Key Points
Question
Can investigating circulating, noncoding, extracellular RNAs (ex-RNAs) associated with heart failure (HF) and myocardial disease in a community-based sample offer insights into pathways of myocardial remodeling?
Findings
In this large epidemiologic cohort with a broad array of ex-RNAs, we found several micro RNAs associated with prevalent left ventricular remodeling. Three of these RNAs were associated with incident HF during long-term follow-up and with regulating genes implicated in hypertension.
Meaning
Circulating ex-RNAs associated with myocardial remodeling and incident HF in this community-based sample regulate biological pathways relevant to cardiomyocyte pathobiology. Further studies integrating the molecular epidemiology of HF with mechanistic studies to inform disease detection and therapy are warranted.
This community-based cohort study measures the association of highly expressed circulating extracellular RNA molecules with incident heart failure and left ventricular remodeling.
Abstract
Importance
Mortality is high among patients heart failure (HF) who are receiving treatment, and therefore identifying new pathways rooted in preclinical cardiac remodeling phenotypes may afford novel biomarkers and therapeutic avenues. Circulating extracellular RNAs (ex-RNAs) are an emerging class of biomarkers with target-organ epigenetic effects relevant to myocardial biology, although large human investigations remain limited.
Objective
To measure the association of highly expressed circulating ex-RNAs with left ventricular remodeling and incident HF in a community-based cohort.
Design, Setting, and Participants
This is a prospective observational cohort study of individuals who were included in the eighth examination of the Framingham Offspring Cohort (2005-2008). Collected data include measurements of the left ventricle via electrocardiography, determination of circulating ex-RNAs in plasma, and incidence of heart failure. Data analysis was completed from December 2016 to June 2018.
Exposures
A total of 398 circulating ex-RNA molecules in plasma were measured by reverse transcription polymerase chain reaction; disease ontology analysis was also performed.
Main Outcomes and Measures
Echocardiographic indices of left ventricular (LV) remodeling and incident heart failure.
Results
A total of 2763 participants of the Framingham Heart Study with measured ex-RNAs (mean [SD] age, 66.3 [9.0] years; 1499 [54.3%] female) were included in this study. Of this sample, 2429 to 2432 individuals had echocardiographic measures recorded (depending on the measurement). A total of 2681 individuals had HF status determined, of whom 116 (4.3%) experienced HF (median [interquartile range] follow-up, 7.7 [6.6-8.6] years). We identified 12 ex-RNAs associated with LV mass and at least 1 other echocardiographic phenotype (LV end-diastolic volume or left atrial dimension). Of these 12 ex-RNAs, 3 micro RNAs (miR-17, miR-20a, and miR-106b) were associated with a 15% reduction in long-term incident HF per 2-fold increase in circulating level during the follow-up period, after adjustments for age, sex, established HF risk factors, and prevalent or interim myocardial infarction. These 3 RNAs shared sequence homology and targeted a shared group of messenger RNAs that specified pathways relevant to HF (eg, transforming growth factor–β signaling, growth/cell cycle, and apoptosis), and shared a disease association with hypertension in disease ontology analysis.
Conclusions and Relevance
This study identifies a group of circulating, noncoding RNAs associated with echocardiographic phenotypes, long-term incident HF, and pathways relevant to myocardial remodeling in a large community-based sample. Further investigations into the functional biology of these ex-RNAs are warranted for surveillance for HF prevention.
Introduction
Despite successful therapies for heart failure (HF) that target remodeling pathways, the mortality rate of patients receiving treatment remains high. Efforts to identify alternative mechanisms that contribute to myocardial remodeling at an earlier stage in disease may improve screening and open new therapeutic avenues for HF. Extracellular ribonucleic acids (termed ex-RNAs) are circulating, stable, noncoding RNA molecules that can serve as biomarkers of cardiovascular diseases and, importantly, may also exert epigenetic control of overexpression of a wide array of genes whose products are implicated in myocardial remodeling and HF. The ability to identify a role for ex-RNAs in human HF has been limited to studies of patients with severe or established HF and small cohorts with limited phenotypic characterization without a prognostic time horizon. Here, we address this important gap by identifying ex-RNAs associated with echocardiographic left ventricular (LV) remodeling phenotypes and incident HF during long-term follow-up in the Framingham Heart Study (FHS).
