Multiomic and Electrophysiologic Analyses Reveal that an Inherited MRC2 Variant Causes Fibroblast Dysfunction and Increased Atrial Fibrillation Susceptibility

Abstract

A recent study identified a rare variant in the MRC2 gene in individuals with familial reentrant supraventricular tachycardia, a Wolff-Parkinson-White (WPW) ECG pattern, and structurally normal hearts. WPW syndrome is associated with atrial fibrillation (AF), and MRC2 was recently proposed as a protective gene for AF. We determined whether the E990G-heterozygous (het) loss-of-function variant in Mrc2 increases AF susceptibility and identified aberrant cellular mechanisms resulting from Mrc2 deficiency in atrial cardiofibroblasts (ACFs) and atrial tissue in mice that may promote AF. Programmed electrical stimulation (PES) was performed to determine AF susceptibility in Mrc2 E990G-het mice and wild-type (WT) controls. ACFs were isolated from these mice and cultured, and transcriptomic profiling by RNA sequencing and secretomic/proteomic profiling by mass spectrometry were performed on ACFs and whole atrial tissue. E990G-het mice exhibited increased susceptibility to pacing-induced AF and had decreased atrioventricular effective refractory periods compared with WT controls. Transcriptomic, secretomic, and proteomic profiling of cultured ACFs and whole-atrial tissue revealed differential expression of several fibrotic regulators in E990G-het versus WT mice, including decreased ACF expression of matrix metalloproteinase 13 (MMP-13), which degrades collagen types I, II, and III; decreased ACF expression and secretion of matrix metalloproteinase 12 (MMP-12), which degrades collagen types I, III, IV, elastin, and fibronectin; and increased tissue levels of cellular communication network factor 2/connective tissue growth factor (CCN2/CTGF), a profibrotic regulator. In conclusion, Mrc2 E990G-het mice exhibit increased AF susceptibility and differentially regulated fibrotic genes and proteins.

New & Noteworthy

Our study reveals a rare MRC2 gene variant (E990G) linked to familial supraventricular tachycardia and Wolff-Parkinson-White syndrome increases atrial fibrillation (AF) susceptibility in mice. The E990G-heterozygous variant disrupts atrial cardiofibroblast function, reducing protective matrix metalloproteinases (MMP-12, MMP-13) and elevating profibrotic CCN2/CTGF levels, as shown through transcriptomic and proteomic profiling. This suggests MRC2 deficiency promotes AF by altering fibrotic regulation in atrial tissue.

Introduction

Increased levels of atrial fibrosis are associated with increased risk of atrial fibrillation (AF) onset(1)(1)(1)(1), AF severity(2), and AF recurrence following ablation surgery(3). During structural remodeling, activated atrial cardiofibroblasts (ACFs) produce excessive atrial fibrosis by depositing collagen I, collagen III, and fibronectin in the extracellular matrix (ECM)(4). Other cellular processes that contribute to AF progression include inflammation, oxidative and metabolic stress, and electrical remodeling(5, 6). AF risk is significantly elevated in patients with Wolff-Parkinson-White (WPW) syndrome(7), a related supraventricular tachycardia that is associated with alterations in the annulus fibrosus(8). Recently, the heterozygous (het) loss-of-function variant c.2969A>G;p.Glu990Gly (E990G) in mannose receptor C type 2 (MRC2) was associated with WPW syndrome in a multigenerational family(9).

In addition to increased rates of collagen synthesis, decreased removal of deposited collagen can also contribute to fibrosis by increasing the net accumulation of ECM collagen(10, 11). MRC2, also known as urokinase plasminogen activator receptor-associated protein (uPARAP), Endo180, C-type lectin E5 (CLECE5), or CD280, is a regulator of collagen turnover encoded by the MRC2 gene and is localized to the cell surface of mesenchymal cells, macrophages, and endothelial cells(12–14). It facilitates cell adhesion to fibrillar collagens and the internalization and degradation of ECM collagen by clathrin-mediated endocytocis(15, 16). In addition, MRC2 reduces levels of extracellular thrombospondin-1, an activator of profibrotic TGF-β signaling, by endocytosis(17, 18). Cultured mouse dermal fibroblasts with targeted deletion of Mrc2 show decreased migration on a fibrillar collagen matrix(15) while human ACFs with Mrc2 knockdown migrate faster(9). This MRC2-dependent dysregulation of fibroblast migration may be explained by MRC2’s contrasting roles in promoting cell chemotaxis toward urokinase-type plasminogen activator (uPA) or random cell migration(19, 20). MRC2 protects against renal and liver fibrosis, and MRC2 RNA and protein expression are upregulated in TGF-β activated cardiac fibroblasts(21–23). MRC2 inhibits human cardiac fibroblast proliferation, and decreased epicardial MRC2 was associated with postoperative AF(24). Similarly, Mrc2 was decreased in an exercise-stressed mouse model of AF(25). Given MRC2’s numerous roles in ECM remodeling and the association between atrial fibrosis and AF, MRC2 may protect against AF by inhibiting the inappropriate accumulation of excessive atrial fibrotic scar tissue via mechanisms such as increased ECM collagen internalization and degradation by ACFs, decreased TGF-β activation by thrombospondin-1 and downstream profibrotic signaling, or regulation of ACF migration patterns. However, the actual mechanisms by which MRC2 regulates atrial structural remodeling are yet to be established.

As a model for the MRC2 E990G variant in humans, mice with knock-in of the orthologous Mrc2 c.2966A>G;p.Asp989Gly (Supplemental Table 1) variant (referred to here as “Mrc2 E990G-het”) had arrhythmogenic accessory pathways and more loosely packed ECM collagen in the cardiac annuli fibrosis while not having significantly altered total atrial MRC2 protein expression levels(9). However, the effects of the E990G-het variant on ACF function, which may contribute to increased susceptibility to AF, have not been determined. Here, we integrated functional and multi-omic data in mice, whole atrial tissue, and isolated ACFs to demonstrate that Mrc2 E990G-het mice are susceptible to pacing-induced AF and dysregulation of atrial fibrosis and cell signaling.

Materials and methods

Study animals.

All studies were performed according to protocols approved by the Baylor College of Medicine Institutional Animal Care and Use Committee, conforming to the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (Publication no. 85-23, revised 1996). The Mrc2 E990G-het mice were generated by knocking in the human nonsynonymous variant c.2969A>G via CRISPR-Cas9 embryo microinjections. All animals included in this study had a C57BL/6J background supplied by Jackson Laboratories (stock number 000664) and were backcrossed for at least 8 generations prior to being used in experiments. 23 Mrc2 E990G-het mice (10 male and 13 female) and 22 wild-type (WT) controls (10 male and 12 female) from 14 litters were used in total with ages ranging from 3 to 10 months old. 10 of the E990G-het mice and 9 of the WT controls were matched by at least one littermate with the other genotype. Mice were randomly selected for each of the procedures described below unless otherwise noted. Sample sizes for each study, including overlaps, are as follows: 15 WT (5 male and 10 female) and 15 E990G-het (4 male and 11 female) for programmed electrical stimulation (PES), 4 WT (3 male and 1 female) and 5 E990G-het (3 male and 2 female) for ACF transcriptomics, 4 WT (3 male and 1 female) and 4 E990G-het (2 male and 2 female) for ACF secretomics, 8 WT mice with no induced AF (2 male, 6 female) and 7 E990G-het mice with induced AF except for one non-responder (7 female) for whole-atrial transcriptomics, and 4 WT (2 male, 2 female) and 4 E990G-het (2 male, 2 female) mice that did not undergo PES for whole-atrial proteomics. All animal studies were performed in a blinded manner when possible. Briefly, after PES of the mouse hearts to determine induced AF incidence, the atria were harvested. Some of the atria pairs were enzymatically dissociated to isolate and culture ACFs for transcriptomic and secretomic profiling, while other atria pairs were immediately snap-frozen and processed for whole-atrial tissue transcriptomic or proteomic profiling.

Programmed electrical stimulation (PES).

Electrophysiology studies were performed as previously described(26). Briefly, mice were anesthetized with 2% v/v isoflurane/oxygen, and a 1.1F octopolar catheter (EPR-800, Millar, Houston, TX) was advanced through the right external jugular vein into the right atrium and ventricle. The octopolar catheter leads were connected to an external stimulator (STG3008-FA, MultiChannel Systems, Reutlingen, Germany) in recording mode, and the signals along with surface electrocardiogram (ECG) leads were acquired using IOX2.4 acquisition software (EMKA Technologies, Sterling, VA). Proper catheter positioning was verified by the waveforms of the four intracardiac ECGs and by looking for the proper deflections in the P wave and QRS complex and atrial and ventricular pacing, respectively. All PES protocols were performed at 1.75% v/v isoflurane/oxygen and a rectal temperature of 37.0 ± 0.5 °C. Following baseline recordings, the sinus node recovery time (SNRT) was calculated as the time to first spontaneous sinus beat after right atrial pacing at a basic cycle length (BCL) of 100 milliseconds (ms) for 15 s. The atrioventricular effective refractory period (AVERP) was determined by applying a series of right atrial pacing trains (i.e., S1) at a BCL of 100 ms, after which a premature stimulus (i.e., S2) was applied. The S1-S2 interval was decreased by 2 ms from 70 ms to 20 ms, and the AVERP was defined as the shortest S1-S2 interval at which S2 successfully depolarized the ventricles. AF inducibility was assessed by performing a series of 2s bursts starting at a BCL of 40 ms. The BCL decreased by 2 ms for each 2s burst, starting at a BCL of 40 ms and ending at 20 ms. Burst pacing was performed in triplicate for each mouse, and AF was defined as the presence of an irregularly irregular rhythm without discernible P waves for at least 1 second on at least two out of three atrial burst pacing protocols.

