A cell autonomous regulator of neuronal excitability modulates tau in Alzheimer’s disease vulnerable neurons

Abstract Neurons from layer II of the entorhinal cortex (ECII) are the first to accumulate tau protein aggregates and degenerate during prodromal Alzheimer’s disease. Gaining insight into the molecular mechanisms underlying this vulnerability will help reveal genes and pathways at play during incipient stages of the disease. Here, we use a data-driven functional genomics approach to model ECII neurons in silico and identify the proto-oncogene DEK as a regulator of tau pathology. We show that epigenetic changes caused by Dek silencing alter activity-induced transcription, with major effects on neuronal excitability. This is accompanied by the gradual accumulation of tau in the somatodendritic compartment of mouse ECII neurons in vivo, reactivity of surrounding microglia, and microglia-mediated neuron loss. These features are all characteristic of early Alzheimer’s disease. The existence of a cell-autonomous mechanism linking Alzheimer’s disease pathogenic mechanisms in the precise neuron type where the disease starts provides unique evidence that synaptic homeostasis dysregulation is of central importance in the onset of tau pathology in Alzheimer’s disease.


Introduction
Early stages of neurodegenerative disorders are characterized by the aggregation of ubiquitous proteins in discrete populations of brain cells and the degeneration of these cells.For most diseases this selective vulnerability pattern is unexplained, yet it could yield major insight into pathological mechanisms.Alzheimer's disease (AD), the world-leading cause of dementia, is defined by the appearance of two hallmark pathological lesions, amyloid plaques [extracellular aggregates of amyloid-β (Aβ) peptides] and neurofibrillary tangles (intracellular aggregates of hyperphosphorylated tau, or NFTs).While plaques are widespread in the neocortex and hippocampus, NFTs follow a well defined regional pattern that is commonly used to stage the disease and that starts in principal neurons from layer II of the entorhinal cortex (ECII) during prodromal AD. [1][2][3] .
Neuroimaging studies have shown that NFTs are the feature that best correlates with regional cortical atrophy 4 and clinical symptoms of the disease, 5,6 suggesting a central role of tau in AD pathogenesis.The mechanisms responsible for the progression of neurodegeneration from the entorhinal cortex (EC) to regions that are affected later in the disease remain to be understood, but the propagation of pathological tau along axonal routes has recently gained attention. 7,80][11] However, the LC may present tau seed propagation activity only after the EC. 12,13ECII neurons might thus form pathological tau de novo at the onset of the disease and could be a reservoir for tau spread to other regions.Understanding why ECII neurons start accumulating tau during the earliest stages of AD could therefore reveal a major intervention point for disease-modifying drugs.
We recently used high quality molecular signatures of ECII neurons, along with a large compendium of genomics data to build maps of functional gene interactions within the context of ECII neurons. 14Contrasting ECII neurons with neurons more resistant to AD, we found that pathways related to microtubule remodelling were particularly salient in ECII neurons.To investigate AD processes in the specific context of ECII neurons, we used the NetWAS 2.0 (Network-Wide Association Study 2.0) algorithm, which leveraged our ECII functional map along with genome wide association study data for NFT formation. 14,15NetWAS 2.0 reprioritized genes based on their association with tau pathology within vulnerable neurons and identified four gene modules-groups of genes with predicted common involvement in biological pathways-potentially contributing to NFT formation in AD.Of of these four modules, one displayed higher connectivity, i.e. higher predicted functional interaction in common cellular processes, in ECII neurons than in any other neuron type, suggesting it might represent cellular processes that are more central in ECII neurons than in other neurons.Furthermore, genes within this module are involved in axonal plasticity and affected by two main AD drivers: ageing and amyloid pathology. 14 the present work, we hypothesized that this module could yield major insight into the cascade cell autonomously leading to tau pathology in ECII neurons in AD and searched for core regulators that perturb it.This led to the identification of the protooncogene DEK as a hub gene of this module.
Establishing a roadmap for the identification of neuron-type specific regulatory processes, we first experimentally validate in ECII neurons the predictions of our ECII-specific in silico models.We thus demonstrate that DEK is indeed a central regulator of the vulnerability module.We further show that perturbations in this module cause tau protein accumulation and somatic redistribution in EC neurons in vitro and in vivo.This is accompanied by epigenetic dysregulation of immediate early gene (IEG) induction, pathological alterations in neuronal excitability and microglia reactivity.Taken together, we give here proof-of-principle that our vulnerable neuron-centric systems-level approach can yield novel drivers of tau accumulation, and identify a link between several pathological processes occurring during incipient AD.

Mouse models
ECII-bacTRAP mice were previously generated in our laboratory. 14hey are transgenic for the BAC #RP23-307B16 where a cDNA encoding eGFP-L10a (enhanced green fluorescent protein fused to the L10a subunit of the ribosome) was integrated before the start codon of Sh3bgrl2.As a result, they express eGFP-L10a under the control of the regulatory regions of the ECII enriched Sh3bgrl2 gene, which leads to specific expression of eGFP-L10a expression in ECII neurons.Human-tau transgenic mice (hMAPT mice) 16 were obtained from Jackson [B6.Cg-Mapt tm1(EGFP)Klt Tg(MAPT) 8cPdav/J, Strain #005491].Experiments were performed in 12-15 month old hMAPT mice and 6-8 month old ECII-bacTRAP mice.The groups were age-matched for each experiment.
Mice were maintained on a 12 h dark/light cycle and provided with food and water ad libitum.All the mouse experiments were approved by the Rockefeller University Institutional Animal care and Use Committee (IACUC protocols #16902 and 19067-H) and Karolinska Institutet (4884-2019) ethical committee.

