Sex- and region-specific cortical and hippocampal whole genome transcriptome profiles from control and APP/PS1 Alzheimer’s disease mice

A variety of Alzheimer’s disease (AD) mouse models has been established and characterized within the last decades. To get an integrative view of the sophisticated etiopathogenesis of AD, whole genome transcriptome studies turned out to be indispensable. Here we carried out microarray data collection based on RNA extracted from the retrosplenial cortex and hippocampus of age-matched, eight months old male and female APP/PS1 AD mice and control animals to perform sex- and brain region specific analysis of transcriptome profiles. The results of our studies reveal novel, detailed insight into differentially expressed signature genes and related fold changes in the individual APP/PS1 subgroups. Gene ontology and Venn analysis unmasked that intersectional, upregulated genes were predominantly involved in, e.g., activation of microglial, astrocytic and neutrophilic cells, innate immune response/immune effector response, neuroinflammation, phagosome/proteasome activation, and synaptic transmission. The number of (intersectional) downregulated genes was substantially less in the different subgroups and related GO categories included, e.g., the synaptic vesicle docking/fusion machinery, synaptic transmission, rRNA processing, ubiquitination, proteasome degradation, histone modification and cellular senescence. Importantly, this is the first study to systematically unravel sex- and brain region-specific transcriptome fingerprints/signature genes in APP/PS1 mice. The latter will be of central relevance in future preclinical and clinical AD related studies, biomarker characterization and personalized medicinal approaches.


Introduction
Alzheimer's disease (AD) is a complex neurodegenerative disorder accompanied by progressive cognitive decline, memory deterioration, neuropsychiatric symptoms and eventually death [1,2].From a pathophysiological point of view, the formation and accumulation of amyloid beta (Aβ) together with the extracellular deposits of amyloid plaques represent one of the hallmarks and devastating histopathological characteristics of AD in the central nervous system (CNS) [3,4].Mechanistically, amyloid peptides of different lengths originate from a common biochemical substrate, i.e., the amyloid precursor protein (APP), upon sequential cleavage steps.Two specific secretases are responsible for these endoproteolysis/cleavage steps, i.e., β-site APP cleaving enzyme 1 (BACE-1), and γ-secretase [5,6].Notably, pro-amyloidogenic APP mutations can enhance the release of cytotoxic Aβ peptide fragments of different length.For example, the Swedish double mutation KM670/671NL aggravates BACE cleavage and increases the abundance of both Aβ 1-40 and Aβ 1-42 , thereby mediating pro-amyloidogenic effects [7].Furthermore, other common mutations exist that affect, e.g., presenilin (PS)1 and PS2.Presenilins act as catalytic sites for the γ-secretase and can severely trigger the Aβ 1-42 formation and aggravate pro-amyloidogenicity [8].The common Aβ theory of AD in humans is complemented by the intraneuronal formation of hyperphosphorylated tau (τ) protein, finally evolving into neurofibrillary tangles [9,10].However, there is increasing evidence from preclinical, clinical and drug research and development studies that cutting the sophisticated etiopathogenesis of AD down to Aβ-and τ-formation might be an oversimplification, misleading and too shortsighted [11][12][13][14][15][16].Indeed, multiple studies in the last decades elicited that AD is a multifactorial disease that can be triggered/modified by highly diverse factors involving numerous molecular, cellular and supra cellular pathways acting on the neuronal nano-, micro-, meso-and macro-scale [17,18].Although some of these results were obtained from human AD patient brain tissue, the majority of related experimental data was obtained from rodents, particularly AD mouse models [19][20][21][22].
More than 100 AD mouse models have been generated so far and many of them exhibit typical age-and AD-related structural and functional abnormalities, e.g., Aβ plaques, axonal and synaptic dystrophy, reduced synaptic plasticity, and disturbance of learning and memory [23][24][25].Importantly, depending on the mutagenesis/transgenesis strategy carried out in the individual mouse lines, these AD mouse models can significantly differ in histopathological features, symptomatic profile, disease progression characteristics and mortality.Recapitulatory, AD mouse models should fulfill fundamental translational categories, i.e., isomorphism, homology, and predictability to a maximum extent [26].Examples of well-established AD mouse lines include, i.a., the 5XFAD, 3xTg and APP/PS1 models.The 5XFAD animals with co-integrated transgenes bred as a single allele, represent a pathophysiologically "aggressive" model of AD exerting juvenile-onset amyloid pathology (* 3 months) with a rapid onset phenotype.However, 5XFAD mice represent a non-physiological combination of familial Alzheimer's disease (FAD) mutations with marked intracellular Aβ accumulation [22].The 3xTg line mimics an early-to mid-life amyloid pathology model that includes hyperphosphorylated tau (τ).It captures both Aβ and phospho-τ features of AD and exhibits variable pathology between colonies and sexes.In addition, genetic drift has been observed [22].The APP/PS1 model exerts early-onset (*6 months) amyloid pathology.It is well-characterized and co-integrated transgenes are bred as a single allele, which however, makes it hard to control for independent transgene effect [22].
In our study, an APPswePS1dE9 (APP/PS1) AD mouse model has been utilized, carrying the human APP with Swedish double mutation (APPswe) co-integrated with human PS1 with exon 9 deletion (PS1dE9) [11-14, 27, 28].These mutations result in an overproduction of APP and PS1 splice variants and a subsequent neural Aβ overload.APPswePS1dE9 mice exhibit a mortality of 10-15% [29].Cognitive/memory impairment and behavioral deficits observed, e.g., in spatial navigation tasks, reference learning, Morris' water maze and radial arm maze, become obvious at an age of 6 to 12 months of age [30].Furthermore, seizure activity is often encountered in AD mouse mutants including APP/PS1 and intracerebral Aβ accumulation was shown to be linked to epileptogenesis [13,14,31,32].Thus, seizure activity might be responsible for sudden death in this mouse line as well [31,33].Overall, the APP/PS1 AD line represents a well-established model of high value in translational terms [22].
We previously demonstrated that neurodegenerative alterations in the motor cortex and septohippocampal system of APP/PS1 mice (but also in other AD lines such as 5XFAD) can result in complex central dysrhythmia [32], particularly in the theta (4-8 Hz) and two distinct gamma frequency ranges  Hz and 50-70 Hz) [13,14].Importantly, these systemic in vivo macro-scale electrophysiological findings in the electrocorticogram (ECoG) and electrohippocampogram (EHipG) of APP/PS1 mice, turned out to be clearly age-, sex-and brain region-specific and demonstrated that subgroup-specific analysis in AD research is indispensable [13,14].The latter has substantial implications for AD drug research and development and individualized/personalized pharmacological AD treatment.Previous studies in APP/PS1 mice further revealed selective electrical activity and individual frequency characteristics in ECoGs and other deflections related to circadian rhythmicity and sleep [6,29,30,[33][34][35].
To get an overall insight into the multifactorial, pathophysiological alterations in APP/PS1 mice, a detailed investigation of genome-wide transcriptional alterations in both sexes and different brain regions of interest (BROIs) is essential.Clearly, transcriptome studies in human AD patient CNS material had been carried out in the past [36] and the same holds true for various rat and mouse AD models [36].However, many of these studies, e.g., in APP/PS1 mice, encounter specific limitations: (i) female and male animals were pooled (in a balanced or unbalanced way) and were not investigated in a sex-specific manner or there was no information provided about the sex distribution [37][38][39][40], (ii) only one sex (generally males) was studied [41][42][43], (iii) the entire brain was investigated and not individual BROIs, although different brain regions can be differentially affected by AD pathogenesis [38].Thus, the complex sex-and region-specific investigation of transcriptomes of AD models such as APP/PS1 mice, is mandatory [36,[44][45][46].
Here we present for the first time, transcriptome results from the retrosplenial (RS) cortex and hippocampus of age-matched (8 months old) female and male wild-type (WT) control and APP/PS1 AD mice.The hippocampus as an interface structure for learning and memory processes [47][48][49] and the RS cortex involved in spatial memory and environmental orientation [50,51] are early affected in mild cognitive impairment (MCI) and manifest AD and therefore represent major BROIs in AD pathophysiology [52,53].Our results provide the first sex-specific and hippocampal/RS cortex-specific analysis of transcriptional alterations in APP/ PS1 AD mice.We enable insight into co-upregulated and co-downregulated (intersectional) gene candidates and those differentially expressed genes (DEGs) that serve as signature/fingerprint gene sets in individual BROIs of males or females.We further demonstrate that DEGs in the individual subgroups are primarily related to the fields of microglia activation, immunological response, inflammation, synaptic integration, learning and memory and ictogenesis in AD.In addition, we also analyzed long (intergenetic) non-coding RNAs (l(i)ncRNA, > 200 nucleotides) that impact on the regulation of (post)transcriptional, translational and epigenetic phenomena in AD [54][55][56][57].
The PCR reaction mix was gently vortexed for 3 sec followed by a brief centrifugation step for 5 sec at 2000xg using a micro centrifuge (ROTILABO, Carl Roth GmbH & Co. KG, Germany).For each genotyping PCR, positive controls (WT DNA and DNA from a validated mutant mouse) and a negative control (no DNA) were added for experimental validation.PCRs were carried out using a Bio-Rad thermal cycler (type C1000, Bio-Rad Laboratories GmbH, Germany) and the following amplification parameters were applied: 3 min pre-incubation at 94˚C, 35 cycles (each cycle composed of 94˚C, 30 sec denaturation; 48˚C, 30 sec annealing; 72˚C, 1 min extension) and finally 10 min incubation at 72˚C.The amplification mix was stored at 4˚C till further use.
PCR products were analyzed by horizontal agarose gel (1.5% in 0.5x TBE buffer, pH8) electrophoresis and visualized by ethidium bromide (0.3 μg/ml).The ChemiDoc Touch System (Bio-Rad Laboratories GmbH, Germany) was used for gel imaging and identification of the amplified DNA fragments (S1 Fig in S1 File).

RS cortex and hippocampus preparation
Experimental animals were deeply anaesthetized using i.p. injection of ketamine (100 mg/kg) / xylazine (10 mg/kg).To ensure that the animals were fully anaesthetized, the absence of tail and foot pinch reflexes was verified.Animals were decapitated, the brain was removed immediately and placed in a clean RNase-free petri-dish kept on ice and filled with pure RNAlater reagent (Qiagen GmbH, Germany).By using a scalpel, forceps and a thin brush, the whole hippocampus and a piece (2-3 mm 3 ) of the RS cortex were dissected from both brain hemispheres of each experimental animal (S2 Fig in S1 File).The tissue fragments were placed in a 2 ml RNase free reaction tube, snap frozen in liquid nitrogen and stored at -80˚C until RNA preparation.To avoid potential interference of transcriptional profiles with circadian rhythmicity, tissue preparation was always performed ante meridiem between 8 am and 11 am.

RNA isolation from the RS cortex and hippocampus
Total RNA was isolated using the RNeasy Lipid Tissue Mini Kit (Qiagen GmbH, Germany) according to the manufacturer's instructions (incl.an optional DNase digestion step).Cortical and hippocampal tissue samples were removed from -80˚C, immediately lysed in QIAzol lysis buffer (Qiagen GmbH, Germany) and homogenized using the Tissue Ruptor (Qiagen GmbH, Germany), a handheld-rotor-stator homogenizer with disposable probes.Following phenolchloroform separation, DNase digestion and three washing steps, total RNA was eluted in 30 μl RNase-free ddH 2 O.The quantity and quality of the eluted RNA was checked using Nano-Drop 1 ND-1000 (Thermo Fisher Scientific, USA) according to the manufacturer's instructions.The ratio of absorbance at 260 and 280 nm (260/280) was used to assess RNA purity with a ratio of * 2.0 being generally accepted as "pure" for RNA related experimental approaches (see NanoDrop 1 user manual).The ratio of absorbance at 260 and 230 nm (260/ 230) is a secondary measure of nucleic acid purity.They are commonly in the range of 1.8-2.2(see NanoDrop 1 user manual).Our RNA probes from APP/PS1 and control BROI samples for microarray and qPCR experiments exhibited absorbance ratios of * 2.0 (for 260 nm/280 nm) and of * 2.05 (for 260 nm/230 nm).

