Genome-Wide Mendelian Randomization Identifies Ferroptosis-Related Drug Targets for Alzheimer’s Disease

Ferroptosis-related genes list

The principle and flow diagram of MR study are depicted in Fig. 1. FerrDb, a central database dedicated to the study of ferroptosis regulators and their disease associations, is accessible via http://www.zhounan.org/ferrdb/. Genes related to ferroptosis were extracted from this database for subsequent analysis.

OPEN IN VIEWER

Fig. 1

The principle and flowchart of MR analysis.

QTL datasets used for genetic instrumental variables selection

We leveraged four QTL datasets of brain to derive genetic instrumental variables for ferroptosis-related genes. Moreover, we utilized eQTL and pQTL data from available blood tissue to exhaustively search for ferroptosis-related targets related to AD. The GTEx eQTL datasets were acquired from the SMR website (https://yanglab.westlake.edu.cn/software/smr/#eQTLsummarydata).

The PsychENCODE eQTL datasets were generated using brain tissues (specifically, the prefrontal cortex) from 1,378 human individuals. 20 We accessed the LD reference panel of the European (EUR) population from the 1000 Genomes project provided by OpenGWAS API. 21 The BrainMeta v2 cis-eQTL summary data, downloaded from the SMR website, underwent eQTL mapping using RNA-seq data from 2,865 brain cortex samples, collected from 2,443 unrelated individuals of European ancestry with genome-wide single nucleotide polymorphism (SNP) data. The pQTL data of brain originated from a recent study, which can be downloaded from Synapse (https://doi.org/10.7303/syn23627957). 22

The summary-level eQTL data of blood tissues were procured from the eQTLGen Consortium, comprising a total of 31,684 individuals (https://www.eqtlgen.org/). The pQTL summary study is grounded in a large-scale GWAS on the blood proteome (https://www.decode.com/summarydata/). We extracted summary-level statistics of genetic associations with levels of 4,907 circulating proteins from a large-scale pQTL study conducted in 35,559 Icelanders.

GWAS summary statistics of AD

The summary statistics for AD from GWAS were retrieved from the MRC IEU Open GWAS Project (https://gwas.mrcieu.ac.uk/). The GWAS results, identified by the GWAS ID ieu-b-5067, served as the outcome data in the MR analysis. This analysis incorporated data from 954 cases and 487,331 controls.

Mendelian randomization

PPI analysis and expression of the identified gene targets in different cell types of the human brain

The online tool, Search Tools for the Retrieval of Interacting Genes (STRING), was used to scrutinize protein interactions. We filtered the Protein-Protein Interaction (PPI) pairs based on a confidence score exceeding 0.40 and visualized the PPI network using Cytoscape V3.8.2 software. 26 We leveraged the Cortical Development Expression Viewer (CoDEx) data portal to investigate the expression patterns of genes significant in the SMR analysis at the single-cell level. 27

Triple-transgenic mouse model of AD

3xTg-AD is an important preclinical animal model used for studying the pathogenesis and potential treatments for AD. This mouse model was created by incorporating three transgenes (APP, PS1, and Tau mutation) into the mouse genome to mimic key features of AD pathology observed in human patients. 28 By combining these three mutations in a single mouse model, the 3xTg-AD model recapitulates various aspects of AD pathology, including Aβ plaque deposition, neurofibrillary tangle formation, synaptic dysfunction, and cognitive decline.

Males of the 3xTg-AD strain, aged between 6-7 weeks, were obtained from the Changzhou Cavens Laboratory Animal Co., Ltd. (Wuhan, China; license number: SCXK (Su) 2016-0010). These mice were maintained in a pathogen-free laboratory at Wuhan Myhalic Biotechnology Co., Ltd. (Wuhan, China), under precise conditions: a temperature range of 20–26°C, 50% ±10% humidity, and a 12-h day-night light cycle. Standard rodent chow and unrestricted access to water were provided until they reached 9 months of age. Males of the C57BL/6J strain, aged 7 months, were purchased from Wuhan Youdu Biotechnology Co., Ltd. (Wuhan, China; license number: SCXK (E) 2021-0025). These mice were housed under identical conditions to those of the 3xTg-AD mice, including comparable dietary regimens up to 9 months of age. All animal research was conducted with the approval of the Animal Ethics Committee of Wuhan Myhalic Biotechnology Co., Ltd. (ethical approval number: HLK-20230518-001). The experimental procedures complied with the ARRIVE guidelines and the American Veterinary Medical Association Guidelines for the Euthanasia of Animals (2020). No interventional experimental procedures were employed on the mice. Following anesthesia with isoflurane, the mice were decapitated, and their entire brains were rapidly harvested for western blot analyses.

