miRNAs, single nucleotide polymorphisms (SNPs) and age-related macular degeneration (AMD)

Introduction

Advanced age-related macular degeneration (AAMD), a complex chronic sight-threatening disease of the neural and vascular retina [1], [2], [3], is the leading cause of irreversible vision loss in elderly people of European ancestry [4]. There are currently an estimated 10.1M people living with AMD – projected global cases are 11.3M, 14.9M, and 18.6M for 2020, 2030, and 2040 [4]. The major clinical subtypes of advanced AMD are geographic atrophy (GA) and neovascular (NV) AMD. GA (a.k.a ‘dry AMD’) is defined by a region of retinal pigment epithelium (RPE) depigmentation or loss ≥175 μm in size, with coexisting degeneration in the choriocapillaris. Photoreceptors enveloped by apical RPE processes are frequently affected in GA. NV AMD (a.k.a. ‘wet AMD’) manifests as abnormal vasoproliferation of nascent vessels that project from the choroidal vascular bed to subretinal and/or sub-RPE regions [5].

AAMD has a strong genetic component [1], [2], [6]. Relative risk of AMD incidence among first-degree relatives is 6–12 fold higher than in the general population [7]. Knowledge on the molecular genetics of AAMD has emerged mainly from analyses of DNA sequence- and structural variation, with large-scale genome-wide association (GWA) studies reporting findings that demonstrate the polygenic and incomplete penetrant nature of many common DNA variants associated with the disease [8], [9]. The International AMD Genomics Consortium (IAGC) Study group examined >12M rare and common DNA variants in 16,144 people with AAMD and 17,832 age-matched AMD-free peers of European ancestry [9] – this effort yielded significant findings on 52 polymorphic AAMD-associated signals from 34 distinct genomic loci. These loci, respectively explained 44% and 53% of variability in the likelihood of having NV AMD and GA. Interestingly, 30 of 34 AAMD-associated loci identified in the IAGC cohort had no association with protein coding regions, yet global profiling of the AAMD-related retinal transcriptome has demonstrated differential expression from over 250 and 500 respective genes within the choroid-RPE (CH-RPE) and neural retinas of people with AAMD [10] GEO:GSE29801. These findings highlight the importance of considering how the actions of regulatory elements in non-coding regions of the AAMD-related genetic architecture may inform us about genetic susceptibility to this disease. This point is relevant for the current report, as micro RNAs (miRNAs) are one form of regulatory element and biomarker with the capacity to interact with mRNA products of disease-associated DNA variants.

miRNAs have been described as the “smallest known carriers of highly specific genetic information” [11] and offer promise as accessible circulating biomarkers for diagnosis and monitoring in health and disease [12]. These evolutionarily conserved single stranded non-coding RNA molecules of ~21–25 nucleotides interact with complementary recognition sequence in the 3′ untranslated regions (3′ UTR) of their target mRNA and alter translational efficiency (mainly via repression) or post-transcriptional stability (via cleavage) [13], [14]. miRNAs usually act to down-regulate production of the target transcript. Conserved miRNA binding sites exist in more than 60% of all human genes [15]. These loci are under strong evolutionary pressure to preserve existing miRNA-mRNA target pairing. Complementarity of miRNAs in target mRNAs can be disrupted by factors that affect: (1) biogenesis and processing to mature (bioactive) miRNA forms; (2) epigenetic factors altering access to consensus sequence in the target mRNA; and (3) sequence and structural DNA variants that have the capacity to alter the structure of consensus binding loci in mRNA. Consequences of single nucleotide polymorphisms (SNPs) or structural variation (deletions, duplications, copy-number variants, insertions, inversions and translocations) in a complementary recognition site may lead to creation, destruction, or alterations in affinity of miRNA-mRNA interactions.

Lukiw et al. present the concept that common immune- and inflammation-based processes act to influence the chronic neurodegenerative and pathophysiologic changes occurring in the progression of both GA and Alzheimer’s disease (AD) [16]. These authors also provide empirical support for the premise that nuclear factor κB (NFkB) has the capacity to transcriptionally regulate pathogenic signaling processes by inducing activity in a number of AMD- and AD-associated miRNAs [16], [17].