AAVs transduction
EC neurons in primary culture at 7 DIV were transduced by a full media change to media containing either control, mDEK-IRES or mDEK-shRNA AAVs at a concentration of 1 × 10 10 genome copies per ml of media (GC/ml).Neurons were maintained in viruscontaining media until the day of the experiment.

KCl stimulation
EC neurons were stimulated by adding either vehicle (PBS) or 50 mM KCl to the culture media.Neurons were incubated with KCl for the whole duration of the experiment.

Epigenomics analysis
Mass spectrometry analysis of histone post-translational modifications (Mod-spec) and H3K36ac cleavage under targets and tagmentation (CUT&Tag) were performed by Activ Motif on primary EC neurons transduced with control or Dek-silencing AAVs (Supplementary material, 'Methods' section).

Resting membrane potential
Resting membrane potential (RMP) was determined in current-clamp mode after breaking the membrane and calculated from the mean of 1 min recording using Clampfit11.2.

Cell capacitance
Cell capacitance (Cm) and Tau (τ) values were taken directly from the Clampex membrane test tool and amplifier readings.

Rheobase
First-spike latency measurements were done by applying a rheobase protocol to PCs in current-clamp whole cell configuration to elicit the firing of a single action potential (AP) (10 pA increments, 300 ms).The current at which the first AP was fired was defined as current threshold.First-spike latency was then calculated as the time between the beginning of the test current pulse and the maximum amplitude of the AP.

Sag
To measure the absolute difference between the steady state decrease in the voltage and the largest decrease in the voltage following a hyperpolarizing step (sag), a hyperpolarizing current pulse was applied to primary cultures (Vh = −60 mV).Amplitude was calculated from the steady state of the current to the peak.

Action potential waveform
Long-lasting current steps to evoke one AP were used to analyse AP properties (10 pA increments, 700 ms).AP parameters [amplitude, firing threshold, half-width, maximum rise slope (MRS), time to MRS, maximum decay slope (MDS) and time to MDS], AP waveform plot and analysis were performed in Clampfit11.2.

Spontaneous excitatory postsynaptic currents
sEPSCs were detected off-line using MiniAnalysis software (Synaptosoft, Decatur, GA, USA).Charge transfer, event amplitude and event frequency were analysed using Microsoft Excel (Microsoft Office) and GraphPad Prism with the result representing average values taken over 1 min periods.

Stereotaxic injections
AAVs were injected in the EC of adult mice using an Angle Two mouse stereotaxic frame with a motorized nanoinjector (Leica).Animals were anaesthetized with xylazine (4.5 mg/kg body weight) and ketamine (90 mg/kg body weight) injected peritoneally.An ophthalmic ointment was applied to the anaesthetized animals to prevent corneal drying during the procedure.AAVs were loaded in a 10 μl syringe (#7653-01, Hamilton) with a 33-gauge needle (#7803-05, Hamilton).AAVs (2 μl, 1 × 10 13 GC/ml) were injected in the mouse EC (AP: −3.70; ML: −4.65; DV: −4.50) with the nanoinjector tilted −4.05°.The same volume and concentration of control AAVs were injected in the contralateral EC (AP: −3.70; ML: + 4.65; DV: −4.50) with a nanoinjector tilt of +4.05°.The injection rate was 0.4 μl/min.Bacitracin antibiotic gel was applied to the surgery wound, which was sutured with a non-absorbable monofilament.To compensate for fluid loss, warm sterile saline solution was injected intraperitoneally (3% of the body weight) and animals were kept on a heated pad and monitored until complete recover from anaesthesia.In cohorts with injection of Dek or control AAVs on contralateral sides, the hemisphere receiving the control AAV was alternated for each mouse to minimize potential confounders.The experimental unit was one EC hemisphere.In cohorts where Dek or control AAVs were injected bilaterally in different mice, mice receiving Dek or control AAVs were alternated.The experiment unit was one mouse (two pooled EC hemispheres).In all cases, the experimenter processing tissue was blinded to the AAV injected in each hemisphere or mouse.No mouse was excluded from the study after injection.The outcome measures were the number of surviving ECII neurons, the levels of Iba-1 around layer II and the total levels of phospho-tau.

Primary culture RNAseq
For RNA sequencing (RNAseq) analysis of EC primary cultures, RNA was extracted 4 days after AAV transduction using the Purelink RNA kit.RNA (200 ng) was used to generate RNAseq libraries using the Truseq RNA sample prep kit.Libraries were sequenced using a NextSeq 500 Illumina sequencer.

PLX5622 treatments
PLX5622 compound was obtained from MedChemExpress.Chow containing PLX5622 at 1200 ppm was manufactured by a trained diet preparation operator at Research Diets Inc in AIN-76A rodent diet.Animals were supplied with AIN-76A diet pellets either alone (control) or containing PLX5622 ad libitum.Diet supplementation was initiated 5 days before stereotaxic injections were performed and was maintained for the whole duration of the experiments.

