2026 年 51 巻 2 号 p. 89-100
Methylmercury (MeHg) is a global pollutant that readily crosses the blood–brain barrier and placenta, posing significant risks to fetal neurodevelopment. While the cerebellum is a recognized target of MeHg toxicity in adults, the effect of fetal exposure remains poorly defined. In this study, we investigated the neurotoxic effects of low-dose MeHg exposure (0.2 ppm via drinking water) on the cerebellums of prenatal C57BL/6 mice using integrated transcriptomic and proteomic analyses. Cerebellar tissues collected from postnatal day 90–120 (P90–120) mice (n = 3/group) were processed for RNA sequencing and proteomics analysis. Differentially expressed genes (DEGs) and proteins (DEPs) revealed significant changes (n = 4/group) in multiple pathways associated with neurodegeneration, including Huntington’s disease, Parkinson’s disease, and amyotrophic lateral sclerosis. Overlapping transcriptomic and proteomic findings identified potential underlying mechanisms such as chemical carcinogenesis driven by reactive oxygen species and retrograde endocannabinoid signaling, underscoring the central role of oxidative stress in MeHg-induced neurotoxicity. Collectively, these results indicate that prenatal MeHg exposure induces persistent molecular alterations consistent with neurodegenerative processes and synaptic dysfunction, despite the absence of overt behavioral changes at the time of sacrifice. The long-term consequences for delayed symptom onset and the potential contribution of these changes to the etiology of neurodevelopmental disorders warrant further investigation.
Methylmercury (MeHg) is a well-known environmental neurotoxicant capable of crossing the blood-brain barrier and the placenta, particularly affecting fetal brain development (Grandjean et al., 1997; Bisen-Hersh et al., 2014; Dack et al., 2022). Prenatal and early life exposure to MeHg has been associated with long-term neurodevelopmental impairments, including cognitive deficits, mood disturbances, and motor dysfunction in children (Grandjean et al., 1997; Yorifuji et al., 2011; Debes et al., 2016). In adulthood, exposure to high doses of MeHg can cause severe neurological symptoms such as dysarthria, ataxia, muscle weakness, tremors, visual impairment, and hearing loss (Harada, 1995). The Minamata disaster of 1956, in which thousands were exposed to lethal concentrations of MeHg, revealed neuropathological lesions in three distinct brain regions: the cerebellum, the calcarine cortex, and the central sulcus. Critically, loss of cerebellar granule cells and structural lesions in both the cortex and cerebellum highlighted the particular vulnerability of these regions, which are essential for coordination and balance in both adults and prenatally exposed infants (Takeuchi et al., 1962). Subsequent rodent studies confirmed that prenatal MeHg exposure induces structural damage and neurodegeneration in the cerebellum (Reuhl et al., 1981). Despite this historical evidence, the molecular and biochemical mechanisms underlying the long-term neurological effects of prenatal MeHg exposure remain incompletely understood.
The cerebral cortex plays an important role in cognitive processing and social behavior. It has been consistently linked to abnormalities associated with neurodevelopmental disorders, particularly autism spectrum disorder (ASD) (Sacai et al., 2020). ASD is characterized by deficits in social communication, impaired interaction, and repetitive behaviors. Our previous work investigated the effects of prenatal low-dose, non-apoptotic MeHg exposure on neurobehavioral outcomes in adult mice. Pregnant dams received 0 or 0.2 ppm MeHg in drinking water from embryonic day 0 (E0) until postnatal day 0 (P0). Offspring were later subjected to behavioral testing, which revealed ASD-like phenotypes, including impaired ultrasonic vocalizations, reduced sociability, and increased repetitive behaviors. At the cellular level, low-dose MeHg accelerated cortical neuron differentiation, suggesting that altered timing of neurodevelopment in the cortex may contribute to these behavioral abnormalities (Loan et al., 2023). While this study focused specifically on cortical development, the effects of low-dose MeHg on other brain regions are yet to be investigated.