One-color microarray-based gene expression analysis
The One-Color Microarray-Based Gene Expression system by Agilent Technologies Germany GmbH & Co. KG (Germany) utilizes the Low Input Quick Amp Labeling Kit to generate cDNA (1 st and 2 nd strand cDNA) via AffinityScript-RT (reverse transcriptase) which is a genetically engineered, highly thermostable version of Moloney Murine Leukemia Virus (MMLV) enzyme reverse transcriptase (RT).Subsequently, the samples were labeled with a T7 RNA polymerase blend.This polymerase incorporates Cyanine 3-CTP during amplification, generating a onecolor fluorescent complimentary RNA (cRNA) as target material.The RNA sample input is supposed to range from 10-200 ng (for details on the procedure, see manufacturer's instructions in the One-Color RNA Spike-In Kit and Low Input Quick Amp Labeling Kit, both from Agilent Technologies Germany GmbH & Co. KG (Germany).The amplification from total RNA to amplified cRNA is typically around 100-fold.Subsequently, the labeled/amplified cRNA was purified using the RNeasy Mini Kit (Qiagen GmbH, Germany) followed by quantification of the cRNA using the NanoDrop ND-1000 UV-VIS spectrophotometer (Thermo-Fisher Scientific, USA).After finalization of cRNA sample preparation, hybridization of samples was carried out for 17 hrs (65˚C) using the Gene Expression Hybridization Kit (Agilent Technologies Germany GmbH & Co. KG, Germany) and the SurePrint G3 Mouse Gene Expression v2 8x60K Microarray Kit (Agilent Technologies Germany GmbH & Co. KG, Germany) according to the manufacturer's instructions.Key features of the SurePrint G3 Mouse Gene Expression v2 8x60K Microarray chip (Agilent Technologies Germany GmbH & Co. KG, Germany) include up-to-date content sourced from RefSeq, Ensembl, RIKEN, UniGene, and GenBank databases to provide full coverage of the mouse transcriptome, wide dynamic range of over 5 logs to ensure detection of low and high expressors for the most biologically relevant data and the possibility to proceed with one or two colors based on experimental needs.The specific SurePrint G3 Mouse Gene Expression v2 8x60K Microarray Kit used in our study is referred to as G4852B (Agilent Product Number).In our study, the microarray setup carries eight arrays on a slide in a 4 x 2 grid, with each array covering 62,976 features arranged in 1,064 rows and 170 columns (8 x 60k format).The biological features include 27457 Entrez Genes (unique) and 4578 l(i)ncRNAs (unique).96 ERCC and 10 E1A served as control probes.To minimize inter-run variability, we always performed hybridization of the maximum microarray load.In addition, we did not process subgroup specific probes as a whole in one hybridization step.Instead, probes from the individual subgroups were randomly distributed on the eight slots hybridization slide to prevent potential subgroup-specific inter-run confounders.
After the washing procedure with Gene Expression Wash Buffer Kit (Agilent Technologies Germany GmbH & Co. KG, Germany) to remove unspecific bindings, the microarrays were prepared for scanning and feature extraction.The microarray scan was performed using the Agilent SureScan Microarray Scanner (Agilent Technologies Germany GmbH & Co. KG, Germany).Feature Extraction Software (Agilent Technologies Germany GmbH & Co. KG, Germany) was used to extract the information from the probe features of the microarray scan data, providing information about gene expression/transcripts for further analysis.Details on the materials and software used here are also listed in S2 and S3 Tables in S1 File.
Please note that for transparency reasons, the raw read out data files were extracted as txt.files/csv.-files and are freely available at the Mendeley Data repository (Weiergra ¨ber, Marco (2023), "Whole genome transcriptome data from the retrosplenial cortex and hippocampus of female and male control and APP/PS1 Alzheimer's disease mice.",Mendeley Data, V2, doi: 10.17632/z9264694b4.2 [59].

Experimental transcriptome study groups
In this transcriptome study, APP/PS1 AD and WT controls were compared in four individual settings, i.e., ♀ Rs cortex, ♀ hippocampus, ♂ RS cortex and ♂ hippocampus.The primary focus is on genotype-, sex-and region-specific alterations in DEGs and l(i)ncRNAs in the individual subgroups.

Transcriptome analysis
Data extraction, DEGs and l(i)ncRNA characterization parameters.The Feature Extraction Software (Agilent Technologies Germany GmbH & Co. KG, Germany) was used to extract relevant information from the probe features of the microarray scan data, providing information about gene expression/transcripts for further analysis.The DEGs were extracted including related probe name, gene symbol (Primary HUGO gene symbol), gene name, gene description, sequence and GO terms, p-value, and fold change (FC).Note that the significance of DEGs was tested using one-way ANOVA and post hoc analysis using Tukey HSD (honestly significant difference) test.In addition, the Benjamini-Hochberg procedure for decrease of the false discovery rate (FDR) was applied [60].
Principal component analysis (PCA) and heatmap chartography/hierarchical clustering.The PCA and heatmap generation/hierarchical clustering was carried out using Gene-Spring GX 14.9 (Agilent Technologies Germany GmbH & Co. KG, Germany).In the PCA, our large multi-dimensional transcriptome data sets are depicted as 3D scatter plots to visualize potential clustering of subgroup triplicates/quadruplicates/quintuplicates.The coordinates of the X-, Y-and Z-axis represent the so-called PC scores.
Gene ontology (GO) and enrichment analysis -protein-protein interaction (PPI) and related pathways.GO analysis (Gene Ontology Consortium; http://www.geneontology.org)including the major fields of biological processes, cellular components, and molecular functions, was utilized to characterize the properties of DEGs and l(i)ncRNAs in the RS cortex and hippocampus of male and female APP/PS1 AD mice in comparison to WT animals and to identify gene sets related to annotated GO terms that are over-or under-represented.For pathway and process enrichment analysis, the Metascape software (https://metascape.org;[61]) had been used.For each list of DEGs extracted via Feature Extraction Software (Agilent Technologies Germany GmbH & Co. KG, Germany), complex pathway and process enrichment analyses have been performed based on the following ontology sources: GO Biological Processes, KEGG Pathway (Kyoto Encyclopedia of Genes and Genomes; http://www.genome.jp/kegg/), Reactome Gene Sets, and WikiPathways [61].The latter were applied to identify biological enriched pathways of AD-related mRNA and l(i)ncRNA alterations detected in the microarray experiments.As enrichment background, the entire genes/genome had been used.Individual terms with a p-value < 0.01, a minimum count of 3, and an enrichment factor > 1.5 were characterized and grouped into clusters due to their membership similarities.Note that the enrichment factor is defined as the ratio between the observed counts and the expected random counts.The p-values were calculated based on the cumulative hypergeometric distribution [62].For further procedural details, please refer to the Metascape software documentation (https://metascape.org;[61]).
Using the Metascape software, top clusters with their representative enriched terms (one per cluster) were defined.Only input genes with at least one ontology term annotation were included in the calculation.Significance is depicted as the log10(P) (https://metascape.org;[61]).
In addition, a PPI enrichment analysis was performed using also Metascape.For each given list of DEGs, the PPI enrichment analysis included the following databases: STRING [63], Bio-Grid [64], OmniPath [65], and InWeb_IM [66].Only physical interactions in STRING (physical score > 0.132) and BioGrid were considered.The resultant network comprises the subset of proteins that establish physical interactions with at least one other member in the list.If the network contains between 3 and 500 proteins, the Molecular Complex Detection (MCODE) algorithm [67] has been applied to identify densely connected network components (https://metascape.org;[61]).
Pathway and process enrichment analysis has been applied to each MCODE component independently, and the three best-scoring terms by p-value have been retained as the functional description of the corresponding components.The related p-values represent the probability of seeing at least x genes out of the total number of n genes in the list annotated to a particular GO term, given the proportion of genes in the whole genome that are annotated to that GO term.Given different p-values and n values, the related annotated number x of genes varies.The closer the p-value is to zero, the more significant is the particular GO term associated with the group of genes.
Note that GO analysis was conducted for upregulated genes with a FC >1.5.For the downregulated genes, a FC < -1.5 did not result in significantly enriched terms due to the low number of significantly downregulated genes.To get an impression of enriched terms we selectively lowered the FC cutoff for downregulated genes in the GO setting to < -1.2.
Venn visualization and analysis.Intersectional, i.e., co-upregulated or co-downregulated DEGs in the RS cortex and hippocampus of female APP/PS1 AD vs. WT mice, the RS cortex and hippocampus of male APP/PS1 AD vs. WT mice, the RS cortex of male and female APP/ PS1 AD vs. WT animals and the hippocampus of male and female APP/PS1 AD vs. WT mice were characterized using the Venn diagram function.The latter was performed using Gene-Spring GX 14.9 (Agilent Technologies Germany GmbH & Co. KG, Germany) or VENNY2.1 (https://bioinfogp.cnb.csic.es/tools/venny/index.html;[68]).A specific focus was set on signature genes/fingerprint gene sets, exclusively up-or downregulated in the individual subgroups.Note that for comparability reasons, the FC cutoff was set > 1.5 for upregulated genes and < -1.5 for downregulated genes.
Statistical analysis -general aspects.Unless otherwise specified, data were analyzed and presented in Microsoft Excel and GraphPad Prism (Version 9.1.2).Statistics were carried out as outlined above for DEGs and l(i)ncRNAs.Unless otherwise stated (see section 2.8.3.), a pvalue < 0.05 was considered significant (*).In many analytical settings (see above), p-values are presented as (-)log10(P) for better visualization.

Quantitative real-time PCR (qPCR)
Two quantitative real-time PCR (qPCR) approaches were carried out: the first was used to unravel whether selected DEGs in another AD mouse model, i.e., 5XFAD, are also differentially up-or downregulated in the APP/PS1 model [13,14,31,32].This approach focusses on AD line-specific and cross-line DEGs.For this purpose, qPCRs were carried out in eight WT controls (four ♂, age 32.71 ± 0.37 wks, four ♀, age 32.93 ± 0.21 wks) and eight APP/PS1 AD mice (four ♂, age 32.53 ± 0.32 wks, four ♀, age 33,29 ± 0.00 wks).Note, that the aforementioned animals were exclusively used for this approach and not for our microarray studies.The second qPCR approach was carried out to validate a selected number of cortical DEGs (Siglech, Ptpn6, Laptm5, Plek, Arpp21, Shisa9) from our transcriptome results.Here, RNA was used from the same animals as utilized for our microarray studies (eight APP/PS1 AD mice (four ♂, four ♀) and eight WT control animals (three ♂, five ♀), see above).
The cDNA synthesis from all RNA samples was performed using anchored-oligo(dt)18 and hexamer primer in a two-step reverse transcriptase (RT) PCR approach (Transcriptor First Strand cDNA Synthesis Kit, Qiagen GmbH, Germany).The qPCR reaction protocol was based on LightCycler 480 SYBR Green I Master (Roche, Roche Life Science, Germany).For details on primers for the specific gene candidates (Cacna1c, Cacna1d, Plcd4, Casp8, Chrm1, Chrm3, Chrm5, Siglech, Ptpn6, Laptm5, Plek, Arpp21, Shisa9) please refer to Siwek et al. (2015) [32] and S4 Table in S1 File.The qPCR was performed in a Light Cycler 480 System (Roche Life Science, Germany) thermocycler.The following cycler protocol was conducted for all primer pairs: 95˚C (10 min, pre-incubation step); 95˚C (10 s, melting step); 60˚C (20 s, annealing step); and 72˚C (30 s, extension step), 35 cycles.Melting curve analysis was performed to evaluate the specificity of the amplification.Product detection and characterization was based on SYBR Green I Master (Roche Life Science, Germany), a ready-to-use hot start reaction mix containing FastStart Taq DNA Polymerase and DNA double-strand-specific SYBR Green I dye.Deionized, nuclease-free water (no cDNA) and total RNA samples (RT excluded) were used as negative controls.The Hprt was used as an internal reference gene (positive control).The Ct (cycle threshold) values were calculated using the LightCycler 480 System software (Roche Life Science, Germany).The FCs of Cacna1c, Cacna1d, Plcd4, Casp8, Chrm1, Chrm3, and Chrm5 gene transcripts for the first qPCR approach and Siglech, Ptpn6, Laptm5, Plek, Arpp21, Shisa9 for the second qPCR approach in APP/PS1 transgenic mice related to WT controls were calculated according to Schmittgen & Livak, 2008 [69].The Cp values provided by the LightCycler 480 software (Roche, Germany) were imported to qBase+ software (Biogazelle, Belgium) and analyzed based on a delta-Cq quantification model with PCR efficiency correction, reference gene normalization considering the reference target stability of the selected housekeeping gene (Hprt) and inter-run calibration.The results were calculated as Calibrated Normalized Relative Quantity (CNRQ) and statistically investigated by the Mann-Whitney test.