Analysis of mouse tissue samples by western blot and RT-qPCR
In order to reduce the number of animals used in this study and to have a consistent representation of the region of interest and injected area for both western blot and RT-qPCR analysis, dissected ECs were initially subjected to a fast homogenization in 100 μl RNAse free H 2 O supplemented with protease (ThermoFisher), phosphatase (PhoSTOP, Merck) and RNAse (40 U/ml RNAsin, Promega and 20 U/ml Superasin, ThermoFisher) inhibitors with a cordless motor pestle homogenizer.Immediately after, 45 μl of the homogenate was transferred to an RNAse-free tube, mixed with RNA extraction buffer and cleaned up using the Purelink RNA kit (see below).The remaining volume was lysed in RIPA protein extraction buffer (ThermoFisher).

Immunofluorescence analysis
Mice were anaesthetized with nembutal and perfused with PBS followed by 4% paraformaldehyde (PFA).Brains were then dissected and post-fixed in 4% PFA for 1 h.Afterwards they were washed with PBS and cryopreserved by incubation in increasing sucrose concentrations (5%, 15% and 30%).They were then embedded in optical coherence tomography (OCT) compound (TissueTek), cut in 40 μm-thick free-floating horizontal sections with a CM3050 Cryostat (Leica) and further preserved in cryoprotectant solution (50% ethylene glycol, 20% glycerol in PBS) at −20°C until further analysis.

Immunofluorescence
EC neurons in primary culture were seeded in poly-D-lysine-coated round glass coverslips.After the treatments, they were fixed in 4% PFA for 10 min.Fixed cells were then blocked for 30 min in 1% bovine serum albumin (BSA), 0.1% Triton X-100, PBS (ThermoFisher) and incubated with the primary antibody overnight at 4°C.The antibodies used were AT8 (1:1000, ThermoFisher) and MAP2 (1:1000, Abcam).The day after, the cells were washed with PBS and incubated with the secondary antibody in PBS for 1 h at room temperature.The secondary antibodies used were Goat antimouse IgG (H + L) cross-adsorbed Alexa fluor 488 (ThermoFisher) and Goat anti-rabbit IgG (H + L) cross-adsorbed Alexa fluor 594 (ThermoFisher).DAPI was used as nuclear staining.Cells were then washed and mounted with ProLong Gold Antifade Reagent (ThermoFisher).

Image acquisition
All images were acquired on a Zeiss LSM 510 META laser scanning confocal microscope.

Iba1 and AT180 fluorescence intensity quantification in mouse sections
A minimum of three pictures were taken from different sections and from each hemisphere from the same mouse (shDEK and Control transduced).The region of interest was established manually by selecting the ECII and on slides where mCherry positive cells (AAV transduced) could be detected.Iba1 and AT180 fluorescence intensity was determined using the threshold selector to identify positive staining signal.

Entorhinal cortex neurons in primary culture
RNA extraction, purification, reverse-transcription and qPCR were performed using the TaqMan Fast Advanced Cells-to-CT kit (ThermoFisher), according to the manufacturer's instructions.

Mouse tissue homogenates
RNA was purified using the PureLink RNA mini kit with in-column DNAse-I digest (Qiagen) following the manufacturer's instructions (ThermoFisher).RNA was then quantified in a nanodrop and ∼400 ng of RNA was reverse-transcribed using the SuperScript III First-Strand Synthesis System with a 1:1 mix of oligo dT and random hexamers oligonucleotides (ThermoFisher), followed by RNAse H digestion. cDNA was then used to run qPCRs using TaqMan Universal PCR master mix.

Network connectivity analysis
To identify genes that are strongly connected to the AD vulnerability module, for each gene g in the ECII functional network, we calculated a z-score for vulnerability module connectivity: , where x g is the mean of posterior probabilities between g and all genes in the AD vulnerability module, and n is the number of genes in the module.μ and σ are the mean and standard deviation, respectively of posterior probabilities of all genes in the network to gene g.Thus, high z connectivity represents stronger connectivity between gene g to the AD vulnerability module than expected based on gene g's general connectivity patterns within the ECII network.

Differential gene expression analysis
Following sequencing, adapter and low quality bases were trimmed by fastp 18 from the raw sequencing files in FASTQ format.Cleaned reads were aligned to the mm10 reference genome using STAR version 2.7.1a. 19 After alignment, the reads per kilobase of transcript per million mapped reads (RPKM) for all genes in each sample were calculated with R package edgeR. 20To analyse differential gene expression between samples, DESeq2 21 was used, applying the standard comparison mode between two experimental groups.

Variant effect prediction
Single nucleotide polymorphisms (SNPs) in linkage disequilibrium (LD) with rs145578678 in populations with European ancestry (based on 1000 Genomes phase 3 data collected from the Finnish populations in Finland, British populations in England and Scotland, Iberian populations in Spain, Toscani populations in Italy, and Utah residents with Northern and Western European ancestry) were obtained from Ensembl (release 110) 22 and analysed with DeepSEA using the Beluga model (https://hb.flatironinstitute.org/deepsea/, hg38 Genome assembly).Only predictions with P-value < 0.05 after Bonferroni correction were considered.

Mass spectrometry analysis of protein abundance
Protein identification and quantification was performed by liquid chromatography-tandem mass spectrometry by the Proteomics Biomedicum Core Facility at the Karolinska Institutet (Supplementary material, 'Methods' section).

