HNRNPC as a Novel Therapeutic Target for Ischemic Heart Disease: Evidence From Mendelian Randomization and Experimental Validation

Abstract

Background

Several studies have suggested that N6‐methyladenosine (m6A) plays an essential role in cardiovascular disease, but the causality of m6A on ischemic heart disease (IHD) remains unknown. Therefore, this study investigated the potential relationship between m6A and IHD using a 2‐sample Mendelian randomization method.

Methods

The publicly available genome‐wide association study data for m6A‐related proteins were obtained from the INTERVAL study, a large population‐based cohort of healthy blood donors in the United Kingdom, whereas the genome‐wide association study database (including 30 952 cases and 187 840 healthy controls) provided the IHD data. We performed a 2‐sample Mendelian randomization analysis to evaluate the potential causal association between HNRNPC (heterogeneous nuclear ribonucleoprotein C) and IHD, followed by experimental validation in vitro and in vivo to confirm the role of HNRNPC in IHD pathogenesis.

Results

There was no indication of pleiotropy or heterogeneity among the 6 m6A‐associated proteins, but Mendelian randomization analysis revealed that HNRNPC (odds ratio [OR], 0.93 [95% CI, 0.88–0.97]; P=0.002) was associated with IHD. When IHD developed, there was a significant upregulation of HNRNPC expression in both animal and cellular tests. HNRNPC knockdown prevented oxidative stress, mitochondrial dysfunction, and cell death.

Conclusions

The Mendelian randomization study suggests a potential causal association of the m6A‐related protein HNRNPC in the cause of IHD and verified the accuracy of the results through a series of experiments, which will help us understand the pathogenesis of IHD and identify potential therapeutic targets in the future.

Nonstandard Abbreviations and Acronyms

ALKBH5

ALKB homologue 5 protein

ELAVL1

ELAV‐like RNA binding protein 1

FS

fractional shortening

HE

hematoxylin/eosin

HNRNPC

heterogeneous nuclear ribonucleoprotein C

IHD

ischemic heart disease

IV

instrumental variable

LRPPRC

leucine rich pentatricopeptide repeat containing

m6A

N6‐methyladenosine

METTL14

methyltransferase‐like 14

METTL3

methyltransferase‐like 3

MR

Mendelian randomization

WTAP

Wilms tumor 1 associated protein

YTHDC1

YT521‐B homology‐domain‐containing protein 1

Clinical Perspective.

What Is New?

• This study identifies HNRNPC (heterogeneous nuclear ribonucleoprotein C), an N6‐methyladenosine RNA‐binding protein, as a novel potential therapeutic target for ischemic heart disease through integrated Mendelian randomization and experimental validation.

• We demonstrate that HNRNPC expression is upregulated during myocardial ischemia and that its knockdown attenuates oxidative stress, mitochondrial dysfunction, and cardiomyocyte death.

What Are the Clinical Implications?

• Targeting HNRNPC or its regulatory pathway may offer a new strategy to protect the heart from ischemic injury by mitigating key pathological processes like oxidative damage.

Ischemic heart disease (IHD), sometimes referred to as atherosclerotic coronary artery disease, is typically caused by atherosclerosis, which results in chronic myocardial ischemia. This causes aberrant myocardial contraction and relaxation. ¹ The primary cause of morbidity and mortality worldwide is IHD. ² , ³ The treatment of IHD is currently mainly drug treatment, and surgical revascularization methods can only relieve the symptoms and cannot solve the problem of permanent myocardial tissue loss. ⁴ Emerging therapeutic drugs, such as stem cells and stem cell‐derived exosomes, can reverse the myocardial ischemic necrosis caused by IHD. ⁵ , ⁶ The current study also found that DNA methylation ⁷ , ⁸ and histone acetylation ⁹ , ¹⁰ play an important role in angiogenesis, indicating that epigenetics has the potential ability to treat ischemic diseases.

N6‐methyladenosine (m6A) is a reversible and heritable epigenetic modification discovered in the 1970s that refers to the substitution of hydrogen at the N6 site of adenosine in RNA by methyl groups. ¹¹ , ¹² , ¹³ , ¹⁴ The m6A modification is a dynamic and reversible process existing in eukaryotic species. It is mainly regulated by 3 proteases, among which methylase (writers) adds a methyl group, demethylase (erasers) removes the methyl group, and RNA‐binding proteins (readers) combine m6A‐methylation sites to regulate a series of biological processes. ¹⁵ WTAP (Wilms tumor 1‐associated protein), METTL3 (methyltransferase‐like 3), and METTL14 (methyltransferase‐like 14) are the primary regulators of the writers. Erasers currently include only the FTO (fat mass and obesity‐associated protein) and ALKBH5 (ALKB homolog 5 protein) in the ALKBH family. ¹⁶ The latest study also found other readers, including HNRNPs (heterogeneous nuclear ribonucleoproteins), IGF2BPs (insulin‐like growth factor 2 mRNA‐binding proteins), ELAVL1 (ELAV‐like RNA binding protein 1), and LRPPRC (leucine rich pentatricopeptide repeat containing). ¹⁷ , ¹⁸ , ¹⁹ , ²⁰

