Alzheimer’s disease (AD) has emerged as a critical public health crisis, progressively impairing cognitive function and diminishing patients’ ability to live independently1. The key pathological characteristics of AD involve the aggregation of amyloid-β plaques and the presence of neurofibrillary tangles within the brain2,3. Among geriatric dementia subtypes, AD accounts for at least two-thirds of dementia cases in individuals aged 65 and older4, imposing significant societal and personal burdens, particularly in aging populations. Although some recent studies have brought new hope for the diagnosis and treatment of AD5,6, the fundamental mechanisms driving AD pathogenesis remain unclear, primarily due to the disease’s complexity and the impact of multiple factors.

So far, single-omics studies have explored AD, revealing notable differences in protein between AD patients and healthy controls7,8,9,10. With the progress of high-throughput sequencing, multi-omics approaches have become valuable tools for deciphering the pathogenesis of AD11. Using integrative multi-omics analyses, researchers have identified novel molecular alterations and dysregulated pathways associated with AD pathology12,13,14. These studies highlight the potential of multi-omics analysis to deepen our understanding of AD at the molecular level. However, despite these advancements, most multi-omics discoveries have yet to translate into clinical applications, such as improved disease prognosis or targeted treatments, an area that warrants further exploration.

In this study, we integrated transcriptomic and proteomic data from brain tissue in an AD mouse model to identify novel molecular and pathway alterations associated with AD. We then evaluated the correlation between their mRNA expression levels and AD pathology, as well as their potential roles in drug treatment.

Methods

Animals

Due to its stable pathological phenotype, the APP/PS1 model has become a classic tool for the discovery of AD biological markers. The use of only male mouse offers advantages in terms of stability and controllability in experimental design, making it particularly suitable for early exploratory studies. Therefore, SPF-grade male APP/PS1 mouse (9 months old, weighing 32 ± 3 g) were used in this study. These mouse were obtained from the Key Laboratory of Animal Models and Human Disease Mechanisms of the Kunming Institute of Zoology, Chinese Academy of Sciences. The animals were housed at the Experimental Animal Center of the Kunming Institute of Zoology, Chinese Academy of Sciences, under controlled conditions. The facility maintained a constant temperature of 23± 2 °C, with relative humidity regulated between 40% and 60%. A 12-hour light/dark cycle was implemented to ensure proper illumination. This study was approved by the Experimental Animal Management Committee of Kunming Medical University (Approval No. Kmmu20231161).

Transcriptomic analysis of hippocampal tissue

Mouse were deeply anesthetized via intraperitoneal injection of sodium pentobarbital (100 mg/kg). Once a surgical plane of anesthesia was confirmed, euthanasia was performed by rapid decapitation. Hippocampal tissues were promptly dissected, snap-frozen in liquid nitrogen, and stored at −80 °C for subsequent analysis. Hippocampal tissue RNA was isolated following the TRIzol® Reagent protocol. RNA integrity and concentration were evaluated using the Agilent 5300 Bioanalyzer and NanoDrop ND-2000, respectively. Libraries were prepared exclusively from RNA samples meeting stringent quality criteria. Subsequent steps, including RNA purification, cDNA synthesis, library preparation, and sequencing, were performed by Shanghai Majorbio Bio-pharm Biotechnology Co., Ltd. The RNA-seq library was constructed from 1 µg of total RNA using the Illumina® Stranded mRNA Prep, Ligation Kit (San Diego, CA). Polyadenylated mRNA was enriched using oligo(dT)-coupled magnetic beads, followed by fragmentation and first-strand cDNA synthesis with random hexamers via the SuperScript kit (Invitrogen, CA). Double-stranded cDNA was generated, subjected to end repair, 5’-phosphorylation, and adapter ligation according to the manufacturer’s instructions. Size selection (300 bp) was performed on a 2% Low Range Ultra Agarose gel, and the library was amplified for 15 PCR cycles using Phusion DNA polymerase (NEB). Quantification was carried out with a Qubit 4.0 fluorometer. High-throughput sequencing was conducted on the NovaSeq X Plus system (PE150) with NovaSeq reagents. Raw paired-end sequencing data were processed with fastp for adapter trimming and quality control. This version maintains the technical details while varying sentence structure and word choice to enhance originality. Clean reads were aligned to the reference genome in orientation mode via HISAT2. Read assembly was conducted with StringTie using a reference-based approach. Differentially expressed genes (DEGs) were determined by comparing transcript abundance (TPM values), applying thresholds of |log2FC| > 1.0 and an adjusted P-value < 0.05. To investigate their biological relevance, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were conducted. Significantly enriched terms and pathways were identified using a stringent cutoff (Bonferroni-adjusted P < 0.05). GO enrichment and KEGG pathway analysis were conducted using Goatools and Python Scipy, respectively.