Data-driven identification of a tau pathology driver in ECII neurons
To determine drivers of tau pathology in ECII neurons, we honed in on the AD vulnerability module identified by the aforementioned NetWAS 2.0 approach. 14We calculated the connectivity of every gene in the ECII functional network to genes in the AD vulnerability module (Supplementary Table 1).Intuitively, high connectivity to genes in the AD vulnerability module indicates a strong functional relationship as a potential regulator of this module.The gene with the highest connectivity to this module was DEK, suggesting that it might be a core regulator of tau pathology within vulnerable neurons.Moreover, in the ECII neuron functional network, DEK was also predicted to be one of the closest functional interaction partners of MAPT, the gene that codes for tau (connectivity score 0.976, ranked 20 out of 23 950; data publicly available at alz. princeton.edu)(Supplementary Table 1), further indicating that DEK might be key to ECII vulnerability.
DEK codes for the proto-oncogene DEK, a nuclear phosphoprotein that was identified for the first time in patients with acute myeloid leukaemia 23 and is frequently upregulated in solid tumours. 24t has also been associated with several autoimmune diseases. 25EK is ubiquitously expressed in the CNS (Supplementary Fig. 1), although its role there remains largely unknown.We therefore aimed to characterize DEK function in ECII neurons, and its potential contribution to pathological processes leading to AD.

DEK regulates vulnerability-associated gene module in the entorhinal cortex
We modulated Dek expression in mouse EC neuron primary cultures by transducing them with AAVs carrying either a cDNA coding for Dek, a silencing small hairpin RNA (shRNA) directed against Dek, or their respective controls.Transduction with these viruses led to a very significant overexpression and silencing of Dek, respectively [overexpression log2FC = 1.79, false discovery rate (FDR) = 4.33 × 10 −80 ; silencing log2FC = −2.18,FDR = 1.55 × 10 −58 ) (Fig. 1B, Supplementary Fig. 2C and Supplementary Table 2).This was accompanied by a significant downregulation of DEK protein levels in Dek-silenced neurons in vitro (Supplementary Fig. 4B).RNAseq followed by differential gene expression (DGE) analysis 4 days after transduction showed that Dek silencing has larger effects on the transcriptional landscape of the cells than DEK overexpression (Fig. 1A and B and Supplementary Fig. 2B and 2C).To study DEK function in vivo, we silenced Dek in ECII neurons of ECII-bacTRAP (bacterial artificial chromosometranslating ribosome affinity purification) mice previously generated in our laboratory (Supplementary Fig. 2A).ECII-bacTRAP mice permit immunoprecipitation and RNAseq profiling of actively translated mRNA from these neurons as previously described. 14,17erturbing Dek expression in the EC of these mice thus allows us to measure ECII-specific gene expression regulation by DEK.BacTRAP-RNAseq analysis 1 week after AAV injection confirmed the successful silencing of Dek in ECII neurons (log2FC = −0.78,FDR = 0.039) (Fig. 1E and Supplementary Table 2).
We used these gene expression data to test the accuracy of the in silico ECII functional network.Genes predicted to be highly functionally connected to DEK in ECII neurons were indeed more affected by Dek silencing in primary EC neurons in vitro [P-value < 2.2 × 10 −16 , one-sided Wilcoxon rank sum test of DEGs FDR < 0.05 (Supplementary Fig. 3) and ECII neurons in vivo (P-value = 2.89 × 10 −11 ), one-sided Wilcoxon rank sum test of DEGs FDR < 0.05, Fig. 1G] than expected by chance.These results confirm the predictive power of our functional network to reveal novel celltype specific hub genes.We next tested whether DEK indeed regulates genes from the AD vulnerability module identified by NetWAS 2.0. 14Genes in the vulnerability module were indeed more differentially expressed after Dek silencing in vivo than expected by chance (P-value = 1.76 × 10 −07 , Fisher's exact test), thereby validating DEK as a regulator of this module, with the potential to affect pathological mechanisms in ECII.

