The HTau (Andorfer et al., 2003) mice (expressing human MAPT with partial deficiency for endogenous mouse Mapt, was generated via targeted insertion of EGFP into the Mapt locus hence referred to as Mapt^egfp/egfp) were bred and maintained in our colony. Mice with a complete knockout of endogenous mouse Mapt (Dawson et al., 2001) (Mapt^-/-) were obtained from Jackson Laboratory (B6.129X1-Mapt^tm1Hnd/J – Stock # 007251). Both HTau and Mapt^-/- mice were maintained in the C57BL/6 background. We outbred HTau to the C57BL/6J (non-transgenic) mice to obtain the 8c line (HTau mice with Mapt^+/+ background). The HTau/Mapt^+/+ were then crossed to Mapt^-/- mice (Dawson et al., 2001) over at least 10 generations to develop hTau-Mapt^-/- mice, which will be referred to as hTau^MaptKO(Duke) mice [‘MaptKO(Duke)’ refers to ‘Mapt knockout mice generated at the Duke University by Dawson’s group [published in Dawson et al. (2001) study]. All comparisons were made to either 6- or 12-month-old non-transgenic mice (referred to as WT). The WT mice were littermates generated during hTau^MaptKO(Duke) and Mapt^-/- breeding, but were without human MAPT transgene and were Mapt^+/+. We utilized mixed gender for all groups of mice in the present study (Table 1). All experimental protocols involving animals were performed in accordance with the US National Institute of Health guidelines on animal care and were approved by the Institutional Animal Care and Use Committee of the University of New Mexico and Cleveland Clinic Foundation.

Microtubule-associated protein tau antibodies: AT8 (pS202/pT205; mouse monoclonal antibody; ThermoFisher Scientific; #MN1020), AT180 (pT231; mouse monoclonal antibody; ThermoFisher Scientific; #MN1040), Tau-5 (or Tau5; mouse monoclonal antibody; ThermoFisher Scientific; #AHB0042), PHF-1 (pS396/pS404; mouse monoclonal antibody provided by Peter Davies, Albert Einstein College of Medicine), Tau12 (human tau specific antibody; mouse monoclonal antibody; Abcam; #ab74137). Other antibodies: GAPDH (rabbit polyclonal antibody; EMD Millipore; #ABS16), T-MECP2 (D4F3; rabbit monoclonal antibody; Cell Signaling #3456); phospho-MECP2 (pSer80; rabbit polyclonal antibody; PhosphoSolutions #p1205-80).

The mice were anesthetized and transcardially perfused with 0.125 M phosphate buffer (PB). Following perfusion, the brains were removed, the left hemisphere was immersion fixed in 4% paraformaldehyde in PB (4% PFA/PB), the right hemisphere was micro-dissected into the cortex and hippocampus, wet weights were recorded, and the tissues were snap frozen in liquid nitrogen for subsequent biochemical analysis. The rest of the right hemispheres were weighed and snap frozen in liquid nitrogen for subsequent mRNA extraction. Hippocampi from 6- and 12-month-old WT and hTau^MaptKO(Duke) mice were used for biochemical (quantitative real-time PCR/qRT-PCR or Western blot) analysis. Sagittal brain sections were used for the immunohistochemical analysis (see below).

RNA from the hippocampus of 6-month-old non-transgenic (or WT) and hTau^MaptKO(Duke) mice (Table 1) was extracted using the TRIzol^® reagent as described by the manufacturer (Thermo Fisher Scientific), quantitated and 3 μg (30 μl at 100 ng/μl) of purified RNA converted to cRNA (High Capacity cDNA synthesis kit, Thermo Fisher Scientific) and hybridized on Illumina^® Mouse WG6 microarray chip (Pool_ID: 11278593_A; Sentrix_ID: 6189772009; see the layout in Table 2). The Illumina Mouse WG6 chip was a single chip that contains six whole genome microarrays. Each of these six microarrays on the WG6 chip contained 30,854 oligonucleotides (Figure 1A; corresponding to over 23,000 well-characterized genes and over 10,000 predicted genes). Microarray chips were scanned using Illumina’s state-of-the-art iScan^®. The raw data from iScan^® were pre-processed in GenomeStudio and exported for further data processing and analysis in R-LIMMA package (Law et al., 2016).

For the gene expression analysis: Total RNA (50 ng/μL) was converted to cDNA using the High Capacity cDNA Reverse Transcription kit and amplified using specific TaqMan probes (gene expression marker: Mecp2 Mm01193535_m1; MAPT Hs00902194_m1) GAPDH was used as a house-keeping gene for normalization. The qPCR assays were run on the StepOnePlus^® Real-Time PCR System (all reagents were purchased from Thermo Fisher Scientific).

