Chamber-enriched gene expression profiles in failing human hearts with reduced ejection fraction

Heart failure with reduced ejection fraction (HFrEF) constitutes 50% of HF hospitalizations and is characterized by high rates of mortality. To explore the underlying mechanisms of HFrEF etiology and progression, we studied the molecular and cellular differences in four chambers of non-failing (NF, n = 10) and HFrEF (n = 12) human hearts. We identified 333 genes enriched within NF heart subregions and often associated with cardiovascular disease GWAS variants. Expression analysis of HFrEF tissues revealed extensive disease-associated transcriptional and signaling alterations in left atrium (LA) and left ventricle (LV). Common left heart HFrEF pathologies included mitochondrial dysfunction, cardiac hypertrophy and fibrosis. Oxidative stress and cardiac necrosis pathways were prominent within LV, whereas TGF-beta signaling was evident within LA. Cell type composition was estimated by deconvolution and revealed that HFrEF samples had smaller percentage of cardiomyocytes within the left heart, higher representation of fibroblasts within LA and perivascular cells within the left heart relative to NF samples. We identified essential modules associated with HFrEF pathology and linked transcriptome discoveries with human genetics findings. This study contributes to a growing body of knowledge describing chamber-specific transcriptomics and revealed genes and pathways that are associated with heart failure pathophysiology, which may aid in therapeutic target discovery.


Tissue Collection
• Zenas Technologies (metadata in Supplemental Table 1) -The postmortem interval for all donors was reported to be within 3 hours. The specimens were removed from the approximate center mass of each chamber and represent transmural sections (e.g. epi to endo through the cardiac wall). All specimens were snap frozen to preserve RNA integrity. The normal (non-failing) donor population was defined as patients without clinical symptoms consistent with or previously diagnosed as heart failure. HFrEF donors were defined as patients with symptoms that categorized them as New York Heart Association (NYHA) functional Class II, III or IV, with LVEF < 50%, and a medical history consistent with heart failure.
• AnaBios Corporation (metadata in Supplemental Table 1) -AnaBios conducted a custom prospective collection in 2016. The warm ischemic time (WIT) was less than 1 hour, any reported downtime was less than 30 minutes, and the cold ischemic time (CIT) was less than 24 hours for all donors. Each region of the heart was submitted as three separate formats; snap frozen, OCT embedded, and formalin fixed paraffin embedded (FFPE). Normal (non-failing) donors excluded donors with any current or history of heart/cardiovascular disease (i.e. MI, CAD, stent(s), pacemaker, ventricular hypertrophy, arrhythmia, high cholesterol, plaques, heart defibrillation, murmur) and patients with inflammatory disease that may affect heart health such as systemic lupus erythematosus. The heart failure donors were defined to have had a history of heart failure (not necessarily cause of death) and excluded patients with inflammatory diseases that affect heart health such as systemic lupus erythematosus (Supplemental Table 1).
• International Institute for the Advancement of Medicine (IIAM, metadata in Supplemental Table 1) -The warm ischemic time (WIT) was less than 1 hour, any reported downtime was less than 30 minutes, and the cold ischemic time (CIT) was less than 24 hours for all donors. Whole hearts were placed in one liter HTK or UW solution on wet ice. Hearts were dissected and each region of the heart was submitted as three separate formats; snap frozen, OCT embedded, and FFPE. The normal (nonfailing) donors excluded donors with any current or past history of cardiovascular disease such as, high cholesterol, plaques, heart defibrillation, heart attack, pacemaker or heart murmur, myocardial infarction, coronary artery disease, stent(s), pacemaker, ventricular hypertrophy, arrhythmia and any inflammatory diseases that affect heart health such as systemic lupus erythematosus. The heart failure donors were defined to have had a history of heart failure (not necessarily cause of death) and excluded patients with inflammatory diseases that affect heart health such as systemic lupus erythematosus.

Frozen heart tissue pulverization and preparation for RNA extraction
Frozen heart tissue biopsies were pulverized using Bussmann Grinders, 100-200 mg of tissue powder was then homogenized for 30 seconds at 6500 rpm in buffer (350 µl of Qiagen RLT buffer with 1% BME; Qiagen, Germantown, MD, USA) using a MagNA Lyser (Roche Diagnostics, Indianapolis, Indiana, USA). After homogenization, 298 µl of RNase-free water and 2 µl of a 50 mg/ml proteinase K solution were added to each tube and mixed well. Samples were incubated at 55°C for 10 minutes, centrifuged, and the supernatant was collected into a new 1.5ml RNase-free microcentrifuge tube.