The cerebellum, which plays a critical role in motor coordination, balance, and emotional regulation, is another major target of MeHg-induced neurotoxicity (Shao et al., 2015). Reduced cerebellar function has been observed in patients with neurodegenerative disorders such as Alzheimer's disease, Parkinson's disease, Huntington's disease, and amyotrophic lateral sclerosis (Iskusnykh et al., 2024). At the biochemical level, prenatal MeHg exposure disrupts multiple pathways in the cerebellum. Reported effects include reduced expression of synaptic proteins such as synaptophysin and PSD-95, suppression of the TrkA signaling pathway, and decreased eEF1A1 expression, all of which may impair synaptic connectivity (Fujimura et al., 2012, 2016). Other studies have reported oxidative stress and redox imbalance in the cerebellum, along with decreased antioxidant enzymes (Huang et al., 2011; Heimfarth et al., 2018). At the cellular level, prenatal treatment has been associated with neuron degeneration, downregulation of myelin, and altered synaptic plasticity (da Silva et al., 2022; Wang et al., 2022). This culminates at the systems level, where cerebellar alterations have been linked to deficits in both short-term and long-term memory (Bittencourt et al., 2022; Fagundes et al., 2022). Collectively, these findings suggest that prenatal MeHg exposure alters neural connectivity and redox homeostasis, potentially driving long-term impairments in motor and cognitive function.
As the global incidence of neurodegenerative diseases continues to rise, it is increasingly important to identify potential environmental contributors in order to mitigate future disease burden (Feigin et al., 2020). Building on previous findings that prenatal MeHg exposure elicits cerebellar alterations consistent with neurodegenerative disease, this study aims to elucidate the mechanisms linking low-dose prenatal MeHg exposure to long-term neurobehavioral dysfunction through an integrated dual-omics approach. We conducted transcriptomic and proteomic analyses of the adult murine cerebellum following prenatal MeHg treatment to identify changes in gene and protein expression and uncover the biochemical pathways disrupted by exposure. While transcriptomics captures gene expression changes, proteomics reveals alterations not always reflected at the mRNA level. Together, this integrated approach provides a comprehensive view of how prenatal MeHg disrupts neurobiological pathways, advancing mechanistic insight into how early-life exposure contributes to lasting neurodegenerative risk.
Pregnant C57BL/6 mice were dosed with 0 ppm or 0.2 ppm MeHg via drinking water starting from the first day of their pregnancy (E0) until birth (P0). The 0.2 ppm MeHg solution was prepared by a 1:5000 dilution of a stock solution of 4 mM MeHg(II)Cl (cat#33553, Alfa Aesar) in drinking water (Loan et al., 2023). At sacrifice, the brains were dissected, and the cerebellum was isolated and immediately frozen with liquid nitrogen and stored at -80°C for analysis.
RNA extractionFrozen brain tissue samples were each suspended in 1 mL of TRIzol reagent (cat#15596026, Thermo Fisher) and were added directly to a microcentrifuge tube. The tissue was homogenized using a tissue homogenizer and incubated for 5 min at room temperature. 200 μL of chloroform (cat#C2432-1L, Sigma-Aldrich) was added to the mixture, which was then vortexed for 15 sec to ensure thorough mixing. The samples were incubated at room temperature for 2-3 min and then centrifuged at 12,000 x g for 15 min at 4°C. The aqueous phase was carefully transferred to a new RNAse-free tube, and an equal volume of 70% ethanol was added. Washes were performed using the PureLink RNA Mini Kit (cat# 12183020, Thermo Fisher). The extracted RNA was stored at -80°C for long-term preservation. The quantity and purity of the extracted RNA were assessed using a spectrophotometer, with all samples showing A260/A280 ratios of ~2.1 and A260/A230 ratios close to 2.0, and RNA integrity was confirmed with RIN values ≥ 9.6.
Bulk-RNA sequencing and bioinformatic analysisRNA samples from biological triplicates (both female and male mice) were isolated as previously described and sent to Sacramento, California, for RNA-Seq analysis at Novogene Bioinformation Technology Co., Ltd. (Sacramento, USA). Sequencing was performed using the Illumina NovaSeq platform with paired-end 150 base pair reads (PE150) and reference genome mm39. DESeq2 was used for differential gene expression analysis (Love et al., 2014). Differentially expressed genes (DEGs) generated via the NovoMagic platform were then filtered and analyzed using the R programming language in RStudio (v.4.3.0) (R Core Team, 2023).