Proteins were homogenized in 10% weight/volume Tissue Protein Extraction Reagent (T-PER^®, Thermo Fisher Scientific) and soluble hippocampal lysates were resolved via SDS-PAGE and immunoblotted as previously described (Bhaskar et al., 2010b). The dilutions of primary antibodies utilized were as follows: GAPDH at 1:20,000; Tau5 at 1:10,000; phospho- and total-MECP2 at 1:1000.

Free-floating sections (30 μm thick) from 12-month-old WT and hTau^MaptKO(Duke) mice as well as formalin fixed autopsy brain sections from cognitively normal (Braak stage 0) and AD (Braak Stage V) were processed for immunohistochemical (IHC) and immunofluorescence (IF) analysis as previously described (Bhaskar et al., 2010b). Briefly, sections were first incubated in 10 mM sodium citrate buffer (pH 6.0) for 10 min at 95°C for antigen retrieval, washed in PBS with 0.1% Tween (PBST), quenched with 0.3% H₂O₂ in PBST for 15 min (only for IHC). Sections were blocked for 1 h at room temperature with the 5% normal sera (goat/donkey – from animal species in which the secondary antibodies were raised). The sections were incubated with phospho-MECP2 (1:250). After washing in PBST, the sections were incubated with biotinylated-secondary antibody (1:250; Jackson ImmunoResearch Laboratories Inc., for IHC) or Alexa 555 conjugated goat anti-rabbit secondary antibody (1:250 in blocking solution; for IF). Sections were then incubated with ABC (Vector Laboratories) reagent for 1 h at room temperature (for IHC). The immunoreactive signals were revealed by developing sections in SigmaFast^® 3,3’-diaminobenzidine (DAB) tablets (Sigma–Aldrich) (for IHC) or the sections were mounted on slides using Hardset (Vector Laboratories) with DAPI (for IF). Bright field images were acquired using Leica DMR upright fluorescence, Leica TCS-SP8 confocal microscope and FLIM system or Olympus bright field microscope.

Mouse neuroblastoma cells (Neuro2a or N2a cells; ATCC^®, Manassas, VA, USA) were cultured in EMEM with L-glutamine media with 10% FBS for up to 10 passages before being expanded and plated at a density of 500,000 cells per well for all the experiments. N2a cells were transiently transfected with human tau plasmid (0.4 μg DNA per transfection), pRC/CMV n123c (0N3R tau) using Effectine^® reagent per manufacturer’s instructions (Qiagen, Cat# 301425). Mock (empty plasmid) or tau transfected N2a cells were nucleofected with siRNA to MECP2. After 24 h, the cells were lysed to detect MECP2, PHF1, Tau5, Tau12, β-actin, and GAPDH via SDS-PAGE and Western blot analysis (see above). Ratios of MECP2/GAPDH were plotted to assess MECP2 knockdown and ratios of PHF1/Tau5 and Tau5/GAPDH were plotted and quantified.

Unless otherwise indicated, comparisons between the two groups were done via unpaired t-test; comparisons between multiple treatment groups were done via one-way or two-way ANOVA with indicated multiple comparisons post hoc tests. All statistical analyses were performed using GraphPad Prism^®.

To identify differentially expressed mRNA levels in the hippocampus of 6-month-old WT and hTau^MaptKO(Duke) mice (n = 3 per genotype; Table 1), we used an Illumina^® MouseWG-6 V2 array chip to probe changes in 30,854 different mRNAs (Figure 1A). A 6-month time point was selected based on, the results from present study and several previous studies where Htau mice have been shown to display significant increase in tau phosphorylation in the hippocampus (Andorfer et al., 2003, 2005; Maphis et al., 2015). Overexpression of human MAPT in hTau^MaptKO(Duke) mice did not alter the overall number of mRNA messages detected in the hippocampus between WT mice and hTau^MaptKO(Duke) mice that were relative to all the probes present on the array (data not shown).