DIA-proteomic analysis of hippocampal tissue

Frozen samples were transferred to MP disruption tubes, lysed with protein lysis buffer, and shaken three times (40s each) using a high-throughput tissue grinder. The lysate was kept on ice for 30 min with intermittent vortex mixing at 5-min intervals. Following centrifugation at 12,000 × g for 30 min (4 °C), the clarified supernatant was carefully collected. Total protein concentration was then determined using a bicinchoninic acid (BCA) assay. Proteins were then analyzed by SDS-PAGE. For proteomic analysis, aliquots containing 100 µg of protein were solubilized in 100 mM triethylammonium bicarbonate (TEAB) buffer. The protein samples were first treated with 10 mM TCEP at 37 °C for 1 h, then alkylated using 40 mM IAA in darkness at ambient temperature for 40 min. Proteins were precipitated with precooled acetone (−20 °C, 4 h), centrifuged (10,000 g, 20 min), and dissolved in 100 µL TEAB. Trypsin digestion (1:50 enzyme-to-protein ratio) was conducted overnight at 37 °C. Digested peptides were dried under vacuum, reconstituted in 0.1% TFA, desalted with an HLB column, and dried again. Peptide concentration was measured using the NanoDrop One UV spectrophotometer. Equal peptide amounts from each sample were dried, reconstituted in 0.1% TFA, desalted, and quantified with the Thermo peptide quantification kit. Peptides were separated using a Vanquish Neo chromatograph with a uPAC High Throughput column. Mobile phases were: A and B, with a 180 SPD chromatography gradient. Data were acquired using Thermo Xcalibur DIA Raw DIA data were processed in Spectronaut™ 18. The quantitative analysis was performed using six representative peptides, with three specific fragment ions analyzed per peptide(Peptide FDR ≤ 0.01, Peptide Confidence ≥ 99%). Differentially expressed proteins were annotated via the GO database. Functional clustering and KEGG pathway analyses identified metabolic pathways. Significant proteins were analyzed in R (criteria: Fold change ≥ 1.2 or ≤ 0.83, p < 0.05).

Combined examination of transcriptomic and proteomic profiles

We conducted correlation studies between DEGs and DEPs, followed by KEGG pathway mapping to identify overlapping signaling pathways.

Experimental verification via quantitative real-time PCR

Three candidate DEGs were subjected to qPCR verification to substantiate the RNA sequencing findings. Hippocampal samples were cryogenically pulverized in liquid nitrogen, followed by RNA isolation with Trizol reagent (1 mL per sample)(15596-018CN, Invitrogen, Waltham, MA, USA). Reverse transcription was carried out employing the PrimeScript™ RT Kit (RR037A, TaKaRa, Japan). Data analysis employed the 2−△△Ct methodology to determine expression differences. All PCR primers were custom-designed and commercially synthesized through Tsingke Biotech.

Western blot (WB) analysis

Total proteins were extracted from mouse hippocampal tissues by homogenization in RIPA lysis buffer, and their concentrations were determined. Protein separation was achieved via SDS-PAGE, after which the samples were transferred to a PVDF membrane. The membrane was subjected to blocking with 5% skimmed milk powder for 2 h at room temperature under gentle agitation. Following blocking, three PBST washes were performed (10 min each). Subsequently, the membrane was incubated with primary antibodies overnight at 4 °C. The membrane was rinsed three times with PBST (10 min each) following primary antibody incubation. The membrane was then incubated with an HRP-conjugated secondary antibody for 2 h at room temperature, followed by three 10-minute PBST washes. Protein detection was performed using an ECL chemiluminescence kit (KF8001, Affinity) according to the manufacturer’s instructions. α-Tubulin (BF9210, Affinity) served as the loading control, and band quantification was conducted using ImageJ software (USA).