The FHS is a community-based observational cohort study of cardiovascular disease in Framingham, Massachusetts, with serial examinations of participants every 4 to 8 years. The study sample for this investigation consists of individuals in the FHS Offspring cohort who had both detailed echocardiography and plasma collected at their eighth examination (in 2005 to 2008). Written informed consent was obtained from all study participants, with institutional review board approval received at Boston University and the University of Massachusetts.
The techniques used for plasma collection and ex-RNA quantification in FHS have been previously published.1 We included ex-RNAs (microRNAs, or miRNAs; small nucleolar RNAs, or snoRNAs; and piwi-interacting RNAs, or piRNAs) expressed within the linear range of a Fluidigm Biomark System thermocycler (Fluidigm; 6 to 23 polymerase chain reaction cycles) in at least 100 FHS participants, yielding 301 miRNAs, 59 piRNAs, and 38 snoRNAs (eTable 1 in the Supplement). We performed echocardiography as previously published2 to measure LV mass (LVM), LV size (by LV end-diastolic volume [LVEDV]), LV function (by LV ejection fraction [LVEF] or fractional shortening [FS]), and left atrial dimension (LAD). We used the Devereux method for end-diastolic LVM,3 and the Teichholz method for LV volume,4 normalized to height to the 2.7th power.
The overall strategy was to measure associations between ex-RNAs and echocardiographic indices of cardiac remodeling as a screening test for those candidate ex-RNAs potentially relevant to preclinical HF phenotypes. Subsequently, we tested the association of these candidate ex-RNAs with incident all-cause HF.
Echocardiographic indices were log-transformed based on visual assessment of distributions to approximate a normal distribution prior to regression. We constructed age-adjusted, sex-adjusted linear regression models to associate each echocardiographic index with each ex-RNA. Given multiple models (1 model per each ex-RNA), we established a 5% false discovery rate (via the Benjamini-Hochberg false discovery rate approach) to screen associations. The ex-RNAs associated with LVM and 1 additional echocardiographic phenotype (LVEDV, LVEF, FS, or LAD) were moved into models for incident HF (which were diagnosed as previously published).5,6 Based on previous large consortium efforts in HF prediction,7 we used Cox proportional hazards regression for incident HF for each candidate ex-RNA, first in a model adjusted for age, sex, prevalent myocardial infarction (MI), and incident MI (modeled as a time-dependent covariate) (model 1); and then in a model further adjusted for diabetes, body mass index (calculated as weight in kilograms divided by height in meters squared), systolic blood pressure, lifetime use of antihypertensive medications, and current smoking (model 2). Survival plots were generated using the %survplot macro in SAS (Ryan Lennon). Finally, we identified networks of genes and pathways regulated by candidate ex-RNAs using previously published techniques (including open-source tools such as Enrichr, Cytoscape, and WikiPathways, among others, as described in eMethods in the Supplement).8 All statistics were performed with SAS software version 9.3 (SAS Institute) with a 2-tailed P value less than .05 considered statistically significant. Data analysis was completed from December 2016 to June 2018.
Results
The baseline clinical and echocardiographic characteristics of our study sample are shown in the Table. There were a total of 2763 participants. The mean (SD) age was 66.3 (9.0) years, with a nearly equal distribution by sex (1499 participants [54.3%] were female). Nearly 50% of individuals (1353 [49.0%]) reported ever having used antihypertensive medication, and mean (SD) systolic blood pressure was 129 (17) mm Hg at the clinic visit. Per measurements of LVEF (mean [SD] values, 67.0% [7.5%]), mean LV systolic function was preserved.
Table. Clinical and Biochemical Characteristics in the Study Cohort.
| Variable | Participants, No. | Mean (SD) Value |
|---|---|---|
| Age, y | 2763 | 66.3 (9.0) |
| Female, No. (%) | 2763 | 1499 (54.3) |
| Diabetes, No. (%) | 2760 | 397 (14.4) |
| Current cigarette smoking, No. (%) | 2759 | 222 (8.0) |
| BMI | 2756 | 28.3 (5.4) |
| Anti-hypertension medication use ever, No. (%) | 2753 | 1353 (49.0) |
| Blood pressure, mm Hg | ||
| Systolic | 2761 | 129 (17) |
| Diastolic | 2759 | 73 (10) |
| Total cholesterol, mg/dL | 2763 | 186 (38) |
| High-density lipoprotein cholesterol, mg/dL | 2762 | 57 (18) |
| Echocardiographic indices | ||
| LV mass, g/m2.7 | 2429 | 42.3 (10.5) |
| LV end-diastolic volume, mL/m2.7 | 2432 | 28.2 (5.8) |
| LV ejection fraction, % | 2431 | 67.0 (7.5) |
| LV fractional shortening, % | 2431 | 37.6 (5.7) |
| Left atrial dimension, cm | 2698 | 4.0 (0.6) |
Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); CVD, cardiovascular disease; LV, left ventricular.
