Fibulin-3 is necessary to prevent cardiac rupture following myocardial infarction

Despite the high prevalence of heart failure in the western world, there are few effective treatments. Fibulin-3 is a protein involved in extracellular matrix (ECM) structural integrity, however its role in the heart is unknown. We have demonstrated, using single cell RNA-seq, that fibulin-3 was highly expressed in quiescent murine cardiac fibroblasts, with expression highest prior to injury and late post-infarct (from ~ day-28 to week-8). In humans, fibulin-3 was upregulated in left ventricular tissue and plasma of heart failure patients. Fibulin-3 knockout (Efemp1−/−) and wildtype mice were subjected to experimental myocardial infarction. Fibulin-3 deletion resulted in significantly higher rate of cardiac rupture days 3–6 post-infarct, indicating a weak and poorly formed scar, with severe ventricular remodelling in surviving mice at day-28 post-infarct. Fibulin-3 knockout mice demonstrated less collagen deposition at day-3 post-infarct, with abnormal collagen fibre-alignment. RNA-seq on day-3 infarct tissue revealed upregulation of ECM degradation and inflammatory genes, but downregulation of ECM assembly/structure/organisation genes in fibulin-3 knockout mice. GSEA pathway analysis showed enrichment of inflammatory pathways and a depletion of ECM organisation pathways. Fibulin-3 originates from cardiac fibroblasts, is upregulated in human heart failure, and is necessary for correct ECM organisation/structural integrity of fibrotic tissue to prevent cardiac rupture post-infarct.


Analysis of online single cell/nuclei and bulk RNA-Seq datasets
Online single-cell (sc)RNA-Seq and single-nuclei (sn)RNA-Seq datasets are available from ArrayExpress (www.ebi.ac.uk/arrayexpress) under accession codes E-MTAB-7376 (scRNA-Seq of interstitial and Pdgfra-GFP + cardiac cells from sham and MI hearts [1]) and E-MTAB-7869 (snRNA-Seq of young and aged healthy mouse hearts [2]).Bulk RNA-Seq datasets are available from the gene expression omnibus (www.ncbi.nlm.nih.gov/geo)under identifiers GSE141929 (Pdgfra-GFP + cardiac cells from uninjured and MI hearts [3]) and GSE114695 (total cardiac cells from the left ventricles of sham and MI hearts [4]).
scRNA-Seq and snRNA-Seq datasets were analysed using the Seurat version 3.1.4R package [5].The scRNA-Seq datasets were processed to log-normalized counts, scaled and principal component (PC) analysis performed on the top 2000 variable genes.For snRNA-Seq, the replicates were batch aligned, similar to previously [2], using the FindIntegrationAnchors and IntegrateData functions in Seurat and scaled prior to PC analysis.For all datasets, UMAP dimensionality reduction was run on the top 25 PCs.
Cell labels were assigned to Pdgfra-GFP + cells using previous characterisations [1].For the corresponding interstitial cell dataset, the FindNeighbors function in Seurat was run on the top 25 PCs, followed by FindClusters with the res parameter set to 0.2.Gene co-expression analysis with Efemp1 was performed by calculating Pearson correlation coefficients between Efemp1 and all other genes on log-normalized counts in the Pdgfra-GFP + dataset and selecting the top 20 most positively and negatively correlated genes.
For the GSE141929 bulk RNA-seq, previous gene counts were available and downloaded for analysis.
For GSE114695, raw Fastq files were downloaded, mapped and processed to gene counts using STAR aligner [6] following removal of Illumina adaptors and trimming of low-quality bases using Trimmomatic [7].Analysis of bulk RNA-Seq data was performed using DESeq2 [8].Differentially expressed (DE) genes were evaluated between MI time-points and undamaged hearts for the Pdgfra-GFP + RNA-Seq, and between MI and sham hearts for matched time-points in the total cardiac cell RNA-Seq.Efemp1 was considered DE if it obtained an adjusted p-value below 0.05.

LC-MS/MS analysis
Failing and non-failing heart tissue specimens were lysed in 200 µl of 100 mM Tris pH = 8.5, containing 1 % sodium dodecylsulfate and 10 mM of NEM, and 100 µg of protein was acetoneprecipitated overnight.Protein pellets were re-dissolved in 87.5 µl of 50 % trifluoroethanol in 50 mM ammonium-bicarbonate, reduced with tris(2-carboxyethyl)phosphine (5 mM final concentration), then realkylated and digested with trypsin overnight at 37°C.The tryptic digest (40 µg) was then fractionated using Pierce High pH Reversed-Phase Peptide Fractionation Kit (Thermo Fisher, USA) according to manufacturer's instructions and subjected to LC-MS/MS analysis.Chromatography was carried out on an Ultimate 3000 RCS Nano Dionex system equipped with an Ionoptiks Aurora Series UHPLC C18 column (250 mm x 75 µm, 1.6 µm) (Bruker Daltonics, Germany).Mass spectrometry was performed on a Maxis II qTOF set to fragment the top 20 most abundant peptides.