Atrial tissue dissociation.

Using aseptic technique in a biosafety cabinet, each excised heart was placed on a separate petri dish in Dulbecco’s phosphate-buffered saline (DPBS) with 0.133 g/L (0.77 mM), CaCl₂ • 2H₂O and 0.1 g/L (0.49 mM) MgCl₂ • 6H₂O (D8662, Sigma-Aldrich, Inc., St. Louis, MO) with 2.5X antibiotic-antimycotic solution (250 units penicillin, 250 μg streptomycin, and 0.625 μg amphotericin B per mL) (A5955, Sigma-Aldrich, Inc., St. Louis, MO), 10 μg/mL Plasmocin^® Prophylactic (ant-mpp, InvivoGen, San Diego, CA), and 100 μM L-Ascorbic acid 2-phosphate (ascorbate) sesquimagnesium salt hydrate (A8960, Sigma-Aldrich, Inc., St. Louis, MO)(27). The tissue surrounding each atrium was trimmed, and the atria were excised and placed in dissociation solution containing 1 mg (280 U) per mL collagenase, type 4 (LS004188, Worthington Biochemical Corp., Lakewood, NJ), and 0.7 mg (1 U) per mL dispase II (D4693, Sigma-Aldrich, Inc., St. Louis, MO) in DPBS with Ca²⁺ and Mg²⁺, 25% 1:1 Dulbecco’s Modified Eagle’s Medium/Ham’s F-12 (DMEM/F12) nutrient mixture (D8062, Sigma-Aldrich, Inc., St. Louis, MO), 1X antibiotic-antimycotic solution (100 units penicillin, 100 μg streptomycin, and 0.25 μg amphotericin B per mL), 5 μg/mL Plasmocin^® Prophylactic, and 100 μM ascorbate(28). Each pair of atria from each mouse heart was placed in 1 mL dissociation solution in separate 2-mL microcentrifuge tubes. All non-sterilized reagents used for tissue dissociation, ACF isolation, and cell cultures were filter-sterilized in the biosafety cabinet before use, and all tissues and ACFs were handled in the biosafety cabinet using aseptic technique.

ACF isolation and culturing.

The atria were incubated at 37°C with gentle agitation for 5 rounds of dissociation, with 25 minutes for the first round and the duration incrementally decreasing by 5 minutes for each subsequent round. After each round of dissociation, the tissues were gently triturated with a 1000 μL micropipette (the pipette tip was cut to enlarge the opening for the first two rounds). The tubes then were held vertically to allow the undigested tissue to settle, and the supernatant in each tube was passed through a 40 μm cell strainer into a 50 mL centrifuge tube on ice (separate pipette tips, strainers, and centrifuge tubes for each ACF line). 1 mL of ice-cold quenching buffer [18% DMEM/F12, 2% fetal bovine serum (FBS) (35-015-CF, Corning, Tewksbury, MA), 1X antibiotic-antimycotic solution, 5 μg/mL Plasmocin Prophylactic, and 100 μM ascorbate in DPBS] was then passed through each cell strainer. 1 mL of fresh dissociation solution was added to each tube, and the procedure was repeated for each dissociation round. After the second round only, the atria were minced into several smaller pieces using scissors. After the fifth round, the 50 mL tubes were centrifuged at 1000 rpm (193 g) for 5 min at 4°C, and the supernatants were discarded. 10 mL of DPBS was then passed through each cell strainer into each 50 mL tube to rinse the strainers and wash the pelleted ACFs, and the centrifugation step was repeated. The supernatants were then discarded, and the ACFs were resuspended in growth medium (DMEM/F12 with 10% FBS, 1X antibiotic-antimycotic solution, 5 μg/mL Plasmocin Prophylactic, and 100 μM ascorbate)(29). 80% of each ACF line was seeded in one tissue culture-treated 6-well plate well. The ACFs were incubated at 37°C for 2 hours, and the media were then aspirated from the adherent ACFs. The adherent ACFs were washed three times with DPBS and cultured at 37°C in growth media(30).

RNA and protein sample processing.

Once the ACFs in the 6-well plates reached ~50% confluency (2–4 days), they were adapted to serum-free media by replacing half of the conditioned media with serum-free media. The following day, all conditioned media were replaced with serum-free media. The serum-free conditioned media were collected the following day, and the ACFs were lysed with TRIzol reagent (15596018, Invitrogen^™, Waltham, MA), and total RNA and cellular protein were isolated using the Direct-zol RNA Miniprep Kit (R2050, Zymo Research, Irvine, CA) according to the manufacturer’s protocol. Total RNA sample quantity, purity, and integrity were determined by spectrophotometric absorbance at 260 nm, the 260/280 nm absorbance ratio, and gel electrophoresis, respectively.

8 frozen atrium pairs from WT mice with no induced AF response and 7 age-matched atrium pairs from E990G-het mice with induced AF (except for one non-responder) were selected for transcriptomic profiling (Supplemental Table 2). The atria were lysed with 250 μL TRI reagent (T9424, Sigma-Aldrich, Inc., St. Louis, MO) per atrial pair. The lysates were partially homogenized by brief mechanical homogenization at 4°C and incubated for 5 minutes at room temperature. Total RNA was then isolated from the supernatant using the Direct-zol RNA Miniprep Kit as described above. 200 μL fresh TRI reagent was added to each leftover unhomogenized tissue sample, and the samples were completely homogenized by additional mechanical homogenization at 4°C. Total RNA was extracted from these homogenates upon the addition of chloroform according to the manufacturer’s protocol. The Direct-zol kit-extracted and Tri Reagent/chloroform-extracted RNA fractions from each sample were combined for RNA-seq and RT-qPCR. RNA sample quantity, purity, and integrity were determined as described above.

Transcriptome profiling.

The total-RNA samples (Supplemental Table 3) were shipped to GENEWIZ (Azenta, South Plainfield, NJ) on dry ice for standard RNA-seq at a targeted depth of 30 million reads per sample. The yield for all ACF samples and 9 of the 10 whole-atrium samples exceeded 30 million reads per sample; one of the atrium samples yielded 24.3 million reads. At Azenta, RNA sample quantity was confirmed using the Qubit 2.0 Fluorometer (Life Technologies, Carlsbad, CA), and RNA integrity was confirmed using the Agilent TapeStation 4200 (Agilent Technologies, Palo Alto, CA). Library preparation with stranded polyA selection was then performed. The libraries were sequenced using a 2×150 bp Paired End (PE) configuration on the Illumina^® NovaSeq^™ or HiSeq^® platform. Raw sequence data (.bcl files) generated from the Illumina platform were converted into .fastq files, de-multiplexed using Illumina’s bcl2fastq 2.17 software, and delivered by File Transfer Protocol (FTP).

RNA-seq read quality from the ACF and whole-atrium total RNA samples was verified with FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/, Simon Andrews, Babraham Institute, Babraham, Cambridgeshire, UK) and MultiQC(31). All samples had mean quality scores > 35, and > 90% of reads had quality scores > 30. Quality and adapter trimming of the read ends was performed using Trim Galore (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/, Felix Krueger, Babraham Institute). Reads were mapped to the GRCm39 mouse reference(32) genome using the STAR package(33), and transcripts were quantified as Ensembl gene IDs using the RSEM package(34). BAM files of the aligned reads were sorted and indexed using SAMtools(35), and the indexed reads were viewed in Integrative Genomics Viewer to determine the relative frequencies of WT and c.2966A>G Mrc2 transcripts(36).

Differential expression analyses on the ACF and whole-atrium samples were performed in R using the EDASeq(37), limma(38), and edgeR packages(39). The ACF read count data were filtered for genes with at least 1 count per million (CPM) in at least 4 samples, with the sum of CPM for all 9 samples being at least 4. The ACF isolation batch and sex were accounted for as explanatory variables in the linear model. Similarly, the whole-atrial tissue data were filtered for genes with at least 1 count per million (CPM) in at least 7 samples with the sum of CPM for all 10 samples being at least 7. The sex and age of the mice from which the atria were harvested were accounted for as explanatory variables in the linear model. The data for both ACFs and whole-atrial tissue were normalized for sample library size using the weighted trimmed mean of M-values method and corrected for changes in variance across different mean CPMs using the voom method(40). For each of the E990G-het vs. WT contrasts, the t-statistic and p-value for each gene were determined with H₀: |log₂(fold change)| < log₂(1.2) and H_a: |log₂(fold change)| > log₂(1.2).

Gene set enrichment analysis (GSEA) was performed for Gene Ontology (GO) gene set sizes from 5 to 500 with genes ranked by t-statistic using the clusterProfiler R package(41). Gene sets with only 1% of genes tagged as “leading edge” were disregarded. After stratifying the results by all genes (excluding genes containing “Gm—” or “—Rik” in their MGI symbols) with log₂(fold change) > 0 and those with log₂(fold change) < 0, overrepresentation analysis for GO terms was performed using the “GOseq” R package to determine Wallenius-approximated p-values for overrepresentation of genes having an unadjusted P<0.01 that are annotated for each GO term(42). Metadata, raw fastq files, and processed data were deposited in Gene Expression Omnibus (accession number GSE281558 for the ACF samples (43) and GSE281559 for the atrial tissue samples, NCBI (43)).