Retinal-, vitreous-, and plasma-resident miRNAs have been proposed as biomarkers for advanced age-related macular degeneration (AAMD). In this report we discuss findings examining the potential for AAMD-associated SNPs to influence the explanatory value of these biomarkers. Figure 1 contains information on our conceptual approach. The organization of text in subsequent sections of the manuscript follows this framework. Our central messages are that: (1) miRNAs are regulatory elements with the capacity to influence health and disease of the retina; (2) a number of miRNAs are informative biomarkers for AAMD; (3) the prognostic utility of miRNA biomarkers for AMD may be modified by the profile of DNA sequence variation in regions encoding miRNA-mRNA pairing sites (thus influencing the existence or strength of binding events); and (4) such knowledge is valuable for refining inferences on both the strength of the circulating miRNA biomarkers for disease risk prediction, and the putative role of miRNAs in altering processes acting in AAMD pathogenesis or pathophysiology.

Figure 1:

Conceptual approach to the current project.

The central aim of our efforts was to determine whether AAMD-associated SNPs have the capacity to influence miRNA-messenger RNA (mRNA) pairing and, if so, what would be the impact on the AAMD transcriptome. Such information may guide inferences on the use of miRNA biomarkers for AAMD across genetically diverse populations. The key premise is that presence of SNPs in regions encoding miRNA consensus recognition sites may alter the capacity or strength of miRNA-mRNA binding events in the genome. All analyses described in this, and following sections are based on public-access data that have been de-identified. All existing studies producing these data complied with the Declaration of Helsinki, regarding the ethical conduct of research involving human subjects.

Signature of miRNAs in the context of retinal health and disease: speciation and abundance of miRNAs in the retina and vitreous humor

miRNA expression is tissue and cell type specific [16]; as such, our primary analyses were limited to miRNA species detected in retina and vitreous.

An integrative approach to study of miRNA-SNP relationships in AAMD

We applied an integrative genomics approach to determine whether AAMD-associated SNPs coding for miRNA target sequence have the capacity to influence miRNA-mRNA pairing and, if so, to what extent such pairing may be manifested in a distinct AAMD transcriptome. Such knowledge may be applied in refining inferences on the use of miRNA biomarkers for AAMD across genetically diverse populations. For example, one may have high circulating levels of hsa-miR-146a-5p (hsa-miR-146a-5p is overexpressed AAMD retina and vitreous), yet complementary recognition sequence in the 3′ UTR of target hsa-miR-146a-5p mRNAs in inflammatory genes may bind with substantially lower affinity in some individuals due to variation in their DNA that codes for the mRNA 3′ UTR. Interpretation of the hsa-miR-146a-5p biomarker for AAMD should consider such conditions.