Gene expression data from Novomagic for the control (0 ppm MeHg) and 0.2 ppm MeHg-treated groups were filtered to include only genes quantified in 100% of the samples per group, which were then used for the analysis. To visualize common and unique genes, a Venn diagram was created using the eulerr package. DEGs generated from Novomagic were filtered based on padj < 0.05 and Log2 fold-change > 0 for upregulated and < 0 for downregulated genes. A heatmap was created from all significant DEGs using the pheatmap package. These DEGs were then used for KEGG (Kyoto Encyclopedia of Genes and Genomes) Pathway Enrichment Analysis using the enrichKEGG and Gene Ontology (GO) Enrichment Analysis using the enrichGO function from the clusterProfiler package. GO terms were classified into three categories: biological process (BP), molecular function (MF), and cellular component (CC).
Digestion and preparation of brain tissue for proteomic analysisFrozen brain tissue samples were each suspended in 0.5 mL of ice-cold lysis buffer and homogenized for 3 min. Lysis buffer consisted of 1.5 M urea, 5% (v/v) glycerol, 1 mM DTT, 1:200 (v/v) protease inhibitor cocktail (cat#78430, Thermo Fisher), and 25 mM HEPES at pH 8.0. The suspension was centrifuged at 12,000 x g for 10 min at 4°C, and the soluble protein supernatants were separated from the insoluble debris. 50 μL of the resulting supernatant from each sample was used for protein quantification, and 150 μL of the supernatant was used for proteomic sample preparation.
Protein quantification by Bradford assayThe protein concentration and quantity of the resulting supernatant were determined by the Bradford assay following a previously described method (Li et al., 2024).
Protein reduction, alkylation, and enzymatic digestionAliquots of 50 µg of protein were further processed using a previously described modified filter-aided sample preparation (FASP) protocol for proteomic analysis (Li et al., 2024). Briefly, the resulting supernatant samples were diluted with a denaturation buffer (8 M urea, 25 mM HEPES at pH = 8.0), vortexed briefly, and transferred to a 10 kDa MWCO filter. The samples were then processed as previously described (Li et al., 2024). A mixture of peptides equivalent to 50 µg of protein was desalted and dried via vacuum centrifugation as previously described.
Untargeted proteomic data acquisition using nano-LC MS/MS Nano-LC-MS/MSA peptide mixture containing 5 µg of total protein content, was analyzed using a system comprising the UltiMate 3000 nanoRSLC (Dionex, Thermo Fisher Scientific, Mississauga, ON, Canada) coupled with an Orbitrap Fusion mass spectrometer (Thermo Fisher Scientific, Mississauga, ON, Canada). The samples were processed using a previously described method (Li et al., 2024). Briefly, the mobile phase consisted of water, acetonitrile (ACN), and 0.1% (v/v) formic acid (FA), delivered at a flow rate of 0.30 µL/min under gradient.
MS Spectra Processing and Bioinformatic AnalysisRaw mass spectrometry data files were processed using MaxQuant (v.2.6.4.0). The dataset included 8 samples, comprising 4 control brain tissue samples and 4 brain tissue samples exposed to MeHg. Peptide identification was performed against the Mus Musculus species (NCBI taxonomy ID: 10090) FASTA database, along with a default contaminants database, using the integrated Andromeda search engine. The raw data files were processed following previously described parameters (Li et al., 2024). Default parameters were applied unless specified otherwise.
Bioinformatics of sample LFQ intensities was conducted using the R programming language in RStudio (v.4.4.1.) (R Core Team, 2023). Potential contaminants and reverse proteins were removed from the MaxQuant data. Only proteins that were quantified in at least 50% of samples per group were used for the analysis. To visualize common and unique proteins, a Venn diagram was created using the eulerr package. To investigate the effect of MeHg on the proteome, a 2-way ANOVA analysis was conducted comparing common Log2 fold-change normalized LFQ intensities of proteins between control (0 ppm MeHg) and experimental (0.2 ppm MeHg-exposed) groups using the aov() function from the stats package. This function performs analysis of variance to test for statistically significant differences among group means. A Tukey’s Honest Significance Difference (TukeyHSD) test was then used to calculate padj. In addition to the adjusted p-values, the Log2 fold-change was also calculated. Overall, differentially expressed proteins (DEPs) were defined as those with a padj < 0.05 and a minimum Log2 fold-change of 0.5.
A heatmap was created from all significant DEPs using the pheatmap package. Lastly, KEGG and GO enrichment were conducted using significantly up- and down-regulated proteins as well as proteins unique to the MeHg exposed samples using the ClusterProfiler package. Uniprot IDs were converted to Entrez IDs using the getBM function, and both analyses were done using Mus Musculus (mmu) as the organism. GO terms were classified into BP, MF, and CC categories.