Principal component analysis
PCA analysis as a data quality control measure was used to reduce the dimensionality of the large multi-dimensional transcriptome data sets, to increase their interpretability and to limit information loss at the same time (see 3D scatter plot in S3A Fig in S1 File).In these complex transcriptome data sets, the genes represent the variables and PCA allows us to get an impression of the similarity between samples.Based on a transformation of the high-dimensional transcriptome data into an orthogonal basis, the first principal component is aligned with the largest source of variance, the second principal component to the largest remaining source of variance and so on.Consequently, the similarities of the individual transcriptome data sets in individual subgroups are more amenable to visual exploration.
The PCA analysis in our study revealed that the individual samples related to the specific subgroups exhibit little separation between the groups of replicates (triplicates, quadruplicates, quintuplicates) and most of them cluster together and separate from arrays of other subgroups (S3A Fig in S1 File).It should be acknowledged here that the experimental animals are prone to increased variability per se.For example, the animals are housed in groups of 3-4 exhibiting characteristic social hierarchy, which can potentially exert substantial influence on CNS transcriptomics [70].Dominant versus subordinate male mice were reported to display distinct transcriptome patterns, particularly related to the immune and inflammatory system [70].The latter had been attributed to the higher corticosterone and stress levels in subordinate mice compared to alpha animals [70].Importantly, hierarchical phenomena are also present in female mice.The female estrus cycle does not seem to exert significant effects on aggressive behavior, although dominant females exhibit an elongated estrus cycle in comparison to subordinate females [71,72].Whereas the plasma estradiol levels were similar between dominant and subordinate females, subordinate females displayed largely elevated basal corticosterone levels compared to dominant females [71].Thus, the biochemical and physiological implications of social hierarchy need to be considered as an inevitable confounding factor.
PCA results were also plotted for genotype comparison (control vs. APP/PS1 samples, S3B Fig in S1

Heatmaps/hierachical clustering -chartography of DEGs
To visualize overall transcriptome readouts and the transcript/expression profile of genes across the individual subgroups, we performed a heatmap presentation.Hippocampus of male APP/PS1 mice.The Feature Extraction and GeneSpring RT software (both Agilent Technologies Germany GmbH & Co. KG, Germany) was utilized to list DEGs based on one-way ANOVA and Tukey post hoc testing followed by the Benjamini-Hochberg procedure for control of FDR.As depicted in S8 Table in S1 File, a total number of 354 gene candidates including l(i)ncRNAs were found including eleven candidates with a FC < -1.5 and 64 candidates with a FC > 1.5.The majority (67.80%) of DEGs was upregulated in this subgroup (240"/114#) (Figs 1A and 2D).(A) Illustration of the total number of DEGs in our four comparative subgroups, i.e., Rs Cx of female APP/PS1 mice, Hip of female APP/PS1 mice, RS Cx of male APP/PS1 animals and Hip of male APP/PS1 mice.All identified genes are based on a one-way ANOVA, Tukey and Benjamini-Hochberg procedure to identify significantly up-or downregulated genes.The total number of DEGs is higher in the hippocampus compared to the RS cortex.For the total number of upregulated genes, there seems to be no clear preference in BROI or sex.The same holds true for DEGs < -1.5 or > 1.5.Note that for most subgroups, the number of downregulated genes is much less compared to upregulated ones.(B) Illustration of the number of DEGs from a meta-analysis from 2114 post mortem AD patient samples performed by Wan et al. (2020).In contrast to our APP/PS1 study, the number of DEGs is persistently higher in post mortem probes from female AD subjects compared to probes from males.Overall, the numbers of DEGs are much higher compared to our APP/PS1 study.The latter might be due to the interindividual variability in the AD patients probes and the fact that Wan et al. (2020) included seven BROIs in their meta-analysis.For methodological details on their meta-analysis and statistics see [73].Volcano plots illustrating statistical significance vs. magnitude of FC.Profiles are depicted for the RS cortex of female APP/PS1 mice (A), the hippocampus of female APP/PS1 mice (B), the RS cortex of male APP/PS1 mice (C) and the hippocampus of male APP/PS1 animals (D).All genes exhibiting differential transcript levels were plotted with each dot representing one gene.The log2-FC in the individual APP/PS1 subgroups vs. WT is represented on the x-axis.The y-axis displays related log10 p-values (t-test procedure without further correction).The p-value of 0.05 and a FC of > 1.5 and < -1.5 are indicated by a horizontal line and vertical lines, respectively.Data points below the horizontal line represent gene candidates that were not significantly altered in transcription.Gene candidates in the different Volcano sectors (significantly up-or downregulated and FC > 1.5 or FC < -1.5) are color-coded.Note that downregulated gene candidates in the upper left sector (FC < -1.5, light orange) and upregulated genes in the upper right sector (FC > 1.5, blue) were further investigated in Venn and pathway studies.In most subgroups, the upregulated gene candidates predominate.Selected DEGs with high FC and statistical significance were individually labelled.Genes with significantly altered FC in the range from -1.5 to 1.5 (green and purple) were not further analyzed in our study.https://doi.org/10.1371/journal.pone.0296959.g002

GO and enrichment studies for upregulated genes (FC > 1.5) in the RS cortex and hippocampus of APP/PS1 AD mice compared to controls
RS cortex of female APP/PS1 mice.GO and enrichment analysis in the individual subgroups was performed using Metascape (metascape.org).Pathway and process enrichment analysis was carried out with the following ontology sources, i.e., GO Biological Processes, KEGG Pathway, Reactome Gene Sets, and WikiPathways.Importantly, all genes in the genome have been used as the enrichment background.Due to the number and proportion of the significantly up-and downregulated genes, the GO analysis and enrichment was performed for gene candidates with an absolute FC > 1.5.The top 20 enrichment clusters for upregulated DEGs in the RS cortex of female APP/PS1 mice compared to controls are depicted in Fig 3A .In addition, a detailed list of genes related to the top 10 identified clusters is given in Table 1A.The top three cluster terms/categories in this subgroup encompass (i) the microglia pathogen phagocytosis pathway (WP3626, Log(10)P: -23.06), (ii) the immune effector process (GO:0002252, Log(10)P: -18.88), and (iii) the innate immune response (GO:0045087, Log(10) P: -16.66).In addition, a PPI enrichment analysis was conducted using the following databases: STRING6, BioGrid7, OmniPath8, InWeb_IM9 (metascape org.).The top three PPI categories included (i) the microglia pathogen phagocytosis pathway (WP3626, Log(10)P: -25.5), (ii) the immune effector process (GO:0002252, Log(10)P: -17.3) and (iii) the neutrophil degranulation (R-MMU-6798695, Log(10)P: -15.4).
Comprehensive enrichment characteristics of upregulated DEGs of the RS cortex and hippocampus in male and female APP/PS1 mice compared to WT controls.Obviously, some cluster terms turned out to be most relevant in all four subgroups, i.e., the microglia pathogen phagocytosis pathway (WP3626), neutrophil degranulation (R-MMU-6798695) and phagocytosis (GO:0006909).Other common categories include the immune effector process (GO:0002252) and the Tyrobp causal network in microglia (WP3625) in three subgroups and the leukocyte activation (GO:0045321), macrophage markers (WP2271) and negative regulation of immune system process (GO:0002683) in two subgroups.Individual, subgroup-specific categories comprised, e.g., the leukocyte mediated immunity (GO:0002443) in the RS cortex of female APP/PS1 mice and the regulation of tumor necrosis factor production (GO:0032680) in the hippocampus of female APP/PS1 mice or sex-specific categories such as the innate immune response (GO:0045087) in the cortex and hippocampal of female APP/PS1 probes.
Overall, the majority of DEGs is related to the activation of microglial cells, leucocytes and macrophages/the immune system which is in accordance with previous studies [74].However, Table 1.Enrichment clusters and related upregulated DEGs.Enriched upregulated gene candidates (FC > 1.5, indicated via gene symbols) are depicted for the top 10 related categories/cluster terms derived from GO analysis (see also Fig 2) for (A) the RS cortex of female APP/PS1 AD mice, (B) the hippocampus of female APP/PS1 AD mice, (C) the RS cortex of male APP/PS1 AD mice, and (D) the hippocampus of male APP/PS1 AD mice.In addition, the Log(10)P values for the individual enriched categories/cluster terms are provided.there are additional, distinct sex-and brain-region related specificities in transcriptional profiles which are elaborated in detail below.GO and enrichment studies for downregulated DEGs (FC < -1.2) in the RS cortex and hippocampus of APP/PS1 AD mice compared to controls.As outlines above, the overall number of downregulated DEGs with a FC < -1.5 was severely lower compared to the upregulated DEGs (FC > 1.5).To carry out GO and enrichment studies with a sufficient number of downregulated gene candidates, the cutoff was set to a FC < -1.2.

Description/Category
RS cortex of female APP/PS1 mice.Analysis of downregulated DEGs of the RS cortex of female APP/PS1 vs. WT mice revealed no GO categories and significant enrichment in this subgroup.In addition, PPI enrichment analysis also revealed no significant categories (Fig 4A and Table 2A).
Hippocampus of male APP/PS1 mice.The three enrichment clusters for downregulated gene candidates of the hippocampus of male APP/PS1 mice compared to WT animals are depicted in Fig 4D .In addition, a detailed list of genes related to the identified clusters is shown in Table 2D.The three cluster terms/categories in this subgroup encompass (i) the centrosome cycle (GO:0007098, Log(10)P: -4.20), (ii) the histone monoubiquitination (GO:0010390, Log(10)P: -4.19), and (iii) endocytosis (GO:0006897, Log(10)P: -2.54).PPI enrichment analysis revealed only one category, i.e., histone modification (GO:0016570, Log (10)P: -2.7).Comprehensive enrichment characteristics for downregulated genes of the RS cortex and hippocampus in male and female APP/PS1 mice.Overall, the number of significantly downregulated DEGs was lower compared to the upregulated genes.Similarly, the number of significantly enriched categories was also reduced for downregulated gene candidates compared to the upregulated DEGs.

