Circulating microRNAs as Biomarkers of Alzheimer’s Disease: A Systematic Review

INTRODUCTION

Approximately 35.6 million people worldwide have dementia, of which 60% is due to Alzheimer’sdisease (AD), with numbers expected to double every 20 years to 65.7 million in 2030 and 115.4 million in 2050 [1, 2]. Although AD is one of the fastest growing causes of major disease burden, currently there are no effective treatments to alter the natural history of this disease. Pathological changes in AD brains are known to begin at least two decades before the estimated onset of clinical symptoms [3, 4]. Barriers to the development of effective therapy include the lack of a specific biomarker for the early identification of the disease prior to the onset of clinically apparent cognitive impairment. The massive neuronal death, characteristic of late stages of neurodegenerativediseases, makes treatment late in the course of the disease extremely difficult [5].

The ideal biomarker would identify the neurodegenerative process before cognitive decline has begun, and be specific, predictive, and easy to measure. One of the most robust AD biomarkers to date is the combined decrease in amyloid β 42-residue (Aβ₄₂) and increase in phosphorylated tau in the cerebrospinal fluid (CSF) [6], but the invasiveness of lumbar punctures limits their clinical use, especially for serial measurements. In recent years, microRNAs (miRNA) have been recognized as novel biomarkers for diseases because of their diverse but tissue and cell specific biological and pathological functions [7]. MiRNAs are a class of non-coding RNA of approximately 22 nucleotides in length, and are known in general to downregulate protein expression post-transcriptionally by binding to complementary sites of the 3’ untranslated region (UTRs) of their target messenger-RNAs (mRNA), resulting in mRNA cleavage or inhibition of translation [8, 9]. MiRNAs can be released into the extracellular environment by binding to RNA-binding proteins or transported through body fluids in vesicles such as exosomes and microvesicles [10]. Circulating miRNAs are considered to be reproducible and consistent among individuals of the same species and exhibit altered expression under certain disease states [11]. They are desirable candidate biomarkers for diseases because they remain stable over time, even after repeated freeze-thaw cycles in plasma and serum [12, 13].

With respect to AD, it has been hypothesized that downregulated miRNAs may lead to pathological upregulation of AD-relevant genes such as amyloid precursor protein (APP), presenilin-1 and presenilin-2 (PSEN1 & 2) [6]. Many studies have identified miRNA expression changes in AD brain and CSF, yielding putative biomarkers and insights into disease pathways [14, 15].

Peripheral blood miRNAs are ideal biomarker candidates as they are easily accessible, non-invasive, and cost-effective, and hence there are a growing number of studies exploring the role of miRNAs in AD. Earlier reviews have evaluated prior studies examining the role of miRNAs in AD [15–20] and in general recognize that the development of miRNA biomarkers of AD is in its early stages, which require further validation [16]. We present here one of the first systematic reviews specifically examining circulating peripheral blood miRNAs in mild cognitive impairment (MCI) and AD using the latest data. The primary endpoints of interest in this review were (i) miRNAs that are differentially expressed and (ii) their sensitivity and specificity in differentiating AD patients from normal controls. After reviewing the current literature, we also propose a research agenda to progress knowledge in this field.

METHODS

RESULTS

The literature search yielded 989 publications that were subjected to title and abstract review. Studiesthat were duplicates, of irrelevant topics, animal studies, or conducted on non-blood human samples were excluded. Ninety-two publications with potential relevancy were subjected to full text reviews. One study was excluded because the control group included patients with neurological conditions [22]. Eighteen studies fulfilled the eligibility criteria (Fig. 1).

DISCUSSION

The search for an effective biomarker for AD is an active area of research with many recently published studies examining the use of blood miRNAs as potential biomarkers for MCI and AD. The results from the present review suggest that miRNAs can differentiate patients with AD or MCI from a healthy comparison group with modest to high accuracy. However, in general, the results have not been replicated in independent cohorts or have shown inconsistent findings across studies.

Of the reviewed studies, the most consistently dysregulated miRNA markers for AD include miR-29b, miR-181c miR-15b, miR-146a, and miR-107, where consistent results have been observed in at least two independent studies. Indeed, several of these miRNAs have also been significantly associated with AD using miRNAs derived from brain and CSF samples (mir29b, miR181c, miR15b) (see Table 4). For MCI, there are fewer studies reported and hence fewer consistent results. Higher expression of miR-132 and miR-107 was consistently reported in MCI cases in at least two studies. In general, these consistent results were observed across different blood sample types such as serum and plasma (see Table 4).

When reported, the majority of miRNA panels as well as individual miRNAs yielded sensitivities and specificities for AD greater than 0.75. As a comparison, the diagnostic accuracy for AD using Pittsburgh Compound-B positron emission tomography (PET), which measures Aβ deposits in the brain, has a sensitivity of 83% –100% and specificity of 46% –88% [43]. Similarly, combined CSF Aβ₄₂ and tau measurements have a sensitivity of 93.5% and specificity of 82.7% in diagnosing AD [44]. Although these tests show good diagnostic and prognostic accuracy, the high costs of PET and invasive nature of CSF collection currently preclude their utility in clinical practice. Consequently, measurement of miRNAs in peripheral blood has potential to be both a cost-effective and non-invasive biomarker for AD.