To identify strongest potential DNA [viz. SNP]-mRNA-miRNA-AAMD relationships for inferences on miRNA biomarkers, we started with the set of 12 miRNAs overexpressed in retinal specimens of people with AAMD (Table 1). Lukiw et al. [16] and Bhattacharjee et al. [13] have reported these findings from a global microchip-based profiling effort followed by RT-PCR and/or Northern dot-blot validations [16]. Next, we identified miRNA-mRNA relationships for target sites of these 12 miRNAs. All target sites were either experimentally validated with immunoprecipitation/reporter assays/qPCR (data from miRTarBase 6.1 and TarBase 7) or strongly predicted through nucleotide alignment with TargetScan 7.1 (context++ score <−0.20). Impact of AAMD-related SNPs on miRNA-mRNA pairing was then characterized using the miRanda 3.3 algorithm with the MirSNP database (bioinfo.bjmu.edu.cn/mirsnp/). MirSNP provides genome coverage from dbSNP 135 (NCBI) and miRBase 18 to search for SNPs encoding miRNA target regions. Polymorphic miRNA-mRNA-SNP binding scores obtained from the miRanda algorithm were computed using strict miRNA-mRNA pairing within the seven nucleotide seed region to ensure uninterrupted matches of at least seven nucleotides from the 5′ end of the miRNA. Findings were then prioritized with information on the biophysical characteristics (nature and affinity) of predicted miRNA-mRNA pairings and the consequences of the variant AAMD-associated allele. We used MirSNP and PolymiRTS 3.0 (compbio.uthsc.edu/miRSNP/) for this purpose. SNPs with a miRanda 3.3 pairing score >150 and a free energy for the miRNA-mRNA duplex <−7.0 were accepted as candidates for follow-up. We then examined these potential miRSNPs for residence in a set of 8854 AAMD-associated SNPs significant at p-values ≤5.0E−7; these were culled from total of 12,023,830 tested SNPs in a 33,976-person study on the genetics of AAMD [9]. Integration of these DNA-AAMD findings with the 12 miRNAs from Lukiw et al. [13], [16] and Ménard et al. [19] was applied to make inferences on the role of miRNA-SNP associations and retinal miRNA-mRNA interactions in AAMD pathogenesis. To permit a more comprehensive assessment of miRNA-mRNA pairings we used resources from ENSEMBL (positional coordinates of miRNA binding sites from TarBase 7) to identify untested SNPs overlapping the seven to eight nucleotide seed pairing sites in genes associated with AAMD at p-values ≤5.0E−7.

Genome-wide identification of AAMD related SNP-mRNA-miRNA-AAMD relationships

From the global analysis on 8854 AAMD-associated SNPs, we identified 456 miRNA-mRNA pairings potentially altered by 96 AAMD-associated SNPs also resident in regions of 46 genes encoding mRNA sequence motifs targeted by 322 miRNAs (Figure 2, Supplemental Table S1). Among these findings, notable variants existed in: (1) regulator of complement activation (RCA) genes [rs390679 (CFHR1, p≤2.14E−214), rs390837 (CFHR3, p≤1.1.4E−20)]; (2) genes of the major histocompatibility loci on chromosome 6p21 [rs4151672 (CFB, p≤8.91E−41), rs115404146 (HLA-C, p≤6.32E−12), rs1055821 (HLA-B, p≤1.93E−9) and rs1063355 (HLA-DQB1, p≤6.82E−14)]; (3) the gene for cytokine receptor TGFBR1 (rs868, p≤4.57E−9); and (4) two genes of the 10q26 locus [rs1045216 (PLEKHA1, p≤4.17E−142) and rs2672603 (ARMS2, p≤7.14E−46)]. Roughly one in every three of the AAMD-associated miRSNPs had the capacity to create or destroy an miRNA-mRNA pairing event, one in every four had the capacity to decrease affinity of binding, and one in every six were predicted to enhance affinity (Supplemental Table S1). Supplemental Tables S2 and S3 contain details on analysis and data resources used in the present report.

Figure 2:

Identification of AAMD-associated miRSNPs.

Integration of data on dysregulated retinal and vitreous AAMD miRNAs with the AAMD retinal transcriptome profile and AAMD-related SNPs

Evidence that both AAMD-associated miRNAs and target mRNAs are co-expressed and co-regulated in retinal specimens from the same subjects will strengthen inferences on the clinical significance of these small RNAs. In the present study we did not have access to data linking variation in retinal miRNAs to the retinal transcriptome in the same people. As the best available means to investigate the potential for such miRNA-mRNA target co-regulation we used a two-sample approach to examine strongly predicted targets (TargetScan 7.1 context++ score <−0.20) of the 12 miRNAs in Table 1 for overlap with the differential expression patterns in the AAMD transcriptomic signature. Public access data on the AAMD retinal transcriptome were available from Newman et al. [10] (GEO:GSE29801). The intersection of these miRNA targets was then made with all AAMD-associated SNPs and those potentially encoding complementary recognition sequence (assigned with MirSNP, bioinfo.bjmu.edu.cn/mirsnp/). We used SNPs significant for AAMD at p≤5.0E−7. Figure 3 contains results on the intersection of pertinent miRNA targets with AAMD-associated genes and transcripts, providing links between the miRNA biomarkers in retina, the related AAMD retinal transcriptome, and specific gene variants with the capacity to alter these biomarker-transcript relationships. Supplemental Table S1 has details for linking these genes to specific SNPs. Intersecting features of the Venn diagrams in Figure 3 are useful for prioritizing efforts to rationally examine the expansive set of putative miRNA-mRNA relationships in the context of findings from AAMD-related GWA studies and mechanistic work on the RCA and MHC systems in AAMD.