RNA‐binding proteins posttranscriptionally regulate cellular homeostasis by controlling RNA abundance and function, and their dysregulation has now been found to be associated with cardiovascular disease and cancer. ²¹ HNRNPC (heterogeneous nuclear ribonucleoprotein C), an RNA‐binding protein, is a recently discovered m6A reader ²² that controls RNA metabolism, including selective splicing, mRNA stabilization, and translation. ²³ , ²⁴ HNRNPC is aberrantly produced in injured human hearts and plays a crucial role in controlling cardiac contractile proteins. Alterations in HNRNPC expression, phosphorylation, and localization can be determined mechanistically and affect selective splicing of mRNAs involved in mechanotransduction and cardiovascular disease. ²⁵ These mechanisms may underlie the pathogenesis of IHD. Although multiple m6A regulators have been implicated in cardiovascular diseases, the precise role of HNRNPC in cardiovascular disorders remains elusive. To date, no studies have integrated population‐based genetic causal inference with experimental validation to demonstrate the association between m6A readers and IHD. Therefore, this study aimed to address this knowledge gap by integrating Mendelian randomization (MR) genetic evidence with experimental verification, thereby establishing HNRNPC as a novel therapeutic target.

MR is an analytic approach used in epidemiological studies to evaluate causal inference. It uses genetic variations that are strongly correlated with exposure factors, usually single‐nucleotide polymorphisms (SNPs), as instrumental variables (IVs) to evaluate the causal effects of exposure on outcomes. ²⁶ In our study, we used the comprehensive European population‐based genome‐wide association study analysis summary statistical data of plasma m6A‐associated proteins to conduct MR research. This MR focuses on the causal associations between IHD and m6A‐associated proteins, such as ALKBH5 (ALKB homologue 5 protein) in erasers and ELAVL1, heterogeneous nuclear ribonucleoprotein A2/B1 (HNRNPA2B1), HNRNPC, LRPPRC, and YTHDC1 (YT521‐B homology‐domain‐containing protein 1) in writers. Despite the obvious advantages of MR, there are still limitations to its accuracy of results. Therefore, we performed partial experiments to verify the accuracy of MR results. Specifically, through cellular and animal experiments, we evaluated the effect of HNRNPC on the occurrence of IHD, and our results showed that the occurrence of IHD was accompanied by an upregulation of HNRNPC expression, which confirmed the results of MR analysis. These results suggest that MR can reasonably infer potential causal association and provide new ideas for exploring the pathogenesis of IHD.

METHODS

Raw data, detailed methods, and materials will be made available to any researcher upon request to the corresponding author.

Research Report Guidelines and Study Design

The MR analysis was performed according to the latest Strengthening the Reporting of Observational Studies in Epidemiology Using Mendelian Randomization (STROBE‐MR) guidelines. ²⁶ Informed consent and ethical standards for each study were available in the original publication and were implemented by the ethical guidelines of the 1975 Declaration of Helsinki. This is a study to assess the causal associations between m6A‐associated proteins and IHD via 2‐sample MR. The whole MR analysis should satisfy the following 3 hypotheses: (1) IVs must be significantly associated with m6A‐associated proteins, (2) IVs are not associated with any confounders, and (3) IVs can only affect IHD through m6A‐associated proteins (Figure 1A).

Figure 1. MR analysis methods and results.

A, A directed acyclic map was used to indicate the causal associations between m6A‐associated proteins and IHD. B, Forest plot of the results of the m6A‐associated proteins MR analysis. ALKBH5 indicates ALKB homologue 5 protein; ELAVL1, ELAV‐like RNA binding protein 1; HNRNPA2B1, heterogeneous nuclear ribonucleoprotein A2/B1; HNRNPC, heterogeneous nuclear ribonucleoprotein C; IHD, ischemic heart disease; LRPPRC, leucine rich pentatricopeptide repeat containing; m6A, N6‐methyladenosine; MR, Mendelian randomization; OR, odds ratio; SNPs, single‐nucleotide polymorphisms; and YTHDC1, YT521‐B homology‐domain‐containing protein 1.

Data Source

The genetic variation in m6A‐associated proteins was derived from the publicly available INTERVAL study, a large population‐based cohort of healthy blood donors in the United Kingdom (https://www.phpc.cam.ac.uk/ceu/proteins/), which measured about 3622 plasma proteins from 3301 healthy European participants. ²⁷ Using a sample size of 218 792 European people, comprising 30 952 cases and 187 840 controls, the Integrative Epidemiology Unit genome‐wide association study provided the summary statistics for the IHD data (https://gwas.mrcieu.ac.uk).

Selection of IVs

First, all IVs we selected were significantly associated with the exposure proteins (P<5e‐6). ²⁸ , ²⁹ Second, we tested the corresponding linkage disequilibrium to determine the SNPs at linkage disequilibrium. The linkage disequilibrium threshold for these SNPs was r ²<0.001, kilobases (kb)>10 000 based on European ancestry reference data. For each SNP, we calculated the individual F statistic using the formula F=(β/SE)². SNPs with an individual F statistic <10 were excluded to avoid weak instrument bias. After selection, all retained SNPs had F>10. To further assess overall instrument strength, we also computed the mean F statistic for each protein's instrument set, all of which substantially exceeded the conventional threshold (F>10), confirming that weak instrument bias is unlikely. In addition, we also used the PhenoScanner V2 (http://www.phenoscanner.medschl.cam.ac.uk/) database to evaluate the secondary phenotype of each SNP, ²⁷ excluding the SNP associated with the outcome (IHD) at P<5e‐8 through alternative pathways, and the remaining SNPs were used for subsequent analyses.