Validation of the gene targets by transcriptome and western blotting

Ferroptosis-related genes list

A total of 369 genes were identified as drivers, 348 as suppressors, 11 as markers, and 116 were left unclassified. Some genes may duplicate across different categories. Consequently, we have gathered 564 genes related to ferroptosis for further analysis (Supplementary Table 1).

SMR to prioritize candidate gene targets in brain tissue

SMR analysis can be used to ascertain causal relationships and investigate whether alterations in gene or protein expression contribute to disease onset. Consequently, the significant genes or proteins identified by SMR can serve as potential therapeutic targets. By using cis-eQTL SNPs from GTEx (spanning 13 brain regions) as genetic instruments, we pinpointed 17 potential gene targets for AD (SMR p < 0.05 and HEIDI test p > 0.05) (Fig. 2A). We also found 13 significant correlations from PsychENCODE, and 22 from BrainMeta (Fig. 2B, C). Additional details can be found in the Supplementary Table 1.

OPEN IN VIEWER

Fig. 2

The Manhattan plots of MR analysis results using QTLs and AD GWAS summary statistics. The gray horizontal line is the Bonferroni corrected significant level. A) The MR result using GTEx brain eQTL as instruments; B) The MR result using PsychENCODE eQTL as instruments; C) The MR result using BrainMeta eQTL as instruments; D) The MR result using ROSMAP pQTL as instruments; E) The MR result using cis-eQTL SNPs from eQTLGen Consortium as genetic instruments; F) The MR result using cis-pQTL SNPs from circulating 4,907 proteins levels as genetic instruments.

When we combined the three gene lists, we were able to identify 30 potential genes. When we found the common genes among the three lists, we discovered six overlapping candidate genes, including PEX10, CDC25A, EGFR, DLD, LIG3, and TRIB3 (SMR p < 0.05 and HEIDI test p > 0.05) (Fig. 3A, Tables 1 and 2). These genes, including PEX10, EGFR, DLD, and LIG3, function as inducers of ferroptosis, promoting its occurrence. Conversely, CDC25A and TRIB3 serve as an inhibitor of ferroptosis, preventing its onset. The detailed SMR locus plot for the six genes can be found in Supplementary Figure 1.

OPEN IN VIEWER

Fig. 3

A) SMR method identified six overlapping ferroptosis-related gene targets associated with AD when using GTEx, PsychENCODE, and BrainMeta brain eQTL as genetic instruments, including PEX10, CDC25A, EGFR, DLD, LIG3, and TRIB3. B) Venn diagram showed the intersection of AD risk genes identified by GTEx, PsychENCODE, and BrainMeta brain eQTL and ROSMAP brain pQTL as genetic instruments. C)Venn diagram showed the intersection of AD risk genes identified by GTEx, PsychENCODE, and BrainMeta brain eQTL and eQTLGen blood eQTL as genetic instruments. D) Venn diagram showed the intersection of AD risk genes identified by eQTLGen blood eQTL and blood pQTL as genetic instruments.

In addition to the gene expression-level candidate targets, we also conducted a two-sample MR analysis by incorporating the pQTL of all proteins from brain tissues. In total, we selected 65 LD-independent pQTL instruments from 65 proteins. We then intersected these 65 potential proteins with 564 ferroptosis-related targets, and finally obtained 5 candidate proteins associated with AD, including CFL1, MGST1, GLS2, DLD, and EIF2AK4 (p < 0.05) (Fig. 2D). When overlapping the results from eQTLs and pQTLs derived from brain tissue, we identified two common genes, DLD and EIF2AK4 (Fig. 3B).