Figure 3:

Intersection of findings on AAMD-related retinal miRNA- and transcriptome profiles in people with AAMD-SNP associations significant at p≤5.0E−7.

(A) Description and enumeration of potentially affected genes in each dataset used in the present study. Target genes in dataset A were identified with TargetScan 7.1 for 12 miRNAs differentially overexpressed in retinal specimens of people with advanced age-related macular degeneration (AAMD), relative to specimens from age-matched controls (references [13], [16], see Table 1). Dataset B was obtained from AAMD-SNP association tables published by Fritsche et al. (reference [9]). After identifying loci significant at p-values <5.0E−7 we used the MirSNP 2 database to annotate SNPs in regions encoding complementary recognition sequence of miRNAs. Dataset C contains information on the differential transcriptomic profile of retinal specimens from people with AAMD and age-matched AAMD-free controls. Data were accessed from the Gene Expression Omnibus (GEO) project files (PMID 22364233; GEO:GSE29801) (B) Legend for data intersections. (C) Intersections of findings from data on AAMD-related miRNA, -SNP, and -transcriptome signatures. Findings are presented for transcript sets of specimens with neovascular AMD (NV AMD) and geographic atrophy (GA). Table S1 contains details on AAMD-associatedSNPs resident within each gene presented for the A-B and B-C intersections.

Data set A in Figure 3 consists of strongly predicted targets of the 12 AAMD related miRNAs in Table 1. There are 4074 strongly predicted targets of these miRNAs. Data set B consists of the genes containing AAMD-associated SNPs – there are two subsets: (1) genes carrying all SNPs significant with AAMD at p-values ≤5.0E−7 (n=227 genes); and (2) genes carrying AAMD-associated SNPs in DNA sequence predicted by the miRanda 3.3 algorithm to encode miRNA binding sites (n=45 genes). Dataset C consists of information on the transcriptomic signature of the AAMD retina – there are two subsets: (1) genes encoding transcripts down-regulated in NV AMD retinal specimens (n=352 genes); and (2) genes encoding transcripts down-regulated in GA retinal specimens (n=219 genes). Panel A enumerates the findings in each of the three data sets. Panel B provides a legend on the nature of information in the intersecting regions. Panel C provides findings for neovascular AMD and geographic atrophy at the gene level. Details on the intersections of findings appear in the sections directly below.

The A-B dataset intersections (

Findings from the A-B intersection can be used with the AAMD-associated miRSNP set to refine inferences on miRNA-AAMD relationships. The nature of influence from miRNA on target mRNA binding is based partially on the magnitude and direction of differential miRNA expression. In an environment of increased miRNA expression (usually linked to transcriptional repression), carriers of a miRSNP allele producing a break in the miRNA-mRNA bond may be less responsive to the altered volume of the miRNA present in the diseased state. Variant alleles in miRSNPs may destroy, create, decrease affinity of, or enhance affinity of miRNA-mRNA interactions. Information on the consequence of the allele substitution in an AAMD-associated miRSNP (viz. change in free energy) is essential for predicting the effect of responsiveness to factors altering AAMD-associated retinal miRNA concentrations and availability.

The A-B dataset intersections of predicted AAMD-related miRNA targets with AAMD-associated SNPs located outside the boundaries of DNA sequence that encodes miRNA target recognition sites may be used to identify candidate regions in the genome for fine sequencing of 3′ UTR DNA variants in loci encoding complementary recognition sequence of AAMD-associated miRNAs. The 30 A-B dataset genes in Figure 3 Panel C may be analyzed with the seed structure of the 12 miRNAs for mRNA nucleotide alignment and then annotated for SNPs or structural variants. Primary candidates for follow-up are hsa-miR-146a-5p and hsa-miR-155-5p.