Association of AD-related gene expression with pathological features in mouse models

Spatiotemporal gene expression profiles and neuropathological characteristics were acquired from both transgenic AD mouse and wild-type (WT) controls through the Mouseac database15. The dataset included 219 brain tissue samples from five different AD mouse models and 114 samples from age-matched WT mouse. Hippocampal gene expression levels of AD-associated genes were compared between AD and WT mouse across four developmental stages (2, 4, 8, and 18 months). Additionally, Aβ plaque pathology was quantified to assess disease progression.

Evaluation of AD genetics in drug treatment

To investigate the role of AD genetics in pharmacological interventions, we compiled the targets of FDA-approved AD drugs from DrugBank 5.016. The STRING database was employed to generate protein-protein interaction networks to assess potential associations between AD risk genes and drug targets17. Protein-protein interaction network visualizations were generated using Cytoscape v3.7.218.

Statistical analysis

GraphPad Prism 9.0 (GraphPad Software, California) was utilized for statistical evaluation and graphical representation. Independent samples t-tests were used when normality and variance homogeneity assumptions were met, while Welch’s correction was applied for normal distributions with unequal variances. All datasets were derived from three or more biological replicates. We examined associations between AD-relevant gene expression and disease pathology using Pearson correlation coefficients, with statistical significance set at p < 0.05.

Results

Analysis of transcription levels in hippocampal tissue

Our analysis confirmed that the RNA samples met the required purity and integrity standards. Using a threshold of |log2FC| > 1.0 with statistical significance (P < 0.05), we detected 263 DEGs, comprising 162 with increased expression and 101 with decreased expression. The distribution of DEGs is visualized in Fig. 1A and B. To further investigate the biological significance of these DEGs, GO enrichment analysis identified their primary associations with biological processes, molecular functions, and cellular components (Fig. 1C). Additionally, the top 20 significantly enriched KEGG pathways are depicted in a bubble chart (Fig. 1D), highlighting key functional pathways.

Fig. 1
Fig. 1
Full size image

Transcriptomic analysis of differentially expressed genes (DEGs) in the hippocampus of APP/PS1 mouse. (A-B) Volcano plots and histograms showing DEGs between the two groups. (C) GO enrichment analysis top20 histogram of DEGs. (D) KEGG enrichment analysis top20 bubble plot of DEGs between two groups43,44,45. N = 3 mouse/group.

Protein expression assessment in the hippocampus

Through comprehensive proteomic screening, we quantified 448 proteins showing significant expression changes, with 178 upregulated and 270 downregulated. A plot of the DEPs is shown in Fig. 2A and B. The GO analysis results for the top 20 DEPs are depicted in Fig. 2C, while Fig. 2D displays the bubble chart visualization of the 20 most significantly enriched KEGG pathways.

Fig. 2
Fig. 2
Full size image

Proteomic analysis of differentially expressed proteins (DEPs) in the hippocampus of APP/PS1 mouse. (A-B) Volcano plots and histograms showing DEPs between the two groups. (C) GO enrichment analysis top20 histogram of DEPs between the two groups. (D) KEGG enrichment analysis top20 histogram of DEPs between two groups43,44,45. N = 3 mouse/group.

Synergistic evaluation of transcriptome and proteome datasets

Integration of mRNA and protein expression profiles uncovered five co-upregulated DEGs/DEPs: EGFL8, LY86, CD180, C1QB, and APP, while ERMN exhibited an opposite trend (Fig. 3B). Pathway analysis of differentially expressed genes and proteins from both datasets revealed that the key signaling pathways primarily converged on five distinct signaling pathways (Fig. 3A). Among all the enriched biological pathways, Metabolic pathways19, spliceosome20, serotonergic synapse21, and complement and coagulation cascades22 have been reported to be associated with Alzheimer’s disease. In addition, since Alzheimer’s disease is a neurodegenerative disorder, biological pathways related to neurodegeneration and Alzheimer’s disease are also generally recognized as being associated with Alzheimer’s disease.