Histology.

Murine hearts were fixed in 10% buffered formalin and dehydrated with ethanol, as described (9). The hearts were embedded in paraffin, and 5 μm tissue sections were cut. To quantify fibrosis, sections were deparaffinized in xylene and rehydrated. Subsequently, sections were stained with a commercial Picrosirius Red staining kit (#ab150681, Abcam, Cambridge, MA), according to the manufacturer’s protocol, with one-hour incubation in Sirius Red F3B solution. The sections were washed, dehydrated, and mounted for visualization. All images were captured using a Zeiss Axio Observer microscope. The percentage of the area stained with Picrosirius Red relative to the total tissue area was calculated using ImageJ software.

Secretome profiling.

Secreted proteins from the serum-free ACF conditioned media were concentrated ~4X using an Amicon^® Ultra-4 mL centrifugal filter, 10 kDa MWCO (UFC8010, Millipore, Burlington, MA) at 4,000 × g and 4°C. The concentrates were precipitated with 4 volumes of acetone pre-chilled at −20°C, and the mixtures were placed at −20°C for at least 2 hours. The precipitated proteins were pelleted by centrifugation at 14,000 × g for 10 minutes at 4°C, eluted in RIPA lysis buffer, and quantified using the Pierce^™ BCA Protein Assay Kit (23227, Thermo Scientific^™, Waltham, MA) according to the manufacturer’s protocol. The proteins were precipitated again the same way, and the protein pellets were resuspended and fragmented by trypsin digestion. The peptide fragments were dissolved in 10 mM ammonium bicarbonate buffer (pH 10) and loaded onto a C18 disk plug (3M Empore C18) within a micro-pipette tip containing a C18 column (Reprosil-Pur Basic, 3 μm, Dr. Maisch GmbH, Germany). The peptides were eluted in a stepwise gradient of acetonitrile (2%, 4%, 6%, 8%, 10%, 12%, 14%, 16%, 18%, 20%, 22%, 24%, 26%, 28%, and 30%), pooled into five fractions (2+12+22; 4+14+24; 6+16+26; 8+18+28; 10+20+30), vacuum-dried, resuspended, and analyzed using a nanoLC-1000 (Thermo Scientific^™, Waltham, MA) coupled to an Orbitrap Fusion mass spectrometer (Thermo Scientific^™, Waltham, MA) with an ESI source. Peptides were loaded onto an in-house trap column and separated on an analytical column over a 75-minute gradient (2–28% acetonitrile/0.1% formic acid).

Spectral data were searched against the Mouse RefSeq database using Proteome Discoverer 2.1 (Thermo Fisher) with the Mascot algorithm (Mascot 2.4, Matrix Science). Dynamic modifications included N-terminal acetylation and methionine oxidation. Precursor mass tolerance was set to 20 ppm with a fragment mass tolerance of 0.5 Da, allowing up to two missed cleavages. Protein identification and quantification were performed using gpGrouper (v1.0.040) with shared peptide iBAQ area distribution(44). Missing values were imputed with half the minimum value of the recovered proteome. Differential secretion analysis was performed with the R packages, as described previously for RNA-seq differential expression analysis, with filtering for proteins that were detected in all WT samples and/or all E990G-het samples. Proteins whose transcripts were not detected in RNA-seq were excluded, and quantile normalization of the protein abundance distributions was then performed using the “preprocessCore” R package (Ben Bolstad)(45). The t-statistic and p-value for each protein were determined with H₀: log₂(fold change) = log₂(1) and H_a: |log₂(fold change)| > log₂(1). GSEA was performed for Gene Ontology (GO) gene set sizes from 5 to 500 with proteins ranked by t-statistic as described for transcriptomic analysis. After stratifying the results by all proteins with log₂(fold change) > 0 and those with log₂(fold change) < 0, overrepresentation for GO terms was further analyzed using the “GOseq” R package to determine Wallenius-approximated p-values for overrepresentation of proteins having an unadjusted P<0.01 that are annotated for each GO term(42). Metadata, raw data, and processed data were deposited via the massIVE repository (MSV000096658) to the ProteomeXchange Consortium (dataset identifier PXD058860).

Proteome profiling.

4 WT and 4 Mrc2 E990G-het snap-frozen atria (2 from each sex for each genotype, Supplemental Table 4) were directly processed for mass-spectrometry analysis as described for secretome profiling previously. The t-statistic and p-value for each protein were determined with H₀: log₂(fold change) = log₂(1) and H_a: |log₂(fold change)| > log₂(1). GSEA was performed for Gene Ontology (GO) gene set sizes from 5 to 500 with proteins ranked by t-statistic as described for transcriptomic analysis. After stratifying the results by all proteins with log₂(fold change) > 0 and those with log₂(fold change) < 0, overrepresentation for GO terms was further analyzed using the “GOseq” R package to determine Wallenius-approximated p-values for overrepresentation of proteins having an unadjusted P<0.01 that are annotated for each GO term(42). Hemoglobin and other proteins found predominantly in blood or plasma were excluded from the dataset before differential expression analysis. Metadata, raw data, and processed data were deposited via the massIVE repository (MSV000096658) to the ProteomeXchange Consortium (dataset identifier PXD058860) (46).

Comparison of differential gene and protein expression in mice and humans.

To determine whether the differentially expressed genes and proteins in E990G-het mice with AF are also relevant in human disease, we retrieved single-cell RNA sequencing data of atrial tissue from patients with and without persistent AF from the publicly available dataset GSE224959 (47). The filtering criteria, normalization, and clustering followed the parameters described in the original study (47). The dot plot was generated using Seurat (v5.1.0) built-in function DotPlot() with modifications to color scales, axis labels, and legend.

Statistics, rigor, and reproducibility.

Sample sizes for WT and Mrc2 E990G-het mice were determined based on 80% power to detect a significant difference at α = 0.05 in AF inducibility between WT and E990G-het mice based on prior studies(9, 48). For the categorical variables in the mice (sex and induced AF), two-tailed Boschloo’s exact tests were performed with fixed sample sizes for the genotypes (15 WT and 15 E990G-het). To compare continuous variables between WT and E990G-het, if n > 10 per group and the distributions for both groups are approximately normal (Shapiro-Wilk test p-values > 0.05), two-tailed Welch’s t-tests were performed, and the means and standard deviations (SDs) of each group were reported; otherwise, two-tailed Mann-Whitney U tests with continuity correction were performed, and the medians and interquartile ranges (IQRs) of each group were reported. Statistical analyses and data plotting were executed in R. Boschloo’s tests were conducted using the “exact2×2” R package (Michael Fay). Data processing and plotting were executed using the tidyverse R packages “dplyr”, “ggplot2”, “readr,” and “tidyr.” The investigator who isolated and cultured the ACFs and prepared the RNA-seq and mass spectrometry samples was blinded to the genotypes of the mice until the samples were submitted. Methods for transcriptomic and proteomic analyses are described in detail in the supplemental material.

Results

E990G-het mice are susceptible to AF induction

Programmed electrical stimulation was performed on 15 Mrc2 E990G-het mice and 15 WT controls to determine AF inducibility. Table 1 displays the baseline characteristics of the mice and the surface ECG intervals. Mrc2-het mice had a greater probability of developing AF after PES (60%) than WT controls (20%; P=0.030) (Fig. 1A–B). Among pacing-induced AF responders, AF episode non-zero durations (averaged across the 2 or 3 trials that showed AF rhythm for each AF responder) do not significantly differ between WT (median 21.4 s, IQR 44.6 s) and E990G-het mice (median 14.7 s, IQR 113.0 s; P=0.853) (Fig. 1C). The mean effective refractory period of the AV node (AVERP₁₀₀) was significantly shorter in E990G-het mice (44.6 ms, SD 6.8) compared with WT controls (56.7 ms, SD 7.0; P<0.001) (Fig. 1D–E). To a smaller degree, the mean AVERP₁₀₀ was significantly shorter in AF-positive mice regardless of genotype (45.6 ms, SD 7.7) compared to AF-negative mice (53.6 ms, SD 8.4; P=0.018) (Fig. 1F). In summary, these data indicate that the Mrc2 E990G-het variant increases AF susceptibility and decreases AVERP₁₀₀. Finally, while females had greater AF inducibility than males (P=0.039), sex was not a significant confounder of AVERP₁₀₀ as described in Supplemental Fig. 1.

Table 1.

Baseline ECG parameters of Mrc2 E990G-het mice and WT controls that underwent PES.

	WT	E990G-het	p-value	Test
N	15	15	N/A	N/A
Sex (M/F)	5/10	4/11	1.000	Fisher’s exact test
Age (months)	4.9 (2.9)	6.6 (3.8)	0.392	Mann-Whitney U test
RR (ms)	137.3 (18.5)	134.5 (16.9)	0.671	Welch’s t-test
PR (ms)	48.8 (5.0)	47.8 (4.5)	0.551	Welch’s t-test
QRS (ms)	12.7 (0.86)	13.6 (0.95)	0.006	Welch’s t-test
SNRT100 (ms)	200.4 (62.1)	194.6 (53.4)	0.790	Welch’s t-test

Age is reported as median (IQR). RR, PR, QRS, and SNRT₁₀₀ are reported as mean (SD).