Proteomic network synthesis and hub node analysisA full STRING network was constructed of DEPs (padj <0.05) using the network function on string db (string-db.org) (Fig. S1). Default settings were used unless otherwise stated. Due to the sparse number of proteins used for network construction <50, the complexity of the network was increased manually in order to perform the Hub Node Analysis, such that the 1st shell interactors were set to 50. Access to the network and all the settings is permanently linked in Fig. S1.
The network was then imported into Cytoscape v.3.10.3, and the CytoHubba plug-in was used to determine the hub nodes. The top 20 hub nodes were calculated using the embedded Degree algorithm in the CytoHubba plugin and highlighted based on the degree of interaction, where higher-ranked nodes have more protein-protein interactions compared to lower-ranked nodes.
Transcriptomic analysis revealed 20,514 shared genes between control and MeHg-treated mice, a 93.5% similarity (relative to control) (Fig. 1A). A total of 1385 genes are differentially expressed, with 542 genes upregulated (padj < 0.05 and Log2 fold-change > 0) and 843 genes downregulated (padj < 0.05 and Log2 fold-change < 0) (Fig. 1B-C).
Transcriptomic alterations in the cerebellum of mice prenatally treated with MeHg. (A) Venn diagram showing the overlap of genes between the control (0 ppm MeHg) and 0.2 ppm MeHg-treated groups. (B) Volcano plot of DEGs between 0 ppm and 0.2 ppm MeHg. Discriminated based on Log2 fold-change > or < 0 and padj < 0.05. (C) Heatmap of DEGs between control and 0.2 ppm MeHg-treated samples. (D-E) Dot plot depicting KEGG pathways that are significantly enriched from both (D) upregulated or (E) downregulated DEGs. Each dot represents a specific KEGG pathway, with the size of the dot corresponding to the number of genes involved in the pathway, and the color indicating the q-value of the enrichment analysis. (F-G) Bar plot depicting GO terms for BP, CC, or MF that are significantly enriched from both (F) upregulated or (G) downregulated DEGs.
The differentially expressed genes (DEGs) are involved in 99 enriched KEGG pathways (q-value < 0.05); 12 pathways from upregulated DEGs and 87 pathways identified from downregulated DEGs. We presented the top 10 KEGG pathways identified from both up- and down-regulated DEGs (Fig. 1D-E). Of the top 10 enriched KEGG pathways from upregulated genes, 4 are related to neurodegenerative disease. These include amyotrophic lateral sclerosis (mmu05014, q-value = 5.63×10-5), Parkinson's disease (mmu05012, q-value = 1.95×10-3), Huntington's disease (mmu05016, padj = 1.18×10-3), and pathways of neurodegeneration (mmu05022, q-value = 5.19×10-3) (Fig. 1D). KEGG pathways identified from downregulated DEGs include long-term potentiation (mmu04720, q-value = 3.92×10-8), cAMP signaling pathway (mmu04024, q-value = 3.76×10-7), glutamatergic synapse (mmu04724, q-value = 2.13×10-6), and cholinergic synapse (mmu04725, q-value = 1.87×10-5) and are all involved in neurotransmission and synaptic plasticity (Fig. 1E).
Using Gene Ontology (GO) Enrichment Analysis, we identified 1425, 219, and 218 significantly enriched GO terms for biological process (BP), molecular function (MF), and cellular component (CC), respectively (q-value < 0.05). We showed the top 5 GO terms identified from both up- and down-regulated DEGs (Fig. 1F-G). GO terms of interest identified from the upregulated DEGs include those related to ribosome dysregulation, including structural constituent of ribosome (GO:0003735, q-value = 2.32×10−15), cytosolic ribosome (GO:0022626, q-value = 9.86×10−22), and ribosome (GO:0005840, q-value = 3.82×10−14). Additional enriched terms include those associated with extracellular matrix remodeling and synaptic structure, such as extracellular matrix structural constituent (GO:0005201, q-value = 4.63×10−7) and cell adhesion molecule binding (GO:0050839, q-value = 8.43×10−6) (Fig. 1F). GO terms of interest identified from the downregulated DEGs include axonogenesis (GO:0007409, q-value = 1.54×10-10), asymmetric synapse (GO:0045211, q-value = 2.49×10-20), postsynaptic density (GO:0014069, q-value = 3.01×10-19), and GABA-ergic synapse (GO:0098982, q-value = 3.07×10-12) (Fig. 1G).