Fig. 3
Fig. 3
Full size image

Joint analysis. (A) Differentials co-mapping pathway histogram43,44,45. (B) Differentials co-mapping gene venn diagram. N = 3 mouse/group.

Experimental validation by qPCR and WB assays

Five genes (EGFL8, CD180, C1QB, LY86, and ERMN) that showed elevated expression in our omics datasets were validated by qPCR using specifically designed primers to assess differential expression between control and disease groups relative to GAPDH expression (Fig. 4). The corresponding primer sequences are provided in Table S1. Protein-level validation was further performed using WB analysis (Fig. 5). For WB, the primary antibodies used were anti-EGFL8 (OM277683, 1:750), anti-CD180 (ab184956, 1:1000), anti-C1QB (30651T, 1:1000), anti-LY86 antibody (PRS3851, 1:1000), and anti-ERMN antibody (OM278175, 1:1000) solutions. The original, uncropped blots are presented in Supplementary Figures S1-S6. The results aligned with the omics sequencing data, apart from minor differences in expression levels, confirming the reliability of the sequencing results.

Fig. 4
Fig. 4
Full size image

qPCR verification of transcriptomic sequencing results. (A-E) The mRNA expression of EGFL8, CD180, C1QB, LY86, and ERMN in the hippocampus of APP/PS1 mouse were evaluated through the qPCR. N = 3 mouse/group; *p < 0.05;** p < 0.01; *** p < 0.001.

Fig. 5
Fig. 5
Full size image

Western blotting validation of proteomics sequencing results. (A, D, E, F) Representative Western blotting images of CD180, EGFL8, C1QB, LY86, and ERMN. Tubulin was used as the internal standard; (B-F) Image J software was used to quantitatively analyze CD180, EGFL8, C1QB, LY86, and ERMN in the hippocampus of two groups. Data are expressed as mean ± SEM; N = 3 mouse/group; * p < 0.05.

Correlation between AD-related mRNA levels and Aβ pathology in experimental mouse

We investigated possible relationships between mRNA abundance of genes linked to AD and disease-related pathological changes by acquiring and processing transcriptional and neuropathology datasets from AD models using the Mouseac platform15. Comparative assessment of hippocampal transcriptomes at various ages showed that the majority of genes implicated in AD’s pathogenesis exhibited increased expression in HO_TASTPM mouse during terminal disease progression (Fig. 6). Additionally, strong positive correlations were found between the mRNA expression levels of C1QB, CD180, and LY86 and the accumulation of Aβ plaques in APP K670N/M671L mutant mouse, whereas EGFL8 showed no correlation (Fig. 6). However, the mouseac dataset did not include measurements of both ERMN expression levels and Aβ pathology burden, making it impossible to evaluate their potential relationship. Collectively, these findings strengthen the evidence for molecular associations between AD-related genetic markers and disease mechanisms.

Fig. 6
Fig. 6
Full size image

The mRNA expression levels of differentially expressed genes (DEGs) were correlated with the level of Aβ plaque in hippocampus tissues of AD mouse models. The gene expression data and pathology data were retrieved from Mouseac [15]. Data shown were the age-related mRNA expression (left) and the correlation between mRNA level and Aβ plaque (right) of each of 4 genes (A-D). ERMN had no data available in Mouseac [15]. The age-related mRNA expression level was measured in hippocampus tissues of wild type (n = 7) and HO_TASTPM mouse (with homogenous APP K670N/M671L and PSEN1M146V mutations, n = 3) at different life stages (2, 4, 8, and 18 months).The correlation between mRNA level and Aβ plaques in mouse with APPK670N/M671L mutation (n = 44) was measured using the Pearson correlation test. A P-value < 0.05 was considered as statistically significant, ns: not significant.

Evaluation of AD genetics in drug treatment

To investigate the role of AD-related genes in drug treatment, we integrated transcriptomic and proteomic data and identified six DEGs. The 27 approved AD drug targets were identified by searching DrugBank 5.0.(Table S2). Notably, four of the six AD risk genes (66.7%) encoded proteins that interact with known AD drug targets (Fig. 7), highlighting their potential relevance in therapeutic strategies.