This primary whole genome microarray screen, which was based on a small group size, resulted in generation of three data sets: (1) the raw expression dataset exported directly from GenomeStudio, referred to as raw data; (2) the normalized dataset, which was obtained based on the raw dataset after force-positive background correction, log₂ transformation and quantile normalization; (3) the QC filtered dataset, which was obtained based on the normalized dataset by selecting the present probes using the detection threshold of p < 0.05. The raw and normalized datasets had 30,854 genes and the QC filtered dataset had 11,619 genes (Figure 1A). Next, QC analysis was also performed on the normalized expression data and the results are plotted as normalized box plot, normalized control, normalized M (log ratios) and A (mean average) scale plot, normalized pairs, and normalized sample relation (Supplementary Figures S1–S4). Overall, the QC analysis of the normalized data suggested that the generated dataset was of good quality for the downstream analysis. Next, hierarchical clustering (unsupervised) was performed on the three data sets. Correlation distance was used in all cases. Results suggested some association between sample conditions and clustered grouping (Supplementary Figure S5). For example, samples from WT-2 and WT-3 (non-transgenic mice #2 and #3) segregated together. Similarly, samples from hTau^MaptKO(Duke)-2 and hTau^MaptKO(Duke)-3 segregated together. However, samples from WT-1 and hTau^MaptKO(Duke)-1 were separate from other samples in their respective groups (Figure 1B and Supplementary Figure S5). Overall, there were no outliers in the sample sets after the QC and clustering analysis.

Next, comparisons were performed via linear modeling to check differences between the two genotypes [WT and hTau^MaptKO(Duke)]. Empirical Bayes and other shrinkage methods were used to borrow information across genes. FDR adjusted p-values were calculated. Genes were defined as significantly altered, if their FDR-adjusted p-value < 0.05. We also calculated raw p-value and fold change. For the initial screening purposes, the differentially altered genes were defined as those genes whose adjusted p-value < 0.05 or raw p-value < 0.001 whichever was smaller, and whose fold change was greater than the threshold or less than 1/threshold in this analysis (Illumina^® recommends 1.35 for the threshold). A volcano plot analysis was performed which showed the relationship between –log₁₀ (p-value) and log₂ (fold change) of all the genes between WT and hTau^MaptKO(Duke) groups. Genes scattered beyond the cut-off window in the volcano plot indicated the statistical significance of gene expression changes and provided a direct visual representation of the genes that were differentially expressed. Based on the fact that only four genes (Gh, Krt12, Mapt, and Rapgefl1; Table 3) were altered between WT and hTau^MaptKO(Duke) mice when we used a strict FDR cut-off (of 2.0), we relaxed our criteria to include genes that met both the raw p-value and fold change (FDR cut-off of 1.5) (Figure 1C). Prior to relaxing our criteria the heat-map analysis of all 30,854 genes showed no obvious pattern of significantly altered genes between WT and hTau^MaptKO(Duke) groups (Figure 1D). However, following relaxation of our criteria (with a raw p-value of <0.001), 64 genes were either up- or down-regulated in hTau^MaptKO(Duke) mice compared to WT group (Figure 1E and Table 3) and only these were utilized for downstream functional pathway annotation and network analysis via IPA^TM, Metacore^TM, and GO (Table 3). Some of the differentially regulated genes, proteins they encode, their function and their relevance to the brain function are shown in Supplementary Table S1.

To obtain an inventory of all the up- and down-regulated genes between WT and hTau^MaptKO(Duke) mice, we plotted all 64 significantly altered (raw p < 0.001) genes displaying a cut-off of +1.5-fold change (Figure 2A). Interestingly, exactly half of the 64 genes (i.e., 32) were down-regulated and the remaining half was up-regulated (Figure 2A). The most down-regulated gene was C330006P03RIK, which is an unclassified gene of unknown function. The most up-regulated gene, yet with only ∼0.5 log fold change, was Vat1L protein, which is a protein-coding gene with functions related to oxidoreductase activity and transferase activity (source: GeneCards and GO annotations) (Figure 2A). Mecp2 and Wdr18 were among the lowest (<0.25-fold) up- or down-regulated genes, respectively (Figure 2A). While Mecp2 was the gene with the lowest level of upregulation in hTau^MaptKO(Duke) mice, it is highly relevant to our analysis because of its involvement in the regulation of DNA methylation. Furthermore, a mutation in the MECP2 gene in humans is the primary cause of most cases of Rett Syndrome (Joyner et al., 2009; Petazzi et al., 2014), a rare genetic postnatal neurological disorder predominantly affecting females. Next, we performed network analysis using Metacore^TM to identify which transcription factor(s) would regulate the expression of the most number of genes from our screening. As shown in Figure 2B, SP1 transcription factor appeared to regulate 9 of the significantly altered genes (Mapt, Strn4, Mecp2, Slc40a1, Fn3K, Pcsk2, Prkca, Lass1, and Pold2) in our panel compared to other transcription factors: NFκB, PXR, ETS, RAR-α/RXR-β, p53, p73 or AP-1, which only appeared to regulate anywhere between 1 and 4 genes (Figure 2B).