The A-C dataset intersections (

The B-C dataset intersections (

Linking miRNAs to AAMD pathogenesis and pathophysiology: AAMD-associated miRNAs and SNPs in MHC genes

It is important to consider the biologic plausibility of the miRNA biomarkers in the context of knowledge on the etiopathology of AAMD. Dysregulated immune response and inflammation have been implicated in AAMD pathogenesis and pathophysiology [20]. In the two sections that follow we discuss SNP-mRNA-miRNA-AAMD relationships and the potential role of AAMD-associated miRNAs in regulation of the RCA complex and MHC genes.

Table 2 contains information on the number of AAMD-associated SNPs also overlapping DNA encoding predicted complementary recognition sequence of miRNAs in RCA- (1q31) and MHC- (6p21) constituent genes. Genes of these regions have been consistently replicated as AAMD-related loci. Also present in this table are findings in the AAMD-associated locus in 10q26. Table 3 contains information on AAMD-associated SNPs resident in DNA sequence encoding mRNA complementary recognition sequence for four of the 12 miRNAs overexpressed in retinal specimens of people with AAMD (hsa-miR-155-5p, hsa-let-7a-5p, hsa-let-7b-5p, hsa-let-7d-5p) and one miRNA underexpressed in vitreous humor specimens of people with AAMD (hsa-miR-152).

Jorgensen et al. [21] present an overview on the biologic plausibility of MHC relationships in AMD pathogenesis, suggesting that progression to AAMD may be the result of dysregulated immune and inflammatory response to RPE damage associated with alterations in MHC and the complement systems (also discussed in Lukiw et al. [16]). These authors cite evidence indicating increased HLA class II immunoreactivity exists in retinal specimens of people with AMD [22] and that HLA class II antigens are constituents of drusen [23]. Drusen are lipid-protein deposits that accumulate under the neural retina between Bruch’s membrane and the RPE. Although not clearly implicated as an etiologic factor in AMD, drusen size, area, and distribution in the macula are useful prognostic factors for progression to AAMD [24], [25]. Details on AMD pathogenesis and pathophysiology exist elsewhere [26], [27], [28]. Table 3 highlights miRNA-SNP-AAMD relationships in genes of the MHC. Findings in Table 3 strengthen inferences on this relationship, as AAMD-associated loci in targets hsa-miR-155-5p (rs1063355, p≤6.82E−14) and hsa-miR-152 (rs9273455, p≤3.18E−12 | rs1770, p≤6.64E−11) exist – both of these miRNAs have capacity to pair with target binding motifs in the 3′ UTR of HLA-DQB1 mRNA. HLA-DQB1 encodes a module of the MHC class II DQ cell surface receptor that acts as an essential macrophage/dendritic cell antigen presentation site for peptides originating from extracellular proteins. The DQB1 complexes with DQA1 to produce the DQαβ heterodimer demonstrated to trigger immune responses via presentation foreign peptides to CD4(+) T cells. Figure 4 contains information on HLA-DQB1 locus SNPs plotted for association with AAMD. In support of these findings, HLA-DQB1 has been strongly implicated in AMD pathogenesis through a recent HLA fine-mapping study in >25,000 people [21] – this work extended findings from two reports published by Goverdhan et al. nearly a decade ago [30], [31]. The key point of this section is that we have identified two miRNA biomarkers (one in retina and one in vitreous) of AAMD specimens that target mRNAs of genes in AAMD-associated HLA-DQB1 in the AAMD-associated MHC complex. HLA-DQB1 contains three AAMD-associated SNPs in regions coding for the pertinent miRNA targets. As such, information about the abundance of these miRNAs may be used with that on the three HLA-DQB1 SNPs cited above to refine inferences on the explanatory power of miRNA biomarkers – two HLA-DQB1 SNPs are predicted to reduce miRNA-mRNA binding affinity and one is predicted to create strong binding site.