Proteomics analysisMass spectrometry analysis revealed 2,666 proteins that were quantified in at least 50% of samples per group. Among them, 50 and 55 proteins were unique to control and MeHg-exposed mouse brain samples, respectively, with 2,561 proteins common to both groups (Fig. 2A). A total of 40 proteins were found to be significantly upregulated in MeHg-exposed samples, with 15 proteins in female samples and 25 proteins in male samples (padj < 0.05 and Log2 fold-change > 0.5). A total of 23 proteins were also found to be significantly downregulated in MeHg (padj < 0.05 and Log2 fold-change < 0.5) (Fig. 2B-C).
Proteomic alterations in the cerebellum of mice prenatally treated with MeHg. (A) Venn diagram showing the overlap of proteins between the control (0 ppm MeHg) and 0.2 ppm MeHg-treated groups. (B) Volcano plot of DEPs between 0 ppm and 0.2 ppm MeHg. Discriminated based on Log2 fold-change > or < 0.5 and padj < 0.05. (C) Heatmap of DEPs between control and 0.2 ppm MeHg-treated samples. (D) Dot plot depicting KEGG pathways that are significantly enriched from upregulated DEPs. Each dot represents a specific KEGG pathway, with the size of the dot corresponding to the number of proteins involved in the pathway, and the color indicating the q-value of the enrichment analysis. (E-F) Bar plot depicting GO terms for BP, CC, or MF that are significantly enriched from both (E) upregulated or (F) downregulated DEPs.
KEGG Enrichment Analysis of significantly upregulated proteins in MeHg-exposed samples showed 13 significant pathways (q-value < 0.05). Of the top 10 upregulated KEGG pathways, 3 are related to neurodegenerative disease; these include Parkinson's disease (mmu05012, q-value = 0.015), Huntington's disease (mmu05016, q-value = 0.004), and Alzheimer’s disease (mmu05010, q-value = 0.009) (Fig. 2D). No significant KEGG pathways were identified from downregulated DEPs.
Additionally, GO Enrichment Analysis of differentially expressed proteins (DEPs) and unique proteins to the MeHg exposed samples identified 144, 95, and 64 significantly enriched GO terms for BP, CC, and MF, respectively (q-value < 0.05) with most of them involved in the sub-cellular respiration and energy metabolism (Fig. 2E-F).
Dual-omics integrationComparison of KEGG pathway enrichment from both transcriptomic and proteomic analyses revealed seven shared pathways (q-value < 0.05) (Fig. 3A-C). Notably, four of these are directly associated with neurodegenerative diseases: amyotrophic lateral sclerosis (mmu05014), pathways of neurodegeneration multiple diseases (mmu05022), Parkinson’s disease (mmu05012), and Huntington’s disease (mmu05016). The remaining three shared pathways, chemical carcinogenesis reactive oxygen species (mmu05208), thermogenesis (mmu04714), and retrograde endocannabinoid signaling (mmu04723), are closely linked to oxidative stress and synaptic dysfunction (Fig. 3B-C).
Comparison between transcriptomics and proteomics data. (A) Venn diagram of KEGG pathways identified from DEGs (orange) and DEPs (maroon) and their overlap. (B-C) Dot plots depicting the 7 significantly enriched overlapping KEGG pathways for both (B) transcriptomics and (C) proteomics. Each dot represents a specific KEGG pathway, with the size of the dot corresponding to the number of proteins/genes involved in the pathway, and the color indicating the q-value of the enrichment analysis in the datasets. (D) Venn diagram of GO terms identified from DEGs (orange) and DEPs (maroon) and their overlap. From left to right each Venn represents a comparison of BP, CC, and MF.
Similarly, GO term enrichment revealed overlaps between transcriptomic and proteomic datasets, with 41.7%, 31.6%, and 15.6% similarity across BP, CC, and MF categories, respectively (q-value < 0.05, relative to the proteomics dataset). Key overlapping terms include response to oxidative stress (GO:0006979, BP), respiratory chain complex (GO:0098803, CC), and antioxidant activity (GO:0016209, MF), highlighting convergence on pathways related to mitochondrial function and redox imbalance (Fig. 3D, Table S1-3).