DEK epigenetically regulates the induction of Egr1 expression
DEK has been previously identified as a chromatin-binding protein that plays a variety of roles in the regulation of chromatin structure and gene expression. 26,27To explore whether the effects observed on neuronal function are initiated by alterations in chromatin accessibility, we performed ATAC-seq (assay for transposaseaccessible chromatin using sequencing) in primary cultures of EC neurons 4 days after Dek silencing.Our results showed differences in chromatin accessibility at the promoter regions of a total of 20 genes (Supplementary Table 2).These alterations were accompanied by differential expression of 13 of these genes, determined by our previous in vitro RNAseq analysis (Fig. 2A).Interestingly, as shown in Fig. 2A, the IEG Egr1 was downregulated (FDR = 4.12 × 10 −48 ), with the largest fold change (log2FC = −1.85)among these genes.Furthermore, as shown in our bacTRAP-RNAseq data (Supplementary Table 2), Egr1 was the only gene that was also differentially expressed (downregulated, log2FC = −0.81) in Dek-silenced ECII neurons 1-week post-transduction in vivo (FDR = 0.013).Counterintuitively, the downregulation of Egr1 was accompanied by increased chromatin accessibility at its proximal enhancer and promoter regions (FDR = 0.0409) (Fig. 2B).
9][30] .To test the effect of DEK on Egr1 expression dynamics, we used 50 mM KCl to cause neuronal depolarization and calcium influx in a manner that is independent of synaptic transmission.We then monitored the expression levels of Egr1 over time using quantitative PCR.As expected, KCl caused a large transient increase in Egr1 expression in control neurons.Strikingly though, KCl led to a 2.5-fold higher induction of Egr1 expression in EC neurons where Dek was silenced compared to control (Fig. 2C).This effect is likely mediated by the higher chromatin accessibility at the proximal enhancer and promoter regions of the gene and suggests that the downregulation of Egr1 at baseline is due to secondary alterations in neuronal activity.Indeed, secondary transcriptional changes follow to maintain homeostasis in the network since we observed decreased expression of all AMPA receptors: Gria2 (FDR = 1.68 × 10 −52 , log2FC = −1.22),Gria3 (FDR = 6.91 × 10 −9 , log2FC = −0.57),Gria1 (FDR = 4.29 × 10 −7 , log2FC = −0.43)and Gria4 (FDR = 1.23 × 10 −4 , log2FC = −0.35)(Fig. 2D), likely curbing intrinsic neuronal activity.As a consequence, we observed a significant overlap between genes differentially expressed after Dek silencing and a published set of rapid primary response IEGs, 31 most of which were downregulated in Dek-silenced neurons (Fig. 2E).Out of all these IEGs, Egr1 was the top downregulated gene (FDR = 4.12 × 10 −48 ) (Fig. 2F).
Epigenetic modifications at the cis regulatory elements of IEGs are essential to regulate their expression. 32,33A number of studies showed a role for DEK as a histone chaperone and modulator of histone modifications, especially by regulating the function of several histone acetyltransferases and deacetylases. 34,35To determine specific histone marks that could be modulated by DEK in an unbiased manner, we conducted a mass spectrometry-based screen of histone post-translational modifications in control and Dek-silenced EC neurons in vitro (Supplementary Table 2).The assay revealed a striking >10-fold increase in H3.1K36ac in Dek-silenced neurons compared to control (Fig. 2G).7][38] To test if the increased chromatin accessibility at the Egr1 locus after Dek silencing could be mediated by H3.1K36ac, we performed a CUT&Tag assay using an antibody against H3K36ac.We found no differences in H3K36ac deposition around the Egr1 locus (Supplementary Table 2) in Dek-silenced neurons compared to control.As mentioned before, currently available antibodies cannot discriminate between H3.1K36ac and the more abundant H3.3K36ac (six times higher levels than H3.1K36ac in our neuron cultures and not changed by Dek silencing), which likely masks H3.1K36ac-specific changes.It is therefore still possible that this histone mark is responsible for the effect on Egr1 expression.Moreover, a reader protein of H3K36ac, Setd7 39 is very significantly induced both in vitro and in vivo after Dek silencing (log2FC = 1.69,P adj = 1.2 × 10 −82 ; log2FC = 0.58, P adj = 0.012, respectively), suggesting an important role of this histone mark in mediating DEK function.

Dek-silenced entorhinal cortex neurons show altered intrinsic excitability
The modulation of IEGs by DEK could have profound consequences for neuronal function.To test this hypothesis, we performed patchclamp recordings in Dek-silenced EC neurons in primary culture 4 days after transduction.Evaluation of the RMP showed that Dek-silenced neurons were significantly hypopolarized compared to controls (Fig. 3A and B).Concurrently, the rheobase (minimal current required to elicit an action potential), was lower, suggesting increased excitability (Fig. 3C).Decreased membrane input resistance and increased firing threshold (Fig. 3D and E) confirmed alterations in the neuronal intrinsic properties (Supplementary Table 3).These results suggest a functional change in Na + currents.The conductance of voltage-gated sodium channels can be further evaluated by determining the kinetics of the AP initiation. 40,41We did not find alterations in the AP waveform (Fig. 3F).However, in agreement with alterations in voltage-gated sodium channels, we observed a lack of the first spike initiation, or 'kink' (Fig. 3G) and a trend for a  decrease in the AP MRS (Fig. 3H).Likely underlying these changes, the expression of voltage-gated sodium and potassium channel subunits was broadly dysregulated (Fig. 3I-N and Supplementary Fig. 4A).
Last, we found a significant increase in the sag amplitude (Fig. 3O and P), the difference between initial and steady state decrease in voltage after a hyperpolarizing current step, a key contributor to the RMP. 42,43Taken together, these data strongly indicate that DEK can cell-autonomously regulate ECII neuron excitability.5][46] Our ECII functional network also predicted a strong connection between DEK and tau.We therefore hypothesized that DEK might be a molecular bridge between synaptic homeostasis and tau proteostasis, and that, like at early stages of AD, Dek-silenced neurons might have elevated tau in addition to electrophysiological abnormalities.