Figure 4:

3′ UTR SNPs encoding miRNA targets of HLA-DQB1.

Association results for HLA-DBQB1 relationships with advanced age-related macular degeneration in people of European ancestry [9]. Labeled SNPs have the capacity to encode mRNA complementary recognition sequence for the miRNAs indicated above the SNP identifiers. Figure generated with LocusZoom [29].

Finding a HLA-DQB1 (rs1063355)-hsa-miR-155-5p relationship with AAMD was intriguing both because of the extended support it provides for the MHC-AAMD link and also the fact that hsa-miR-155-5p is an NFkB-inducible miRNA previously identified by Lukiw et al. [13], [16] to share a miRNA regulatory control region with hsa-miR-146a-5p (also NFkB-inducible and elevated in AAMD) in the CFH mRNA 3′ UTR. The implications of this these latter points for AAMD biomarker research are discussed in the next section.

miRNAs differentially expressed in AAMD, RCA genes, and SNPs with the capacity to alter binding events in transcription

Our strongest finding involved an AAMD-associated SNP resident in HLA-DQB1 that has the capacity to produce complementary binding sequence for hsa-miR-155-5p, a miRNA overexpressed in retinal specimens of people with AAMD (Table 3). In addition to a potential interaction with this MHC II target, potential regulatory control of the immune response in AAMD by hsa-miR-155-5p miRNA is linked to the RCA cluster genes on 1q31-q32 through co-residence in an miRNA regulatory control region in the CFH mRNA 3′ UTR.

Products of the RCA cluster genes are thought to operate in AAMD as negative regulators of complement activation [20], [32]. Here we focus on the three 1q31-q32 RCA members with AAMD-associated SNPs predicted to reside in DNA tracts with capacity to encode targets of miRNAs. CFH, CFHR1, and CFHR3 all encode fluid phase regulators of the alternative complement system. CFH is a soluble hydrophilic glycosylated 155 kDa protein that, in systemic deficiency states, is associated with over-activation of complement proteins that drive the immune response – in some cases CFH deficits result in a chronic sustenance of pro-inflammatory signaling. CFH is circulating In blood and is detectable in serum and plasma[33] – it acts in an inhibitory capacity to downregulate the processes of complement component 3 (C3) activation and amplification (blocking the C3b transition to C3Bb). CFHR1 and CFHR3 are able to function independently of CFH. The CFHR1 protein blocks formation of C5 convertase (C3bBb3b) thereby preventing the transition from C5 to C5b. CFHR3 acts with CFI in degradation of C3b – this event reduces the amount of C3Bb produced. CFHR3 also inhibits the C5 convertase in the transition step described directly above. We have demonstrated that AAMD-associated RCA gene loci in CFHR1 and CFHR3 contain putative miRSNPs (Table 2, Supplemental Table S1). While these findings support the role of AAMD-associated RCA gene regulation by miRNAs, none of the four AAMD-associated SNPs in CFHR1 or CFHR3 also have the capacity to encode recognition sequence for any of the 12 miRNAs overexpressed in AAMD retina (Table 1). We have constrained our commentary to the connection of AAMD-associated SNPs with targets of the 12 miRNAs dysregulated in AAMD retina – viz. both the miRNA and DNA sequence profiles in people with AAMD are distinct from their AAMD-free peers. In cases where only the DNA sequence profile varies by disease status, we may also see AAMD-specific miRNA-SNP associations as a result of SNP-associated alterations in mRNA volume; this matter is a topic of ongoing research in our laboratory.