Fig. 7
Fig. 7
Full size image

Protein–protein interaction (PPI) network of putative AD risk genes and genes targeted by AD drugs. Red nodes represent AD risk genes, green nodes indicate genes targeted by AD drugs.

Discussion

Numerous studies confirm that genetic predisposition is a key determinant in the pathology of AD23,24. Transcriptomics enables a comprehensive analysis of gene expression in specific tissues at defined time points25, while proteomics provides detailed insights into protein expression and its dynamic changes within tissues26. Integrating these two approaches allows for a deeper understanding of the interplay between genes and proteins and overcoming the limitations of traditional research. In this study, we performed an integrative analysis combining transcriptomics, proteomics, protein-protein interaction and pathology data, which revealed the active involvement of several novel biomarkers in AD. As far as we are aware, this study represents the first in-depth analysis of the link between these genetic factors and AD.

A total of 263 DEGs and 448 DEPs were identified through transcriptomic and proteomic analyses. A major focus of this work was to analyze critical signaling pathways and identify molecular targets associated with AD. By integrating transcriptomic and proteomic data, we identified six genes with statistically significant differences in expression levels. The accuracy of the sequencing results was further validated through WB and qPCR experiments. Additionally, a comparative KEGG pathway analysis of DEGs and DEPs revealed that immune-related signaling pathways were a commonly enriched category. (Fig. 3b), reinforcing their well-established role in AD pathophysiology27,28,29.

Among the six identified DEGs/DEPs, APP is a well-known AD gene involved in Aβ and tau pathologies, neuronal death, and cognitive impairment in AD patients30,31,32. Additionally, LY86 has been suggested as a potential AD risk gene33. Our integrated transcriptomic and proteomic analysis corroborated these findings and further demonstrated that LY86 expression is positively correlated with Aβ pathology. Another key gene, C1QB, plays a crucial role in immune and inflammatory dysregulation, contributing to AD onset and progression34,35. Our transcriptomic and proteomic integration analysis showed that C1QB was significantly upregulated in the AD hippocampus, which was further validated by qPCR and WB. Enrichment analysis indicated that C1QB is involved in immune-related pathways, and further correlation analysis confirmed its positive association with brain pathology.

While EGFL8 has been linked to various cancers36,37 and brain white matter hyperintensities38, its role in AD has not been previously reported. Our study provides the first evidence of EGFL8’s involvement in AD, as revealed through transcriptomic and proteomic integration as well as enrichment analysis. However, no correlation was found between EGFL8 mRNA expression and AD brain pathology, suggesting that EGFL8 may contribute to AD pathogenesis through other mechanisms, such as by binding to EGFR or other tyrosine kinase receptors, thereby activating downstream PI3K/Akt or MAPK signaling pathways. In addition, our integrated analysis found that ERMN expression was significantly downregulated in the AD hippocampus. Previous studies have linked ERMN to major depressive disorder39 and autism spectrum disorder40, both of which share common pathophysiological features with AD. Genetic and epigenetic methylation defects may be one way the ERMN influences susceptibility to psychiatric disorders41, while another could be through its impact on immune infiltration39. We speculate that ERMN may be involved in AD pathogenesis through mechanisms similar to those observed in psychiatric disorders. Regrettably, our analysis of mouseac revealed no available data on both ERMN expression levels and Aβ pathology burden. Consequently, we were unable to assess their relationship, making further investigation of this potential association necessary. CD180, a newly identified AD-associated gene, was validated at multiple levels, including transcriptomic, proteomic, and brain pathology analyses. Interestingly, CD180 has primarily been associated with inflammatory diseases42, and given the pivotal role of inflammation in AD pathogenesis, our findings further support the link between CD180 and AD progression. These novel discoveries enhance our understanding of the complex molecular landscape of AD.

