermore, tau within cell bodies and dendrites of FTD-MAPT neurons was hyperphosphorylated, as detected by AT8immunoreactivity (phospho-S202/T205; Fig. 1D). Actively extending microtubules were live-imaged in iPSC-derived neurons of each genotype by expression of GFP-tagged EB3 ( Fig. 2; MovieS1), the microtubule plusend binding protein (+TIP; (Akhmanova and Steinmetz, 2008)). Average microtubule dynamics 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).

Microtubules invade the nucleus in FTD-MAPT neurons
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 ( Fig. 2A, B; Movie S1). In contrast, many FTD-MAPT neurons -8 -had microtubules with plus ends projecting into the nucleus (15/26 FTD-MAPT neurons; Fig. 2), an event that was infrequently detected in both groups of control neurons (2/19; Fig. 2). Notably, those microtubules that abnormally projected into the nucleus in FTD-MAPT neurons frequently originated from a pronounced focus that resembled a microtubule organising centre ( Fig. 2A, B; Movie S1).

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 (Fig. 3). that approximately 25% of MAPT P301L and 40% of MAPT IVS10+16 neurons have deep nuclear invaginations, compared with less than 10% of control neurons (Fig. 3A,   B).
-9 -Confirming that microtubules actively deformed the nucleus in FTD-MAPT neurons, we found that acute microtubule depolymerisation (with the small molecule nocodazole) reduced the number of nuclear invaginations and restored round nuclear morphology (Fig. 3C, D). 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 analysis of the neuronal nucleus in iPSC-derived 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 (Fig. 4B). Both tau protein and neuronal tubulin were found within nuclear lamina invaginations in FTD-MAPT neurons (Fig. 4B).
STED imaging demonstrated that tau within nuclear membrane invaginations is within 100s of nanometres of the nuclear lamina (Fig. 4C). Given that laminB1 filaments line -10 -the inner surface of the nuclear envelope, which is typically of the order of 15-60nm 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.

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 of two individuals diagnosed with FTD due to MAPT IVS10+16 mutations. We quantified the fraction of all nuclei with nuclear invaginations, in both neuronal and non-neuronal cells, within each brain region. Invaginations of the nuclear lamina, with intranuclear inclusions of laminB1, were found in both non-demented control individuals, but were more frequent in cells of the deep cortical layers in postmortem cerebral cortex from individuals with FTD due to the MAPT IVS10+16 mutation ( Fig. 5A, B; Fig. S4). This was the case in both frontal and temporal cortex (Fig. 5B).
Our data from iPSC-derived FTD-MAPT neurons predicted 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 -11 -

Disrupted nucleocytoplasmic transport in FTD-MAPT neurons
Abnormalities of the nuclear lamina such as those reported here are also found in ageing 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 contained nuclear pores within these membrane folds, with nuclear pores (labelled by NUP98) co-localising with laminB1positive invaginations (Fig. 6A).
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 localisation/accumulation and cytoplasmic retention/nuclear exclusion within individual neurons (Fig. 6B & S5). Control iPSCderived neurons had discrete cellular distributions of each protein, with prominent nuclear RFP and cytosolic GFP (Fig. 6B). In contrast, localisation of NLS:RFP was altered in FTD-MAPT neurons such that there was a marked decrease in the nuclear:cytoplasmic RFP ratio (Fig. 6B, C). Conversely, nuclear exclusion of NES:GFP -12 -was reduced in FTD-MAPT neurons, with an increase of GFP within the nucleus (Fig.   6B, C). 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.

DISCUSSION
The cell and molecular biology of the pathogenesis of frontotemporal dementia due to MAPT mutations are 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 only represent 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, combined with MAPT mutations, to study the effects of those mutations on neuronal cell biology, finding that tau-mediated dementias lead to 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 neurons. We find that both missense and splicing mutations in MAPT alter tau protein localisation and phosphorylation within iPSC-derived neurons within four months in cell culture, recapitulating a well-described aspect of early FTD pathology in vivo (Götz et al., 1995;Hoover et al., 2010;Kowall and Kosik, 1987).
Mislocalisation 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.
-14 -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 mislocalisation of tau from axons to cell bodies and dendrites, and this key stage in disease progression is also an early event in iPSCderived models of FTD-MAPT. In vivo, mislocalisation of tau is typically associated with tau hyperphosphorylation (Götz 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 mislocalisation 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 which 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 (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 favours the -15 -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 mislocalisation 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 depolymerisation of microtubules. Tau has multiple roles in stabilising microtubules (Y. Wang and Mandelkow, 2015), 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 (Hernández-Vega et al., 2017), accumulation of tau in the cell body may also lead to a net increase in microtubule growth events and increased pushing forces on the nucleus by facilitating microtubule nucleation.
-16 -Alterations in nuclear shape and nuclear membrane function are a common feature of cellular ageing, 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 . 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 Microtubule deformation of the nucleus is a phenotype also seen in classic laminopathies such as the accelerated aging disorder Hutchinson-Gilford Progeria Syndrome (HGPS), where the primary defect is due to mutant lamin A/C protein (Capell 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 -17 -and Huntington's disease, have recently been reported (Freibaum et al., 2015;Grima et al., 2017;Jovičić et al., 2015;K. Zhang et al., 2015). Our finding here of disruption of the neuronal nuclear membrane as a consequence of MAPT mutations in frontotemporal dementia extends this pathogenic mechanism to dementias where 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.
Differentiation of iPSCs to cortical neurons was carried out as described, with minor modifications (Shi et al., 2012b;2012a). For nocodazole (Tocris) treatment, neurons were grown for 120 days in vitro (DIV) and compound was added at 33 µM for 4 hr before imaging.