Abbreviations: F – female, E990G-het – Mrc2 Glu990Gly heterozygous, M – male, PES – programmed electrical stimulation, SNRT₁₀₀ - sinus node recovery time 100 ms, WT - wild-type.

Figure 1. AF inducibility was increased in Mrc2 E990G-het mice.

(A) Representative ECG traces after atrial burst pacing in WT and E990G-het mice. (B) Induced AF incidence was three-fold greater in Mrc2 E990G-het versus WT littermates. p-value for two-tailed Boschloo’s exact test is indicated. (C) Among induced AF responders, mean AF durations (across the 2 or 3 trials that showed AF rhythm for each AF responder) do not significantly differ between WT and E990G-het mice. The central crossbar in each boxplot indicates the median, and the edges indicate the first and third quartiles. p-value for two-tailed Mann-Whitney U test with continuity correction is indicated. (D) Representative ECG traces for determination of AVERP₁₀₀. Blue arrows indicate atrial pacing at a BCL of 100 ms followed by a decremental decrease in a premature S2 stimulus. Red bars indicate the shortest S1-S2 interval that resulted in propagation of atrial depolarization into the ventricles after right atrial pacing, which was defined as the AVERP₁₀₀. (E) AVERP₁₀₀ was decreased in Mrc2 E990G-het mice. p-value for two-tailed Welch’s t-test is indicated. (F) AVERP₁₀₀ was less significantly decreased in induced AF responders vs. non-responders independently of genotype. The central crossbars (E-F) indicate means, and the error bars indicate means ± 1 SD. p-value for two-tailed Welch’s t-test is indicated. Abbreviations: AF – atrial fibrillation, AVERP₁₀₀ – atrioventricular effective refractory period 100 ms, BCL – basic cycle length, Het – Mrc2 Glu990Gly heterozygous, WT – wild-type.

Mrc2 expression was slightly greater in E990G-het ACFs than in WT ACFs (8.9% median increase in transcripts per million in E990G-het vs. WT; Mann-Whitney U test P=0.020; Supplemental Fig. 2A) but did not significantly differ by genotype in whole-atrial tissue (11.0% median decrease; P=0.603; Supplemental Fig. 2B). As expected, there was a significant difference in the E990G-het transcript detected in Het vs. WT ACFs (Supplemental Fig. 2C) and atrial tissue (Supplemental Fig. 2D). In the E990G-het ACFs, the abundance of c.2966A>G transcripts was only slightly lower than that of WT transcripts, suggesting that the slightly increased Mrc2 expression may be due to compensatory upregulation of the WT allele (median relative read frequency = 48% c.2966A>G, 52% WT; Wilcoxon signed rank test P=0.058; Supplemental Fig. 2C). However, this difference in allele expression was not identified in E990G-het whole atrial tissue (median 48% c.2966A>G, 52% WT; P=0.667; Supplemental Fig. 2D). RNA-seq of both the ACFs and whole-atrial tissue confirmed that the Mrc2 E990G knock-in transcripts contain the c.2969A>G substitution (c.2966A>G or D889G based on the mouse sequence) surrounded by six synonymous mutations (Supplemental Table 5).

Differential gene expression of fibrotic regulators in cultured ACFs isolated from MRC2 E990G-het mice

Transcriptome profiling of the isolated ACF cultures did not indicate any significantly differentially expressed genes (DEGs) by Mrc2 genotype after Benjamini-Hochberg adjustment for false discovery rate, accounting for sex and cell-isolation batch. However, based on unadjusted limma(38) p-values < 0.05 (for the null hypothesis of |log₂(|fold change)| < |log₂(1.2)), most of the upregulated and downregulated genes are involved in fibrosis (Fig. 2A). Supplemental Table 6 describes 20 of the most significant DEGs with published roles in fibrosis, as shown in Fig. 2A. These fibrotic DEGs include activating transcription factor 3 (Atf3, 43% decrease in E990G-het vs. WT), erythroid differentiation regulator 1x (Erdr1x, 48% decrease), matrix metalloproteinase 12 (Mmp12, 39% decrease), matrix metalloproteinase 13 (Mmp13, 42% decrease), and neuronal regeneration-related protein (Nrep, 54% increase) (Fig. 2B). Supplemental Table 2 displays the fold-change confidence intervals and p-values for these select genes. Fig. 2C displays a heatmap of relative expression levels by sample for all DEGs with unadjusted P<0.016. We previously showed that E990G-het mice exhibit reduced collagen organization and a disordered extracellular matrix within the annulus fibrosus region (9). An expanded analysis of histological sections stained with Picrosirius Red revealed a 2.4-fold increase in fibrosis level within the right atrial free wall of E990G-het mice compared with WT littermates (Supplemental Fig. 3).

Figure 2. Differential expression analysis of RNA-seq transcriptomic profiling of cultured Mrc2 E990G-het and WT-control mouse ACFs.

(A) Volcano plot of the log₂ fold changes in gene expression and limma unadjusted p-values for E990G-het vs. WT ACF total-RNA samples, accounting for sex and cell-isolation batch variability. Vertical gray lines indicate FCs of ± 1.2. Black: p-value > 0.05; orange: 0.01 < p-value ≤ 0.05; red: p-value ≤ 0.01; magenta: p-value with published roles in fibrosis (described in Supplemental Table 6). (B) Boxplots of transcript abundances of select genes by genotype in the ACF total-RNA samples. Mann-Whitney U-test p-values with continuity correction on sex- and batch-adjusted transcripts per million are indicated. The central crossbars indicate medians, and the edges indicate the first and third quartiles. (C) Heatmap of relative transcript abundances of differentially expressed genes (P<0.016, excluding genes containing “Gm—” or “—Rik” in their MGI symbols) in the ACF total-RNA samples, ordered by descending estimated FC. (D-E) Top significantly enriched GO terms by clusterProfiler GSEA on genes ranked by limma t-statistic among upregulated (D) and downregulated (E) genes. Gene symbols enriched in each gene set’s “leading edge” are listed for each GO term ordered by descending t-statistic. Bars are colored by GO term domain. n = 4 WT (1 F, 3 M) and 5 E990G-het (2 F, 3 M) ACF lines. Abbreviations: ACF – atrial cardiofibroblast, BP – biological process, FC – fold change, GO – Gene Ontology, GSEA – gene set enrichment analysis, Het – Mrc2 Glu990Gly heterozygous, limma – Linear Models for Microarray Data R package(38), MF – molecular function, RNA-seq – RNA sequencing, WT – wild-type.

Gene set enrichment analysis (GSEA) using clusterProfiler(41) on these transcriptomic data ranked by limma t-statistic revealed several significantly enriched Gene Ontology(49) (GO) terms after Benjamini–Hochberg p-value adjustment for false discovery rate, including “tRNA metabolic process” (GO:0006399, B-H adjusted P=0.001), “mitochondrial respiratory chain complex assembly” (GO:0033108, B-H adjusted P=0.022), and “aerobic electron transport chain” (GO:0019646, B-H adjusted P=0.022) among upregulated genes (Fig. 2D) and “transferase complex, transferring phosphorous-containing groups” (GO:0061695, B-H adjusted P<0.001), “ubiquitin binding” (GO:0043130, B-H adjusted P=0.001), and “regulation of TOR signaling” (GO:0032006, B-H adjusted P=0.001) among downregulated genes (Fig. 2E). Overrepresentation analysis revealed that the 21 upregulated genes with unadjusted P<0.05 were most significantly overrepresented among the GO terms “bone mineralization” (GO:0030282), “retinal rod cell differentiation” (GO:0060221), and “ossification” (GO:0001503) (Supplemental Fig. 4A); the 23 downregulated genes with unadjusted P<0.01 were most significantly overrepresented among the GO terms “import into cell” (GO:0098657), “import across plasma membrane” (GO:0098739), and “regulation of cell activation” (GO:0050865) (Supplemental Fig. 4B).

Other GO terms that were overrepresented among these 44 DEGs include “regulation of cell population proliferation” [GO:0042127; 4 up (overrepresented P=0.114); 9 down (P=0.002)] and “regulation of cell migration” [GO:0030334; 3 up (P=0.039); 7 down (P=0.001)]. Erdr1x, the most statistically significant DEG, is annotated with both of these categories. Additionally, “regulation of cell communication” (GO:0010646; up P=0.016; down P=0.007) and “regulation of signaling” (GO:0023051; up P=0.016; down P=0.007) were significantly overrepresented, both containing 7 upregulated DEGs: fibrillin 2 (Fbn2, P=0.035), glypican 3 (Gpc3, P=0.033), glycosylphosphatidylinositol specific phospholipase D1 (Gpld1, P=0.032), opioid related nociceptin receptor 1 (Oprl1, P=0.044), PEAK1 related, kinase-activating pseudokinase 1 (Prag1, P=0.046), pleiotrophin (Ptn, P=0.038), and Nrep; and 10 downregulated DEGs: Atf3, arginase 1 (Arg1, P=0.008), C-type lectin domain containing 7A (Clec7a, P=0.008), cytochrome P450 family 26 subfamily B member 1 (Cyp26b1, P=0.008), fibroblast growth factor 10 (Fgf10, P=0.005), interleukin 1 receptor antagonist (Il1rn, P=0.001), MCF.2 cell line-derived transforming sequences like (Mcf2l, P=0.007), Mmp12, nucleotide-binding oligomerization domain-containing 2 (Nod2, P=0.008), and retinol binding protein 4 (Rbp4, P=0.001). Of these DEGs, Gpc3, Gpld1, Prag, Ptn, Atf3, Arg1, Clec7a, Fgf10, Mmp12, Nod2, and Rbp4 are also annotated with “regulation of cell population proliferation,” and Fbn2, Gpld1, Ptn, Clec7a, Fgf10, Il1rn, Mmp12, and Nod2 are also annotated with “regulation of cell migration.” This suggests that altered autocrine signaling among ACFs and/or paracrine signaling between ACFs and atrial myocytes may contribute to the phenotypic differences between the MRC2 E990G-het variant and WT. Lastly, “immune response” (GO:0006955) was significantly overrepresented among downregulated DEGs with unadjusted P<0.01 (overrepresented P<0.001) and includes 8 genes: Arg1, Clec7a, Il1rn, lipocalin 2 (Lcn2, P=0.003), Mmp12, Nod2; 2 upregulated DEGs are also in this category (overrepresented P=0.306): Gpld1 and collectin subfamily member 11 (Colec11, P=0.036). Atf3 is the only DEG annotated with “response to unfolded protein” (GO:0006986; down genes overrepresented P=0.166), and Lcn2 is the only DEG annotated with “response to oxidative stress” (GO:0006979; down genes overrepresented P=0.458). Supplemental Fig. 5A displays the estimated expression fold changes of the DEGs in these GO categories.