We also conducted similar explorations in blood tissues. Using cis-eQTL SNPs from the eQTLGen Consortium as genetic instruments, we identified 20 potential AD targets (SMR p < 0.05 and HEIDI test p > 0.05) (Fig. 2E). We observed that the candidate genes identified from brain and blood tissue screenings have a high degree of overlap. The overlapping genes between the two tissues include PEX10, CDC25A, PARP14, PARP11, CDH1, ACSF2, EGLN2, TRIB3, and ATF4 (Fig. 3C). We also identified 2 potential ferroptosis-related proteins using cis-pQTL SNPs from circulating 4,907 proteins levels, including BCAT2 and ECH1 (p < 0.05) (Fig. 2F). When comparing the results of eQTLs and pQTLs obtained from blood tissue, we found no overlapping genes (Fig. 3D). More details can be found in Supplementary Table 1.

PPI analysis and expression of the identified genes in different cell types of the human brain

We carried out a PPI analysis of the 30 genes obtained from the brain eQTLs results, using the STRING database (https://string-db.org/). The PPI network exhibited significantly more connections than anticipated (p < 0.001). The top 10 genes (SLC7A11, SLC40A1, SLC3A2, RPTOR, GABARAPL2, EGFR, ATM, ATF4, TXN, and TRIB3) were identified as potential hub genes, based on their Degree with CytoHubba (Supplementary Figure 2). We examined the expression patterns of these prioritized genes using single cell RNA sequencing data. Among the genes associated with AD, PEX10, CDC25A, EGFR, DLD, LIG3, and TRIB3, were observed to be widely expressed across various brain cell types at relatively high levels (Supplementary Figure 3).

Validation of gene targets by transcriptome and western blotting

Using cis-eQTL SNPs from GTEx, PsychENCODE, and BrainMeta as genetic instruments, we identified six putative gene targets associated with AD. The expression levels of these genes were further validated by analyzing in external transcriptomic datasets. We found that EGFR, DLD, PEX10, and LIG3 were significantly expressed compared to the control group (p < 0.05), while TRIB3 and CDC25A did not appear to show statistical significance (Fig. 4). To enhance validation at the protein level, brain tissue was procured from 3xTg-AD mice, followed by western blotting experiment.

OPEN IN VIEWER

Fig. 4

The external transcriptome dataset corroborated the identification of six gene targets through SMR analysis. Notably, EGFR, DLD, PEX10, and LIG3 demonstrated significant expression differences relative to the control group (p < 0.05). Conversely, TRIB3 and CDC25A failed to reach statistical significance.

In the 3xTg-AD, western blotting demonstrated that both EGFR (p < 0.05) and TRIB3 (p < 0.05) were downregulated in the whole brain tissues compared with the control group (Fig. 5). Transcriptome and Western blotting analyses both revealed significant dysregulation of EGFR, underscoring its potential as a therapeutic target for AD.

OPEN IN VIEWER

Fig. 5

Validation of the gene targets by western blotting using 3xTg-AD mice. Both EGFR (p < 0.05) and TRIB3 (p < 0.05) were downregulated in the whole brain tissues compared with the control group. WT, wild type; MOD, model of 3xTg-AD mice.

The ferroptosis-related gene targets associated with AD

Existing research supports the involvement of the ferroptosis-related gene targets that we identified in the pathological process of AD. EGFR signaling has been linked to various cellular processes, such as neurodevelopment and neurodegenerative process. 30 Recent research has demonstrated that EGFR signaling can influence ferroptosis by adjusting the expression of key genes involved in ferroptosis, such as GPX4 and SLC7A11. 31 Some studies have indicated the potential involvement of TRIB3 in neurodegenerative disorders through the modulation of ferroptosis.32,33, 32,33

The PEX10 encodes a peroxisomal membrane protein involved in peroxisome formation. 34 Peroxisomal dysfunction is linked to neurodegenerative disorders such as AD, which shows impaired peroxisomal function and lipid metabolism.34,35, 34,35 CDC25A regulates cell cycle and various cellular processes, which are interconnected with ferroptosis. 36 LIG3, an enzyme crucial for DNA repair, has been linked to impaired DNA repair mechanisms in the process of ferroptosis in AD.37,38, 37,38 Mitochondrial dysfunction has been associated with both ferroptosis and AD, indicating that DLD may indirectly contribute to the dysregulation observed in both conditions.39,40, 39,40

