Three main types of cells are present in the retina: photoreceptor cells (RPE, cone, and rod), bipolar cells, and ganglion cells. Although each region of the retina shares all cell types, the regional composition of these cells varies considerably. The macula is the most color-sensitive part of the retina and has the highest density of optic cone cells; however, it lacks optic rod cells and bipolar cells [13]. Gene expression in the retina was also region-specific[14]. The regional distinctions in retinal cell types and gene expression may explain the region-specific preferential ocular pathologies of the retina. Whitmore SS et al. [1] investigated expressional profiles in the nasal, macular and temporal regions of the RPE-choroid complex and retina. They found significantly different gene expression between the macula and extra-macular (nasal or temporal) tissues, while no apparent difference was observed in nasal and temporal regions. Subsequently, a high-throughput sequencing study involving 537 samples yielded similar results[15]. Additionally, the expression of multiple candidate genes for AMD differed between the macula and non-macular tissues. For example, Prolactin, which exhibits anti-angiogenic effects in the retina[16], had a reduced impact in the macular nerve retina compared to the peripheral region. Iron dyshomeostasis has been implicated in the development of AMD. Voigt AP et al. reported that transferrin (TF) has a peripheral enrichment in the retina[17]. Very few differentially expressed genes were screened out in the previous study, involving all layers of retinal tissue. Hunter et al. identified 46 different expression genes associated with DNA methylation with the criterion of P-value < 0.1 and log2FC > 0.25 in healthy and AMD patient’s retina tissues[3]. No significant variation was observed for ARMS2 and HTRA1 in the human retina in Kanda et al. study[3]. Since gene expression in the macular and extramacular regions of retinal tissue is inherently very different, many AMD-associated differential expressed genes would likely be missed by treating them as a whole.
In our study, we investigated the genetic differences between the macular and extramacular regions of the retina in patients with AMD. A total of 4509 different expression genes were screened out. The involvement of pathological apoptosis in AMD has been widely reported[18, 19].GO enrichment analysis indicated that regardless of up-regulated and down-regulated genes, DE-mRNAs from the macular region unanimously enriched the apoptotic signaling pathway. Furthermore, DE-lncRNAs within the macula were enriched in several diseases, such as Alzheimer’s disease, cardiovascular disease, and diabetic retinopathy, which had been confirmed to share the overlapping pathophysiology with AMD[20].
AMD is an aging disorder. Its typical symptoms include the formation of drusen, global atrophy, and proliferation of new blood vessels, which usually emerge in the middle age of patients. These collective symptoms indicate the critical role of the acquired environment. Given that only a few studies have identified the specific different expression genes in AMD and control tissues, researchers have recently turned the spotlight on epigenetic modification. The epigenetic modification could modulate the gene expression pattern [21, 22], with the entire transcriptional sequence having no change based on the acquired environment. The non-coding RNAs are one model of epigenetic modification. Only a small proportion of the human transcribed genome could encode a protein called messenger RNA (mRNA). Therefore, most transcripts cannot be translated into proteins which are defined as ncRNAs. The ncRNAs contain various lncRNAs and miRNAs with a length of more than 200nt[23] and 22nt[24], respectively. miRNA could combine with the 3’UTR of target mRNA specifically and thus suppress the translation of mRNA at the post-transcriptional level [25]. Therefore, miRNAs get significantly altered in the blood of dry or neovascular AMD patients compared to controls. Thus, they may act as diagnostic and prognostic biomarkers to predict the process of AMD. According to a previous study, the level of miR-27a-3p, miR-195-5p, and miR-29b-3p in whole blood of AMD patients was increased compared to controls[26]. Additionally, various pathological processes associated with AMD, like inflammation and angiogenesis, were also mediated by several miRNAs. Moreover, the involving mechanism has been widely studied in recent years[22]. However, the studies about lncRNAs are few, with most focusing on the single lncRNA. Importantly, lncRNAs undertake a wide range of regulatory functions. Depending on its subcellular localization, one lncRNAs could influence the production of several proteins at the transcriptional, epigenetic, or post-transcriptional levels [27].
The hypothesis of the ceRNA network was presented to describe the regulatory relationship among 3 or more types of RNAs, including lncRNA, circRNA, mRNA, and miRNA. This hypothesis unraveled a possible regulation mode of lncRNAs. In this hypothesis, RNAs are divided into ceRNAs and miRNAs. RNAs (mRNAs, lncRNAs, pseudogenes, and circRNAs) contain regions that can be bound by miRNAs, called miRNA response elements (MREs), collectively termed ceRNAs. MiRNAs that are competitively bound by ceRNAs cannot combine with appropriate mRNAs. However, these mRNAs could be successfully translated and can be functional [12]. LncRNAs are a kind of ceRNA, having the function of a miRNA sponger. A lncRNA-miRNA-mRNA network including 18 lncRNAs, 41 miRNAs, and 404mRNAs was constructed in the current research. Previous studies have shown that MIAT identified in our study was very closely related to the pathological processes of AMD. MIAT regulates microvascular dysfunction[28] and promotes neurovascular remodeling in the retina[29].MEG3 could modulate RPE differentiation[30], and its silencing could protect the retina from light-induced degeneration[31]. The expression of SNHG5 decreased in both atrial fluid and plasma in diabetic macular edema patients. The finding presented a negative correlation with some clinical indicators, including disease duration, body mass index, and fasting blood glucose levels [32]. MALAT1 could protect the retina from oxidative damage via regulating the Keap1-Nrf2 axis[33]. Another study indicated that Knockdown of MALAT1 could modulate endoplasmic reticulum stress and attenuate angiogenesis and inflammation induced by high glucose in human retinal vascular endothelial cells [34]. Although not reported in AMD, other differential expression lncRNAs identified in our study, like SNHG11, DLEU2, GAS5, RFPL3S, and DIO3OS, were widely identified in cancer[35–37]. Their correlation with AMD and underlying mechanisms should be investigated in the future. Based on the theory of the ceRNA network, lncRNA and mRNA in the ceRNA network linked by the same miRNA should be a positive correlation. Co-expression analysis indicated that 51 pairs of lncRNAs and mRNAs shared the positive co-expression relationship, suggesting our lncRNA-miRNA-mRNA network's reliability.
Our study constructed the lncRNA-miRNA-mRNA network based on the latest and most reasonable algorithms and rigorous processes. The lack of validation experiments was the limitation of this research. However, both lncRNAs and mRNAs within the lncRNA-miRNA-mRNA network were derived from DE-mRNAs in the macular region of AMD patients. Furthermore, the prediction of lncRNA-miRNA pairs and miRNA-mRNA pairs was performed using experiment-supported databases, such as miRcode, TargetScan, miRTarBase, and miRDB. These two guaranteed that our research had a degree of experimental support.