CFH polymorphisms encoding mRNA seed region targets for hsa-miR-155-5p and hsa-miR-146a-5p loci in the 232 nucleotide CFH 3′ UTR have yet to be tested for association with AAMD. In the set of 12 million probes to which we had access, no SNPs were in near-complete linkage disequilibrium with three published variants shown in ENSEMBL 85 to be resident in the overlapping hsa-miR-155-5p/hsa-miR-146a-5p miRNA regulatory control region (5′–TTTAGTATTAA–3′) in CFH 3′UTR. Lukiw et al. [16] provide a detailed description of this region. As part of the present work we examined the most current genome build for SNPs in this region. There are seven SNPs published in ENSEMBL 85 (GRCh37.p13) for this tract of DNA – among these, two (rs459598 and rs766666504) are recognized by the hsa-miR-146a-5p seed sequence – both SNPs are resident in the area of the miRNA control region that does not overlap with the hsa-miR-155-5p target. These are our most promising candidates for testing DNA-miRNA-AAMD relationships and should be prioritized due to their positions in miRNA seed regions. Figure 5 contains a nucleotide map of CFH 3′UTR DNA and mRNA with details on the positions of SNPs and putative miRNA binding sites. The central messages of this section are that hsa-miR-146a-5p is dysregulated in retina, vitreous, and plasma specimens of people with AAMD – and that untested SNPs with the capacity to affect hsa-miR-146a-5p-mRNA pairing exist in the AAMD-associated CFH gene.

Figure 5:

miRNA regulatory control region in CFH 3′ UTR and resident single nucleotide polymorphisms (SNPs) linked to complementary recognition sequence of hsa-miR-146a-5p, -155-5p and 125b-5p.

(A) DNA sequence in CFH 3′ UTR with highlighted regions capable of perturbing consequent miRNA pairings in CFH mRNA. (B) Putative miRNA-mRNA pairing in CFH mRNA 3′ UTR. There are two polymorphic sites in the seed region of hsa-miR-146a-5p – these are consequent to presence of rs766666504 and rs459598. SNPs were identified with ENSEMBL GRCh37.p13. Sequence in miRNA seed regions is underlined. Nucleotides formatted in red are polymorphic.

Common characteristics of miRNA biomarkers for AAMD and AAMD pathogenesis: NFkB-inducible AAMD-associated miRNAs that target high-affinity binding sites in the CFH 3′UTR

AAMD is thought to develop, in part, through chronic inflammatory- and immune-mediated processes – here we expand discussion from the immediately preceding section to the common properties of three retinal miRNA biomarkers of AMD with targets in CFH. Three of the 12 miRNAs elevated in AAMD retinal specimens (hsa-miR-125b-5p, hsa-miR-146a-5p and hsa-miR-155-5p) show highly sensitive binding sites in the CFH mRNA 3′ UTR. The evolutionarily conserved miRNA regulatory control region in the CFH 3′UTR contains overlapping domains for hsa-miR-155-5p and hsa-miR-146a-5p miRNA-mRNA high-affinity pairing sites that have been validated with a reporter gene expression assay. We consider this locus as a high priority region of interest to investigate miRNA-based regulatory control of processes that may lead to a pathologic state of CFH deficiency in the AAMD disease profile – which is widely believed to be an innate immune system response and subsequent chronic inflammatory degeneration of the neural and vascular retina. Lukiw et al. propose that hsa-miR-155-5p and hsa-miR-146a-5ps may work in a redundant manner to control expression of CFH in its key role as a major regulator of innate immune and chronic inflammatory response [16]. hsa-miR-155-5p and hsa-miR-146a-5p may influence CFH transcription in concert, or individually, and the pattern of DNA sequence variation in this regulatory control region carried by an individual would presumably lead to specific response profiles after exposure to the miRNAs. Along these lines, we must ask what is known about the common underlying biology of upstream regulators for these two miRNAs. hsa-miR-155-5p, hsa-miR-146a-5p and two other AAMD-associated retinal miRNAs from Table 1 (hsa-miR-125b and hsa-miR-9) are inducible and regulated by the pro-inflammatory transcriptional regulator, NFkB. It is important to note that a rigorous reporter assay has been used to confirm that the gene encoding hsa-miR-146a-5p (MIR146A, 5p34) contains three tandem canonical NFkB binding sites in DNA sequence of the pre-miRNA-146a promoter region [34]. We examined these regions in the MIR146A promoter for AAMD-associated SNPs, and while there were none on the testing panel with with p-values ≤5.0E−7, at least three lines of additional evidence that supported the importance of NFkB with miRNA-driven processes in AAMD; these are: (1) our pathway analysis showed a corrected p-value for enrichment in the NFkB signaling pathway of 0.028, and indicated that all 12 miRNAs elevated in AAMD retina also had predicted targets in at least one constituent gene of this pathway; (2) a SNP in NFKB1, the gene encoding nuclear factor kappa B subunit 1 of the NFkB protein complex, is associated with AAMD at a p-value ≤8.59E−6 (rs1005819); and (3) another AAMD-associated SNP in NFKB1 (rs1599961, p-value ≤2.94E−5) showed strong effects on chromatin state and histone modifications in the gene. Also, Kutty et al. demonstrated a retinal cell specific expression of hsa-miR-146a-5p in human primary cultures of RPE cells – in this model system expression of hsa-miR-146a-5p was strongly induced by the pro-inflammatory cytokine IL-1ß and the authors suggest that hsa-miR-146a-5p may subsequently negatively regulate the NFkB pathway via IRAK1 [35]. The key message of this section is that there is a common underlying biology linking a number of NFkB-inducible retinal miRNA biomarkers to immunomodulatory processes implicated in AAMD development.