To assess potential horizontal pleiotropy, we queried the PhenoScanner V2 database for each candidate SNP. We examined associations with any trait at genome‐wide significance (P<5e‐8). Specifically, we searched for associations with known IHD risk factors (eg, systolic blood pressure, low‐density lipoprotein cholesterol, smoking, type 2 diabetes) that could represent alternative pathways to IHD independent of the protein of interest. SNPs showing such associations were flagged and manually reviewed. Because all of our instruments were selected as cis‐protein quantitative trait loci (cis‐pQTLs) (located within ±500 kb of the gene encoding each m6A‐associated protein), the 6 m6A‐associated proteins are encoded by distinct genes located on different chromosomes, thus the cis‐pQTLs for each protein map to separate genomic regions, which explains their cross‐chromosomal distribution (Supplemental Tables). No trans‐pQTLs or instruments affecting multiple proteins were included. A direct association with IHD at P<5e‐8 is likely to reflect vertical pleiotropy (ie, the effect is mediated through the protein) rather than horizontal pleiotropy. Therefore, we did not automatically exclude SNPs solely because they were associated with IHD. Only SNPs that were associated with a potential confounder at genome‐wide significance and where the confounder is unlikely to be mediated by the protein were removed. In the case of HNRNPC, no such SNPs were identified.

MR Statistical Methods

Using inverse‐variance weighted analysis, which offers a precise causal design, our MR primarily assessed the association between m6A‐associated proteins and IHD. To further infer potential causal association, we also used the weighted median approach and the MR‐Egger regression method. Furthermore, if the results of the inverse‐variance weighted method were significant (P<0.05), and the results of other methods were not significant, and there was no heterogeneity or pleiotropy, as long as the β values of MR Egger and weighted median were consistent with the direction of the β values of the inverse‐variance weighted analysis, the results were reliable. ³⁰ Next, we proceeded with the next step of sensitivity analysis using the Cochran Q to detect heterogeneity. Additionally, we also used the leave‐1‐out sensitivity test to check whether the results were caused by a single SNP. Finally, we also used the MR‐Egger intercept method to test the horizontal pleiotropy of IVs, detecting the pleiotropy effect based on whether the intercept is zero. If the intercept was close to zero, then there was no horizontal pleiotropy in this study. All statistical analyses were conducted using the 2‐sample MR (version 0.5.7) package in R (version 4.3.1).

Generation of the Myocardial Infarction Mouse Model

All mouse procedures were approved by the Peking University People's Hospital Animal Care and Use Committee. All mice were 6‐ to 8‐week‐old C57BL/6 male mice weighing ≈20±2 g. C57BL/6J male mice were purchased from the Huafukang Biotechnology (Beijing, China), and all mice were raised (up to 5 per cage) under the standard conditions of the Animal Center of Peking University People's Hospital with ad libitum access to food and water (temperature was 23 °C±1 °C and humidity was 55±5% with a 12/12 hour light/dark cycle). All male mice were acclimatized for 1 week in the quarantine room before the following experiments. The experimental unit was a single animal. A computer‐generated random number sequence was used to assign mice to the sham or myocardial infarction (MI) Wu group. The randomization list was prepared by an investigator not involved in surgery or outcome assessment. To minimize confounding, surgeries for both groups were performed in a randomized order on the same days. Mice in the sham group (n=6) underwent the same surgical procedure without left anterior descending (LAD) artery ligation. Experimental mice (n=6) underwent ligation of the LAD artery to induce IHD. Mice were anesthetized using intraperitoneal injection of 2,2,2‐tribromoethanol (Sigma, United States) at a concentration of 12.5 mg/mL, followed by intubation and ventilation with a small animal ventilator (ALCBIO). Make an incision between the third and fourth ribs, expose the mouse heart with a chest expander, and then LAD coronary artery with 8‐0 silk sutures, and perform sutures on the mouse skin with 4‐0 needle belt sutures. Adequate LAD artery was assessed by echocardiography analysis, and mice with unsuccessful surgery were removed from the experimental group. Anesthesia was used to euthanize the mice after 4 weeks, and heart tissue was taken for further research (Masson's trichrome staining, hematoxylin/eosin [HE] staining, Western blot). All procedures were performed with blinding to group allocation during outcome assessment. Specifically, the surgeon performing LAD artery ligation was aware of the intervention, but echocardiography acquisition and analysis, tissue processing, Masson's trichrome staining, HE staining, and Western blot quantification were conducted by investigators who were blinded to the experimental groups.

Echocardiography

Four weeks after LAD artery surgery, mice were anesthetized with 1.5% to 2% isoflurane, and the mouse heart rate was stabilized at 450 to 550 beats per minute, followed by measurement of mouse cardiac function using the M‐mode of a small animal ultrasound imaging system (Vevo 3100, Fujifilm VisualSonics). Echocardiographic parameters, including left ventricular ejection fraction and left ventricular fractional shortening, were assessed to confirm the successful induction of IHD‐like cardiac impairment after LAD artery ligation.

Masson's Trichrome Staining

Hearts were embedded and sectioned with 4% neutral buffered formalin and stained with Masson's trichrome staining according to the manufacturer's instructions and observed under a microscope (Olympus, Japan), followed by Image J to analyze the extent of collagen deposition.

HE Staining

Hearts were embedded and sectioned with 4% neutral buffered formalin and stained with HE according to the manufacturer's instructions and observed under a microscope (Olympus, Japan).