Dek silencing leads to tau accumulation, microglia reactivity and ECII neuron loss
To assess tau accumulation, we measured total tau protein levels and its phosphorylation by western blot at different time points following Dek silencing in EC neurons in primary culture.As shown in Fig. 4A, Dek silencing led to increased tau levels 4 days after transduction, accompanied by its redistribution to the somatodendritic compartment (Fig. 4C), an early event on AD progression. 47,48hese alterations were not accompanied by increased tau phosphorylation at Thr231, generally associated with early AD pathology, or at Ser202/Thr205, associated with later stages of AD pathology 49,50 (Supplementary Fig. 4C ad D).But a significant decrease in phosphorylation at another early AD epitope, Thr181 (Supplementary Fig. 4E) suggested a subtle imbalance in the phosphorylation profile of tau.RT-qPCR analysis showed no differences in tau mRNA levels (Fig. 4B), suggesting that Dek silencing triggers tau accumulation by modulating its translation, release or degradation.Pathway analysis on our in vitro RNAseq data showed autophagy alterations in Dek-silenced primary EC neurons (Fig. 1C).To test if Dek silencing indeed affects autophagy in these neurons, we performed western blot analysis of the protein LC3, the lipidation of which initiates autophagosome formation.Dek-silenced neurons showed increased levels of the lipidated form LC3-II, which supports the notion of an altered autophagic flux (Fig. 4D).Interestingly, Dnajc6, which codes for auxilin, a protein necessary for lysosomal degradation of autophagosome content with mutations associated with Parkinson's disease, 51 was among the most significantly downregulated genes in Dek-silenced EC neurons both in vitro and in vivo (log2FC = −2.1,P adj = 4.9 × 10 −145 ; log2FC = −0.6,P adj = 0.008, respectively).Mass spectrometry proteomics analysis of Dek-silenced neurons in vitro further showed a significant decrease of auxilin protein abundance in these cells (log2FC = − 2.25, P-value = 0.003), confirming its regulation by DEK (Supplementary Table 2).
To test whether these alterations also lead to tau pathology in ECII neurons in vivo, we performed stereotaxic injections of a control and a Dek silencing AAV in ECs in hMAPT mice.This mouse model is knocked-out for mouse tau while expressing the full length human MAPT gene. 16Immunofluorescence labelling of ECII neurons using the Reelin marker at 4 days, 1 week and 2 weeks after injection (Supplementary Fig. 5A) revealed a highly significant loss of these neurons after 2 weeks (Fig. 5A).Iba1 immunofluorescence staining showed increased reactive microglia specifically in the layer II of the EC at 2 weeks post-injection (Fig. 5A).At 1-week post-injection (before neuron loss), we found no differences in tau protein levels in ECII by immunofluorescence against phospho-tau (Thr231) (Supplementary Fig. 5B).Accumulation of reactive microglia around transduced neuron cell bodies (Supplementary Fig. 5C) suggested that microglia might contribute to the neuron loss observed at 2 weeks post-injection.Indeed, RT-qPCR on EC lysates showed increased expression of the microglia markers Cx3cr1, Itgam (that codes for CD11b) and complement C1qa (associated with synaptic pruning) 52,53 (Supplementary Fig. 5D-F).We also found an increase in the disease-associated microglia genes Tyrobp and Spp1 54,55 (Supplementary Fig. 5G and H).Considering that our previous data suggest that Dek-silenced neurons exhibit increased excitability, we measured the levels of Ccl3, a cytokine shown to be upregulated in microglia in response to neuronal activation 56 (Supplementary Fig. 5I).The significant upregulation of Ccl3 suggested that ECII hyperexcitability contributes to microglial reactivity.
To test the possibility that a longer time is necessary for progressive build-up of tau, we eliminated microglia to increase ECII neuron survival.We fed mice with chow containing PLX5622, an antagonist of colony-stimulating factor 1 receptor (CSF1R), that leads to microglia depletion after 5 days. 57At Day 5, we injected control or shDek AAVs in the EC of opposite hemispheres and analysed ECII neuron survival and tau accumulation at 2 weeks postinjection (Supplementary Fig. 6A).As shown in Supplementary Fig. 6B and C, PLX5622 successfully ablated microglia and rescued the loss of Dek-silenced ECII neurons.We next compared the somatic levels of phospho-tau (Thr231) in layer II of the EC between control and Dek silenced entorhinal cortices in PLX5622-treated mice and found a significant accumulation of tau following Dek silencing (Fig. 5B).

Dysregulation of DEK in human Alzheimer's disease
To test if DEK is dysregulated at the transcriptional level in early AD in EC neurons, we consulted a recently published single-nucleus RNAseq dataset, but this study did not detect differential expression of DEK.We then mined the NHGRI-EBI GWAS catalogue for genetic variations in DEK that could be associated with AD.Interestingly, an SNP in a DEK intron, rs145578678, was recently found to have a suggestive association with the rate of cognitive decline in AD (P-value = 3 × 10 −6 ). 58To assess the functional effect that this putative disease-associated allele may have, we used DeepSEA, a deep learning-based algorithm that predicts the chromatin effects of sequence alterations, 59 using rs145578678 and SNPs in LD with it as input.After filtering for observations specific to brain, H3K36me3 was predicted to be increased in the hippocampus for the disease-associated allele of rs182353546, within an intron of DEK (Supplementary Table 2).H3K36me3 and H3K36ac are mutually exclusive. 36,60As H3K36ac is the histone modification most affected by DEK, it is likely that, in a feedback loop, DEK regulates H3K36ac deposition in its own loci to regulate its transcriptional level.The DEK allele putatively associated with cognitive decline in AD might therefore improperly regulate DEK levels and perturb proper synaptic homeostasis and tau regulation.