Venn analysis of upregulated DEGs (FC > 1.5) of the individual subgroups
Cross-regional and region-specific upregulation of DEGs in the RS cortex and hippocampus of female APP/PS1 mice vs. controls.We first investigated DEGs (FC > 1.5) in both the RS cortex and hippocampus of female APP/PS1 vs. WT mice (Fig 5A).In total, 22 candidates were upregulated solely in the hippocampus, 36 genes were upregulated only in the RS cortex, but 55 candidates exhibited an increased, intersectional transcription profile in both RS cortex and hippocampus.The individual gene candidates are also listed in S9A Table in Table 2. Enrichment clusters and related downregulated DEGs.Enriched downregulated gene candidates (FC < -1.2, indicated via gene symbols) are depicted for the related categories/cluster terms derived from GO analysis (see also Fig 3) for (A) the RS cortex of female APP/PS1 AD mice, (B) the hippocampus of female APP/PS1 AD mice, (C) the RS cortex of male APP/PS1 AD mice, and (D) the hippocampus of male APP/PS1 AD mice.Note that GO analysis and enrichment did not reveal significant results for the RS cortex of female APP/PS1 AD animals.The Log(10)P values for the individual enriched categories/cluster terms are also provided.Importantly, the number of enriched clusters and related gene candidates is considerably reduced for downregulated compared to upregulated gene candidates (see also Table 1  S1 File.The cross-area GO and enrichment profiling revealed the following top three categories for the 55 upregulated, intersectional gene candidates: (i) the microglia pathogen phagocytosis pathway (WP3626, Log(10)P: -15.07), (ii) the innate immune response (GO:0045087, Log (10)P: -12.94), and (iii) the immune effector process (GO:0002252, Log(10)P: -11.81).Notably, these co-upregulated candidates represent the majority of upregulated genes in both subgroups.Although it is interesting to point out gene candidates that are co-upregulated in both the RS cortex and hippocampus of female APP/PS1 mice, it is also important to consider the region-specific candidates that are exclusively upregulated in either the RS cortex or the hippocampus of female AD mice.
Interestingly, the number of upregulated, intersectional DEGs in both RS cortex and hippocampus of female APP/PS1 mice vs. controls was much higher than the related number in male APP/PS1 mice compared to WT animals (55 vs. 33).
Sex-and region-specific transcriptome fingerprints / signature genes in APP/PS1 AD mice vs. controls.A Venn analysis including DEGs in all four APP/PS1 subgroups is depicted in Fig 5E .Note that there are sets of DEGs that are exclusively upregulated in only one of these APP/PS1 subgroups (vs.controls) and can thus serve as molecular/transcriptome fingerprint genes/signature genes for BROI-and sex-specific alterations in this APP/PS1 AD model.In the RS cortex of female APP/PS1 this includes 25 out of 91 DEGs, 17 out of 77 DEGs in the hippocampus of female APP/PS1, only two out of 57 DEGs in the RS cortex of male APP/PS1 mice and finally, 18 out of 57 DEGs in the hippocampus of male APP/PS1 animals.
The two upregulated signature genes in the male APP/PS1 RS cortex (Derl3 and Mki67, Fig 5E ) revealed no significant enrichment terms.
In addition, we performed pathway studies using the Reactome Pathway Database (for details, please refer to "reactome.org").We first analyzed intersectional gene sets for upregulated DEGs (FC > 1.5).Notably, for all intersectional genes sets (RS cortex and hippocampus of female APP/PS1 mice, RS cortex and hippocampus of male APP/PS1 mice, male and female RS cortex of APP/PS1 mice, male and female hippocampus of APP/PS1 mice, see also Fig 5A -5D), the top three pathways were related to (i) the innate immune system, (ii) the immune system and (iii) neutrophil degranulation (S6 Fig in S1 File).Thus, the transcriptional overlay between the individual subgroups is very much conserved.We then investigated pathway characteristics for the significantly upregulated signature genes of the individual subgroups (Fig 5E).The signature gene sets again turned out to exhibit pathway specificities: In the female APP/PS1 hippocampus, the top three pathways included (i) interleukin-4 and interleukin-13 signaling, (ii) interferon signaling, and (iii) aggrephagy.In the female APP/PS1 RS cortex, the top three pathways were related to (i) hyaluronan uptake and degradation, (ii) hyaluronan metabolism, and (iii) the immune system.The most relevant signature pathways in the male APPPS1 hippocampus included (i) the transcriptional regulation of granulopoiesis, (ii) cytokine signaling in the immune system, and (iii) interferon gamma signaling.In the male APPPS1 RS cortex, the latter comprised (i) defective CFTR causes cystic fibrosis, (ii) ABC transporter disorders, and (iii) ABC-family proteins mediated transport (S6 Fig in S1 File).

Venn analysis of downregulated DEGs (FC < -1.5) in the individual subgroups
We first investigated DEGs (FC < -1.5) in both the RS cortex and hippocampus of female APP/PS1 vs. WT mice (Fig 6A).In total, 14 candidates were exclusively downregulated in the female hippocampus of APP/PS1 mice, two were exclusively downregulated in the cortex of female APP/PS1 mice, but only one candidate was co-downregulated in both the cortex and hippocampus of female APP/PS1 AD vs. WT mice, i.e., Pisd-ps3 (S9B Table in S1 File).Candidate gene transcripts that were selectively downregulated in the RS cortex of female APP/PS1 mice included LOC105247294 and Ptpn6.The DEGs that were selectively downregulated in the hippocampus of female APP/PS1 mice comprised Asic4, Pcsk1, Etnppl, Myh7b, Spag5, Arpp21, Gpr1, Myh8, Ccdc184, Pdzph1, Trim66, BC030499, Rims3, Prr16.
Secondly, we investigated the DEGs (FC < -1.5) in both the RS cortex and hippocampus of male APP/PS1 vs. WT mice (Fig 6B).Six candidates were exclusively downregulated in the hippocampus of male APP/PS1 mice, four were exclusively downregulated in the cortex of male APP/PS1 mice, but only one candidate was co-downregulated in both regions, i.e., Arpp21 (S10B Table in S1 File).Candidate gene transcripts that were selectively downregulated in the RS cortex of male APP/PS1 mice included Scyl2, Ctr9, Scyl2, and 1700001L05Rik.Candidate gene transcripts that were selectively downregulated in the hippocampus of male APP/PS1 mice comprised 4933407I18Rik, Hlcs, Hdac9, Pla2g4e, Wnt9a, Shisa9.
Next, we investigated the DEGs (FC < -1.5) in the RS cortex of both male and female APP/ PS1 vs. WT mice (Fig 6C).Three candidates were exclusively downregulated in the RS cortex of female APP/PS1 mice, five were exclusively downregulated in the cortex of male APP/PS1 mice, but no intersectional gene candidates were detected in the RS cortex of both sexes (S11B Table in S1 File).In the RS cortex of male APP/PS1 mice only the following genes were downregulated: 6530402F18Rik, Ctr9, Arpp21, Scyl2, 1700001L05Rik.In the RS cortex of females, only the following genes were downregulated: LOC105247294, Ptpn6, Pisd-ps3.
A Venn analysis including intersectional and signature DEGs in all four APP/PS1 subgroups compared to WT animals is depicted in Fig 6E .Note that there are again sets of DEGs that are exclusively downregulated in only one of these subgroups and can thus serve as molecular/transcriptome fingerprints/signature gene sets for BROI-and sex-specific alterations in this APP/PS1 AD model.In the RS cortex of female APP/PS1 mice, this includes two out of three DEGs, 13 out of 15 DEGs in the hippocampus of female APP/PS1, four out of five DEGs in the RS cortex of male APP/PS1 mice and finally, six out of seven DEGs in the hippocampus of male APP/PS1 animals.Interestingly, although the number of downregulated DEGs is significantly lower compared to the upregulated candidates, the majority of downregulated DEGs contributes to their related BROI-and sex-specific transcriptome fingerprints.Gene ontology and enrichment studies using Metascape did not reveal significantly enriched terms/ categories.
Due to the minimal or lacking number of intersectional downregulated genes, pathway analysis was not applicable here (Fig 6A -6D).The downregulated signature gene sets however, revealed interesting specificities (Fig 6E): In the female hippocampus of APP/PS1 mice, the top three signature pathways included prostanoid ligand receptors, peptide hormone biosynthesis, and the synthesis, secretion, and inactivation of Glucose-dependent insulinotropic polypeptide (GIP).In the female RS cortex of APP/PS1 mice, the top signature pathways encompassed interleukin-37 signaling, interleukin-1 family signaling, and PECAM1 interactions.In the hippocampus of male APP/PS1 mice, authenticator genes were related to the following pathways: defective HLCS related multiple carboxylase deficiency, defects in biotin metabolism, and hydrolysis of LPC.Finally, top signature pathways in the RS cortex of male APP/PS1 mice were related to RNA polymerase II transcription elongation, formation of RNA Pol II elongation complex, and E3 ubiquitin ligases ubiquitinate target proteins (S6 Fig in S1 File).

Differentially regulated l(i)ncRNAs in the cortex and hippocampus of male and female APP/PS1 mice compared to WT controls
Long (intergenic/intervening) non-coding RNAs (l(i)ncRNAs) represent a class of non protein-coding RNAs (> 200 nt) that serve important regulatory functions in transcriptional silencing or activation, chromosomal modification and intranuclear trafficking.L(i)ncRNAs are predominantly classified as antisense l(i)ncRNAs, intronic transcripts, large intergenic noncoding RNAs, promoter-associated l(i)ncRNAs and UTR associated l(i)ncRNAs.
Using the Feature Extraction tool (Agilent Technologies Germany GmbH & Co. KG, Germany), we selectively evaluated differentially regulated l(i)ncRNA entities in all four subgroups (S13 Table in S1 File illustrates the significantly altered l(i)ncRNAs, the related FCs, p-values and sequence characteristics).Statistics are based on one-way ANOVA, Tukey post hoc analysis and the Benjamini-Hochberg procedure.With only one exception (hippocampus of female APP/PS1 mice), all l(i)ncRNA related FCs were in the range of -1.5 to 1.5.Notably, the majority of differentially regulated l(i)ncRNAs was downregulated compared to WT controls.In RS cortex probes of female APP/PS1 mice, only two lincRNA candidates were identified (S13A Table in S1 File), whereas 14 l(i)ncRNAs were found in hippocampal probes of female APP/ PS1 (S13B Table in S1 File).13 l(i)ncRNAs were found in the RS cortex of male APP/PS1 (S13C Table in S1 File) and 13 l(i)ncRNAs in the hippocampus of male APP/PS1 mice (S13D
All l(i)ncRNAs differentially transcribed in the individual subgroups in our study turned out to be annotated by the NONCODE database.One l(i)ncRNA sequence was annotated by five databases with known presence in the hippocampus and reports about involvement in epigenetic effects [79].In most cases, database information about l(i)ncRNA tissue distribution nicely paralleled our findings.However, detailed functional implications of up-or downregulated l(i)ncRNAs in the pathophysiology of the individual APP/PS1 subgroups remain to be determined in the future.