Although miRNAs appear to be able to differentiate patients with AD/MCI from healthy controls, their ability to differentiate from patients with otherpathologies is less clear. One study showed that a 12-miRNA panel could differentiate AD from cognitively normal controls with 93% accuracy, whereas the accuracy was between 74% and 78% for differentiating AD from other central nervous system illnesses such as multiple sclerosis, Parkinson’s disease, and major depression [41]. Therefore, future studies are needed to validate results in different disease populations. AD patients, especially due to their age, will often have diseases of multiple organs, and an effective biomarker will need to be accurate despite the presence of co-morbid pathologies.

Of the reviewed studies, there was only one longitudinal study, with a short follow-up period (2–5 years) [34]. In this study, 10 out of 19 participants with normal cognitive function progressed to MCI, and three biomarker pairs (miR128/miR-491-5p, miR-132/miR-491-5p, and miR-874/miR-491-5p) were able to identify seven of the 10 subjects who progressed to MCI at the asymptomatic disease stage, preceding their MCI diagnosis by six to 61 months. Among the nine patients who remained MCI free, none were classified as disease-positive by the miRNAs. Further studies with longer duration of follow-up including predementia syndromes such as MCI are needed to determine the prognostic accuracy of miRNAs and how they change with normal aging, disease severity, and progression.

The dysregulated miRNAs identified by these studies may provide further insight into the pathobiology of AD given the large number of downstream targets of miRNAs. Notably, a number of miRNAs were found to have discordant results in different studies in this review. For example, miR-9 has been shown to be both up- and downregulated in blood [30, 36], CSF [45, 46], and brain [14, 47–50]. It is postulated that these discrepancies may be due to the different predicted targets of miR-9 and its roles in the pathological development of the disease. Downregulation of miR-9 increases BACE1 expression activity and increases Aβ₄₂ production [48]. MiR-9 is also involved in the AD inflammatory and oxidative stress pathways, and may be upregulated in response to interleukin 1β [50].

The current studies of miRNAs in AD are limited by differences in methodology and sample heterogeneity. Circulating miRNA levels may be affected by various intrinsic and extrinsic factors. In addition to donor-related factors such as genetic variation, race, gender, inflammatory status, and lifestyle factors [51], differences in methodology, including varied sample types (serum, plasma, whole blood, exosomes), processing techniques and methods for miRNA extraction, quantification, and normalization make it difficult to compare and contrast studies. It has been shown that miRNA concentrations in serum samples and plasma samples differ within the same individual [52]. In a more recent study, miRNAs were compared across whole blood, plasma, serum, and exosome samples from three age-matched young adults, which showed miRNA differences in their profiles and concentrations [53]. In addition, there is currently no commonly accepted single normalization method of measuring miRNA concentrations. Studies have normalized their qRT-PCR reactions using synthetic miRNAs such as Caenorhabditis elegans miRNA (cel-miR-39) or various ubiquitous and least variable circulating miRNAs such as miR-16, miR-191, miR-126, and miR-159a.

How data is analyzed will also influence which miRNAs are considered to be of statistical significance. For example, using a discovery approach, Kumar et al. chose candidate miRNAs for replication which had at least a 1.5 fold difference between average AD/MCI and normal control samples [9]. In comparison, Tan et al. chose candidate miRNAs with at least a 2-fold change in probable AD patients versus the controls [39]. The impact on results due to differences in methodology is further exemplified by a study [54], which examined the original miRNA-sequencing dataset from the study by Leidinger et al. [41] using different software tools for data processing and statistical analysis. They found that only two (miR-151a-3p and let-7f-5p) of the twelve miRNAs in the original study were identified as significant. These differences highlight the need for a standardized and commonly accepted miRNA measurement platform, including uniform methods for sample preparation, analysis of circulating miRNAs, and statistical approaches, so that cross-study comparisons and data validation can be performed. Without standardized study designs, there is a risk of amassing data with limited research and clinical utility.

Another limitation of current studies is the heterogeneity of study samples, with phenotypic variability in disease stage and severity. The current standard for a definitive AD diagnosis is brain postmortem examination. The reference standard for AD used in studies in this review was based on the NINCDS-ADRDAcriteria [24] to define ‘probable’ AD, which has high sensitivity (91% to 98% ), but poor specificity (40% to 61% ) compared to postmortem diagnosis [55]. Studies included in this review had no postmortem confirmation of diagnosis, therefore we cannot confidently ascertain if changes in miRNA expression were indeed due to AD pathology. There are considerable logistical difficulties in conducting miRNAs studies in biological samples such as plasma or CSF during life and then confirming the AD diagnosis at autopsy. Therefore, the use of other biomarkers such as amyloid PET imaging and CSF Aβ₄₂ and tau measurements in conjunction with miRNA analysis may strengthen the diagnosis. Furthermore, the cognitively normal control groups in the reviewed studies were selected based on normal clinical cognitive testing, but these patients may have subclinical incipient AD pathology. No follow-up was performed to ensure the cognitively normal controls did not develop clinical cognitive impairment. Future AD biomarker studies will therefore need to have their cohorts strictly phenotyped with stringent follow-up.

DISCLOSURE STATEMENT

Authors’ disclosures available online (http://j-alz.com/manuscript-disclosures/15-0619r2).