Discussion
The molecular cascade by which ADdrivers and genetic susceptibility factors result in early tau pathology in specific neuronal subpopulations remains to be understood.In the present work, we set out to identify these molecular processes by using a cell type-specific functional genomics approach.2][63] It is, thus likely to identify genes that are indeed regulating tau during the natural course of human AD within ECII neurons, especially during early disease stages.
By using this approach, we take a significant step towards understanding the neuronal arm of AD pathogenesis: we identify perturbations in DEK as a novel molecular bridge between a wide range of processes that take place during early stages of AD: dysregulation of neuronal excitability, tau accumulation and alterations in microglia activity (Supplementary Fig. 7).While we did not find altered DEK expression in neurons in a recent single cell transcriptomics study of the EC in human AD, 54 its function can be regulated by post-translational modifications 64 and by degradation. 65Further studies will be necessary to assess DEK activity levels during prodromal AD, as well as to elucidate the functional impact of genetic variation in DEK, like the rs145578678 polymorphism.It will also be important to validate the cell type specificity of the pathological cascade mediated by ubiquitously-expressed DEK and determine if DEK also regulates tau accumulation in LC or in hippocampal neurons.Whether the perturbation of DEK activity by prodromal AD or its downstream consequence occur in a cell type-specific manner is an open question.Here we perform an in-depth characterization of DEK function in vulnerable neurons and report that its silencing potentiates activity-regulated gene expression.At steady state, this results in decreased IEG expression, likely due to homeostatic changes aimed to normalize what is sensed as aberrant network activity following chronic upregulation of IEGs.The existence of these compensatory events is supported by our data showing a downregulation of all AMPA glutamate receptors subunits that would, in turn, decrease IEG expression.Together with the perturbation of genes involved in microtubule dynamics, this suggests that DEK regulates synaptic homeostasis in ECII neurons, a central process in their associative memory-forming function.
Our results further show that the alterations in neuronal activity mediated by Dek silencing trigger neuron-to-microglia signalling mechanisms that steer microglia towards a pro-inflammatory and phagocytic state that contributes to neurodegeneration, as demonstrated by the rescue of neuron loss by microglia depletion.Microglia reactivity preceded accumulation of pThr231-tau in vivo.Although we cannot exclude that neuronal accumulation of another tau proteoform is responsible for altering microglial state, our data indicate that it might be due, instead, to the detection of other neuronal dysfunction, like excessive firing.This may represent an extreme case of the spine phagocytosis described after neuronal hyperactivation. 66In agreement with this notion, we found elevated levels of Ccl3, a cytokine upregulated by microglia in response to neuronal activity 56 and of known mediators of synaptic spine loss (Itgam, C1qa). 52,53,67,68While we also found increased expression of Tyrobp and Spp1, both upregulated in disease-associated microglia in AD, 54,55,69 a more comprehensive analysis of gene expression changes at single cell resolution will be necessary to determine which microglia states are triggered by Dek silencing in ECII neurons.This neuron-glia crosstalk regulated by DEK in neurons might take centre stage during early AD pathology in the EC.Follow-up studies to identify the neuronal signal regulated by DEK that drives the elimination of hyperexcitable ECII neurons, will be of great relevance, as these signalling molecules may serve as potential therapeutic targets for early intervention in AD.
1][82][83][84][85] However, the connection between AD lesions and the disruption of synaptic homeostasis in the entorhinal-hippocampal network has remained elusive.It is possible that the regulation of tau levels is an integral part of a physiological programme of structural and synaptic plasticity.Indeed, a recent transcriptional analysis of NFT-bearing neurons in human AD at single-cell resolution suggested that the build-up of NFT is a cell type-specific response to alterations in synaptic function. 86Improper fine-tuning of these crucial mechanisms for ECII function might contribute to the onset of a deadly cascade in ECII neurons.This notion is supported by the progressive tau build-up that we observe in Dek-silenced neurons both in vitro and in vivo.In vitro, tau accumulation and proteoform imbalance is accompanied by an increase in LC3 lipidation suggestive of alterations in the autophagy flux, and a downregulation of Dnajc6/auxilin both at the transcriptional and protein levels, which might cause an impairment in autophagosomal content degradation.While these data point to autophagy alterations as contributing factor to the build-up of tau in Dek-silenced neurons, other mechanisms could further participate in this process.For instance, neuronal activity has been shown to affect tau translation, degradation and secretion, 83,85,87,88 suggesting that the electrophysiological phenotype caused by impaired DEK signalling might contribute to tau accumulation through different mechanisms that will need to be explored in future studies.0][91][92] According to our data, if ECII neurons adjust DEK activity to re-establish proper excitability, this might consequently lead to impairments in tau proteostasis.Understanding how this novel axis is affected by amyloid accumulation and ageing will be crucial to dissect the molecular steps that take place specifically in these neurons during prodromal AD.

Figure 1
Figure 1 Gene expression changes in EC neurons triggered by downregulation of Dek in vitro and in vivo.(A) Heat map of the differentially expressed genes (DEGs) between control (CONT) and Dek-silenced primary neurons (n = 3 wells for each group).(B) Volcano plot of DEGs between control and Dek-silenced primary neurons.(C) Pathways that are altered in Dek-silenced primary neurons compared to control.(D) Heat map of the DEGs between control and Dek-silenced ECII neurons in vivo (n = 5 mice per group).(E) Volcano plot of DEGs between control and Dek-silenced ECII neurons in vivo.(F) Pathways that are altered in Dek-silenced ECII neurons in vivo compared to control.(G) Enrichment analysis of DEGs [false discovery rate (FDR) < 0.05, denoted in the rug plot, bottom] between control and Dek-silenced ECII neurons in vivo demonstrates strong enrichment within genes ranked by probability of functional interaction with DEK in the EC network (blue).Dashed line represents expected enrichment if given random predictions.(H) Venn diagram that shows a significant overlap of up-and downregulated DEGs between Dek-silenced neurons in vitro and in vivo.Fisher's exact test P-value < 0.001.EC = entorhinal cortex; sh = small hairpin.