Protein extraction and western blot analysis
Total cell protein was extracted using RIPA buffer (Sigma; R0278) supplemented with protease inhibitors (Sigma; 4693159001) and Halt phosphatase inhibitors (Thermo Scientific; 78420). Immunoblots were detected using LI-COR Odyssey CLx Infrared Imaging System and processed with the Image Studio Software (LI-COR).

Immunoprecipitation and mass spectrometric analysis of intracellular tau
Tau was immunoprecipitated from 1 mg of total protein extracted from iPSC-derived neurons (120 DIV) using a polyclonal anti-tau antibody (Dako Cytomation; A0024). Immunoprecipitated samples were analysed by western blot using a monoclonal tau antibody (MN1000; ThermoFisher Scientific) or stained with colloidal blue (Invitrogen; -19 -LC6025). Bands that corresponded to tau by western blot analysis were excised from the colloidal blue SDS-PAGE. Peptide masses of digested protein samples were determined using a Bruker ultrafleXtreme Maldi mass spectrometer in reflectron mode and ms/ms fragmentation performed in LIFT mode. Data analysis was with FlexAnalysis, BioTools and ProteinScape software (Bruker). Database searches of the combined mass fingerprint-ms/ms data were performed using Mascot (http://www.matrixscience.com).

Standard confocal images of fixed and immunostained cells were acquired with an
Olympus Inverted FV1000 confocal microscope and processed using Fiji software (Schindelin et al., 2012). 3D-STED imaging was performed on a custom built, dual color, beam scanning system with gated detection optically identical to the instrument described in (Bottanelli et al., 2016). For image analysis of colocalisation of tau and MAP2, Pearson's R correlation was calculated using the Coloc2 plugin for Fiji software (https://imagej.net/Coloc2). To quantify nuclear invaginations, at least 5 imaging fields from three independent experiments for genotype were analysed using a custom plugin for the Fiji. See Fig S3 for details.

Live imaging of microtubule dynamics
Neurons were grown to 120 DIV in individual µ-Dish 35 mm dishes (Ibidi; 80136) and transfected with a plasmid encoding for GFP-EB3 (Addgene plasmid # 56474). 48 hours after transfection, neurons were imaged live using a Leica SP5 microscope equipped -20 -with a controlled environment chamber (37 C; 5% CO2). Images were acquired at resonant scanning with a 63x objective (1frame/sec). Movies were analysed using the plusTipTracker software (Applegate et al., 2011).

Analysis of human post-mortem cerebral cortex
Human brain sections were obtained from the Queen's Square Brain Bank, Institute of Neurology, University College London. 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). Nuclear lamina invaginations were quantified after 3,3′-Diaminobenzidine (DAB) staining of LaminB1. Nuclei were scored from 20 randomly acquired imaging fields, from each post-mortem sample, by three observers blinded to sample identity. The percentage of nuclei with invaginations relative to total nuclei was calculated for each imaging field (see Fig. S4C). For fluorescent immunostaining of laminB1 and tau, following incubation with secondary antibodies and DAPI nuclear counterstain, images were acquired using an Olympus Inverted FV1000 confocal microscope .

Nucleocytoplasmic transport assay
Nucleocytoplasmic trafficking was analysed by infection of 120 DIV human iPSCderived neurons with the pLVX-EF1alpha-2xGFP:NES-IRES-2xRFP:NLS construct (Addgene #71396; Mertens et al., 2015). After 6 days, neurons were fixed and immunostained for β3-tubulin and GFP. Only cells positive for neuron-specific β3-tubulin were considered. The nuclear to cytoplasmic ratios of both GFP and RFP -21 -(nucRFP:cytRFP and nucGFP:cytGFP) were calculated separately using the integrated density of ROIs drawn within and outside the nucleus (see Fig.S5 for details). See also Figure S1.  Table S1.
See also Figure S2. 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 less than 10% of control neurons. 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; pair-wise comparison of the MAPT P301L line and its -36 -isogenic control was carried out using Student's t test; *=P<0.05; error bar represents s.e.m.; n = 3 independent experiments.