Differential secretion of fibrotic regulators from cultured ACFs isolated from Mrc2 E990G-het mice

Secretomic profiling of conditioned media samples from cultured ACFs isolated from 4 WT and 4 E990G-het mice shows that MMP-12, which was downregulated in the E990G-het ACF transcriptomic profiling, was also downregulated in the E990G-het ACF secretome, accounting for sex and cell-isolation batch (45% decrease in E990G-het vs. WT) (Fig. 3A,B). Additional differentially secreted proteins involved in fibrosis with unadjusted P<0.05 include fatty acid binding protein 4 (FABP4, 1248% increase), insulin-like growth factor binding protein 3 (IGFBP3, 78% decrease), insulin-like growth factor binding protein 4 (IGFBP4, 98% decrease), and periostin (POSTN, 118% increase) (Fig. 3B). Supplemental Table 7 displays the fold-change confidence intervals and p-values for these select proteins. Fig. 3C displays a heatmap of relative secreted protein levels by sample for differentially secreted proteins with unadjusted P<0.05. MRC2 was detected at low levels in the secretome of two WT samples and one E990G-het sample and not detected in the other five samples (not analyzed due to filtering).

Figure 3. Mass-spectrometry secretomic profiling of cultured Mrc2 E990G-het and WT-control mouse ACFs.

(A) Volcano plot of the log₂ fold changes in protein abundance and limma unadjusted p-values for E990G-het vs. WT ACF conditioned media samples, accounting for sex and cell-isolation batch. Vertical gray lines indicate FCs of ± 1.2. Black: p-value > 0.05; orange: 0.01 < p-value ≤ 0.05; red: p-value ≤ 0.01. (B). Boxplots of abundances of select proteins by genotype in the ACF conditioned media samples. Mann-Whitney U-test p-values with continuity correction on quantile-normalized protein levels adjusted for sex and batch are indicated. The central crossbars indicate medians, and the edges indicate the first and third quartiles. (C) Heatmap of relative abundances differentially secreted proteins (P<0.05) in the ACF conditioned media samples, ordered by descending estimated FC. (D-E) Top significantly enriched GO terms by clusterProfiler GSEA on proteins ranked by limma t-statistic among increased (D) and decreased (E) secreted proteins. Protein symbols enriched in each gene set’s “leading edge” are listed for each GO term ordered by descending t-statistic. Bars are colored by GO term domain. n = 4 WT (1 F, 3 M) and 4 E990G-het (2 F, 2 M) ACF lines. Abbreviations: ACF – atrial cardiofibroblast, BP – biological process, CC – cellular component, FC – fold change, GO – Gene Ontology, GSEA – gene set enrichment analysis, Het – Mrc2 Glu990Gly heterozygous, limma – Linear Models for Microarray Data R package(38), MF – molecular function, WT – wild-type.

GSEA on these secretomic data ranked by limma t-statistic did not reveal significantly enriched GO terms after Benjamini–Hochberg p-value adjustment for false discovery rate, but GO terms with unadjusted P<0.001 enriched with upregulated proteins include “extracellular matrix structural constituent” (GO:0005201), “G2/M transition of mitotic cell cycle” (GO:0000086), and “regulation of cell division” (GO:0051302) among upregulated genes (Fig. 3D), and GO terms enriched with downregulated proteins include “protein-lipid complex” (GO:0032994, B-H adjusted P=0.001), “dendritic tree” (GO:0097447, B-H adjusted P=0.002), and “leukocyte mediated cytotoxicity” (GO:0001909, B-H adjusted P=0.003) (Fig. 3E). The 8 proteins that were secreted more by E990G-het ACFs with unadjusted P<0.05 were most significantly overrepresented among the GO terms “oxoacid metabolic process” (GO:0043436), “organic acid metabolic process” (GO:0006082), and “carboxylic acid metabolic process” (GO:0019752) (Supplemental Fig. 3C); the 20 proteins that were secreted less with unadjusted P<0.05 were most significantly overrepresented among the GO terms “regulation of response to stimulus” (GO:0048583), “regulation of signal transduction” (GO:0009966), and “negative regulation of signaling” (GO:0010648) (Supplemental Fig. 3D).

Similar to the ACF transcriptomic analysis, “regulation of cell communication” (GO:0010646; up P=0.717; down P=0.005) and “regulation of signaling” (GO:0023051; up P=0.717; down P=0.005) were significantly overrepresented among significantly downregulated secreted proteins, both containing 10 downregulated proteins: clusterin (CLU, P=0.026), histidine triad nucleotide binding protein 1 (HINT1, P=0.021), IGFBP3, IGFBP4, MMP12, nucleobindin 2 (NUCB2, P=0.024), protein disulfide isomerase associated 3 (PDIA3, P=0.049), protein disulfide isomerase associated 6 (PDIA6, P=0.028), reticulocalbin 3 (RCN3, P=0.035), and tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein beta (YWHAB, P=0.047); but not among upregulated proteins (POSTN being the only significantly upregulated protein). Among these secreted cell-communication and signaling regulators that were downregulated, MMP-12 was also downregulated at the transcript level.

Among non-overrepresented pathways that are relevant to cardiac remodeling, 1 upregulated protein is annotated with “regulation of cell population proliferation” (GO:0042127; overrepresented P=0.747): FABP4; 6 downregulated proteins are also in this category (overrepresented P=0.189): C-type lectin domain containing 11A (CLEC11A, P=0.043), CLU, IGFBP3, MMP12, syndecan 4 (SDC4, P=0.048) and vimentin (VIM, P=0.049). Similarly, 1 upregulated protein is annotated with “regulation of cell migration” (GO:0030334; overrepresented P=0.788): POSTN; 6 downregulated proteins are also in this category (overrepresented P=0.135): IGFBP3, MMP12, phospholipase A2 group VII (PLA2G7, P=0.029), SDC4 thymosin beta 4 X-linked (TMSB4X, P=0.049), and VIM. The upregulated tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein epsilon (YWHAE, P=0.018) and downregulated complement C1q B chain (C1QB, P=0.004), complement C1q C chain (C1QC, P=0.039), MMP12, PDIA3, and VIM are annotated with “immune response” (GO:0006955; up proteins overrepresented P=0.822, down proteins overrepresented P=0.185); the upregulated mesencephalic astrocyte-derived neurotrophic factor (MANF, P=0.012) and downregulated PDIA6 are annotated with “response to unfolded protein” (GO:0006986; up proteins overrepresented P=0.156; down proteins overrepresented P=0.367); and the upregulated thioredoxin reductase 1 (TXNRD1, P=0.023) is the only significantly differentially secreted protein annotated with “response to oxidative stress” (GO:0006979; up proteins overrepresented P=0.405). Supplemental Fig. 5B displays the estimated secretion fold changes of the differentially secreted proteins in these GO categories.

Transcriptomic profiling of atrial tissue from Mrc2 E990G-het mice

Like transcriptomic profiling of ACFs, transcriptomic profiling of whole-atrial tissues from 8 WT mice without induced AF response and 7 E990G-het mice with induced AF response (except for one non-responder, Supplemental Table 4), accounting for sex and age of the mice at harvesting, did not indicate any DEGs after false discovery rate adjustment. However, based on unadjusted P<0.05 (for the null hypothesis of log₂(|FC)| < log₂(1.2)), the analysis suggests differential expression of several genes involved in cardiac fibrosis and arrhythmia, including bromodomain-containing protein 4 (Brd4, 106% increase in E990G-het vs. WT), ciliary neurotrophic factor (Cntf, 66% decrease), Kruppel-like factor 7 (Klf7, 123% increase), SRY (sex determining region Y)-box 4 (Sox4, 83% increase), and SRY-box 9 (Sox9, 154% increase) (Fig. 4A,B). Supplemental Table 8 displays the fold-change confidence intervals and p-values for these select genes. Fig. 4C displays a heatmap of relative expression levels by sample for DEGs with unadjusted P<0.0022.