For the protein-level validation, WB results for both proteins(LY86 and ERMN) showed the same directional trend as the DIA proteomics data. However, while the DIA proteomics results reached statistical significance, the WB results did not achieve formal significance between the two groups.We would like to explain the observed consistency in trends and the nuanced difference in statistical significance for the protein-level data: (1) Consistency in Biological Trend: The core biological signal, the direction of differential expression of LY86 and ERMN proteins between groups, was consistent across both DIA proteomics and WB. This alignment supports the robustness of the underlying biological phenomenon, as trends are less likely to be technical artifacts when replicated across distinct methods. (2) Factors Contributing to Statistical Significance Differences: ①DIA proteomics quantifies proteins based on multiple peptide signals, providing higher throughput and sensitivity for detecting subtle changes in complex samples. ②WB, while a gold standard for protein validation, relies on single antibody binding and densitometric analysis, which may be more susceptible to subtle variations in sample processing, antibody efficiency, or loading normalization, especially when the magnitude of protein change is moderate.

Bridging the gap between genetic findings and pharmaceutical targets is essential for translating these findings into clinical applications. Our study demonstrated that AD risk genes (CD180, LY86, C1QB, and APP) are more likely to interact with targets of approved AD drugs (Fig. 7), providing insights into potential novel drug targets. LY86 and CD180 are key regulatory factors in the TLR4 pathway. Targeting these two proteins may provide a new anti-inflammatory strategy for the treatment of AD by inhibiting TLR4-mediated excessive neuroinflammation. Regarding C1QB, as a key initiator of complement activation, the development of its specific inhibitors is expected to inhibit complement-mediated synaptic toxicity while potentially preserving part of the physiological functions of the complement system, thereby providing better safety. However, the validity of these newly identified targets remains dependent on our current understanding of existing drug mechanisms. Therefore, further investigations are needed to develop new therapeutic strategies and gain deeper insights into AD pathogenesis.

The limited overlap between DEGs and DEPs is attributed to the complexity of post-transcriptional regulation, differences in protein stability and turnover, and spatiotemporal heterogeneity of samples. However, our functional enrichment analysis revealed that both DEGs and DEPs are co-enriched in AD-related pathways, suggesting that differential regulation at the transcriptional and translational levels may collectively contribute to AD pathogenesis rather than being a random phenomenon. The roles of EGFL8, ERMN, and CD180 have not been fully reported in existing human AD genetic data. In this study, we are the first to identify their abnormal expression in AD models and suggest their potential functions, which confers a certain degree of novelty to our findings.

While we have identified several AD biological markers by integrating multi-omics and brain pathology data, this study has notable limitations that warrant acknowledgment: First, we utilized APP/PS1 transgenic mouse, a widely adopted model in AD research due to its robust amyloid-β pathology, yet it has critical drawbacks. By overexpressing mutant human APP and PSEN1 genes, this model exhibits early and aggressive amyloid plaque deposition but fails to recapitulate key features of sporadic late-onset AD, the most prevalent form in humans. Specifically, it lacks tau pathology, significant neuronal loss, and the complex interplay between aging, environmental exposures, and polygenic risk factors. Consequently, molecular signatures identified in this model, particularly those related to gene expression and immune processes, may not fully translate to human AD. This underscores the need for future research to incorporate additional models or human-derived data to enhance findings’ relevance. Second, existing evidence indicates sex-specific differences in neuroinflammation, synaptic plasticity, and metabolic markers in APP/PS1 mouse. For instance, female mouse show higher brain levels of proinflammatory cytokines, which may confound the detection of inflammation-related markers. Moreover, AD incidence and pathological progression differ between female and male patients, meaning the exclusive use of male mouse could introduce bias in marker discovery. Given that females constitute the majority of AD patients and estrogen level fluctuations represent a key risk factor, marker studies relying solely on male mouse may fail to fully capture the disease’s characteristics in humans.

Conclusions

Through an extensive multi-level association analysis, encompassing transcriptomics, proteomics, pathology, and protein-target interactions, we provide compelling evidence supporting the significant role of three novel genes (EGFL8, ERMN, and CD180) in AD. Further examination through population-based replication studies and functional analyses will be crucial to substantiate these findings and deepen our understanding of these genes’ involvement in AD pathology and potential interventions.