Conclusions

miRNAs are genomic regulatory elements with the capacity to affect AAMD pathogenesis and pathophysiology. miRNAs have been proposed as biomarkers for AAMD. Our findings suggest that the prognostic utility of miRNA biomarkers for AMD may be modified by the profile of DNA sequence variation in genomic regions encoding miRNA-mRNA pairing sites, and that knowledge on the potential for such variants to alter miRNA target existence or binding strength is valuable for refining inferences on the potential influence of miRNAs biomarker in AAMD. A more sophisticated knowledge on the actions of regulatory elements affecting the genetic architecture of AAMD will guide us farther in identifying biomarkers with prognostic utility, as well as developing preventive and therapeutic interventions for this sight-threatening disease. We applied a rational approach to integrate robust data on AAMD-associated retinal miRNA, -SNP, and -retinal transcriptomic profiles for the purpose of making inferences on the capacity of SNPs to influence miRNA-mRNA pairings in loci implicated in pathogenesis and pathophysiology of the disease. The foundation of this approach was a set of 12 miRNAs empirically determined to be differentially elevated in retinal and vitreous specimens of people with AAMD. Convergent evidence on AAMD-associated polymorphisms in DNA encoding targets of these miRNAs, and decreased transcript levels of these targets in AAMD retina, led to identification of multiple associations in the RCA and MHC systems. These findings reinforce the idea that miRNAs elevated in retinas of people with AAMD may affect processes implicated in the control of immune response and inflammation. The present work emphasizes strength of evidence on the roles of hsa-miR-152 and hsa-miR-155-5p in the MHC system and hsa-miR-155-5p & hsa-miR-146a-5p in the RCA system. We believe that hsa-miR-146a-5p may have great relevance as biomarker for AAMD, as: (1) in specimens from people without eye disease this miRNA shows a nearly 100-fold enrichment in the choroid-RPE, relative to its expression the neural retina[18]; (2) in AAMD retinal specimens there is an overexpression of this miRNA by 2.1- to 6.3-fold, in comparison to retina from AAMD-free age-matched controls [13]; (3) levels of this miRNA in vitreous humor are >5-fold higher than in age-matched AMD-free controls [19]; (4) plasma concentrations are 2.5-fold higher than in age-matched healthy controls [19]; (5) a conserved, high-affinity, polymorphic seed pairing site for hsa-miR-146a-5p exists in the 3′UTR of CFH, the most strongly and consistently AAMD-associated gene.

For the purposes of guiding research to refine inference on miRNA biomarkers for AAMD, we have used information on the intersection of findings from AAMD-associated miRNAs, SNPs, and retinal transcripts to generate sets of high-quality miRNA-associated genomic loci with the capacity to disrupt miRNA-mRNA pairing. It is our intention that such work may increase explanatory power of miRNA biomarkers in genetically diverse populations while yielding information to develop site-specific small molecules (synthetic mimetics or anti-miRNAs) with preventive or therapeutic efficacy for AAMD.