Figure 4. Differential expression analysis of RNA-seq transcriptomic profiling of atrial tissue from Mrc2 E990G-het mice with induced AF response and WT non-responders.

(A) Volcano plot of the log₂ fold changes in gene expression and limma unadjusted p-values for whole-atrium total-RNA samples from E990G-het induced-AF responders (and one E990G-het non-responder) vs. WT non-responders, accounting for sex and age. Vertical gray lines indicate FCs of ± 1.2. Black: p-value > 0.05; orange: 0.01 < p-value ≤ 0.05; red: p-value ≤ 0.01. (B) Boxplots of transcript abundances of select genes by genotype in the whole-atrium total-RNA samples. Mann-Whitney U-test p-values with continuity correction on sex- and age-adjusted transcripts per million are indicated. The central crossbars indicate medians, and the edges indicate the first and third quartiles. (C) Heatmap of relative transcript abundances of differentially expressed genes (P<0.0022, excluding genes containing “Gm—” or “—Rik” in their MGI symbols) in the whole-atrium total-RNA samples, ordered by descending estimated FC. (D-E) Top significantly enriched GO terms by clusterProfiler GSEA on genes ranked by limma t-statistic among upregulated (D) and downregulated (E) genes. Gene symbols enriched in each gene set’s “leading edge” are listed for each GO term ordered by descending t-statistic. Bars are colored by GO term domain. n = 8 whole-atrium pairs from WT mice with no induced AF (6 F, 2 M) and 7 whole-atrium pairs from E990G-het mice with induced AF except for one non-responder (7 F). Abbreviations: BP – biological process, CC – cellular component, FC – fold change, GO – Gene Ontology, GSEA – gene set enrichment analysis, Het – Mrc2 Glu990Gly heterozygous, limma – Linear Models for Microarray Data R package(38), MF – molecular function, RNA-seq – RNA sequencing, WT – wild-type.

GSEA on these transcriptomic data ranked by limma t-statistic revealed no significantly enriched GO terms among the upregulated genes after Benjamini–Hochberg p-value adjustment for false discovery rate, but GO terms with unadjusted P<0.001 among upregulated genes include “forebrain development” (GO:0030900), “ATP-dependent activity” (GO:0140657), and “cytosolic ribosome” (GO:0022626) (Fig. 4D). Three GO terms were significantly enriched among downregulated genes after Benjamini–Hochberg p-value adjustment: “mitochondrial translation” (GO:0032543, B-H adjusted P=0.004), “fatty acid beta-oxidation” (GO:0006635, B-H adjusted P=0.013), and “estrogen metabolic process” (GO:0008210, B-H adjusted P=0.034) (Fig. 4E). Overrepresentation analysis revealed that the 287 upregulated genes with unadjusted P<0.01 were most significantly overrepresented among GO terms such as “chromatin binding” (GO:0003682), “somatic stem cell population maintenance” (GO:0035019), and “multicellular organism growth” (GO:0035264) (Supplemental Fig. 4E); the 41 downregulated genes with unadjusted P<0.05 were most significantly overrepresented among the GO terms “U4 snRNP” (GO:0005687), “U5 snRNP” (GO:0005682), and “U2-type prespliceosome” (GO:0071004) (Supplemental Fig. 4F). Other GO terms that were significantly overrepresented among these 328 DEGs include “regulation of cell proliferation involved in heart morphogenesis” [GO:2000136; 2 up (overrepresented P=0.036): hes family bHLH transcription factor 1 (Hes1, P=0.002) and Sox9; 0 down (P=1.000)].

Among non-overrepresented pathways that are relevant to cardiac remodeling, 2 out of the 287 upregulated DEGs with unadjusted P<0.01 are annotated with “fibroblast proliferation” (GO:0048144; overrepresented P=0.606): lysine methyltransferase 2A (Kmt2a; P=0.005) and lysine methyltransferase 2C (Kmt2c; P=0.001); no downregulated DEGs with unadjusted P<0.05 are in this category (underrepresented P=0.739). 4 upregulated DEGs are annotated with “cation channel complex” (GO:0034703; overrepresented P=0.223): A-kinase anchoring protein 9 (Akap9, P=0.002), potassium voltage-gated channel subfamily A member 1 (Kcna1, P<0.001), potassium voltage-gated channel subfamily A member 2 (Kcna2, P=0.003), and phosphodiesterase 4D (Pde4d, P=0.002); no downregulated DEGs are in this category (underrepresented P=0.817). 10 upregulated DEGs are annotated with “immune response” (GO:0006955; overrepresented P=0.907), including ankyrin repeat and KH domain containing 1 (Ankhd1, P=0.001), Brd4, eukaryotic translation initiation factor 2-alpha kinase 2 (Eif2ak2, P=0.003), FYN binding protein 1 (Fyb1, P=0.001), Pde4d, protein kinase C epsilon (Prkce, P=0.008), and tripartite motif containing 5 (Trim5, P=0.005); no downregulated DEGs are in this category (underrepresented P=0.040). Prkce is also the only DEG annotated with “macrophage activation” (GO:0042116; overrepresented P=0.682). 2 upregulated DEGs are annotated with “response to unfolded protein” (GO:0006986; overrepresented P=0.274): Eif2ak2 and ubiquitin thioesterase OTU1 (Yod1, P=0.008); no downregulated DEGs are in this category (underrepresented P=0.734). No DEGs are annotated with “response to oxidative stress” (GO:0006979; up genes underrepresented, P=0.025; down genes underrepresented, P=0.355) or “sarcoplasmic reticulum” (GO:0016529; up genes underrepresented, P=0.214; down genes underrepresented, P=0.837). None of the 6 connexin genes detected by RNA-seq (Gja1, 3, 4, and 5 and Gjc1 and 2) or sodium channel genes were significant DEGs (P>0.05) except for Scn3a, which was marginally upregulated (98% increase in E990G-het vs WT, P=0.037). Supplemental Fig. 5C displays the estimated expression fold changes of the DEGs in these GO categories.

Proteome analysis of atrial tissue from Mrc2 E990G-het mice

Proteomic profiling of additional atrial tissue samples from mice (that did not undergo PES, Supplemental Table 9) suggests that expression of cellular communication network factor 2 (CCN2), also known as connective tissue growth factor (CTGF), is increased by 877% in E990G-het vs. WT (Figs. 5A, B). Additional differentially expressed proteins involved in fibrosis or arrhythmia with unadjusted P<0.05 include asporin (ASPN, 849% increase), CD151 antigen (CD151, 216% increase), myristoylated alanine-rich protein kinase C substrate (MARCKS, 59% decrease), and trans-2,3-enoyl-CoA reductase-like (TECRL, 608% increase). Supplemental Table 10 displays the fold-change confidence intervals and p-values for these select proteins. None of the MMPs were detected in sufficient quantity to be analyzed. Fig. 5C displays a heatmap of relative secreted protein levels by sample for differentially expressed proteins with unadjusted P<0.005. MRC2 was detected at low levels in the proteome of three WT samples only and not detected in the other five samples (not analyzed due to filtering).

Figure 5. Mass-spectrometry proteomic profiling of Mrc2 E990G-het and WT-control mouse atrial tissue.

(A) Volcano plot of the log₂ fold changes in protein abundance and limma unadjusted p-values for E990G-het vs. WT whole-atrium total-protein samples, accounting for sex and age, from mice that did not undergo PES. Vertical gray lines indicate FCs of ± 1.2. Black: p-value > 0.05; orange: 0.01 < p-value ≤ 0.05; red: p-value ≤ 0.01. (B) Boxplots of abundances of select proteins by genotype in whole-atrium total-protein samples. Mann-Whitney U-test p-values with continuity correction on quantile-normalized protein levels adjusted for sex and age are indicated. The central crossbars indicate medians, and the edges indicate the first and third quartiles. (C) Heatmap of relative abundances of differentially expressed proteins (P<0.005) in the whole-atrium total-protein samples, ordered by descending estimated FC. (D-E) (D-E) Top significantly enriched GO terms by clusterProfiler GSEA on proteins ranked by limma t-statistic among increased (D) and decreased (E) secreted proteins. Protein symbols enriched in each gene set’s “leading edge” are listed for each GO term ordered by descending t-statistic. Bars are colored by GO term domain. n = 4 WT (2 F, 2 M) and 4 E990G-het (2 F, 2 M) whole-atrium pairs. Abbreviations: FC – fold change, GO – Gene Ontology, GSEA – gene set enrichment analysis, Het – Mrc2 Glu990Gly heterozygous, limma – Linear Models for Microarray Data R package(38), PES – programmed electrical stimulation, WT – wild-type.

GSEA on these proteomic data ranked by limma t-statistic revealed one significantly enriched GO term after Benjamini–Hochberg p-value adjustment for false discovery rate: “nucleosome” (GO:0000786, B-H adjusted P=0.032). Other GO terms with unadjusted P<0.001 enriched with upregulated proteins include “anchoring junction” (GO:0070161) and “anatomical structure morphogenesis” (GO:0009653) (Fig. 3D). GO terms enriched with downregulated proteins with unadjusted P<0.001 include “cellular respiration” (GO:0045333), “organelle inner membrane” (GO:0019866), and “mitochondrial respiratory chain complex assembly” (GO:0033108) (Fig. 3E). The 37 proteins that were upregulated in E990G-het atria with unadjusted P<0.01 were most significantly overrepresented among the GO terms “catalytic activity, acting on DNA” (GO:0140097), “calcium ion transmembrane import into cytosol” (GO:0097553), and “maintenance of location in cell” (GO:0051651) (Supplemental Fig. 4G); the 20 proteins that were downregulated with unadjusted P<0.01 were most significantly overrepresented among the GO terms “presynaptic cytosol” (GO:0099523), “cytosolic region” (GO:0099522), and “germinal vesicle” (GO:0042585) (Fig. 5H). Other GO terms that were significantly overrepresented among these 57 differentially expressed proteins include “fibroblast proliferation” [GO:0048144. 2 up (overrepresented P=0.029); 0 down (P=1.000)].