Next we performed process and phenotype annotations for the genes that are significantly altered in the hTau^MaptKO(Duke) mice using Metacore^TM, IPA^TM, and GO. There were overlaps in certain functional terms between these analyses. For example, the enrichment analysis from these publicly available software packages showed that the following process/functions were pronounced in hTau mice: “Regulation of protein modification and metabolic processes, acetylation, protein-protein interaction, protein ubiquitination, regulation of apoptosis, cell surface receptor linked signal transduction,” etc., while “alternative splicing, metal binding, presence of WD40 motif proteins, secretory pathway” were relatively suppressed in hTau^MaptKO(Duke) compared to WT. The protein modification, ubiquitination, protein transport, cytoskeletal remodeling, inflammation, etc., were some of the key process/function terms that surfaced from the IPA/Metacore analysis. To identify the most relevant set of genes that overlapped with more than one functional GO terminology, and the set of genes that would be most relevant to neurological diseases, we plotted the top six biological process categories from the GO analysis and mapped genes that were up- (red) or down-regulated (blue) within each of these biological processes. The top six GO biological process and the gene representations were; (a) anatomical structure development (17/47 genes; 36.17%); (b) cellular nitrogen compound metabolic process (15/47 genes, 31.91%); (c) biosynthetic process (13/47 genes, 27.66%); (d) cell death (9/47 gene, 19.15%); (e) signal transduction (8/47 genes, 17.02%); and (f) response to stress (8/47 genes; 17.02%) (Figure 2C). Differentially altered genes associated with these six biological processes were split into two groups and plotted as two Venn diagrams (Figure 2C). The intersection included Mecp2↑ and Prkca↑ (arrows indicate these are up-regulated) as key genes common for both sets of Venn diagrams. The overlapping region also included Foxk2↓, Med7↓, Ybx3↓, and Tsc22d3↓ (arrows indicate these are down-regulated). However, these were absent in the second set of Venn diagram (Figure 2C). We also plotted a third Venn diagram with ‘neurological system process’ (5/47 genes; 10.64%), ‘immune system process’ (5/47 genes; 10.64%), and ‘cell cycle’ (3/47 genes; 6.38%) as three additional biological/disease-relevant processes. While Prkca was the only gene present in the overlapping region, Mecp2 was clearly unique to ‘neurological system process’ (Figure 2C). Taken together, these results suggest that Prkca and Mecp2 are the most common genes relevant to several biological processes, which may have been altered in the hippocampus of hTau^MaptKO(Duke) mice compared to WT controls.

Whereas Prkca is a generic protein kinase that affects multiple cellular pathways, the Mecp2 is one of the altered genes present in the interface of all three biological processes from the GO analysis and is directly implicated in neurological diseases (Rett Syndrome). Therefore, we decided to determine the total protein levels and phosphorylation status of MECP2 in the hippocampus of non-transgenic and hTau^MaptKO(Duke) mice. To confirm our microarray data, we first assessed Mecp2 mRNA levels via real-time quantitative PCR analysis and observed an expected increase in the MECP2 mRNA levels in the hemi-brains of 6-month-old hTau^MaptKO(Duke) mice compared to age-mated WT controls (Figure 3A). Next, we determined the regional differences in the MECP2 mRNA levels with hTau^MaptKO(Duke) mice at 6 months of age. Interestingly, we observed cortex displaying significantly highest levels of MECP2 mRNA compared to hippocampus and rest of the brain (ROB), which are enriched in striatum, diencephalic structures, brain stem but lacked cortex/hippocampus (Figure 3B). We also assessed the mRNA levels of Vat1l, which was up-regulated in the microarray analysis. There was an increased trend for Vat1l, but the statistical significance was at p = 0.06 (Figure 3C). Next, we prepared hippocampal lysates from 6-month-old WT and hTau^MaptKO(Duke) mice. The levels of both phosphorylated at Ser80 (p)MECP2/total (t)MECP2 and tMECP2/GAPDH appeared elevated in the hippocampus of 6-month-old hTau^MaptKO(Duke) mice compared to 6-month-old WT controls (Figures 3D,E). However, only the total (t)MECP2 levels appeared increased more than three folds compared to WT controls (Figures 3D,E), yet statistical analysis suggested p = 0.08 despite having n = 10 mice in hTau^MaptKO(Duke) group (Figures 3D,E). This could be due to high variability in the levels of both pMECP2 and tMECP2 at 6-months of age in both WT and hTau^MaptKO(Duke) groups. Finally, we performed double IF analysis for pMECP2 and DAPI to determine if the pMECP2 levels reveal elevated labeling within specific cellular population in the hippocampus. Both 6-month-old WT and hTau^MaptKO(Duke) mice showed basal level of, mostly nuclear, labeling for pMECP2, which appeared slightly increased as a heterochromatic speckles within the nucleus of the CA3 region of hTau^MaptKO(Duke) mice compared to WT mice (Figure 3F). Together these results suggests that because the MECP2 mRNA levels are only marginally up-regulated in the whole genome gene expression and mRNA analysis, the MECP2 protein levels did not reach desired statistical significance in the hippocampus of hTau^MaptKO(Duke) mice.