Cell Culture

H9C2 rat cardiomyocytes were obtained from the Shanghai Institute of Biochemistry and Cell Biology (China) and cultured in high‐glucose DMEM (high glucose; VivaCell, China) supplemented with 10% fetal bovine serum, 100 U/mL penicillin, and 100 μg/mL streptomycin at 37 °C in a humidified atmosphere containing 5% CO₂. We exposed adherent H9C2 cells to 600 μM H₂O₂ for 4 hours to induce cell damage and create in vitro models of IHD. ³¹

Dot Blot

Trizol reagent (Lablead, China) was used to extract the total RNA from H9C2 cells following a standard protocol. The concentration and purity of the extracted RNA were then measured using a NanoDrop 2000 spectrophotometer. After being spotted on N⁺ nylon membranes (FFN10; Beyotime, China), the RNA samples were cross‐linked with the membrane using ultraviolet light, stained with methylene blue, and then washed with double‐distilled H₂O. The m6A antibody (1:1000; Synaptic Systems) was then incubated overnight at 4 °C after blocking with 5% skim milk for 1 hour. After 1 hour of room temperature incubation with secondary antibodies, the membrane was analyzed using an enhanced chemiluminescence (ECL) chemiluminescence kit (Affinibody, China) in a chemiluminescent imaging system (Bio‐Rad, United States).

Western Blot

The total protein was extracted from H9C2, and protein concentration was determined by the bicinchoninic acid (BCA) kit (Lablead, China). Following the unification of the protein content across samples, the protein samples were denatured for 5 minutes at 100 °C using a metal bath and 5× loading buffer. After being separated using 10% SDS‐PAGE, protein samples (30 μg per group) were transferred to nitrocellulose membranes (Pall, United States). The membrane was then coincubated with the secondary antibody for 1 hour at room temperature after blocking with 5% skim milk for 1 hour and incubating with the HNRNPC antibody (1:1000; Proteintech, China) overnight at 4 °C. Membranes were visualized in a chemiluminescence imaging system (Bio‐Rad, United States) and quantified using Image Lab software.

Transfection Plasmid

H9C2 cells were seeded in DMEM supplemented with 10% fetal bovine serum in dishes. According to the manufacturer's instructions, we mixed 1 μg/mL of the HNRNPC plasmid or negative control plasmid with P3000 and Opti‐MEM (Gibco, United States) for 5 minutes. Then, we added it to the Opti‐MEM and Lipo 3000 mixture for another 5 minutes. Finally, we gently pipetted the mixtures and let them sit at room temperature for 20 minutes. The plasmid was cocultured with the cells for 48 hours before subsequent experiments. The HNRNPC‐shRNA‐1 sequence was GCTCTGTGCATAAGGGCTTTG. The HNRNPC‐shRNA‐2 sequence was GGATGGCAGAATGATTGCTGG. The HNRNPC‐shRNA‐3 sequence was GCGATTATTATGACAGGATGT. The short hairpin negative control (shNC) sequence was TTCTCCGAACGTGTCACGT.

Cell Viability Assay

A 100‐μL cell suspension was seeded in a 96‐well plate, transfected overnight for 48 hours, and then treated with H₂O₂ for 4 hours. Subsequently, 10 μL of Cell Counting Kit‑8 (CCK8) solution (Beyotime, China) was added to each well and incubated at 37 °C for 0.5 to 1 hour in the dark, and the absorbance was obtained at 450 nm.

Reactive Oxygen Species Assay

Intracellular reactive oxygen species (ROS) levels were measured using a dihydroethidium (DHE) fluorescent probe (Abmole, United States). Next, 2.5 μM DHE was added to each well and incubated at 37 °C for 30 minutes, followed by observation under a fluorescence microscope (Olympus, Japan).

JC‐1 Staining

5,5′,6,6′‐tetrachloro‐1,1′,3,3′‐tetraethylbenzimidazolylcarbocyanine iodide (JC‐1) (Beyotime, China) was used to determine changes in mitochondrial membrane potential. The JC‐1 staining solution was added to the treated cells and incubated in a cell culture incubator for 30 minutes. We added a prechilled JC‐1 staining buffer to the wells and washed twice. The JC‐1 monomers were captured first at the 475/35 nm channel and then the JC‐1 aggregates at the 530/43 nm channel using a fluorescence inverted microscope (Olympus, Japan). The degree of mitochondrial depolarization was determined by the ratio of red to green fluorescence intensity.

Mitochondrial Superoxide Measurement

MitoSOX Red (M36007; Thermo, United States) was used to probe intracellular mitochondrial ROS levels. After H₂O₂ administration, Hoechst 33342 solution was first added to each well and incubated for 10 minutes. After washing with PBS, 5 μM MitoSOX Red was cocultured with the cells at 37 °C for 30 minutes. The fluorescence was observed using a fluorescence inverted microscope (Olympus, Japan).

Determination of Related Indicators

The levels of malondialdehyde (MDA), oxidized nicotinamide adenine dinucleotide (NAD+), and reduced nicotinamide adenine dinucleotide (NADH) in H9C2 cells were quantified using a lipid peroxidation MDA assay kit (S0131; Beyotime, China) and a NAD+/NADH assay kit (S0175; Beyotime, China) according to the respective manufacturer's protocols. Absorbance was measured at 532 nm for MDA and 405 nm for NAD+/NADH.

Statistical Analysis

All studies were replicated at least 3 times, and all data were analyzed using GraphPad Prism 9.5 software and presented as mean±SEM. We used t tests between the 2 groups, whereas 1‐way ANOVA was followed by the Tukey test among the multiple groups. P<0.05 was considered statistically significant.