Among non-overrepresented pathways that are relevant to cardiac remodeling, 2 out of the 37 upregulated proteins with unadjusted P<0.01 are annotated with “cation channel complex” (GO:0034703; overrepresented P=0.103): AKAP9 (P=0.002) and ryanodine receptor 1 (RYR1, P=0.007); no downregulated proteins are in this category (underrepresented P=0.873). 4 upregulated proteins are annotated with “immune response” (GO:0006955; overrepresented P=0.488): including phosphohistidine phosphatase 1 (PHPT1, P=0.009), phospholipase C gamma 1 (PLCG1, P=0.002), ras-related protein Rab-2B (RAB2B, P=0.003), and raftlin (RFTN1, P=0.003); no downregulated proteins are in this category (underrepresented P=0.275). Of these proteins, none are annotated with “macrophage activation” (GO:0042116; underrepresented P=0.818), but PHPT1, PLCG1, and RFTN1 are annotated with “immune response-activating cell surface receptor signaling pathway” (GO:0002429), which is significantly overrepresented among upregulated proteins (overrepresented P=0.023). Two downregulated proteins are annotated with “response to oxidative stress” (GO:0006979; overrepresented P=0.280): ARF-like GTPase 6 interacting protein 5 (ARL6IP5, P=0.003) and coiled-coil-helix-coiled-coil-helix domain containing 4 (CHCHD4, P<0.001); no upregulated proteins are in this category (underrepresented P=0.238). Two upregulated proteins are annotated with sarcoplasmic reticulum” (GO:0016529; overrepresented P=0.221): phospholamban (PLN, P=0.003) and RYR1); no downregulated proteins are in this category (underrepresented P=0.748). No differentially expressed proteins are annotated with “response to unfolded protein” (GO:0006986; up protein underrepresented P=0.754; down protein underrepresented P=0.840) or “regulation of cell proliferation involved in heart morphogenesis” GO:2000136; up and down protein underrepresented P=0.962). Connexin-43 (GJA1) was the only connexin detected by mass spectrometry, and it was not significantly differentially expressed (17% decrease in E990G-het vs. WT, P=0.588). No sodium channel proteins were detected. Supplemental Fig. 5D displays the estimated expression fold changes of the differentially expressed proteins in these GO categories.

Comparison of omics analyses

Table 2 summarizes the genes and proteins that were significantly differentially expressed or secreted (limma P<0.05) in more than one of the transcriptomic, secretomic, and proteomic analyses previously described. Mmp12 was downregulated in both the ACF transcriptome and secretome; Cas scaffold protein family member 4 (Cass4) was downregulated in both the ACF transcriptome and atrial tissue transcriptome; X-linked retinitis pigmentosa GTPase regulator-interacting protein 1 (Rpgrip1) was upregulated in both the ACF transcriptome and atrial transcriptome; dipeptidase 2 (Dpep2), Fgf10, Gm28439 pseudogene, and tropomodulin 1 (Tmod1) were downregulated in the ACF transcriptome but upregulated in the atrial transcriptome; family with sequence similarity 174, member B (Fam174b) was downregulated in both the ACF transcriptome and atrial proteome; Akap9, PDGFA associated protein 1 (Pdap1), proline-rich coiled-coil 2C (Prrc2c), treacle ribosome biogenesis factor 1 (Tcof1), xin actin-binding repeat containing 2 (Xirp2), and YLP motif containing 1 (Ylpm1) were upregulated in both the atrial transcriptome and atrial proteome; and serine dehydratase (Sds) was downregulated in both the atrial transcriptome and atrial proteome. These transcriptomic, secretomic, and proteomic data suggest moderate changes in gene expression, protein expression, and/or secretion in several regulators of fibrosis, inflammation, cell signaling, and electrical remodeling in the MRC2 E990G-het ACFs and atria: specifically, upregulation of profibrotic molecules and downregulation of antifibrotic molecules.

Table 2.

Log₂(FC) 99% CI and limma unadjusted p-values of genes and proteins that were differentially expressed or secreted (P<0.05, bold) in more than one of the transcriptomic, secretomic and proteomic analyses (Mrc2 E990G-het vs. WT).

	ACF transcriptome		ACF secretome		Atrial transcriptome		Atrial proteome
Gene/protein	Log2(FC) CI	p-value	Log2(FC) CI	p-value	Log2(FC) CI	p-value	Log2(FC) CI	p-value
Mmp12	(−1.09, −0.33)	0.002	(−2.00, 0.25)	0.031	(−1.80, 1.22)	0.626	Not detected
Cass4	(−3.45, 0.49)	0.040	Not detected		(−1.90, 0.14)	0.049	Not detected
Dpep2	(−1.43, −0.08)	0.019	Not detected		(0.05,1.86)	0.020	Not detected
Fgf10	(−3.02, −0.33)	0.004	Not detected		(0.07, 1.88)	0.019	Not detected
Gm28439	(−2.11, −0.02)	0.016	Not detected		(0.12, 2.97)	0.010	Not detected
Rpgrip1	(−0.19, 2.27)	0.033	Not detected		(0.08, 1.85)	0.018	Not detected
Tmod1	(−1.76, −0.25)	0.006	Not detected		(0.12, 1.19)	0.023	(−0.84, 0.38)	0.587
Fam174b	(−2.19, 0.18)	0.034	Not detected		(−0.45, 0.34)	0.945	(−7.08, −0.09)	0.009
Akap9	(−1.33, 0.61)	0.395	Not detected		(0.37, 1.55)	0.002	(1.26, 7.67)	0.002
Pdap1	(−0.38, 0.30)	0.984	Not detected		(0.03, 1.20)	0.049	(0.35, 2.67)	0.004
Prrc2c	(−0.21, 0.28)	0.995	Not detected		(0.15, 0.98)	0.025	(−0.46, 4.17)	0.031
Sds	(−0.76, 0.86)	0.913	Not detected		(−1.76, 0.03)	0.034	(−4.15, −0.39)	0.005
Tcof1	(−0.35, 0.27)	0.986	Not detected		(0.10, 1.17)	0.029	(0.21, 5.48)	0.008
Xirp2	Not detected		Not detected		(−0.05, 1.58)	0.046	(−0.30, 5.87)	0.018
Ylpm1	(−0.21, 0.38)	0.967	Not detected		(0.22 1.21)	0.009	(−0.06, 6.76)	0.012

Abbreviations: ACF – atrial cardiofibroblast, CI – confidence interval, FC – fold change, Het – heterozygous, limma – Linear Models for Microarray Data R package, WT – wild-type

To determine whether the candidate genes identified from murine studies (Table 2) are relevant to human disease, we analyzed a publicly available single-cell RNA sequencing dataset from atrial tissues of patients with or without persistent AF (47). Examination of the fibroblast population revealed that all genes except Gm28439 were expressed in human fibroblasts (Supplemental Fig. 6). Notably, PRRC2C, PDAP1, and AKAP9 are broadly expressed in fibroblasts, with marked upregulation in patients with persistent AF relative to controls (Supplemental Fig. 6). These findings align with the elevated expression of these genes in E990G-het mice and underscore their translational relevance as potential therapeutic targets.

Discussion

Increased atrial fibrosis is associated with increased AF incidence, severity, and recurrence, and we hypothesize that impaired atrial collagen turnover due to Mrc2 loss-of-function contributes to excessive atrial fibrosis and AF progression. To identify electrophysiological and ACF functions that may be altered with the Mrc2 E990G-het variant identified in patients with WPW syndrome, we compared the ECG characteristics and ACF cultures from mice with knock-in of the orthologous variant to WT controls. Key differences with the E990G-het mice include increased susceptibility to PES-induced AF and decreased AVERP₁₀₀ and altered expression or secretion of fibrotic regulators, such as decreased gene expression and/or secretion of Atf3, Mmp12, and Mmp13 and increased protein expression or secretion of CCN2/CTGF, FABP4, and POSTN. While there was little overlap in the sets of differentially expressed ACF genes, secreted ACF proteins, atrial tissue genes, and atrial tissue proteins, Mmp12 was significantly decreased in the ACF transcriptome and secretome, while not being detected in the atrial proteome, and being detected at very low levels in the atrial transcriptome (likely due to the preponderance of non-expressing myocytes). In summary, our studies revealed that the Mrc2 E990G-het variant identified in patients with WPW syndrome causes mildly profibrotic effects in isolated mouse ACFs in the form of differential regulation of some profibrotic gene transcripts and secreted proteins. When combined with decreased collagen internalization and degradation with the Mrc2 loss-of-function variant,(9) these molecular changes may favor inappropriate accumulation of arrhythmogenic fibrotic scar tissue in the atria. This may also be accompanied by changes in the expression or secretion of genes and proteins involved in inflammation, cell signaling, and electrical remodeling, which, together with profibrotic remodeling, could promote atrial arrhythmogenesis.