The hub node network analysis of significant DEPs revealed two major clusters of interacting proteins, pointing to distinct biological processes potentially perturbed by prenatal MeHg exposure (Fig. 4). The first cluster included several members of the NADH:Ubiquinone Oxidoreductase (Nduf) family, specifically Ndufb10, Ndufa7, Ndufs5, and Ndufa2. These ranked among the top 20 hub proteins and suggest disruption of mitochondrial Complex I function (Kahlhöfer et al., 2021). The second cluster featured Low-Density Lipoprotein Receptor (Ldlr) and Clathrin Heavy Chain 1 (Cltc), implicating pathways related to lipid metabolism and vesicle trafficking (Fig. 4).
Proteomic Hub Node Analysis Data. Protein network of significant DEPs and identified hub nodes created using CytoScape and the CytoHubba plugin. Each square represents one node in the network, identified with its corresponding gene name. Blue squares represent non-hub nodes within the network, and red, orange, and yellow squares represent hub nodes ranked based on the number of protein-protein interactions. Red nodes are the highest ranked, and yellow nodes are the lowest ranked. The top 20 hub nodes are presented.
Our current research employed a dual-omics approach integrating transcriptomics and proteomics to provide a comprehensive understanding of the long-term effects of prenatal MeHg exposure on the brain. Using this extensive methodology, we found that prenatal MeHg exposure induces long-lasting molecular changes in the adult cerebellum, despite the absence of overt behavioral deficits in early adulthood.
This study builds on our previous work, where we investigated the effects of prenatal MeHg exposure on the developing cerebral cortex (Loan et al., 2023). Pregnant mice were treated with 0 or 0.2 ppm MeHg in drinking water from E0 to P0. We then assessed behavioral changes in adult offspring beginning at P60. While we observed significant alterations in communication, sociability, and repetitive behaviors, there were no changes in locomotor activity, learning, or memory, as measured by the open-field and Morris Water Maze tests, respectively (Loan et al., 2023). Despite this, our current omics analysis revealed that MeHg exposure regulates the expression of genes and proteins associated with several neurodegenerative diseases, including Huntington’s disease, Parkinson’s disease, and Alzheimer’s disease. Importantly, a direct comparison of KEGG pathways enriched in both the transcriptomic and proteomic datasets revealed seven overlapping pathways. Of these, four are directly implicated in neurodegenerative disease. This strong overlap across omics layers reinforces the specificity of our findings.
Subsequent protein network analysis of significant DEPs identified many mitochondrial proteins in the Nduf family, including Ndufb10, Ndufa7, and Ndufs5, Ndufa2 as hub nodes, key proteins within a network. Nduf proteins are subunits that make up the mitochondrial complex I enzyme within the electron transport chain (ETC), which takes electrons from NADH and transfers them to ubiquinone (Kahlhöfer et al., 2021). Dysregulation of these proteins has been shown to occur during periods of oxidative stress and has been linked to neurodegenerative disorders, particularly Alzheimer’s and Parkinson’s (Grünewald et al., 2016; Adav et al., 2019). This finding further supports the dual-omic KEGG pathway findings of this analysis, which relate MeHg exposure to neurodegeneration in the cerebellum. In addition to Nduf, two other hub proteins, Clatherin heavy chain-1 (CLTC) and low-density lipoprotein receptor (Ldlr), have also been implicated in neurodegeneration (Kim et al., 2009; Nardecchia et al., 2023). This is further supported by overlapping GO terms. Both the transcriptomic dataset and proteomics dataset identified response to oxidative stress (GO:0006979) as the most significant GO term following MeHg treatment.
Our previous publication also assessed mercury (Hg) accumulation in multiple brain regions at sacrifice and found no significant difference in the level of Hg in the cerebellums of MeHg and control mice (Loan et al., 2023). Here, our results suggest that the molecular and transcriptional changes observed in the cerebellum are not simply a result of persistent Hg accumulation into adulthood. Instead, it underscores the particular vulnerability of the prenatal period as a critical window for cerebellar development, during which even transient MeHg exposure can induce long-lasting effects.
These findings are supported by existing literature linking mercury exposure to increased neurodegenerative disease risk. For instance, elevated blood and brain mercury levels have been associated with an increased risk of Alzheimer’s disease (Mutter et al., 2004; Gerhardsson et al., 2008). Similarly, studies suggest that occupational and environmental exposure to MeHg causes mitochondrial dysfunction, α-synuclein aggregation, and neural loss (Yamin et al., 2003; Pamphlett and Bishop, 2022), potentially contributing to an increased risk of Parkinson’s disease (Ngim and Devathasan, 1989; Peterson and Nutt, 2008; Bittencourt et al., 2022). Many of these same mechanisms were reflected in our dual-omics analysis, which showed consistent enrichment of pathways related to Alzheimer’s, Parkinson’s, Huntington’s, and amyotrophic lateral sclerosis. These results suggest that MeHg may play a role in molecular cascades associated with neurodegeneration.
