HFpEF vs. HFrEF: can microRNAs advance the diagnosis?

This editorial refers to 'Circulating microRNAs in heart failure with reduced and preserved left ventricular ejection fraction'†, by L. Wong et al. and 'MicroRNA signatures differentiate preserved from reduced ejection fraction heart failure'‡, by C.J. Watson et al., published in this issue, European Journal of Heart Failure (2015); 17: 393–415. Although heart failure with preserved ejection fraction (HFpEF) and heart failure with reduced ejection fraction (HFrEF) have similar morbidity and mortality risks, both types of heart failure (HF) require specific disease management due to their distinctly different pathophysiology. In addition, the diagnosis of HFpEF requires often complex additional evidence for a functional or structural cardiac abnormality, together with a preserved LVEF. This makes the diagnosis of HFpEF complex. Therefore, tools to improve the diagnosis and prognostication of HFpEF are of great interest and are key towards targeted treatment.1, 2 MicroRNAs (miRNAs) are conserved small non-coding RNA molecules (21–25 nucleotides in length) that regulate gene expression by base-pairing to the complementary mRNA sequences resulting in translational repression or mRNA degradation3 (Figure 1). MiRNAs are implicated in cardiac development and have been linked to the pathogenesis of a plethora of diseases and conditions, including HF.4, 5 The exact downstream molecular mechanisms by which miRNAs could influence disease development and progression are poorly understood. The current and ongoing research efforts suggest added value of miRNAs as diagnostic and prognostic tools for HF.6-8 Wong et al.9 and Watson et al.10 studied the role of miRNAs to improve the diagnosis of HF and to differentiate between HFpEF and HFrEF. Wong et al. identified 12 miRNAs that significantly differed between HF and non-HF controls. Four of these (miR-125a-5p, miR-550a-5p, miR-190a, and miR-638) could differentiate patients with HFpEF vs. HFrEF. The area under the curve (AUC) values of individual miRNAs had a lower discriminative power in HFpEF vs. HFrEF than NT-proBNP, but the miRNA panel in combination with NT-proBNP achieved maximal diagnostic accuracy (AUC 1.0; Table 1). Similarly, Watson et al. discovered five circulating miRNA candidates (miR-375, miR-146a, miR-30c, miR-328, and miR-221) all with reduced levels in HF patients and able to distinguish HFpEF from HFrEF with a diagnostic potential that was superior for miR-375 and two miRNA combinations compared with log(BNP). Again, the highest intergroup distinction was achieved by combining the miRNA panel with log(BNP) (AUC 0.86; Table 1). Both studies align in presenting promising data supporting added value of miRNAs in diagnosing HF and in distinguishing HFpEF from HFrEF. Interestingly, except for miR-221, there was no overlap in the identified miRNAs between the two studies. However, while Watson et al. found miR-221 to differentiate HFpEF from HFrEF, miR-221 was only able to distinguish between HF and control patients in the study of Wong et al. In addition, the two current studies did not show any overlap with the miRNAs identified but not validated in the earlier work of Ellis et al. (Table 1).11 The lack of overlap is intriguing. There are several potential explanations. First, there were major variances in methodology. Differences in sample collection and preparation may influence the results. Wong et al. performed miRNA profiling in whole blood and corresponding plasma samples but did not determine which sample type is more appropriate for circulating miRNA detection and quantification. Each study used different methods for detecting miRNAs. Microarrays have been widely and extensively used as an efficient method for circulating miRNA profiling. Although considered less sensitive than quantitative reverse transcription–PCR (qPCR) quantification, microarray analysis is an essential tool for the discovery phase, while qPCR is the gold standard for miRNA quantification and validation of microarray data. An important difference of the Watson study is the analysis of pooled samples with only one array readout per cohort during the discovery phase. Differential levels of some miRNAs may not have been detected by screening a pool, and outliers could impact the generated data. Quantitative PCR validation in an independent cohort would have strengthened the findings. Furthermore, the low amount of total RNA in blood demands a precise normalization of the detected miRNAs. Any pre-analytical variations such as sample storage, degree of haemolysis, and extraction efficiency could affect miRNA normalization and quantification. Additional methods could be employed, including spiking-in oligonucleotides, standardization against stably detectable miRNAs, and/or the mean of all measured miRNAs to validate the results further. Secondly, more large-scale studies with well-defined control, external validation cohorts, and cohorts with non-cardiac conditions, e.g. including, for example, dyspnoeic/breathless COPD patients, are required in order to understand these findings further and to limit the noise caused by different HF aetiologies, concomitant diseases, and treatments. The results of both studies presented in this issue require further external validation in larger independent cohorts to assess the utility of the discovered circulating miRNA biomarker candidates. Finally, when interpreting the data, confounders such as medication and common cardiovascular risk factors, e.g. diabetes, hypertension, and body mass index, must be considered. With our current knowledge about circulating miRNAs and the variability of the available research results in HF, how can we clinically use miRNAs in the future? With the advances in technology it soon will become feasible to measure a large number of biomarkers including miRNAs simultaneously in blood samples at relatively low costs. The development of next- and third-generation sequencing technologies offers an increasingly efficient alternative for the discovery of novel or low-expressed miRNAs and will overcome the limitations of array- and PCR-based technologies.12 In line with advances in technology, further 'harmonization' of data and analyses, including the isolation and handling of clinical samples, normalization methods, and analytical standards, will improve interstudy comparability. This will also allow cross-referencing and large-scale global collaborations in a variety of cohorts. To support the exciting hypothesis of a causal role of miRNAs, Wong et al. performed a pathway analysis based on the identified HFpEF- and HFrEF-specific miRNAs. Differences in miRNA levels could be linked to several pathways of potential relevance in the HF disease process such as neurotrophin and Wnt signalling, as well as p53 signalling. The identification of regulatory and disease-specific miRNAs that not only mark but also have a causal association with HF might provide a foundation for targeted therapeutic approaches and personalized medicine (Figure 1). In theory, miRNAs can be targeted in vivo by using synthetic molecules (anti-miRs and miRNA mimics) that result in changes of expression of the mRNA and hence influence entire signalling pathways.13 Currently, the field of miRNA research in HF is only in its infancy, but it is surrounded by exciting data and novel hypotheses that might, eventually, lead to improvements in the diagnosis, prognostication, and personalized treatment of HF. We acknowledge the support from the Netherlands CardioVascular Research Initiative: the Dutch Heart Foundation, Dutch Federation of University Medical Centres, the Netherlands Organisation for Health Research and Development, and the Royal Netherlands Academy of Sciences. This study was supported by a Grant from the Dutch Heart Foundation: Approaching Heart Failure By Translational Research of RNA Mechanisms (ARENA). Conflict of interest: none declared.