The association of EGFR with ferroptosis

EGFR activation leads to downstream signaling through pathways such as the Ras-MAPK pathway, PI3K-AKT pathway, and STAT3 pathway. These pathways regulate various cellular processes, including proliferation, survival, and differentiation. 41 However, they also influence the expression of genes involved in iron metabolism and oxidative stress, thereby indirectly affecting ferroptosis.42,43, 42,43 EGFR signaling can modulate iron uptake and utilization within cells. For instance, EGFR activation can increase the expression of divalent metal transporter 1 (DMT1), which facilitates iron uptake. 44 Additionally, EGFR signaling can affect the activity of enzymes involved in iron processing, such as ferritin and hephaestin, influencing intracellular iron levels. 43

The accumulation of ROS due to increased iron availability can lead to lipid peroxidation, a hallmark of ferroptosis. EGFR signaling can enhance ROS production through various mechanisms, including upregulation of NADPH oxidases (NOXs).45,46, 45,46 This increased oxidative stress can damage cellular membranes and organelles, promoting ferroptotic cell death.

Interestingly, therapeutic agents targeting EGFR, such as tyrosine kinase inhibitors (TKIs), can also modulate ferroptosis by altering iron metabolism and oxidative stress.47,48, 47,48 For example, erlotinib, an EGFR TKI, has been shown to induce ferroptosis by disrupting iron uptake and enhancing lipid peroxidation. The interaction between EGFR and ferroptosis represents a promising area of research with potential clinical applications. 49 Further elucidation of the underlying molecular mechanisms will facilitate the development of targeted therapeutic strategies aimed at leveraging ferroptosis for AD treatment.

EGFR serving as a therapeutic target for AD

In recent times, there’s been growing interest in repurposing existing anti-cancer drugs that target the EGFR for AD research, given the link between EGFR overactivity and several neurodegenerative conditions, including AD.50,51, 50,51 Research has highlighted that blocking EGFR can reduce the activity of reactive astrocytes, lessen Aβ toxicity and inflammation, and encourage the regrowth of axons. 52 Furthermore, other investigations have noted positive changes in behavior and cognitive functions, alongside decreased amyloid formation and enhanced autophagy, as a result of EGFR blockade in different animal models of AD. 53

Mansour and his colleagues conducted a detailed review of past research, pinpointing overlooked neuroprotective routes and underscoring the importance of exploring EGFR inhibitors as possible treatments for AD. 53 They found that Aβ peptides intensify oxidative stress mediated by NADPH oxidase (NOX).53,54, 53,54 Activating EGFR leads to an increase in ROS via the PI3K pathway, which then sets off the Ras-Rac1 sequence, culminating in NOX activation. 55 Interestingly, NOX appears to be closely associated with AD, suggesting it could serve as a valuable indicator for developing new medications.

Additionally, another EGFR-driven process that contributes to oxidative stress and axon loss involves the ZNRF1–AKT–GSK3β–CRMP2 pathway. 56 ZNRF1, a ubiquitin ligase, is prevalent in neurons throughout the central nervous system (CNS). When neurons are damaged, ZNRF1 breaks down AKT, blocks GSK-3β, and disrupts microtubule stability, leading to axonal damage in the Wallerian model. 56 More research is needed to understand the relationship between NOX-mediated oxidative stress, EGFR, and ZNRF1 in live models of AD.

Limitations

Our study acknowledges several limitations. Initially, despite a notable correlation between eQTLs in blood and brain, these observations are confined to genes expressed in both tissues and possessing eQTLs in them. While these studies can be beneficial for estimating the long-term effects of exposure beyond the scope of randomized controlled trials, they fail to directly reveal the impact of short-term drug treatments on disease risk. Secondly, the complete removal of confounding bias and/or horizontal pleiotropy remains a formidable challenge. Lastly, the discovery of additional candidate genes in the 3xTg-AD mice does not negate their association with AD. Additional laboratory and clinical data are necessary to substantiate these findings.

Conclusions

This study provides genetic evidence supporting the potential therapeutic benefits of targeting the six druggable genes for AD treatment, especially for EGFR (validated by transcriptome and 3xTg-AD), which could be useful for prioritizing AD drug development in the field of ferroptosis.