The expression data of individual genes and proteins should be interpreted with caution in these discovery-driven omics studies since the small sample sizes and effect sizes prevented any hits after a p-value adjustment for the false discovery rate. Further validation of gene and protein expression levels is required to distinguish true positives from false positives. GSEA offered greater power by accounting for all detected genes and proteins regardless of significance, and it revealed several significantly enriched GO terms after Benjamini-Hochberg p-value adjustment.

In the context of fibrotic remodeling, plasmin cleaves and activates the matrix metalloproteinases MMP-3, MMP-9, MMP-12, and MMP-13(50). Of these MMPs, MMP-13 cleaves and degrades collagen type I, the major constituent of fibrotic scar tissue(51, 52), while MMP-12 cleaves collagen types I, III, and IV, elastin, and fibronectin in the ECM(53, 54). Mmp12 and Mmp13 were notably also downregulated at the gene transcription level according to our ACF transcriptomic analysis. Furthermore, secreted MMP12 levels were decreased in the E990G-het ACF conditioned media. In addition to decreased expression of Mmp12 and Mmp13, we hypothesize that the Mrc2 E990G-het variant leads to less uPA activation of plasminogen mediated by ACFs and consequently less activation of pro-MMPs. Since full-length collagen must be degraded before Mrc2 can mediate its internalization, reduction in MMP-13 activation could contribute to reduced collagen turnover with the Mrc2 variant. MMP-12 has been shown to be protective against interstitial cardiac fibrosis while simultaneously promoting perivascular cardiac fibrosis(55), while MMP-13 was shown to be protective against lung fibrosis(56, 57). Treatment with an MMP-1-sparing inhibitor in dogs decreased AF inducibility, which could potentially be explained by collagen degradation products of MMPs paradoxically stimulating increased collagen deposition in the atria(58).

Among the other downregulated genes identified in E990G-het ACFs with transcriptomic profiling, Atf3 expressed specifically in cardiac fibroblasts, has been shown to protect against angiotensin II-induced cardiac fibrosis(59). Upregulated proteins in the conditioned media of E990G-het ACFs according to secretomic profiling include FABP4, which is positively associated with AF recurrence after catheter ablation and increases migration of ACFs(60, 61), and POSTN, which is also positively associated with AF, AF recurrence after catheter ablation, and atrial fibrosis(62, 63). Of note, altered protein levels detected in this secretomic study may be due to either altered trafficking and secretion of the proteins by ACFs into the extracellular space or altered ratios of ECM-bound to conditioned media-soluble protein levels, as only the soluble fraction was sampled for mass spectrometry. The secreted protein levels may also be confounded by inadvertent differences in ACF handling during isolation and culturing that affect protein binding to the ECM that are unrelated to the MRC2 genotype. Overall, our in vitro studies suggest that genes and secreted proteins that regulate cell signaling and communication were significantly overrepresented among the DEGs and differentially secreted proteins in ACF cultures.

One of the most significantly upregulated proteins in E990G-het atria was CCN2/CTGF, which has been shown to be an autocrine profibrotic regulator of the heart involved in post-myocardial infarction fibrosis(64, 65). Similarly upregulated proteins in the atria include ASPN, which promotes cardiac fibroblast proliferation, migration, and collagen production(66). Lastly, upregulated genes in the atrial tissue transcriptome of E990G-het mice include Brd4, which activates cardiac fibroblasts as an effector of profibrotic TGF-β signaling(67), Klf7, which protects against cardiac fibrosis in WT male mice but promotes cardiac fibrosis when overexpressed(68), Sox4, which is involved in cardiac hypertrophy-associated fibrosis(69), and Sox9, which is positively associated with AF severity and atrial fibrosis, and which increases α-Smooth muscle actin (α-SMA, a marker of myofibroblast differentiation) expression and migration of ACFs(70). Downregulated genes in the atria include Cntf, which, like Atf3, has been shown to protect against angiotensin II-induced cardiac fibrosis(71).

The membrane-bound MRC2-uPAR-uPA complex promotes cell chemotaxis up a uPA gradient(19). In contrast, internalized endosomal MRC2 promotes random migration(20). This raises the question of whether the disordered collagen deposition in the annuli fibrosis of Mrc2 E990G-het hearts can be explained by altered activation of random fibroblast migration vs. chemotaxis toward uPA, leading to a more diffuse pattern of fibrosis (e.g., increased fibrosis throughout the atria with decreased annular fibrosis) that contributes to arrhythmogenesis. Whether this variant affects MRC2 affinity to uPA/uPAR in the ACF membrane, localization to the membrane vs. endosomes, and/or activation of signaling proteins promoting random ACF migration [Rho1-GTP, Rho kinase (ROCK), and MLC2-P] vs. proteins promoting uPA chemotaxis (Rac1-GTP and Cdc42-GTP) remains to be determined.

A limitation of the studies on isolated 2D ACF cultures is that they could not detect any effects of the MRC2 variant that are dependent on ACF interactions with their native atrial microenvironment and non-fibroblast cells, such as immune cells and atrial myocytes(72). We also observed strong clustering of gene expression and protein secretion levels in the ACF isolation batch, indicating the potential effects of unintended slight differences in cell isolation and culture conditions. Our transcriptomic and proteomic analyses of atria that were immediately frozen after excision helped to mitigate this issue, but the atrial tissue transcriptomic and proteomic profiles differed greatly from those of ACF cultures, likely due to the predominance of atrial myocytes over fibroblasts in the tissue samples. Future studies could profile the transcriptome and proteome of these other atrial cell types in isolation to determine whether they may contribute to arrhythmogenesis in MRC2 deficiency.

Published studies on associations between MRC2 and AF thus far have only measured induced AF (postoperative AF for humans and exercise-induced and pacing-induced for mice). Future longitudinal studies could therefore investigate the association between MRC2 deficiency and spontaneous AF. Furthermore, functional validation studies are required to identify any causal relationships between the differentially expressed fibrotic regulators identified here and AF susceptibility. These may involve experimentally altering the expression of these fibrotic regulators in ACFs or atrial tissue and measuring changes in AF inducibility by PES and structural remodeling by cardiac MRI and Masson trichome or Picrosirius red histological staining.

In the previous study of the MRC2 E990G-het variant in WPW syndrome, ex vivo optical mapping of isoproterenol-stimulated mouse hearts revealed a higher incidence of supraventricular tachycardia and aberrant retrograde accessory pathways in E990G-het vs. WT hearts, and Picrosirius red staining suggested more disorganized collagen deposition in the annulus fibrosis between the left atrium and left ventricle(9). A similar study of the right atrial free wall revealed increased interstitial fibrosis levels in E990G-het mice compared with WT controls.

Since we detected a higher incidence of pacing-induced AF in Mrc2 E990G-het females than in E990G-het males, future studies could also investigate the underlying molecular or structural differences in the atria that underlie this sex difference. To determine whether our findings are generalizable to humans with the MRC2 E990G-het variant or MRC2 deficiency not caused by this variant, future studies could replicate our experiments on other mouse models of MRC2 deficiency (e.g., atrial-specific Mrc2 knockdown) and human cultured ACFs with MRC2 E990G knock-in or MRC2 knockdown. If additional human patients with the E990G-het variant who undergo cardiac procedures are identified, these studies could also profile biopsied atrial tissue from them compared to wild-type human tissue and determine whether this replicates our findings in mice. Older E990-het patients may show more long-term structural remodeling effects due to this variant. Finally, future transcriptomic and proteomic profiling studies on healthy and diseased human cardiac tissue specimens may reveal additional arrhythmias or cardiomyopathies that are associated with MRC2 deficiency.

Conclusion and clinical implications

Excessive cardiac fibrosis contributes to atrial fibrillation progression, but current treatments do not target fibrosis at the cellular level (ACFs)(5). MRC2 promotes collagen removal from the ECM, and Mrc2 deficiency is associated with AF susceptibility. We showed here that the Mrc2 E990G-het loss-of-function variant modestly alters mouse ACF function toward a more profibrotic phenotype. Mrc2 deficiency and its consequences in ACFs may be targetable mechanisms in designing antifibrotic therapies for AF.

Supplementary Material

Supplemental Tables S1–S10 and Supplemental Figures S1–S6: https:/doi.org/10.6084/m9.figshare.30053662

Abbreviations

ACF

Atrial cardiofibroblast

AF

Atrial fibrillation

CI

Confidence interval

ECM

Extracellular matrix

FC

Fold change

GO

Gene Ontology

GSEA

Gene set enrichment analysis

Het

Heterozygous

IQR

Interquartile range

MRC2

Mannose receptor C type 2

PES

Programmed electrical stimulation

SD

Standard deviation

WT

Wild-type

DATA AVAILABILITY

The RNAseq data have been submitted to GEO (Gene Expression Omnibus) with accession number GSE281558 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE281558) for the ACF samples and GSE281559 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE281559) for the atrial tissue samples. The mass-spectrometry data have been submitted to the ProteomeXchange Consortium with the dataset identifier PXD058860 (https://proteomecentral.proteomexchange.org/cgi/GetDataset?ID=PXD058860) via the MASSIVE repository (MSV000096658, https://massive.ucsd.edu/ProteoSAFe/dataset.jsp?task=b8bf7f038bab44bc84120be92727d8f5). Samples corresponding to the deposited mass spectrometry files are shown in Supplemental Tables 3 and 4.