Aside from neurodegenerative diseases, it is interesting to note that the chemical carcinogenesis - reactive oxygen species KEGG pathway was also enriched in both transcriptomics and proteomics datasets. This finding highlights a well-documented mechanism of MeHg-induced oxidative stress, where mercury’s affinity for thiol groups allows it to bind and inactivate antioxidant enzymes, depleting glutathione, promoting excessive ROS generation, and impairing mitochondrial function (Aschner et al., 2007; Farina et al., 2011; Pyatha et al., 2022). Additionally, prolonged exposure to heavy metals, including MeHg, has also been shown to trigger metal-induced carcinogenesis through oxidative damage (Buha et al., 2021).
One likely explanation for the disconnect between molecular changes and behavioral outcomes is the timing of our assessments. Behavioral testing in our study began at P60, a time point generally considered early adulthood in mice. However, many neurodegenerative phenotypes in mouse models typically do not manifest until 6 to 12 months of age. It is therefore possible that our behavioral assays occurred too early to detect the functional consequences of the observed molecular changes. This hypothesis is further supported by other rat studies, where prenatal MeHg exposure resulted in impaired learning and memory in adulthood, accompanied by disrupted redox homeostasis (Fagundes et al., 2022). Historical human data also reinforce this connection. In Minamata, Japan, where industrial wastewater contamination caused widespread MeHg poisoning, new cases of neurological impairment continued to emerge more than a decade after exposure had ceased, with incidence peaking in 1975—15 years after the acute exposure ended. Similarly, the grain contamination disaster in Iraq revealed latency periods for neurological symptoms ranging from 16 to 38 days (Weiss et al., 2002). These examples illustrate the often-delayed onset of MeHg-induced neurological symptoms, suggesting that early behavioral assessments may underestimate long-term risk.
Taken together, our findings suggest that while behavioral abnormalities consistent with neurodegenerative disease were not evident at the time of sacrifice, the cerebellum of MeHg-exposed mice harbors persistent molecular signatures of neurodegeneration. These changes may precede or predict future behavioral deficits. Follow-up studies are crucial for determining whether these molecular alterations translate into clinical phenotypes later in life. Extending the behavioral assessments into mid- to late adulthood will be critical to capture delayed functional impairments. Additionally, using disease-relevant genetic models could help determine whether prenatal MeHg exposure exacerbates disease progression or lowers the threshold for phenotypic onset.
The authors would like to thank the University of Ottawa Animal Care Veterinary Services (ACVS) at the University of Ottawa for technical support, advice, and discussions. The authors would also like to thank Novogene (Sacramento, California, Novogene Bioinformation Technology Co., Ltd.) for RNA-seq library preparation and high-throughput sequencing.
FundingThis work was supported by NSERC Discovery Grant (06605/RGPIN/2019) and Canada Research Chair Grant (950-225645) to H.M. Chan and a CIHR Project Grant (PJT1958223A). A. Loan was supported by an NSERC-CREATE grant (CREATE-449153) and the Queen Elizabeth II Graduate Scholarship in Science and Technology.
Conflict of interestThe authors declare that there is no conflict of interest.
Data availabilityThe RNA-seq count data are available on NCBI’s Gene Expression Omnibus (GEO) under the accession number GSE308270. Code available at https://github.com/allisonloan/MeHg-Cerebellum
All MS raw data and Maxquant processed results were submitted to the PRIDE repository under the accession number PXD068503, at the European Bioinformatics Institute.
Author contributionsData curation: A.L., R.D., Y.L., T.N., and Z.M.
Formal analysis: A.L. and R.D.
Funding acquisition: J.W. and H.M.C.
Methodology: J.W., H.M.C., A.L., and R.D
Supervision: J.W. and H.M.C.
Visualization: Y.L., T.N., and Z.M.
Writing – original draft: J.W., H.M.C., A.L., and R.D.
Writing – review & editing: Y.L., T.N., Z.M., J.W., H.M.C., A.L., and R.D.
Ethical approval and consent to participateNot applicable.
Patient consent for publicationNot applicable.