qPCR analysis of selected gene candidates relevant in AD etiopathogenesis
Comparison of selected DEGs between APP/PS1 and 5XFAD mice.Our own previous radiotelemetric EEG studies from the cortex and hippocampus of 5XFAD mice, another model of AD disease, had revealed complex alterations in CNS rhythmicity [31,32].Similarly, our electrophysiological studies in APP/PS1 mice had showed age-and sex-dependent differences in the relative power of the theta and gamma frequency range in the cortex and hippocampus [13,14].Our own previous transcriptome studies in 5XFAD mice had revealed some gene candidates that might be functionally relevant for the electrophysiological alterations observed and some had been validated using qPCR [31,32].As APP/PS1 mice exhibit similar phenotypical features as 5XFAD animals, we analyzed the same set of critical 5XFAD related genes via qPCR in the APP/PS1 model, i.e., Cacna1c, Cacna1d, Plcd4, Casp8, Chrm1, Chrm3, and Chrm5, to elaborate whether there are common, relevant DEGs in both AD mouse models (first qPCR approach).In female APP/PS1 mice, Caspase 8 (encoded by Casp8) was significantly upregulated compared to WT controls (FC: 1.3674, p = 0.028), but not the other candidates (S14A Table and S7A and S7B Fig in S1 File).Interestingly, Caspase 8 did not exhibit altered expression in male APP/PS1 mice (FC: 1.0892, p = 0.685).In contrast, the high-voltage activated L-type Ca 2+ channel Ca v 1.3 (encoded by cacna1d) was significantly downregulated in male APP/PS1 mice (FC: -1.2122, p = 0.0285) but not in female APP/PS1 animals (FC: 1.1681, p = 0.114) (S7E and S7F Fig in S1 File).This sex-specific difference in Ca v 1.3 transcript levels in APP/PS1 mice was statistically confirmed (FC: 1.4113, p = 0.0285) (S7E and S7F Fig in S1 File).For the high-voltage activated L-type Ca 2+ channel Ca v 1.2 (encoded by cacna1c), no genotype-and sex-specific difference in transcript levels was detected.A significant difference in transcript levels between male and female APP/PS1 mice was also observed for Chrm1, the gene encoding the muscarinic receptor type 1 (M1) (S7I and S7J Fig in S1 File).These findings further stress the fundamental relevance of sex-specific analysis as carried out in our transcriptome studies.
Our transcriptome data revealed no significant up-or downregulation of Casp8, Plcd, Chrm1, Chrm3 and Chrm5.In the RS cortex and hippocampus of female APP/PS1 mice and the RS cortex of male APP/PS1 mice, PLCgamma was altered (FC: 1.3, FC: 1.4 and FC: 1.3, respectively).In the hippocampus of male APP/PS1 mice, PlCd3 was downregulated (FC: -1.2).Importantly, cacna1d encoding the Ca v 1.3 HVA Ca 2+ channel was listed as a DEG in the RS cortex of male APP/PS1 mice, but not in the hippocampus.This phenomenon is due to the applied strict statistical procedure including one-way ANOVA, Tukey post hoc test and the application of the Benjamini-Hochberg procedure to decrease the FDR.The DEG extraction without application of post-hoc correction again included cacna1d as a differentially regulated gene candidate.These qPCR findings underline that different AD mouse lines, i.e., 5XFAD and APP/PS1, can clearly exhibit different transcriptome characteristics.
Validation of selected DEGs in APP/PS1 mice.In the second qPCR approach, we validated a selected number of DEGs from our APP/PS1 transcriptome study, i.e., four upregulated (Siglech, Ptpn6, Laptm5, Plek, Fig 7A -7H) and two downregulated (Arpp21, Shisa9, Fig 7I -7L) genes.These DEGs were analyzed in the RS cortex of both male and female APP/PS1 animals (same mice (RNA) as utilized for our transcriptome studies).As qPCR analysis of all relevant DEGs, including intersectional and signature genes, would be beyond the scope of our study, we focused on a selected number of candidates in the RS cortex.The latter is early and severely affected in AD pathogenesis.Importantly, the four DEGs Siglech, Ptpn6, Laptm5, and Plek which were suggested to be significantly upregulated in female APP/PS1 mice in our transcriptome study, were also significantly upregulated in the qPCRs (Fig 7B , 7D, 7F and 7H and S14B Table in S1 File).The same held true for the qPCR validation of Siglech, Ptpn6, Laptm5, and Plek in the male RS cortex of APP/PS1 animals (Fig 7A, 7C and 7E) with Siglech exhibiting a statistical trend (S14B Table in S1 File).In addition, the FC-values from the transcriptome study and the qPCR validation nicely correlate.Note that many DEGs (such as Ptpn6) exhibit different transcript variants which can be up-or downregulated.For Ptpn6, the upregulated transcript variants were validated.
The two DEGs that were suggested to be downregulated in our microarray analyses (Arpp21, Shisa9), were also confirmed in our qPCR approach in the RS cortex of female APP/ PS1 mice (Fig 7J and 7L and S14B Table in S1 File).In the RS cortex of male transgenics however, significant downregulation of Shisa9 could not be confirmed (S14B Table in S1 File).Overall, our qPCR findings underline the validity of our transcriptome studies and the detected DEGs.

The role of APP/PS1 mice in AD research and transcriptomics-previous results and open challenges addressed in our study
The APP/PS1 AD mouse model investigated in our study exhibits moderate severity and disease progression that well resembles AD in humans based on the categories of homology, isomorphism, and predictability [19,22].We previously performed complex electrophysiological in vivo analysis of APP/PS1 mice using implantable EEG radiotelemetry to elucidate defined alterations in hippocampal and cortical rhythmicity via FFT based frequency analysis [13,14].We detected region-, age-and sex-specific differences in EEG characteristics that stresstogether with findings of other groups -that functional alterations in AD pathophysiology clearly require subgroup-specific analyses [13,14,80].Although transcriptome studies have previously been carried out in APP/PS1 AD mice, these studies exhibit some limitations in design that affect their informative value.The latter is mainly related to the combined analysis of sexes [37,38,81], the selective analysis of males only [41][42][43], an unspecified sex distribution [39,82], analysis of only one brain region [37,39,40,43,81], investigation of no distinct brain region (e.g., cortex in total [41,42] or the analysis of entire brain probes [38] (S15 Table in S1 File).Based on our previous findings on distinct electrophysiological alterations in the APP/ PS1 AD model related to sex, age and brain region [13,14], we performed transcriptome studies in the RS cortex and hippocampus of both sexes in WT controls and APP/PS1 mice.

Role of the RS cortex and hippocampus in AD etiopathogenesis and disease progression
The cortex is known to be early affected in AD pathophysiology both on the structural and functional level [52,83].Interestingly, progression of AD-specific lesions severely differs between individual cortical regions.Whereas the (pre)frontal cortex is early involved, alterations in the motor cortex seem to be delayed [84,85].We have chosen the RS cortex for our transcriptome studies, as its early and severely affected in AD pathology and symptomatology [50,86].As another predilection area in the etiopathogenesis of AD, the hippocampus was chosen as a second BROI for our transcriptome analysis.The hippocampus is judged to be a key interface structure in the brain and involved in the consolidation of memory engrams and their ecphoria [47,87].Impaired and dysfunctional hippocampal activity together with complex structural alterations related to extracellularly localized Aβ plaque formation and intracellular τ-deposits in APP/PS1 AD mice are typically accompanied by sophisticated deficits in memory formation, and cognitive/behavioral tests [88,89].

Functional relevance of selected DEGs in APP/PS1 AD mice
Our data illustrate, that key gene candidates in AD etiopathogenesis are differentially regulated in brain regions of male and female APP/PS1 mice compared to WT controls (Tables 3 and  4).We identified some gene candidates that had already been related to AD, whereas others have not yet been attributed in detail or at all to its pathophysiology.An overall discussion of the entire set of high fold change DEGs in the individual subgroups is beyond the scope of our study.However, we have pointed out some functional aspects of selected DEGs and how they might influence the etiopathogenesis of AD with a focus on sex-and region-specific distribution.As outlined below, DEGs are predominantly related to the neuroinflammatory response, the activation of microglia, macrophages and neurophil granulocytes [37,90].Many DEGs further contribute to innate immune reactions and immune effector processes that are largely triggered by Aβ [90,91].
Intersectional DEGs in microglial/neutrophilic activation.Microglial activity seems to play a critical role in the etiopathogenesis of AD and the devastating structural and functional alterations that encompass progression of the disease [92][93][94].Microglia represent the resident immune cells within the CNS and share some phenotypic markers with myeloid cells of the macrophage/monocyte lineage [95].Microglia have been functionally segregated into proinflammatory M1 and anti-inflammatory M2 subgroups based on their molecular and functional profile [96,97], though the entire activation spectrum between the maxima seems to a continuum [74,98].Strong efforts have been made to decipher the phenotypic diversity of microglial cells in the healthy, aging, and AD brains of mice and humans.AD-dysprogrammed microglia are often referred to as disease-associated microglia (DAM) [99], microglia associated with neurodegenerative disease (MGnD) [100], or activated response microglia (ARM) [101].The delineation of transcriptome profiles in these microglia has led to the specification of gene candidates related to the homeostatic signature of microglia, differentiation of microglia vs. myeloid populations, microglia vs. macrophages, the microglial surfaceome, and the microglial sensome [74,102].Detailed comprehensive lists of microglia associated transcripts are provided by Boche and Gordon (2020) [74] and Crotti A et al. (2016) [74,103].Table 4. Overview of the sex-and region-specific downregulation of DEGs in the four APP/PS1 subgroups.This table illustrates all downregulated DEGs (FC < -1.5) from APP/PS1 microarray analysis.DEGs that are downregulated only in one APP/PS1 subgroup are highlighted in red, those exhibiting enhanced transcript levels in two subgroups are marked in green.Increased transcript levels in three subgroups are marked in ocher, those in all four subgroups are highlighted in yellow.The table elicits the complex sex-and brain region specific profile of downregulated DEGs.Recent advances in single-cell-omic methodologies have dramatically increased the availability of gene expression data from microglia in healthy and aged brains, and AD [74].Previous studies by Mukherjee et al. (2019) described DEGs in microglia related to neurodegenerative diseases, including Tyrobp, Fcer1g, Itgb2, Myo1f, Ptprc, Trem2 and C1qa [104].We were able to confirm many of these gene candidates in our experimental subgroups and most important, in both sexes.The Fcer1g gene for example, plays an important role in IgE-specific FC gamma receptor mediated phagocytosis and microglial activation [105].Importantly, upon activation of Trem2, Syk is recruited by Fcer1g and Tyrobp [106,107].Notably, Tyrobp and Trem2 were upregulated in all four APP/PS1 subgroups in our study, similar to Fcer1g which however, remained unaltered in the hippocampus of male APP/PS1 mice.Some DEGs that were upregulated in our study are functionally related to leucocyte migration, e.g., Itgb2 [108].Myo1f is important in neutrophil degranulation and regulation of cell migration as well [108].Rac2 encodes a protein with GTPase and protein kinase regulator activity and is involved in the regulation of leukocyte activation and leukocyte chemotaxis.It belongs to the Rho family GTPases which are master regulators of actin cytoskeletal organization and associated with a number of neuropsychiatric and neurodegenerative diseases, including AD [109].