RESULTS

Basic Information of IVs

The connection between m6A‐associated plasma proteins and IHD was examined in this work using MR analysis. Figure 1A shows 3 hypotheses that must be true. As previously described, we obtained m6A‐associated plasma proteins from the INTERVAL study. After a series of quality control steps, the remaining 7 to 11 SNPs (P<5e‐6, r ²<0.001, kb=10 000) were associated with IHD (Tables S1 through S6). The mean F statistics for the instruments corresponding to each protein were as follows: ALKBH5 (18.37), ELAVL1 (52.71), HNRNPA2B1 (105.52), HNRNPC (21.90), LRPPRC (124.28), and YTHDC1 (24.24). All individual SNPs had F>10, and the mean F statistics were well above the conventional threshold, confirming adequate instrument strength and minimal risk of weak‐instrument bias.

MR Estimates the Relationship Between m6A‐Associated Plasma Proteins and IHD

The results of the conventional MR analysis are presented in Figure 1B. The inverse variance weighted results showed a correlation between HNRNPC protein levels and a lower risk of IHD (odds ratio [OR], 0.93 [95% CI, 0.88–0.97]; P=0.002), whereas weighted median (OR, 0.92 [95% CI, 0.86–0.98]; P=0.013) also showed a significant correlation. However, the estimated values of MR‐Egger (OR, 0.92 [95% CI, 0.86–1]; P=0.077) showed no significant correlation. No causal association between IHD and other proteins (ALKBH5, ELAVL1, HNRNPA2B1, LRPPRC, YTHDC1) was found in this study.

Sensitivity Analyses Result

To assess the robustness of MR results, we used the Cochran Q to detect heterogeneity and the MR‐Egger intercept method to test the horizontal pleiotropy of IVs, with detailed outcomes presented in the Table. The Cochrane Q test revealed no heterogeneity in HNRNPC (Q=7.378, P _Q=0.598) or horizontal pleiotropy (MR‐Egger regression intercept=0.0009; P=0.920). Our analysis demonstrated significant pleiotropic effects in LRPPRC (MR‐Egger regression intercept=0.028; P=0.048) but no heterogeneity (Q=11.942, P _Q=0.0633). However, YTHDC1 exhibited significant heterogeneity (Q=11.942, P _Q=0.0633) and no pleiotropy (MR‐Egger regression intercept=0.028; P=0.019). Notably, neither protein showed causal associations with IHD, which did not affect the overall results. Furthermore, after excluding 1 SNP from HNRNPA2B1, ELAVL1, and LRPPRC in the analysis of the leave‐1‐out method, the risk estimation did not show significant changes, proving that no specific SNPs were crucial for causal associations. The findings of the sensitivity analysis are shown in Figures 2, 3, and 4.

Table 1.

Sensitivity Analysis for the Association Between m6A‐Associated Proteins and IHD

Exposure	Pleiotropy test		Heterogeneity test
	Intercept	P Intercept	Q value (IVW)	P Q (IVW)	Q value (MR‐ER)	P Q (MR‐ER)
ALKBH5	−0.008	0.683	6.45	0.597	6.269	0.509
ELAVL1	0.001	0.942	10.34	0.242	10.331	0.171
HNRNPA2B1	−0.010	0.424	4.535	0.605	3.778	0.582
HNRNPC	0.0009	0.920	7.378	0.598	7.368	0.498
LRPPRC	0.028	0.048	11.942	0.063	5.069	0.407
YTHDC1	0.109	0.596	21.315	0.019	20.624	0.014

ALKBH5 indicates ALKB homologue 5 protein; ELAVL1, ELAV‐like RNA binding protein 1; HNRNPA2B1, heterogeneous nuclear ribonucleoprotein A2/B1; HNRNPC, heterogeneous nuclear ribonucleoprotein C; IVW, inverse variance weighted; LRPPRC, leucine rich pentatricopeptide repeat containing; MR‐ER, Mendelian randomization‐Egger regression; P _Intercept, P value corresponding to MR‐ER intercept test; Q value, statistics of Cochrane Q test; and YTHDC1, YT521‐B homology‐domain‐containing protein 1.

Figure 2. Scatter plots of the effects of m6A‐related proteins on IHD were analyzed by MR.

A, ALKBH5. B, ELAVL1. C, HNRNPA2B1. D, HNRNPC. E, LRPPRC. F, YTHDC1. ALKBH5 indicates ALKB homologue 5 protein; ELAVL1, ELAV‐like RNA binding protein 1; HNRNPA2B1, heterogeneous nuclear ribonucleoprotein A2/B1; HNRNPC, heterogeneous nuclear ribonucleoprotein C; IHD, ischemic heart disease; LRPPRC, leucine rich pentatricopeptide repeat containing; m6A, N6‐methyladenosine; MR, Mendelian randomization; SNP, single‐nucleotide polymorphism; and YTHDC1, YT521‐B homology‐domain‐containing protein 1.

Figure 3. MR analysis of funnel plots of the effects of m6A‐related proteins on IHD.