We constructed age-adjusted and sex-adjusted linear regression models to identify associations between ex-RNAs and each echocardiographic parameter (eTable 2 in the Supplement). We found 29 ex-RNAs associated with LVM (with associations ranging from β = −0.025; P = 4.52 × 10−4 for miR-339-5p to β = 0.051; P = 3.27 × 10−3 for piRNA-40766), 11 ex-RNAs associated with LAD (with associations ranging from β = −0.015; P = 5.43 × 10−4 for miR-33a-3p to β = −0.006; P = 2.60 × 10−4 for piRNA-54042), and 7 associated with LVEDV (at a false detection rate threshold of 0.051; associations ranged from β = −0.015; P = 4.39 × 10−4 for miR-20b-5p to β = −0.011; P = 7.90 × 10−4 for miR-26b-5p and β = −0.011; P = 8.97 × 10−4 for miR-106b-5p). Twelve ex-RNAs were associated with LVM and at least 1 other echocardiographic phenotype (with LAD: miR-126-5p, piRNA-54043, miR-26a-5p, piRNA-54042, miR-23a-3p, and miR-126-3p; with LVEDV: mIR-20a-5p, miR-93-5p, miR-20b-5p, miR-106b-5p, and miR-17-5p; with both LAD and LVEDV: miR-26b-5p). Negative β-coefficients signify that increased plasma abundance of each ex-RNA was associated with a decrease in LV mass.
Of the initial 2763 participants in the study cohort, 72 (2.6%) had a history of HF diagnosed before the eighth examination cycle and were excluded from the analysis, leaving 2691 individuals (97.4%). Of these 2691 individuals, HF status was adjudicated on 2681 study participants (99.6%), with 116 cases diagnosed at a median follow-up of 7.7 years (interquartile range, 6.6-8.6 years). Notably, 9 of these cases (7.8%) had a previous diagnosis of myocardial infarction; of the 107 individuals without a prevalent MI before the eighth examination, 17 participants experienced an incident MI before the time of HF diagnosis in follow-up (15.9%). The number of FHS participants included in each survival analysis varied based on overlap between participants with HF status available, RNAs quantified, and adjustments (eTable 3 in the Supplement). The results of Cox proportional hazards regression for the association of the 12 candidate ex-RNAs with incident HF are shown in eTable 3 in the Supplement. Lower plasma abundance of miR-106b-5p, miR-17-5p, and miR-20a-5p were associated with a nearly 15% higher hazard of incident all-cause HF after adjustment in model 1 (miR-20a-5p: HR, 0.84 [95% CI, 0.72-0.98]; P = .03; miR-106b-5p: HR, 0.82 [95% CI, 0.71-0.96]; P = .01; and miR-17-5p: HR, 0.82 [95% CI, 0.70-0.96]; P = .02) and model 2 (miR-20a-5p: HR, 0.86 [95% CI, 0.73-1.00]; P = .047; miR-106b-5p: HR, 0.85 [95% CI, 0.73-0.99]; P = .04; and miR-17-5p: HR, 0.84 [95% CI, 0.72-0.99]; P = .03; Figure 1).
Figure 1. Kaplan-Meier Cumulative Incidence Curves for 3 miRNAs Associated With Heart Failure.
Kaplan-Meier cumulative incidence curves for 3 miRNAs associated with heart failure, after multivariable adjustments (miR-20a: HR, 0.86 [95% CI, 0.73-1.00]; P = .047; miR-17: HR, 0.84 [95% CI, 0.72-0.99]; P = .03; and miR-106b: HR, 0.85 [95% CI, 0.73-0.99]; P = .04). Each hazard ratio is expressed as a function of a 2-fold increase in RNA.