Figure 2
Figure 2 Effect of Dek silencing on chromatin accessibility, immediate early gene expression and histone marks.(A) Volcano plot of the differentially expressed genes (DEGs) in vitro that also showed alterations in chromatin accessibility at their loci (n = 3 wells per group).(B) Representative ATACseq tracks of Dek-silenced (yellow), control (CONT) neurons (blue) and their overlay.(C) Expression levels of Egr1 after 1, 2, 3 and 4 h of induction with 50 mM KCl in control and Dek-silenced neurons.Two-way ANOVA Sidak's multiple comparisons test P-value = 0.0006 (1 h), 0.0001 (2 h) (two independent experiments with three replicates per time point and group; additional experiments with consistent results were performed).(D) AMPA receptor subunits expression levels in control and Dek-silenced neurons in vitro.P-value = 0.0070 (Gria1), P < 0.0001 (Gria2), P = 0.0036 (Gria3), P = 0.0446 (Gria4).(E) Venn diagram of the overlap between rapid primary response immediate early genes (IEGs) and DEGs in Dek-silenced neurons in vitro.Fisher's exact test P-value = 0.0095.(F) Volcano plot of all the IEGs that are differentially expressed in Dek-silenced neurons compared to controls.(G) Relative abundance of acetylated histone marks in control and Dek-silenced neurons (n = 1 well per group).FDR = false discovery rate; RPKM = reads per kilobase of transcript per million mapped reads.

Figure 3
Figure 3 Electrophysiological properties of Dek-silenced EC primary neurons.(A) Representative traces of first spike and resting membrane potential from: control (CONT, blue) and Dek-silenced neurons (yellow) (n = 10-11 neurons per group).(B) Bar graph of the resting membrane potential in control versus Dek-silenced neurons.Unpaired t-test P-value = 0.023.(C) Bar graph of the rheobase in control versus Dek-silenced neurons.Mann-Whitney test P-value = 0.037.(D) Bar graph of the membrane input resistance in control versus Dek-silenced neurons.Mann-Whitney test P-value = 0.034.(E) Bar graph of the firing threshold in control versus Dek-silenced neurons.Unpaired t-test P-value = 0.0046.(F) Representative action potential (AP) waveform in control and Dek-silenced neurons.(G) Representative phase plot in control and Dek-silenced neurons.(H) Bar graph of the AP maximum rise slope in control versus Dek-silenced neurons.Unpaired t-test P-value = 0.07.(I-N) Expression levels of the differentially expressed potassium and calcium channel subunits between control and Dek-silenced entorhinal cortex (EC) primary neurons (n = 3 wells for each group).Unpaired t-test P-value = 0.0001 (Scn1a), P < 0.0001 (Scn2a), P = 0.0001 (Scn8a), P < 0.0001 (Scn1b), P = 0.0013 (Scn3b), P = 0.024 (Scn4b).(O) Representative traces of 100 pA hyperpolarizing current from control and Dek-silenced neurons.(P) Bar graph of Sag amplitude in control and Dek-silenced neurons.Unpaired t-test P-value = 0.050.RPKM = reads per kilobase of transcript per million mapped reads.

Figure 4
Figure 4 Effect of Dek silencing in EC neurons in vitro.(A) Western blot analysis of the levels of total tau (green) in entorhinal cortex (EC) neurons in primary culture at 2, 3 and 4 days post-transduction (one representative experiment is shown with n = 2 replicates per time point per group).The graphs show protein levels in % relative to 2 days post-transduction control neurons.Band intensity was normalized with actin for total tau measurements and with total tau for phospho-tau measurements.Two-way ANOVA Sidak's multiple comparisons test P-value = 0.0019 (three independent experiments with two replicates per time point and group).(B) RT-qPCR analysis of Mapt expression levels in control and Dek-silenced neurons 4 days after transduction (three independent experiments with three replicates per group).(C) Confocal microscopy images of immunofluorescence staining of MAP2 (red) and AT8 (greyscale) in control (CONT) and Dek-silenced neurons at 4 days post-transduction.Scale bar = 100 μm.The graphs show AT8 fluorescence intensity levels in % relative to control-treated neurons.Unpaired t-test P-value = 0.017 (total AT8), P = 0.0055 (somatic AT8) (AT8 total fluorescence graph shows the relative fluorescence intensity per field from one experiment and 4-5 replicates per condition.Additional experiments with consistent results were performed.Distribution of AT8 fluorescence in the soma was determined in two independent experiments with 4-5 replicates each, n = 40-60 neurons per group in total).(D) Western blot analysis of LC3 (three independent experiments and 3-4 replicates per experiment and group).P-value = 0.0006.Dnajc6 expression levels in control and Dek-silenced neurons in vivo and in vitro.Unpaired t-test P-value = 0.0014 (in vivo), P < 0.0001 (in vitro).(F) Dnajc6/auxilin relative protein levels in control and Dek-silenced neurons in vitro (n = 3 wells per group).Unpaired t-test P-value = 0.0028.LC = locus coeruleus; MW = molecular weight; RPKM = reads per kilobase of transcript per million mapped reads; sh = small hairpin.