A, ALKBH5. B, ELAVL1. C, HNRNPA2B1. D, HNRNPC. E, LRPPRC. F, YTHDC1. ALKBH5 indicates ALKB homologue 5 protein; ELAVL1, ELAV‐like RNA binding protein 1; HNRNPA2B1, heterogeneous nuclear ribonucleoprotein A2/B1; HNRNPC, heterogeneous nuclear ribonucleoprotein C; IHD, ischemic heart disease; LRPPRC, leucine rich pentatricopeptide repeat containing; m6A, N6‐methyladenosine; MR, Mendelian randomization; and YTHDC1, YT521‐B homology‐domain‐containing protein 1.

Figure 4. Leave‐1‐out analysis for m6A‐related proteins on IHD.

A, ALKBH5. B, ELAVL1. C, HNRNPA2B1. D, HNRNPC. E, LRPPRC. F, YTHDC1. ALKBH5 indicates ALKB homologue 5 protein; ELAVL1, ELAV‐like RNA binding protein 1; HNRNPA2B1, heterogeneous nuclear ribonucleoprotein A2/B1; HNRNPC, heterogeneous nuclear ribonucleoprotein C; IHD, ischemic heart disease; LRPPRC, leucine rich pentatricopeptide repeat containing; m6A, N6‐methyladenosine; MR, Mendelian randomization; and YTHDC1, YT521‐B homology‐domain‐containing protein 1.

HNRNPC Correlates With IHD

Through MR, we discovered a potential causal association between HNRNPC and IHD in m6A‐associated proteins. To confirm the findings of MR, we performed a range of in vitro and in vivo experiments. The effectiveness of the IHD model was evaluated in vivo. Cardiac function was assessed via echocardiography, whereas pathological changes were examined using Masson's trichrome staining and HE staining. According to the results of HE staining, the mice in the control group had neatly arranged cardiomyocytes, uniform cytoplasm, and a normal interstitium. The MI group of mice had disordered cell arrangement, a large number of inflammatory cells that might be infiltrated, and some missing cardiomyocyte nuclei and vacuole‐like alterations. In addition, Masson staining revealed that mice in the MI group had more collagen than those in the control group (Figure 5A and 5B). Echocardiographic assessment showed that ejection fraction (%) and fractional shortening (%) decreased by ≈10% to 20% in the MI group (Figure 5C and 5D), verifying successful induction of myocardial ischemia. To exclude the factors of the growth cycle of mice, the tibia length was standardized to obtain heart weight/tibia length (HW/TL), a cardiac function indicator, and the results showed that the ratio of heart weight to tibia length caused by myocardial infarction was up‐adjusted (Figure 5E). To investigate whether IHD is correlated with HNRNPC, Western blot analysis showed significant upregulation in both in vivo and in vitro models of IHD, and dot blot results showed that H₂O₂ (600 μM) on H9C2 cells for 4 hours reduced m6A levels (Figure 5F through 5H). The results mentioned above suggest a strong correlation between HNRNPC and the onset and progression of IHD (n=6 per group for all measurements in this figure).

Figure 5. Relationship between HNRNPC and IHD.

A, Representative HE and Masson's trichrome staining of mouse myocardial tissue. Red: muscle fibers; blue: collagen fibers. B, Statistical plot of collagen fiber area in Masson staining (n=6). C, Cardiac ultrasound M mode was used to detect changes in cardiac function. D, Echocardiographic measurements showing statistical plots of EF, FS, ΔEF, and ΔFS between groups (n=6). E, HW/TL ratio (n=6). F, Western blot detection of HNRNPC expression in MI mice (n=6). G, Increased protein expression levels of HNRNPC in H₂O₂‐treated H9C2 cells (n=6). H, Dot blot to detect the expression of each group of m6A in H₂O₂‐treated H9C2 cells. MB staining was used as a top control (n=6). The results are expressed as mean±SEM. **P<0.01, ***P<0.001. ΔEF, change in ejection fraction; ΔFS, change in fractional shortening; EF indicates ejection fraction; FS, fractional shortening; HE, hematoxylin/eosin; HNRNPC, heterogeneous nuclear ribonucleoprotein C; HW/TL, heart weight/tibia length; IHD, ischemic heart disease; m6A, N6‐methyladenosine; MI, myocardial infarction and MB, methylene blue.

HNRNPC Knockdown Inhibits Cell Death, Oxidative Stress, and Mitochondrial Dysfunction

We investigated HNRNPC knockdown in cardiomyocytes H9C2 to confirm the biological function of HNRNPC in IHD. The results of the Western blot analysis used to evaluate the transfection efficiency demonstrated that short hairpin RNA targeting HNRNPC (sh‐HNRNPC‐2) was successfully knocked down (Figure 6A). To investigate the effect of HNRNPC on cardiomyocytes, we knocked down HNRNPC in H9C2 cells. Subsequently, we used CCK8 to study the viability of the cells, and the results showed that knockdown of HNRNPC significantly increased the viability of cardiomyocytes (Figure 6B). Meanwhile, dot blot results showed that knockdown of HNRNPC increased the expression of m6A (Figure 6C). Oxidative stress is closely related to the development of IHD, ³² so we detected the expression levels of MDA and ROS in cardiomyocytes in HNRNPC by MDA detection kit and DHE fluorescent probe, and the results showed that knockdown of HNRNPC reduced the levels of MDA and ROS in cardiomyocytes (Figure 6D through 6F). In the meantime, we examined how HNRNPC affected mitochondrial function. Using the MitoSOX fluorescent probe, NAD+/NADH kit, and JC‐1 fluorescent probe, we were able to identify changes in mitochondrial ROS, NAD+, and NADH levels as well as changes in mitochondrial membrane potential. The results revealed that after using H₂O₂, mitochondrial ROS were much higher than in the control group. At the same time, the amount of NAD+ and NADH, as well as the level of mitochondrial membrane potential, decreased, which could be reversed by knocking down HNRNPC (Figure 6G through 6L). The above results indicate that HNRNPC knockdown inhibits cell death, oxidative stress, and mitochondrial dysfunction, further confirming that upregulation of HNRNPC induces IHD.