RNA isolation and sequencing analysis of the heart
RNA extraction was performed using the RNeasy Micro Kit (Qiagen) with on-column DNase treatment (Qiagen) according to the manufacturer's instructions. RNA concentration and integrity were assessed using a NanoDrop 8000 (Thermo Fisher, Waltham, MA, USA) and a Bioanalyzer (Agilent, Santa Clara, CA, USA). Samples with ≥100 ng total RNA and RNA integrity numbers (RIN) ≥7 were used for sequencing. After passing the QC, total RNA (100 or 500 ng) was used for cDNA library preparation using a modified protocol based on the Illumina Truseq RNA sample preparation kit and the published method for strand-specific RNA-Seq 45,46 . After poly-A selection, fragmentation, and priming, reverse transcription was carried out for 1 st strand cDNA synthesis in the presence of RNaseOut (Invitrogen) and actinomycin-D (MP Biomedicals). The synthesized cDNA was purified by using AMPure RNAClean beads (Beckman Coulter) following the manufacturer's protocol. A modified method by incorporation of dUTP instead of dTTP was prepared and used for the second strand synthesis according to previously published protocols 45,46 . After AMPure XP bead purification (Beckman Coulter), following the standard protocol recommended by the Illumina Truseq RNA kit, end repairing, A-tailing, and ligation of index adaptors were sequentially performed for generation of cDNA libraries. After size selection of libraries using Pippen Prep (SAGE Biosciences), the dUTP-containing strands were destroyed by digestion with USER enzymes (New England Biolabs) followed by a step of PCR enrichment for the introduction of strand specificity. After cleaning up, the enriched cDNA libraries were analyzed in an Agilent Bioanalyser and quantified by Quant-iT TM Pico-Green assays (Life Technologies) before being loaded onto the HiSeq platform (Illumina).

Heart tissue specificity
Heart tissue specificity was calculated using RNA-Seq of 30 different normal tissues in the GTEx database 22 . RNA-Seq datasets for normal tissues from GTEx were processed by Omicsoft based on human genome version GRCm38 and gene model GENCODE v24. Gene expression was represented as FPKM and normalized with a refinement of the commonly employed upper-quartile method 49 that sets the FPKM to a value of 10 at the 70 th percentile 50,51 . In GTEx, different tissues include tissue subcategories, for example, uterus tissue include ectocervix, endocervix, and general uterus. To maintain gene expression profile in the tissue subcategories, median FPKM values were calculated for different tissue subcategories. Tissue gene expression used in the tissue specificity calculation was the highest median normalized FPKM value of tissue subcategories for that tissue. Heart enriched genes were selected as gene expression in heart as top 3 out of 30 normal tissues with normalized FPKM >=1.

Definition of subregion enriched gene lists
Based on the significance tests from DESeq2, genes enriched in a single chamber were significantly upregulated in that chamber in three pairwise comparisons with the other three heart chambers. For example, for LV enriched genes, they were significantly up-

Supplemental Table 4. Cell type enriched genes in NF heart tissue defined by snRNA-Seq (Excel Data Supplement S4)
The list is generated using the Seurat Package FindAllMarkers function aimed to compare major cardiac clusters. "pct.1": The percentage of cells where the feature is detected in the specific cardiac cluster; "pct.2": The average percentage of cells in all the other clusters where the feature is detected; "avg_logFC": log fold-change of the average expression between the specific cluster and the rest of clusters. "p_val" and "p_val_adj": raw and Bonferroni corrected p value using all features in the dataset based on Wilcoxon rank sum test.  Table 6. Significantly differentially expressed genes (DEGs) in HFrEF vs NF in LA heart tissue (Excel Data Supplement S6) DEGs in LA were generated by DESeq2 Wald test to compare HFrEF vs NF two groups; "LA.HFrEF.log2FoldChange" indicates the log2 transformed fold change of gene expression comparing HFrEF vs NF in left atrium; "LA.HFrEF.padj" represents the BH (Benjamini-Hochberg) corrected p value based on the test.

Supplemental Table 7. Significantly differentially expressed genes (DEGs) in HFrEF vs NF in LV heart tissue (Excel Data Supplement S7)
DEGs in LV were generated by DESeq2 Wald test to compare HFrEF vs NF two groups; "LV.HFrEF.log2FoldChange" indicates the log2 transformed fold change of gene expression comparing HFrEF vs NF in left ventricle; "LV.HFrEF.padj" represents the BH (Benjamini-Hochberg) corrected p value based on the test.