APP/PS1 DEGs (FC
We further identified several cluster of differentiation (Cd) encoding genes to be upregulated in APP/PS1 AD mice.Most of these Cd genes, i.e., Cd48, Cd52 and Cd68, are involved in antigen binding, signaling and receptor activity.Some, like Cd68, were reported to be involved in aging and cellular response to organic substances with critical localization in lysosomes within the brain.The gene product of Trem2 forms a receptor signaling complex that activates myeloid cells, including dendritic cells and microglia.In humans and mice, mutations in Trem2 may serve as risk factors for the development of AD [110].
Further relevant genes are related to the interleukin, interferon, and cytokine system, e.g., the C-C motif chemokine 6, encoded by Ccl6, one of the top FC candidates in our study.The latter displays expression in microglia and astrocytes.Interestingly, treatment of neurons with CCL6 revealed protective effects against glutamate neurotoxicity.This neuroprotective effect of CCL6 is mediated via the chemokine receptor CCR1, and downstream by PI3K and likely to be another protective mechanism in AD pathophysiology [111].For additional description of the DEGs, please refer also to S16 Table in S1 File.
Intersectional genes in phagocytic activation.Phagocytosis is an important field in AD pathogenesis [112] and numerous DEGs in our study groups are related to this process.Capg (encoding for capping actin protein, gelsolin like), for example, is a member of the gelsolin/villin family of actin-regulatory proteins and was reported to play an important role in phagocytic vesicles and actin-related intracellular processes [113].Others, like Fcgr (FC gamma receptor) are also involved in endocytosis similar to Fcgr2b [114].The latter is located in dendritic spines and involved in phagocytosis, cytotoxicity and cellular response to Aβ [114].Studies in both mice and humans suggest that FcγR mediates a pro-inflammatory response which could also be responsible for vascular side effects following application of therapeutic antibodies against Aβ [115].Increasing evidence has emerged that FcγR expression in microglia and neurons is enhanced upon aging and involved in the etiopathogenesis of AD [115].Glial fibrillary acid protein encoded by Gfap is another example of a structural constituent of the cytoskeleton that is involved in chaperone-mediated autophagy and intracellular protein transport.It was reported to be involved in long-term potentiation (LTP) and neurogenesis.The Gfap gene product is currently under discussion for its utility as a plasma biomarker for AD-related pathologies [116].Itg2 (coding for integrin beta2), Laptm5 (encoding for lysosomal-associated protein transmembrane 5) and Lcp1 (encoding for lymphocyte cytosolic protein 1) are involved in cargo receptor processes and endocytosis, lysosomal transport and filament bundle assembly located in phagocytic cups [117].Notably, we were able to validate Laptm5 transcript upregulation via qPCR in the RS cortex of both male and female APP/PS1 mice.
The prion protein encoded by Prnp exhibited one of the highest fold-changes in our study.It serves several functions, including Aβ binding activity, regulation of protein localization to membranes, protein phosphorylation, aspartic-type endopeptidase inhibitor activity, negative regulation of apoptotic processes and modulation of phagocytosis [118].There is likely to be an association between AD and cellular prion protein (PrPC) levels.PrPC exerts high affinity for oligomeric Aβ (particularly Aβ42) and mediates Aβ-induced neurotoxicity in AD pathogenesis [119].Synaptotoxic Aβ assemblies further mediate long-term depression (LTD) via action on PrPC [120].PrPC is decreased in the hippocampus and temporal cortex in aging and sporadic AD but not in familial Alzheimer's disease (FAD) [121].Another study also suggested a reduced expression of PrPC in AD patients compared to healthy individuals, which pointed to a potential protective role of PrPC expression in AD [122], although this remains controversial [123].However, with symptomatic clinical manifestation, PrPC expression seemed to decrease again [123].There is clearly an ambiguous picture of PrPC expression levels in AD that needs future clarification [118].Interestingly, Prnp was strongly co-upregulated in the RS cortex and hippocampus of both female and male APP/PS1 mice in our study.
One of the highest FC candidates in our study turned out to be glycosidase lysozyme 2 (Lyz2).Alzheimer's disease patients were reported to exhibit increased lysozyme levels in the cerebrospinal fluid.In addition, lysozyme 2 co-localizes with Aβ and exerts protective effects against Aβ associated neurodegenerative action.Importantly, lysozyme was proposed as a potential biomarker and therapeutic target in AD [124].
The gene product of Siglech (sialic acid binding Ig-like lectin H, possibly an ortholog of CD33 in humans) has cargo receptor activity and acts upstream of or within receptor-mediated endocytosis and was reported to serve as a risk factor for late-onset AD [125].Importantly, Siglech was upregulated in each APP/PS1 subgroup in our study despite the hippocampus of male AD animals.Our qPCR experiments further validated the upregulation of Siglech in the cortex of APP/PS1 mice.For additional DEG profile information see also S16 Table in S1 File.
Intersectional genes encoding for endopeptidases in APP/PS1 AD mice.Endopeptidases play an important role in AD pathology and are potential drug target candidates and biomarkers with high predictive efficacy [126].We identified several endopeptidases to be upregulated in the different APP/PS1 subgroups.The latter include Bcl2a1d (encoding for B cell leukemia/lymphoma 2 related protein A1d), a cysteine-type endopeptidase involved in apoptotic processes and Ctsd, an aspartic-type endopeptidase (cathepsin D) localized in the extracellular space and lysosomes and involved in autophagosome assembly [127].Bcl2a1d turned out to be a top fold-change candidate in our study.Other upregulated endopeptidases include Ctsh, a cysteine protease (cathepsin H) that can form both aminopeptidase and endopeptidase activity depending on its posttranslational processing and assembly.Ctss represents another upregulated peptidase C1 family member of cysteine protease (cathepsin S, like Ctsh).Naip2 (NLR family, apoptosis inhibitory protein 2) encodes a protein with ATP binding activity and cysteine-type endopeptidase inhibitor activity involved in apoptotic process [128].Naip2 was upregulated in all four APP/PS1 subgroups in our study.
Again, our study revealed sex-specific characteristics in differential regulation: the top foldchange gene candidate Bcl2a1d was upregulated in all APP/PS1 groups, whereas Ctsd and Ctss were not upregulated in the female hippocampus, and Ctsh was not upregulated in the RS cortex of male and female APP/PS1 animals.For further details please refer to S16 Table in S1 File.
Oxidative stress and reactive oxygen species in APP/PS1 AD mice.Reactive oxygen species (ROS) play an important role in AD pathogenesis and microglia-associated neuroinflammation [129].DEGs such as Cyba and Ncf are part of the NADPH oxidase complex.They are involved in ROS metabolic processes, Ca 2+ homeostasis and AD pathogenesis [130].The Nrf2 encoded NF-E2-related-factor 2 was reported to suppress inflammation and oxidative stress in an APP knock-in AD mouse models and could thus exert protective effects [131].For additional information see also S16 Table in S1 File.