Figure 6. Effect of HNRNPC knockdown on H₂O₂‐treated H9C2 cells.

A, Western blot confirming HNRNPC knockdown efficiency (n=6). B, Cell viability assessed by CCK8 assay after HNRNPC knockdown and H₂O₂ treatment (n=8). C, Dot blot to determine the effect of HNRNPC knockdown on m6A content (n=6). D, Changes in MDA content in cardiomyocytes H9C2 (n=6). E and F, DHE fluorescence staining and fluorescence intensity in H9C2 cells. Scale bar=100μm (n=6). G and H, MitoSOX fluorescence staining and fluorescence intensity in H9C2 cells. Scale bar=100μm (n=6). I, Changes in NADH content in cardiomyocytes H9C2 (n=5). J, Changes in NAD+ content in cardiomyocytes H9C2 (n=6). K and L, Representative image of H9C2 cells stained with JC‐1. The red‐green fluorescence ratio reflects the change in mitochondrial membrane potential. Scale bar=100μm (n=6). The results are expressed as mean±SEM. *P<0.05, **P<0.01, ***P<0.001. HNRNPC indicates heterogeneous nuclear ribonucleoprotein C; and ns, no statistical significance. CCK8, Cell Counting Kit‐8; MDA, malondialdehyde; DHE, dihydroethidium; NADH, reduced nicotinamide adenine dinucleotide; NAD+, nicotinamide adenine dinucleotide; JC‐1, 5,5′,6,6′‐tetrachloro‐1,1′,3,3′‐tetraethylbenzimidazolylcarbocyanine iodide; sh, short hairpin RNA; NC, negative control.

DISCUSSION

In this study, we suggested for the first time the potential causal association between the levels of m6A‐related proteins and IHD through 2‐sample MR analysis, further demonstrating that HNRNPC is closely related to the development of IHD. To ensure that the results of MR were reliable, we found no heterogeneity or pleiotropy in this study using a sensitivity analysis assessment. In addition, to further validate that MR can indicate the relationship between HNRNPC and IHD, we conducted a series of in vitro and in vivo experiments to provide a theoretical basis for providing clinical treatment for IHD.

Cardiomyocytes experience starvation and a lack of energy resources in the early stages of IHD, which cause them to undergo autophagy and activate cellular endogenous defensive mechanisms. ³³ Autophagy is overactive during the myocardial ischemia–reperfusion injury phase, which raises ROS generation and mitochondrial permeability and ultimately results in cardiac cell death. ³⁴ It may counteract the effects of METTL3 on cardiomyocytes treated with hypoxia/reoxygenation when overexpressed as ALKBH5 in m6A, and this study further found that transcription factor EB (TFEB) induces ALKBH5 and suppresses METTL3, suggesting that m6A methylation may regulate the ischemic heart through the autophagy–lysosomal pathway, ³⁵ which presents an interesting contrast to our findings. Although some studies demonstrated that HNRNPC increased in hepatocellular carcinoma, cervical cancer, and papillary thyroid carcinoma, ³⁶ , ³⁷ , ³⁸ MR data suggested that HNRNPC protein levels were related to a decreased risk of IHD. In the meantime, we discovered through animal and cell experiments that HNRNPC expression increased at the onset of IHD. An important nuance arises from the apparent discrepancy between our MR and experimental findings. First, the MR estimate reflects the lifelong effect of genetically determined differences in circulating HNRNPC on disease susceptibility, which may be distinct from the acute, dynamic changes in cellular HNRNPC expression in response to stress. The observed upregulation in diseased tissue could represent a compensatory cellular stress response that, although associated with the pathological state, may initially be adaptive; its sustained or dysregulated increase, however, might transition to contribute to maladaptive processes such as the oxidative stress and mitochondrial dysfunction we observed. Second, although our sensitivity analyses did not detect significant pleiotropy, we cannot entirely rule out that the genetic variants influencing plasma HNRNPC levels affect IHD risk through other closely linked biological pathways (horizontal pleiotropy). This complex interplay highlights the challenge of integrating population‐level genetic epidemiology with mechanistic biology. At the same time, the small number of plasma proteins in the databases we included may also have contributed to bias in the results, which is why both of the results mentioned above differed from the actual results. Although there was a slight bias in our results, we also succeeded in validating the usefulness of MR to infer a potential causal association between exposure factors and disease.

M6A modifications are widely involved in a variety of biological processes through the regulation of RNA metabolism and have been associated with a variety of disease states, ³⁹ but there is no study reporting the relationship between HNRNPC and IHD, so exploring how HNRNPC induces the development of IHD is a topic that remains important today. Our experimental data clearly demonstrate that HNRNPC knockdown mitigates these key pathological features. It reduces ROS, MDA, and mitochondrial damage while improving cell survival. This positions HNRNPC as a critical regulator in the stress‐response network of cardiomyocytes. From a translational perspective, our findings highlight HNRNPC as a novel potential therapeutic target for IHD. Pharmacological or genetic strategies targeting HNRNPC expression or function could offer a rational approach. Such strategies may attenuate oxidative stress and preserve mitochondrial integrity in the ischemic heart, thereby protecting against cardiomyocyte death. The precise upstream and downstream molecular cascades require further elucidation. This includes potential links to specific death pathways such as ferroptosis. Nevertheless, our study provides a strong foundation for future investigations. Targeting m6A readers like HNRNPC may open new avenues for cardioprotective drug development.