RNA isolation and analysis of day 3 post-MI tissue
Quantitative polymerase Reaction (qPCR) was performed to determine the expression levels of a subset of genes belonging to four genes associate with scar formation (Collagen-I alpha-1, Collagen-III alpha-1, Mmp2 & Mmp9).RNA isolation was conducted using TRIzol® Plus RNA purification reagents (Life Technologies) as per the manufacturer protocol.RNA samples were subject to DNase I treatment using a DNase I, Amplification Grade Kit (ThermoFisher scientific) as per the manufacturer protocol.
Total RNA was quantified using the NanoDrop 1000 then DNAse 1 treated (Invitrogen).One µg of RNA was used for the synthesis of cDNA using Oligo (dT)18 primers (Bioline), dNTP (Bioline) and a SuperScript® III (Invitrogen) First Strand Synthesis System for RT-PCR.
Assays were validated by generating standard curves to evaluate the efficiency of each primer set.The specificity of PCR products was analysed via melt curve and gel electrophoresis.All target and reference genes from cDNA transcripts were measured using quantitative real-time polymerase chain reaction (qPCR), using the 7500 SDS software v2.0.6 (Applied Biosystems, Australia), and were performed on the ABI 7500.The geomean of housekeeping genes Hprt and TPt1 mRNA abundance was used as the reference, and final values calculated using the ΔΔCt method.The following primers were used:

Second harmonic generation two-photon imaging
Imaging was carried out on a Leica TCS SP8 MP microscope using 800 nm excitation laser wavelength.Collagen fibre orientation and length was quantified on Fiji Image J [9].Brightness and exposure threshold settings were applied to all imaged samples, and 40 collagen fibres were traced per image.The proportion of fibres within ±30° of the median angle for each animal was assessed to determine the degree of fibre-alignment.Analysis was performed by two blinded investigators with strong interassessor reliability (length: r=0.917, p<0.0001; angle: r=0.963, p<0.0001).

Gene set enrichment analysis
GSEA pre-ranked run parameters used the base settings with geneset permutation and 1000 permutations.Prior to analysis, a ranked list was calculated with each gene assigned a score based on the FDR and the direction of the log fold-change ("+" or "−").Results of the GSEA were then loaded into Cytoscape version 3.7.2for visualisation and further analysis using the EnrichmentMap plugin.The settings used in Cytoscape were Jaccard Overlap Combined: 0.375 -Test used: Jaccard Overlap Combined Index (k constant = 0.5).A permutation-based P value that was corrected for multiple testing to produce a permutation-based false discovery rate (FDR) Q-value with a cutoff of 0.01 for initial analysis and increased to 0.001 for image visualisation.Nodes were coloured to reflect the different groups in question, and networks of related ontologies were circled and assigned a group label using the AutoAnnotate application.The network map layout algorithm used was prefuse force-directed weighted using the gene set similarity coefficient followed by manual adjustments, separating networks and nodes by direction of fold-change for clarity purposes.Supplementary Figure 2. Fibulin-3 deficient male mice had significantly higher rates of cardiac rupture than females.Higher rates of rupture in male mice has been reported in the literature [11][12][13][14], with one paper demonstrating 2-3 fold increased rupture in 3 strains of male mice [11].Intramural haemorrhage severity, hematoma formation and inflammatory cell accumulation were suggested potential mechanisms.Another study indicated estrogenic as a potential mechanism, demonstrating that estrogenictreated male mice had reduced cardiac rupture prevalence, which was associated with reduced MMP-9 activity, and AKT pathway apoptosis inhibition [12].A further study suggested increases in the severity of inflammation, MMP-9 activation and damage to collagen matrix accounted for the male bias in cardiac rupture [13].The mechanisms governing the gender differences seen in rupture prevalence was outside the scope of this study, however investigation into the inflammatory, MMP, and collagen matrix differences are warranted.Supplementary Figure 3. Gender differences in WT and Efempl -/-mice post-MI.Histology and echocardiography data show there was no difference between genders for collagen deposition (a), LVEF (c), IVS (d) or LVID (e), however males had significantly higher EDV and HW:TL (b, e).Heart weight to tibia length ratio (HW:TL) and collagen deposition: n=16 Efemp1 -/-, n=28 WT.Ejection fraction (LVEF), intraventricular septum thickness (IVS), end diastolic volume (EDV), left ventricular internal diameter (LVID): n=14 Efemp1-/-, n=12 WT.Mean ± SD.Supplementary Figure 4. Pulmonary Analysis in WT and Efempl -/-mice post-MI.To assess the development of lung congestion as a marker of heart failure, lungs were weighed (a), and assessed with echocardiography (b).MoLUS -Mouse Lung UltaSound [15].Mean ± SD.
Supplementary Table 1.Cell number and proportion metadata of interstitial cell populations.Data is separated by time-point/condition post-MI, from the mouse scRNAseq dataset [16].

Table 2 .
[16]p1 expression level metadata of interstitial cell populations.Data is separated by time-point/condition post-MI, from the mouse scRNAseq dataset[16].

Table 4 .
Echocardiography parameters following experimental myocardial infarction.

Table 5 .
Full list of differentially expressed genes from bulk RNA-seq of infarct zone tissue of Efemp1 -/-and WT mice at day-3 post-MI, ranked by fold change of expression in Efemp1 -/-mice relative to WT.

Table 6 .
Full list of enriched/depleted pathways from GSEA analysis of bulk RNA-seq data of infarct zone tissue of Efemp1 -/-and WT mice at day-3 post-MI (data of Supplementary Table4), ranked by normalised enrichment score (NES) of pathways in Efemp1 -/-mice relative to WT.