Sex-and region-specific alterations in gene transcript levels in APP/PS1 mice
In some AD mouse models and humans, sex-specific differences in transcriptome profiles have been partially documented.Transcriptional human-mouse intersections are mainly related to Aβ/τ pathology and microglia and exhibit sex-dependent spatiotemporal signatures upon disease progression.The results from RNA-seq of human post mortem brain tissue stressed the relevance of microglial and inflammatory mechanisms in AD pathogenesis, among other pathways [73,[132][133][134].Notably, different from transcriptome profiles in human brains, mouse models allow more controlled experimental conditions that can be used to investigate causation, isolated effects of specific molecular or pharmacological interventions, and elaborate spatiotemporal alterations.Although transcriptional overlaps have been identified between AD mouse models and human brains in some studies [73,[135][136][137], other reports have questioned the overall degree of conservation [113,138].Regardless of these discrepancies, sex/gender has been suggested as a strong modulator in AD etiopathogenesis [73,139,140].
Most studies carried out in APP/PS1 mice and other AD models do not provide sex-specific analysis of transcriptome data (see S15 Table in S1 File).In the following, we have elaborated the sex-specific profile of the major DEGs detected in our study.Importantly, these gene candidates turned out to be signature genes (Fig 5E).In addition, many of them contribute to the devastating AD pathophysiology.However, some (as outline below) also exert counteracting, autoprotective effects.
Functional relevance of signature genes in the hippocampus of female APP/PS1 mice.In our study design, 17 gene candidates turned out to be exclusively upregulated in the hippocampus of female APP/PS1 AD mice, i.e., Prss23, Slc38a5, Ifitm3, Icam1, Dclk1, Rfx3, Tubb6, Slfn5, Flnc, Apobec1, Lsp1, Myo1g, Vim, 5830408C22Rik, Ptgr1, Pla2g4a, Rab32.The latter are related to the following GO categories: epithelial cell development, membrane trafficking, supramolecular fiber organization, and the adaptive immune system.Two related signature genes are highlighted below: Intercellular adhesion molecule-1 (Icam).An interesting candidate in our study turned out to be Icam1, encoding for the intercellular adhesion molecule-1, a transmembrane glycoprotein that is reported to be overexpressed in many neuropathological settings.Icam1, which is known to play an important role in the immune response, is expressed in the CNS, particularly in astrocytes, microglial and endothelial cells in the white and gray matter of the human forebrain.Icam1 has been attributed to the etiopathogenesis of neurodegenerative, e.g., AD and neuropsychiatric disorders, due to its functional involvement in blood-brain barrier function and as a marker for inflammation [141].Interestingly, the Icam1 gene product was suggested to be neuroprotective and a potential candidate in cytokine-mediated therapy of AD [141].Studies in humans revealed that the Icam1 gene product was increased in the cerebrospinal fluid (CSF) in preclinical, prodromal, and dementia stages of AD, associated with cortical thinning and cognitive deterioration and could act as a biomarker for early neuroinflammation and an AD risk assessment factor [142].Although Icam1 was previously suggested to serve as a CSF biomarker for AD, we found Icam1 to be significantly upregulated (FC > 1.5) only in the hippocampus of female APP/PS1 mice and it turned out to be one of the differentially expressed fingerprint genes in this subgroup.
Tubulin beta 6 (Tubb6).Tubb6 encodes for a tubulin isotype, i.e., tubulin beta 6 class V with GTP binding activity, a constituent of the cytoskeleton being involved in microtubule and actin dynamics.Microtubule dysfunction is a prominent feature in many neurodegenerative diseases including AD [143].Tubb6 mRNA was previously detected in the hippocampus of C57BL/6 mice and mRNA levels were reported to be low and to further decrease in early age [144].Proteome studies indicated an overrepresentation of tubulin beta 6 in mouse models of AD [145].Importantly, in our transcriptome studies, Tubb6 was only upregulated in the hippocampus of female APP/PS1 mice, not in the RS cortex and not in male APP/PS1 mice.
Functional relevance of signature genes in the RS cortex of female APP/PS1 mice.In the RS cortex of female APP/PS1 AD mice, 25 candidates exhibited signature character, i.e., Gusb, Slc14a1, Mir142hg, Cyth4, Gpr34, Plek, Ccl9, Aif1, Psmb9, Lair1, Hk3, Csf2rb, Tlr12, Adora3, Apbb1ip, Pycard, Vav1, Oas1a, Ifi27l2a, Hexb, Ncf4, Fcrls, Irf5, Havcr2, Mlxipl.These specific DEGs turned out be related to the fields of inflammatory response, Tyrobp causal network in microglia, neutrophil degranulation, phagocytosis and the RAF/MAP kinase cascade.Some of the related gene candidates and their implications in AD are further illuminated below: Pleckstrin (Plek).A most interesting DEG selectively upregulated in the RS cortex of female APP/PS1 mice turned out to be Plek.Pleckstrin is an important protein in cytoskeleton reorganization and neurite outgrowth and is supposed to be of functional relevance in AD pathophysiology as the structural and functional integrity/dynamics of the cytoskeleton are a prerequisite for neuronal function and survival [146,147].Pleckstrin (Plek) turned to be a negatively regulated target of miR-409-5p, a miRNA that impairs neurite outgrowth.Interestingly, Aβ 1-42 peptides significantly decrease cortical expression of miR-409-5p, resulting in less decrease of neuronal viability, and reduction in subsequent Aβ 1-42 -induced pathologies [147].In contrast, overexpression of miR-409-5p exerts neurotoxic effects in neuronal cells and impairs differentiation.Overexpression of Plek, on the other hand, harbors the capability to rescue neurite outgrowth from this neurotoxicity [147].It has previously reported that miR-409-5p is early downregulated in APP/PS1 mice and that Plek is upregulated at 9 and 12, but not 4 and 6 months-old APP/PS1 animals [147].Our study reveals that Plek is upregulated in 8 months old animals, and according to our transcriptome results, a signature gene in the RS cortex of female APP/PS1 mice.
Strikingly, the previously reported early downregulation of miR-409-5p in combination with the observed upregulation of Plek in AD progression might thus point to a counteracting, compensatory and autoprotective reaction to alleviate the structural and functional synaptic impairment induced by Aβ.Plek and its gene product Pleckstrin could thus serve as potential early biomarker in AD.
Adenosine receptor type 3 (Adora 3).Some upregulated genes, such as Adora 3, are not only involved in microglia activation but also neutrophil degranulation and leucocyte migration/chemotaxis. Adora3 encodes a G-protein coupled adenosine receptor type 3, which is supposed to mediate inhibition of neutrophil degranulation and therefore potentially prevents cell damage.However, it was also reported to exhibit ambiguous effects in neuroprotection and neurodegeneration [108,148,149] stressing that modulation of the adenosinergic system in AD, in terms of pathology and therapeutics, is a sophisticated issue [150].Despite this potential intractability and the lack of medicines targeting adenosinergic receptors for AD treatment, substantial efforts are made in drug research and development.Importantly, Adora3 turned out to be a unique fingerprint candidate for the RS cortex of female APP/PS1 mice in our study.This sex-specific and potentially neuroprotective profile of Adora3 warrants future attention also in drug development and potential individualized pharmaceutical treatment in AD.
Vav guanine nucleotide exchange factor 1 (Vav1).Vav1 encodes a guanine nucleotide exchange factor (GEF) for Rho family GTPases that activate pathways leading to actin cytoskeletal rearrangements and transcriptional alterations.It serves as a functional regulator of Rac2.It acts upstream of or within several processes, including integrin-mediated signaling pathway and neutrophil chemotaxis [151].Vav1 also seems to promote learning by activating HIF-1 and GLUT-1 and thereby contributing to glucose distribution to the brain, another important factor in AD [152].Interestingly, Vav1 turned out to be a unique fingerprint DEG for the RS cortex of female APP/PS1 mice.
Apoptosis inducing factor (Aif).The apoptosis inducing factor (Aif) is responsible for caspase-independent programmed cell death [153].Aif was reported to mediate cell death in the TgCRND8 AD mouse model in a region-specific manner with aging-related cell death in the cortex but not the hippocampus [153].Notably, this holds true for our findings in APP/PS1 AD mice as well.Indeed, Aif turned out to be a unique fingerprint gene in the cortex of APP/ PS1 mice, however, only in females and not in males-a phenomenon that has not been specified before [153].
Hexosaminidase subunit beta (Hexb).The gene product of Hexb (hexosaminidase subunit beta) has been associated with AD based on genome-wide association studies (GWAS) [154].In addition, β-hexosaminidase can lead to a reduction of Aβ complexes, the aggregation and accumulation of which has been accelerated by exposure to gangliosides [155].Importantly, Hexb was unique for the RS cortex of female APP/PS1 mice.
Proteasome 20S subunit beta 9 (Psmb9).The gene products of Psmb9 and Psmb8 (encoding for proteasome 20S subunit beta 9 and 8 respectively) form part of a proteasome (prosome, macropain) core complex which plays an important role in AD [156].The ubiquitin-proteasome system (UPS) is a major regulator of protein homeostasis.In contrast, the immunoproteasomes constitute a specialized form of proteasomes, which modulate inflammatory processes in AD via clearance of oxidant-damaged proteins.In APP/PS1 mice, Aβ-deposition was reported to parallel upregulation of immunoproteasomes.The latter turned out to be upregulated by pro-inflammatory cytokines, e.g., type I and type II interferons (IFNs).Wagner et al. (2017) suggested that immunoproteasomes are upregulated in the innate immune response towards extracellular Aβ-accumulation, although previous rodent studies revealed inconsistent data with decreased, unchanged, and increased proteasomal activity with aging [156][157][158].Wagner et al. (2017) used both males and females.However, they did not analyze for sex-specific differences [156].Strikingly, Psmb8 was upregulated only in the RS cortex and hippocampus of females and not in males in our study.In contrast, Psmb9 evolved as a unique fingerprint gene for the RS cortex of female APP/PS1 mice.The potential neuroprotective upregulation of Psmb9 thus seems to be attributable to the predilected brain region and sex in AD.
PYD and CARD domain containing gene (Pycard).Pycard (PYD and CARD domain containing gene, encoding for adapter protein apoptosis-associated speck-like protein containing a CARD (ASC)) codes for an inflammasome component and is involved in regulation of autophagy.Previous reports demonstrated that released ASC specks can bind to Aβ, enhance its aggregation, and increase its toxicity [159].In our study, Pycard turned out to be a unique fingerprint for the RS cortex of female APP/PS1 mice.As Pycard was recently suggested to be a promising diagnostic target in early AD patients [160], sex-specific characteristics require further attention.
Functional implications of signature genes in the hippocampus and RS cortex of male APP/PS1 mice.In the hippocampus of male APP/PS1 AD mice, exclusive upregulation was observed in 18 gene candidates, i.e., Sgk1, Ncf1, Adamtsl4, Hck, St8sia6, Gm20743, Irf9, Pfkfb3, B2m, Arrdc2, Cebpa, Synpo2, Cep126, Rhoh, Hist1h2be, Clec5a, Cd86, Ptpn18.Some are discussed below in more detail: Hematopoietic cell kinase (Hck).The hematopoietic cell kinase encoded by HcK serves as a factor of the innate immune system.It has previously been demonstrated that the Hck pathway exerts important functions in the regulation of microglial neuroprotective functions during the early stage of AD [162].Strikingly, hematopoietic cell kinase deficiency aggravated cognitive impairment along with elevated Aβ levels and plaque formation.It further attenuated Aβ phagocytosis and enhanced iNOS expression in microglia [163].Thus, Hck is likely to exert a prominent neuroprotective function via modulation of microglial function and could attenuate early AD development.
Serum-and glucocorticoid-inducible kinase 1 (Sgk1).Serum-and glucocorticoid-regulated kinase 1 (encoded by Sgk1) is a serine/threonine protein kinase that activates certain K + , Na + , and Cl -channels, involved in cell survival and neuronal excitability [164].SGK1 is engaged in various neurodegenerative pathways related to, e.g., apoptosis, autophagy, neuroinflammation, and ion channel regulation.Interestingly, hippocampal overexpression of Sgk1 improved spatial memory, reduced the devastating impact of Aβ accumulation and rescued actin cytoskeleton polymerization in middle-aged APP/PS1 mice [164,165].It has been speculated that Sgk1 could exert a protective role against oxidative stress and play an antiapoptotic role [166].Whether pharmacological stimulation of serum-and glucocorticoid-regulated kinase 1 is protective in AD as well needs to be determined in the future.
Synaptopodin (Synpo2).Synaptopodin serves as an actin-binding protein that is tightly associated with the spine apparatus and plays an important role in synaptic plasticity [167].Synaptopodin forms clusters in spines and regulates Ca 2+ release from internal stores via ryanodin receptors in dendritic spines [167,168].Deficiency of synaptopodin resulted in LTP deficits, impaired spatial memory and a lack of synaptic plasticity [167,169,170].Interestingly, synaptopodin expression was reported to be downregulated in the hippocampus of patients suffering from dementia with Lewy bodies, Parkinson's disease, mild cognitive impairments, and AD [170,171].However, we observed a selective upregulation of synaptopodin in the hippocampus of male APP/PS1 mice.The later might point to a potential autoprotective role of Synpo2 in this experimental setting.
It should be emphasized that impairment of the structural and functional integrity of synapses is a hallmark in AD pathophysiology.Aβ causes disruption of NMDA and AMPA receptors, Ca 2+ dyshomeostasis, reduced synaptic plasticity, suppression of LTP and aggravated LTD.Microglia/astrocyte activation, cytokine release, mitochondrial disruption, impairment of the cytoskeleton organization and deficits in energy metabolism further enhance synaptic dysfunction [120].Mechanistically, a number of synapse-related target molecules and signaling pathways are related to these phenomena, e.g., Wnt/β-catenin, IKK/NF-κB, JAK2/STAT3, JNK, Akt, MAPK, caspase-3, GSK-3β and CDK-5 [120].Many DEGs identified in our study (see above) enhance synaptic dysfunction, e.g.PrPC, whereas others, such as synaptopodin might exert synaptoprotective effects.
Interestingly, only two candidates, Derl3, and Mki67, exhibited unique upregulation in the RS cortex of male APP/PS1 AD animals (Fig 5E).
Derlin (Derl3).Derlin 3 (encoded by Derl3) resides in the endoplasmic reticulum (ER) and is involved in the degradation of mis-folded glycoproteins in the ER.Dysfunction of Derlin 1 and 2 have previously been related to neurodegenerative diseases [75].
Ki-67 (Mki67).Mki67 encoding for Ki-67 serves as cell proliferation marker and is maximally expressed in G2 phase and mitosis.Interestingly, previous studies identified MKi67 (and also Top2a, Mcm2, Tubb5) to be enriched in DNA replication and chromatin rearrangement in a specific subgroup of microglial cells, i.e., cycling/proliferating microglia (CPMs), that contributes only 0.3%-1.2% of the total microglial population [76].It has further been suggested that Ki-67 is involved in the pathogenesis of neurofibrillary degeneration in AD [172].
Notably, the number of intersectional, upregulated genes was much higher in the RS cortex of APP/PS1 mice compared to the hippocampus (52 versus 32, Fig 5C and 5D).Many of these additional, intersectional candidates in the RS cortex (Fig 5C) are recruited from gene candidates that turned out to be selectively expressed in the male or female hippocampus (Fig 5D).Consequently, more gene candidates are specifically upregulated in the female hippocampus compared to the RS cortex (45 vs. 39) and the male hippocampus compared to the RS cortex (25 vs. 5).Thus, sex-specific differences in DEGs are more overt in the hippocampus than in the RS cortex.
It turned out that several DEGs are region-specific as well.It has been shown for example, that microglia in different brain regions can differ in gene expression, particularly in genes related to bioenergetic and immunoregulatory pathways [173].Interestingly, the number of co-upregulated DEGs in the RS cortex and hippocampus of females was much higher than the related number in male APP/PS1 mice (55 vs. 33,Fig 5).
These findings are supported by symptomatic sex dimorphism profiling in 12 months old male and female APP/PS1 AD mice performed by Jiao et al. (2016) [139].Indeed, there is increasing evidence, that sex-biased and BROI-specific patterns may originate from differences in vulnerabilities/susceptibilities and/or resilience in different brain regions at different AD progression states.Female APP/PS1 mice exhibited a higher parenchymal Aβ load compared to male APP/PS1 animals.The latter was most prominent in the hippocampus [139].In addition, cerebral amyloid angiopathy and related microhemorrhage was more frequent in female APP/PS1 mice.Although top-ranked, intersectional genes associated with neuroinflammation, microglia activation and immune response are found in both sexes and both the RS cortex and hippocampus, specific gene sets exert fingerprint character in the individual subgroups.Notably, related parameters, including levels of phosphorylated τ, proinflammatory cytokines, microgliosis, astrocytosis, and synaptic/neuronal disintegration were particularly enhanced in female APP/PS1 animals [139].Some DEGs that predominate in females are related to mitochondria and ROS.In females, mitochondrial function seems to be more resistant against Aβ mediated neurotoxicity.This phenomenon is probably due to a reduction of ROS and suppression of apoptogenic signals via estrogen [174].This may be one reason why females suffer more from AD with decreasing estrogen levels following menopause.Brainregion und sex-related specificities in DEGs, as observed in our study, might also be due to differences in brain region connectivity/architecture between males and females.The latter suggests that network characteristics in females favor a more rapid spread of neurofibrillary tangles in the CNS [175].In addition, different BROIs are disproportionally modulated via complex spatiotemporal activity of sex hormones [176][177][178].
In summary, the determination of sex-and brain region-specific transcriptional profiles as presented here, is highly necessary for future characterization and validation of DEGs as potential biomarkers and personalized medicinal approaches.
Disease aggravating versus neuroprotective DEGs in APP/PS1 mice.Our transcriptome studies in APP/PS1 AD mice have revealed a high number of significantly upregulated and downregulated gene candidates.Whereas many of them turned out to be intersectional, some exhibited signature characteristics.It's interesting to note that not all DEGs contribute to the devastating complex pathophysiology and pathobiochemistry of AD.Indeed, some candidates, such as Plek, Adora3 and Psmb9 in the female RS cortex, Icam1 in the female hippocampus, and Hck and Sgk1 in the male hippocampus of APP/PS1 mice seem to exert counteracting, autoprotective functions in the individual subgroups.Some intersectional genes such as, Ccl6, Lyz2 and Nrf2 also exhibit self-protective effects.They clearly deserve special attention in drug research and development in the future.