Our study has the following advantages. The present study is the first to use MR to investigate the potential causal link between HNRNPC and IHD. MR has previously been used to analyze the genetic susceptibility of indoleamine 2,3‐dioxygenase and IHD, but no MR studies have examined the causal relationship between m6A‐related proteins and IHD. Moreover, compared with observational studies, MR analysis can overcome confounding factors and reverse causality in epidemiologic studies. At the same time, this study successfully validated the results of MR through in vitro and in vivo experiments. Finally, we used the leave‐1‐out method as a sensitivity analysis, which increased the robustness of the conclusions.

There are several limitations to this study. First, MR data were all derived from genome‐wide association studies and are not validated by sufficient clinical data. Second, because all of our data came from European populations, we were unable to demonstrate whether the same is true for populations of other racial or ethnic groups. Third, although we used a variety of methods to examine multiplicity, it was not possible to entirely rule it out. Fortunately, several analytical methods yielded consistent results, and no evidence of horizontal pleiotropy or heterogeneity was discovered, supporting the results of this study. Fourth, the genome‐wide association study of plasma proteins only included 3301 individuals, because plasma protein measurement techniques are expensive. ⁴⁰ Thus, the database was unable to identify all major genetic variations across the genome. Based on this rationale, and in accordance with analogous pQTL‐MR studies, we defined the instrumental variable selection threshold as P<5e‐6, ²⁸ , ²⁹ , ⁴¹ as opposed to the conventional genome‐wide significance level (P<5e‐8), to ensure the procurement of an adequate and valid set of genetic instruments under conditions of limited sample size. Application of this threshold resulted in 7 to 11 independent SNPs per protein, each exhibiting an F statistic >10, thereby confirming strong instruments and a minimal risk of weak‐instrument bias. Fifth, the primary pQTL data set (INTERVAL) is relatively modest in size (n=3301), which limited the power to detect weak genetic instruments and precluded independent replication. To mitigate this concern, we selected instruments at a relaxed threshold (P<5e‐6) and confirmed all F statistics >10. Nonetheless, we acknowledge that replication in larger, independent pQTL cohorts (eg, UK Biobank Pharma Proteomics Project) would substantially strengthen the generalizability of our findings. Future studies should perform such replication analyses when feasible. Sixth, although we increased replicate numbers to n=6 for key end points, the sample sizes in our experimental animal study (n=6 per group) remain modest. Seventh, although our instrumental variables are predominantly cis‐pQTLs and exhibit minimal cross‐protein linkage disequilibrium, we cannot formally exclude the possibility that genetic instruments for different m6A readers influence IHD through correlated pathways. Multivariable MR could theoretically disentangle such pleiotropic effects, but its application requires a sufficient number of independent, strong instruments for each exposure. Given the modest sample size of our pQTL discovery panel and the resulting limited SNP counts (7–11 per protein), we were unable to perform a robust multivariable MR analysis. Future studies with larger pQTL data sets should consider multivariable MR to further dissect the specific contribution of HNRNPC relative to other m6A regulators. Although this is common in exploratory mechanistic studies, it limits the statistical power and generalizability of our conclusions. Future studies with larger sample sizes are needed to confirm these findings.

The absence of statistical evidence for horizontal pleiotropy is further supported by biological considerations. HNRNPC is an RNA‐binding protein that primarily functions in nuclear pre‐mRNA processing, with no established direct role in canonical cardiovascular pathways such as lipid metabolism, blood pressure regulation, or inflammation. All genetic instruments used for HNRNPC are located in cis (within 500 kb of the HNRNPC gene) and are not known to associate with other circulating proteins in large pQTL cross‐trait analyses. Moreover, our experimental data demonstrate that HNRNPC knockdown attenuates oxidative stress and mitochondrial dysfunction in cardiomyocytes, a protective phenotype consistent with the direction of the MR estimate. This convergence of genetic and mechanistic evidence further supports that the observed association is mediated specifically through HNRNPC.

CONCLUSIONS

In summary, the present study is the first to explore the causal relationship among 6 m6A‐associated proteins (including ALKBH5, ELAVL1, HNRNPA2B1, HNRNPC, LRPPRC, and YTHDC1) and IHD by MR, and the findings support a potential causal link between HNRNPC and IHD within the m6A regulatory network. Additionally, the experimental data indicated that HNRNPC is significantly increased during the development of IHD and that knocking down HNRNPC in the IHD cellular model inhibits oxidative stress, mitochondrial dysfunction, and cell death. These findings provide a crucial foundation for the prevention and treatment of IHD in the future.

Sources of Funding

This work was supported by National Natural Science Foundation of China (82104156) and Peking University People's Hospital Scientific Research Development Funds (RDJP2023‐32). The open data are supported by the Medical Research Council (MRC) Integrative Epidemiology Unit Genome‐Wide Association Study and the INTERVAL trial to determine whether intervals between blood donations can be safely and acceptably decreased to optimize blood supply.

Disclosures

None.

Supporting information

Tables S1–S6

ARRIVE Checklist

Unedited Gels