These 3 miRNAs (miR-20a, miR-106b, and miR-17) have a similar sequence and are members of a tight cluster, regulating 883 messenger RNAs in common among all 3 miRNAs, plus another 54 messenger RNAs targeted by 2 of the 3 miRNAs. We found significant homology between these 3 miRNAs in seed sequence. Figure 2 shows the results of an enrichment analysis performed using Enrichr9 with the 937 genes (total) targeted by at least 2 of miR-20a-5p, miR-17-5p, and miR-106b-5p. Several key pathways mechanistically implicated in cardiomyocyte biology in HF were represented in Gene Ontology and WikiPathways10 results by the common target messenger RNAs, including transforming growth factor–β (TGF-β) signaling, growth/cell cycle, apoptosis and other signaling pathways. In addition, an Online Mendelian Inheritance in Man analysis (also performed using Enrichr) demonstrated a disease association between the miRNA gene targets and hypertension (z = 1.95; adjusted P = .01). Data overlays on pathway and network representations of the top TGF-β Signaling Pathway hit from WikiPathways and connections to the Online Mendelian Inheritance in Man enriched hypertension target genes based on GeneMANIA11 were generated using Cytoscape12 (eFigures 1 and 2 in the Supplement).
Figure 2. Gene and Pathway Targets for miR-17, miR-20a, and miR-106b.
A, Gene Ontology (blue) and WikiPathways (light blue) functional enrichment results (Enrichr) based on 937 genes targeted by the 3 miRNAs. Pathway and ontology term titles label each bar, the length of which is the product of the z score and negative base-10 logarithm of the false detection rate–adjusted P value calculated based on accumulative hypergeometric distribution provided by Enrichr. B, Venn diagram of common gene targets. EGF/EGFR indicates epidermal growth factor/epidermal growth factor receptor; ErbB, erythroblastic leukemia viral oncogene; FGF, fibroblast growth factor; MAPK, mitogen-activated protein kinases; TGF-β, transforming growth factor β.
Discussion
We found 12 circulating ex-RNAs in plasma associated with cardiac remodeling, 3 of which were associated with incident HF over a period longer than 7 years in FHS. Three noncoding ex-RNAs associated with cardiac structure and HF (miR-17-5p, miR-20a-5p, and miR-106b-5p) exhibited significant sequence homology and targeted a set of genes involved in processes potentially relevant to cardiomyocyte biology in HF. Each specified a disease association with hypertension, which is a leading cause of HF. Previous animal models of HF support a functional role for the miR-17-92 cluster,13 miR-20a,14 and miR-106b.15
Limitations
These findings are exploratory and limited by lack of external validation, a more precise characterization of subtype of HF (as preserved vs reduced LVEF), and additional biomarker measurements. We focused on echocardiographic endophenotypes that reflect early markers of LV remodeling relevant to HF (eg, LV mass), given their universal acquisition and load-independence; further investigation of more specific serologic or imaging-based markers of earlier alterations in the course of HF (eg, fibrosis) are warranted. Technically, methods for RNA isolation and quantification are variable across different laboratories, and the sequence specificity of our polymerase chain reaction probes is an important limitation: the miR-106b-5p and miR-20a-5p primers may uniquely quantify these miRNAs, while the miR-17-5p primer detected both miR-17-5p and miR-106a-5p.
We did not observe associations between the 3 miRNAs and HF with preserved or reduced LVEF individually, owing to a lack of power for each disease subtype; further studies in these specific, pathophenotypically distinct patient subsets are warranted. Finally, contemporaneous high-sensitivity measurements of troponin or natriuretic peptide were not available at the time of ex-RNA quantification.
Conclusions
In conclusion, we found several noncoding ex-RNAs associated with LV phenotypes and incident HF in a community-based prevention cohort. To our knowledge, no large primary prevention cohorts currently include sufficient follow-up, measurements of noncoding RNAs, echocardiographic phenotypes, and HF events for validation. Further investigations into the relevance of circulating ex-RNAs are warranted to improve detection and therapy in HF.
eMethods. Pathway analysis.
eReferences.
eTable 1. List of circulating extracellular RNAs included in the study.
eTable 2. Age- and sex-adjusted linear models for the association between cardiac structural parameters and extracellular RNAs.
eTable 3. Survival analysis for incident HF, Models 1 and 2.
eFigure 1. Data overlay on pathway representation of top TGF-beta Signaling Pathway hit.
eFigure 2. Data overlay on network representation of top TGF-beta Signaling Pathway hit extended with interactions to hypertension genes.