To determine if the differences in MECP2 between WT and hTau^MaptKO(Duke) mice stands out to be statistically significant with age, we performed Western blot analysis in 12-month-old WT and hTau^MaptKO(Duke) mice. There was no significant alteration in the levels of T-MECP2 protein at 12-months of age (Figures 4A,B). However, the level of pMECP2 was significantly up-regulated in the hippocampus of hTau^MaptKO(Duke) mice compared to WT mice (Figures 4A,B). Elevated levels of pMECP2 were also observed within the nuclei of CA3 pyramidal neurons in hippocampus of hTau^MaptKO(Duke) mice compared to WT group, which again appeared more like speckle staining mostly at heterochromatin regions of the nucleus (Figure 4C). We confirmed this by performing double IF followed by confocal microscopy analysis. Twelve-month-old hTau^MaptKO(Duke) mice displayed markedly elevated pMECP2 labeling that overlapped with nuclear staining DAPI in the CA3 regions of hippocampus (Figure 4D). Finally, to determine if the pMECP2 is relevant to human AD, we probed autopsied brain sections from a human AD patient and compared that from a non-demented control subject. Confocal microscopy analysis revealed numerous cells positive for pMECP2 in the layer III (temporal cortex) of human AD brain compared to non-AD controls (Figure 4E). Interestingly, we observed significantly more pMECP2 immunoreactive structures, which appeared to be peri-vascular and picnotic (Figure 4F arrows) and sometimes nuclear/peri-nuclear in layer III of the temporal cortex (Figures 4F,G). Confocal microscopy did show presence of somewhat cytosolic or diffused nuclear staining for pMECP2 in non-AD controls (Figure 4G), which was readily visible in the IHC via bright field microscopy (Figure 4F). These results reveal higher levels of pMECP2 in both the CA3 hippocampal neurons of hTau^MaptKO(Duke) mice and in the temporal lobe of human AD patients.

Since the MECP2 levels are elevated in hTau^MaptKO(Duke) mice and in human AD, we wanted to determine if MECP2 could directly regulate the total and phosphorylated levels of tau protein. Therefore, we transiently transfected N2a cells with either human tau (non-mutant 0N3R isoform) or control plasmid. After 24 h, the N2a cells were nucleofected with either MECP2 siRNA (siMECP2) or scrambled siRNA (siScr). As expected, human tau transfected N2a cells expressed human tau (positive for Tau12 antibody) and displayed hyperphosphorylation on the T231 (AT180) and Ser396/Ser404 (PHF1) sites (Figures 5E,F). Further, siRNA against MECP2 showed >80% reduction in the T-MECP2 levels compared to siScr control group (Figures 5A,B). Interestingly, both hyperphosphorylated (on PHF1 site) and total tau (Tau5/GAPDH and Tau12/GAPDH) levels were significantly reduced following siRNA-mediated knockdown of MECP2 in tau transfected N2a cells (Figures 5C,D). On the other hand, most intriguingly, there was an increase in AT180/Tau5 ratio with siRNA-mediated knockdown of MECP2 was observed (Figures 4A, 5E). To determine if MECP2 knockdown specific to reducing tau levels or could act as a suppressor of generic transcription; we assessed the immunoreactivity for β-actin. While there was small reduction in β-actin/GAPDH ratio in with siMECP2 in both ‘+/- Tau’ conditions (Figures 5A,G,H), however, the effects were not statistically significant as the one observed for Tau5/GAPDH and Tau12/GAPDH (Figures 5A,C,D). Together, these results suggest that MECP2 may directly or indirectly regulate expression of tau and its phosphorylation state, which may be relevant to AD and related tauopathies.