L(i)ncRNAs in APP/PS1 mice
In the mammalian genome, tens of thousands of l(i)ncRNAs were identified, with up to 40% of these l(i)ncRNAs being specifically expressed in the brain [179,180].These l(i)ncRNAs without apparent protein-coding influence metabolism, proliferation, differentiation, and survival of neuronal cells [180].L(i)ncRNA expression has been associated with many neurodegenerative diseases [181][182][183] in patients or animal models and therefore, could serve as biomarkers or potential treatment targets, e.g., in AD.There is evidence that l(i)ncRNAs are aberrantly expressed in a complex spatiotemporal pattern during AD disease progression.L(i)ncRNAs are involved in the regulation of Aβ production, τ hyperphosphorylation, oxidative stress, mitochondrial dysfunction, impairment of synaptic transmission and neuroinflammation, cell death etc. and are likely to affect clinical diagnosis, disease monitoring and therapy [57].These regulatory actions of l(i)ncRNAs are mediated on the transcriptional and posttranscriptional levels.L (i)ncRNAs can affect many cellular processes, including chromatin and DNA modification, RNA transcription, pre-RNA splicing, mRNA stability, and translation [180,184,185].
The SurePrint G3 Mouse Gene Expression v2 8x60K Microarray which we used in our study is not specifically designated to analyze l(i)ncRNAs.We identified a smaller number of l (i)ncRNAs in our individual experimental subgroups, but we were able to characterize, e.g., two l(i)ncRNAs differentially transcribed in the hippocampus of male APP/PS1 mice (l(i) ncRNA: chr18:38776580-38841080 reverse strand) and the hippocampus of female APP/PS1 animals (l(i)ncRNA: chr8:122920901-123008463 forward strand).This sex-specificity in l(i) ncRNA transcription has not been reported before [40].Our transcriptome study is the first to reveal sex-and region-specific l(i)ncRNAs in the individual subgroups.Although these l(i) ncRNA candidates are annotated in related databases (S13 Table in S1 File), further studies are necessary to elucidate and validate their detailed functional implications in AD.

Social hierarchy and potential inference with cortical and hippocampal transcriptome data.
In the interpretation of transcriptome data, the functional interdependence with social hierarchy is often not addressed, although this aspect must be taken into account as a potential confounding factor.It has been demonstrated that behaviors and brain gene expression in WT, e.g., C57BL/6 male mice, is severely affected by different social network sizes and hierarchy in the home cage [187,188].
Alterations in gene expression mainly affected the serotonergic system in dominant and subordinate mice.In addition, subordinate mice exhibited significantly higher corticosterone concentration than dominant males revealing increased stress in subordinate males.Increased chronic stress can lead to downregulation of Bdnf and increased expression of the BDNF receptor gene, Trkb, in the hippocampus [188].Interestingly, these candidates were not detected as DEGs in our transcriptome studies, neither in females nor in males, which might indicate only limited interference of our results with social hierarchy phenomena.

Influence of APP overproduction and transcriptome profiles
Overexpression of WT APP and mutant APP variants has been used to establish many wellcharacterized transgenic AD mouse lines, including APP/PS1 [189].It's noteworthy that APP overexpression in these first-generation transgenic mouse models can cause overproduction of different APP fragments in addition to Aβ.The latter can make it challenging to distinguish the functional impact of these fragments and might affect the translational relevance of such models [190,191].Potentials limitations of mutant APP-and APP/PS-overexpressing mouse models, such as APP/PS1 include, e.g., the lack of non-coding regions of the APP gene that affect splicing of APP mRNA and transcriptional regulation, unphysiological interaction of overexpressed APP and overproduced non Aβ fragments with cellular proteins such as kinesin via JIP-1, non-specific ER impairment upon APP/PS-overexpression, appearance of Aβ entities that are different from those identified in clinical AD brain, different region specificity of Aβ pathology, and inconsistent drug effects [192].Notably, second-generation mouse models are likely to overcome these intrinsic limitations of the APP overexpression paradigm as they utilize an APP knock-in strategy in order to selectively overexpress, e.g., Aβ 42 , but not APP.The latter was achieved, e.g., in single APP knock-in mouse models with a murine Aβ sequence carrying three humanized amino acids that differ between mice and humans [192].However, these models are facing numerous limitations as well [192].Thus, it's important to be aware of the potential interference of APP-or APP/PS-overexpression with AD disease pathology in first generation AD mouse models [192].

Conclusions
Our APP/PS1 AD model-based data support the hypothesis that a major response in the brain to Aβ accumulation is related to microglia activation, immune response and neuroinflammation.Early and profound neuroinflammatory/immune response seems to be a fundamental driving force in AD pathogenesis.Many DEGs were claimed to be involved in these processes, however, sex-and region-specific differentiation of their functional relevance is rare.Our data clearly illustrate that sex-specific differences in the transcriptome profiles are of major relevance and that some DEGs, i.e., signature genes, are exclusively upregulated or downregulated in males or females (Tables 3 and 4).So far, numerous defective genes, transcriptional and translational alterations were shown to contribute to AD pathogenesis, further influenced by diverse environmental factors and epigenetic phenomena.The consideration of sex, age, comorbidity factors, disease progression state and most important, the individual genomic profile, are mandatory for a valid, promising, personalized pharmacotherapeutic approach in the future.The latter will require distinct protocols and strategies in drug research and development, the planning of clinical trials to optimize therapeutics and the establishment of new diagnostic approaches.The analysis of the patient's individual genome including gene sets associated with disease pathogenesis, mechanisms of drug action, drug metabolism, drug transporters and multifaceted cascades and metabolic pathways, will dramatically facilitate the personalized therapeutic approach in AD.
File), for sex comparison (male vs. female samples, S3C Fig in S1 File) and for brain region comparison (Rs cortex vs. hippocampus samples, S3D Fig in S1 File).Notably, there is a grouping tendency based on genotype, sex and BROI.

Fig 1 .
Fig 1. Quantitative distribution of DEGs in sex-and brain-region specific subgroups of APP/PS1 mice.(A)Illustration of the total number of DEGs in our four comparative subgroups, i.e., Rs Cx of female APP/PS1 mice, Hip of female APP/PS1 mice, RS Cx of male APP/PS1 animals and Hip of male APP/PS1 mice.All identified genes are based on a one-way ANOVA, Tukey and Benjamini-Hochberg procedure to identify significantly up-or downregulated genes.The total number of DEGs is higher in the hippocampus compared to the RS cortex.For the total number of upregulated genes, there seems to be no clear preference in BROI or sex.The same holds true for DEGs < -1.5 or > 1.5.Note that for most subgroups, the number of downregulated genes is much less compared to upregulated ones.(B) Illustration of the number of DEGs from a meta-analysis from 2114 post mortem AD patient samples performed byWan et al. (2020).In contrast to our APP/PS1 study, the number of DEGs is persistently higher in post mortem probes from female AD subjects compared to probes from males.Overall, the numbers of DEGs are much higher compared to our APP/PS1 study.The latter might be due to the interindividual variability in the AD patients probes and the fact that Wan et al. (2020) included seven BROIs in their meta-analysis.For methodological details on their meta-analysis and statistics see[73].

Fig 2 .
Fig 2.Volcano plots illustrating statistical significance vs. magnitude of FC.Profiles are depicted for the RS cortex of female APP/PS1 mice (A), the hippocampus of female APP/PS1 mice (B), the RS cortex of male APP/PS1 mice (C) and the hippocampus of male APP/PS1 animals (D).All genes exhibiting differential transcript levels were plotted with each dot representing one gene.The log2-FC in the individual APP/PS1 subgroups vs. WT is represented on the x-axis.The y-axis displays related log10 p-values (t-test procedure without further correction).The p-value of 0.05 and a FC of > 1.5 and < -1.5 are indicated by a horizontal line and vertical lines, respectively.Data points below the horizontal line represent gene candidates that were not significantly altered in transcription.Gene candidates in the different Volcano sectors (significantly up-or downregulated and FC > 1.5 or FC < -1.5) are color-coded.Note that downregulated gene candidates in the upper left sector (FC < -1.5, light orange) and upregulated genes in the upper right sector (FC > 1.5, blue) were further investigated in Venn and pathway studies.In most subgroups, the upregulated gene candidates predominate.Selected DEGs with high FC and statistical significance were individually labelled.Genes with significantly altered FC in the range from -1.5 to 1.5 (green and purple) were not further analyzed in our study.

Fig 3 .
Fig 3. GO analysis of DEGs (FC > 1.5, p < 0.05).Bar graphs of enriched terms across the input gene list and related p-values were obtained from the Metascape gene list analysis report for (A) the RS cortex of female APP/PS1 AD vs. WT mice, (B) the hippocampus of female APP/PS1 AD vs. WT mice, (C) the RS cortex of male APP/PS1 AD vs. WT mice and (D) the hippocampus of male APP/PS1 AD vs. WT animals.The top 20 categories/clusters are itemized together with their related -Log(10)P values in a descending fashion.https://doi.org/10.1371/journal.pone.0296959.g003

Fig 4 .
Fig 4. GO analysis of DEGs (FC < -1.2, p < 0.05).Bar graphs of enriched terms across the input gene list were obtained from the Metascape gene list analysis report for (A) the RS cortex of female APP/PS1 AD vs. WT mice, (B) the hippocampus of female APP/PS1 AD vs. WT mice, (C) the RS cortex of male APP/PS1 AD vs. WT mice and (D) the hippocampus of male APP/PS1 AD vs. WT mice.The individual categories are provided together with their related -Log(10)P values in a descending fashion.Note that no GO analysis results could be obtained in the RS cortex of female APP/PS1 AD mice vs. WT controls.Overall, GO analysis for downregulated gene candidates revealed a limited number of categories compared to upregulated gene candidates (see also Fig 2).https://doi.org/10.1371/journal.pone.0296959.g004

Fig 6 .
Fig 6.Venn diagrams of downregulated DEGs (FC < -1.5) in APP/PS1 AD vs. WT control mice related to either both brain regions (RS cortex and hippocampus) or both sexes.Intersectional DEGs in (A) the RS cortex and hippocampus of female APP/PS1 AD mice, (B) the RS cortex and hippocampus of male APP/PS1 AD mice, (C) the RS cortex of female and male APP/PS1 AD mice and (D) the hippocampus of female and male APP/PS1 mice.Intersectional gene symbols (in black) are listed below each diagram, the other gene symbols are depicted next to the diagram (in orange and blue).E) Venn-diagram including all four study groups.Note that there are specific DEGs (signature genes, listed here) that are selectively downregulated only in one of the four study groups and thus represent a unique transcriptome profile for sex and BROI.Strikingly, the number of downregulated gene candidates is severely lower compared to the upregulated ones.https://doi.org/10.1371/journal.pone.0296959.g006

Fig in S1 File). DEGs in the RS cortex and hippocampus of APP/PS1 AD vs. WT mice of both sexes RS cortex of female APP/PS1 mice. Using
The latter was complemented by hierarchical clustering to elicit similarities and dissimilarities of transcription profiles between the individual subgroups (S4 the Feature Extraction and GeneSpring RT software (both Agilent Technologies Germany GmbH & Co. KG, Germany), DEGs and l(i) ncRNAs were characterized based on one-way ANOVA testing with Tukey post hoc analysis and the application of the Benjamini-Hochberg procedure for FDR determination.As depicted in

S5 Table in S1 File, a
total number of 273 genes and l(i)ncRNAs were found including four candidates with a FC < -1.5 and 104 candidates with a FC > 1.5.The majority (85.35%) of DEGs was upregulated in this subgroup (233"/40#) (Fig 1A).The predominance of significantly upregulated gene candidates is also confirmed in the Volcano plot (Fig 2A).

Table 3 . Overview of the sex-and region-specific upregulation of DEGs in the four APP/PS1 subgroups.
This table illustrates all upregulated DEGs (FC > 1.5) from APP/PS1 microarray analysis.DEGs that are upregulated only in one APP/PS1 subgroup are highlighted in red, those exhibiting enhanced transcript levels in two subgroups are marked in green.Increased transcript levels in three subgroups are marked in ocher, those in all four subgroups are highlighted in yellow.The table elicits the complex sex-and brain region specific profile of upregulated DEGs. https://doi.org/10.1371/journal.pone.0296959.t003