References
- 1.Freedman JE, Gerstein M, Mick E, et al. Diverse human extracellular RNAs are widely detected in human plasma. Nat Commun. 2016;7:11106. doi: 10.1038/ncomms11106 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Kaess BM, Gona P, Larson MG, et al. Secular trends in echocardiographic left ventricular mass in the community: the Framingham Heart Study. Heart. 2013;99(22):1693-1698. doi: 10.1136/heartjnl-2013-304600 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Devereux RB, Alonso DR, Lutas EM, et al. Echocardiographic assessment of left ventricular hypertrophy: comparison to necropsy findings. Am J Cardiol. 1986;57(6):450-458. doi: 10.1016/0002-9149(86)90771-X [DOI] [PubMed] [Google Scholar]
- 4.Teichholz LE, Kreulen T, Herman MV, Gorlin R. Problems in echocardiographic volume determinations: echocardiographic-angiographic correlations in the presence of absence of asynergy. Am J Cardiol. 1976;37(1):7-11. doi: 10.1016/0002-9149(76)90491-4 [DOI] [PubMed] [Google Scholar]
- 5.McKee PA, Castelli WP, McNamara PM, Kannel WB. The natural history of congestive heart failure: the Framingham study. N Engl J Med. 1971;285(26):1441-1446. doi: 10.1056/NEJM197112232852601 [DOI] [PubMed] [Google Scholar]
- 6.Velagaleti RS, Gona P, Pencina MJ, et al. Left ventricular hypertrophy patterns and incidence of heart failure with preserved versus reduced ejection fraction. Am J Cardiol. 2014;113(1):117-122. doi: 10.1016/j.amjcard.2013.09.028 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Ho JE, Enserro D, Brouwers FP, et al. Predicting heart failure with preserved and reduced ejection fraction: the international collaboration on heart failure subtypes. Circ Heart Fail. 2016;9(6):e003116. doi: 10.1161/CIRCHEARTFAILURE.115.003116 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Shah R, Murthy V, Pacold M, et al. Extracellular RNAs are associated with insulin resistance and metabolic phenotypes. Diabetes Care. 2017;40(4):546-553. doi: 10.2337/dc16-1354 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Kuleshov MV, Jones MR, Rouillard AD, et al. Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic Acids Res. 2016;44(W1):W90-W97. doi: 10.1093/nar/gkw377 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Slenter DN, Kutmon M, Hanspers K, et al. WikiPathways: a multifaceted pathway database bridging metabolomics to other omics research. Nucleic Acids Res. 2018;46(D1):D661-D667. doi: 10.1093/nar/gkx1064 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Montojo J, Zuberi K, Rodriguez H, Bader GD, Morris Q. GeneMANIA: fast gene network construction and function prediction for Cytoscape. F1000Res. 2014;3:153. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Shannon P, Markiel A, Ozier O, et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. 2003;13(11):2498-2504. doi: 10.1101/gr.1239303 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Chen J, Huang ZP, Seok HY, et al. mir-17-92 cluster is required for and sufficient to induce cardiomyocyte proliferation in postnatal and adult hearts. Circ Res. 2013;112(12):1557-1566. doi: 10.1161/CIRCRESAHA.112.300658 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Frank D, Gantenberg J, Boomgaarden I, et al. MicroRNA-20a inhibits stress-induced cardiomyocyte apoptosis involving its novel target Egln3/PHD3. J Mol Cell Cardiol. 2012;52(3):711-717. doi: 10.1016/j.yjmcc.2011.12.001 [DOI] [PubMed] [Google Scholar]
- 15.Deiuliis J, Mihai G, Zhang J, et al. Renin-sensitive microRNAs correlate with atherosclerosis plaque progression. J Hum Hypertens. 2014;28(4):251-258. doi: 10.1038/jhh.2013.97 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods. Pathway analysis.
eReferences.
eTable 1. List of circulating extracellular RNAs included in the study.
eTable 2. Age- and sex-adjusted linear models for the association between cardiac structural parameters and extracellular RNAs.
eTable 3. Survival analysis for incident HF, Models 1 and 2.
eFigure 1. Data overlay on pathway representation of top TGF-beta Signaling Pathway hit.
eFigure 2. Data overlay on network representation of top TGF-beta Signaling Pathway hit extended with interactions to hypertension genes.
