Since miRNAs can regulate the expression of multiple target genes, miRNA molecules and their target genes are linked in an intricate network of signaling pathways. Some miRNAs are tissue-specific, while others can be packaged into vesicles and move through the body in the peripheral blood in freely circulating form (aggregated with Argonaute proteins) or as part of exosomal vesicles.66 Thus, pathologically altered tissue can use miRNA to influence other tissues. Numerous studies on pathologies associated with MetS suggest that miRNA is a suitable and fairly accurate diagnostic and prognostic biomarker,16 making the search for specific miRNAs associated with disease development relevant today.
Although the role of AGEs and their receptors in the context of MetS is quite well defined, as described above, their epigenetic miRNA dysregulation in MetS, which may be important for both therapy and diagnostics of this syndrome and its comorbidities, has not yet been described in detail. Therefore, we were confronted with the task of systematizing the disparate data on miRNA–messenger RNA (mRNA) interactions available in the world literature and databases and compiling them in the most informative way possible. We opted for two approaches: inductive “bottom-up” (AGE/RAGE axis genes – miRNAs targeting them in the context of MetS comorbidities) and deductive “top-down” (miRNAs significantly differentially expressed in MetS comorbidities – their (experimentally proven) target genes – AGE/RAGE axis genes among the target genes and the degree of miRNA influence on them).
Inductive “bottom–up” approach: Key genes of the AGE/RAGE axis and the miRNAs targeting them in the context of MetS comorbidities
In this section, we focus on what we believe to be the most important target genes of the AGE/RAGE axis, whose functions and roles in MetS have already been mentioned: RAGE ligands (HMGB1, various S100s), the RAGE receptor, and other receptors for AGEs (DDOST, PRKCSH, LGALS3, STAB1, STAB2, SR–AI, SR–BI, CD36, OLR1). This section is a classic descriptive review, so the criteria for article selection were relatively soft:
The search for articles on the topic was performed using the following queries (variable-specific queries are in italics): “(gene of interest) and (miRNA) and (MetS comorbidity)” in the PubMed database.
Original research articles were selected in which the data were obtained from human biomaterials or human cell models.
All retracted articles were excluded.
The year of publication was not taken into account.
The description of the studies presented below is also supplemented by Table 1, in which the studies are described in a concise and model-oriented form.67–85
Table 1miRNA regulation of ligands and receptors in the AGE/RAGE signaling pathway
| microRNA–№ | Target | Experimental model of MetS comorbidities | Changes in the level of miRNA expression in the experiment | References |
|---|
| hsa–miR–126–3p and mmu- miR–126–3p | HMGB1 | Endothelial dysfunction in atherosclerosis: in vitro HUVECs treated with high–glucose medium; in vivo high–fat diet fed diabetic ApoE–/– mice with induced pancreatic injury | Downregulated in HUVECs upon high–glucose treatment; reduced levels in aortic tissues of diabetic ApoE–/– mice | 67 |
| hsa–miR–320a–3p | HMGB1 | Hepatocellular carcinoma: hepatic tissue from hepatocellular carcinoma patients; in vitro HepG2 and SK–hep–1 hepatocytes | Aberrant (mostly reduced) levels in tumoral tissues from hepatocellular carcinoma patients; reduced levels in the plasma of MetS patients | 68,69 |
| hsa–miR–328–3p | HMGB1 | Endothelial dysfunction in atherosclerosis: HUVECs treated with oxLDL | Downregulated in HUVECs upon oxLDL treatment | 70 |
| hsa–miR–5591–5p and mmu– miR–5591–5p | AGER | Diabetic wound as a T2DM complication: in vitro primary human ADSCs treated by AGEs; in vivo diabetic mice wound model | Downregulated in ADSCs upon AGEs treatment | 71 |
| hsa–miR–21–3p | ADAM10 | Endothelial dysfunction in atherosclerosis: human aortic VSMCs and human VECs treated by high–glucose medium | Upregulated in VECs (but not in VSMCs) upon high–glucose treatment | 72 |
| hsa–miR–128–3p | LGALS3 | T2DM patients with or without MCI: PBMCs | Reduced levels in PBMCs of T2DM patients with MCI | 73 |
| hsa–miR–204–5p | MSRI | Foam cells in atherosclerosis: BMDMs, HMDMs, TEPMs and Raw 264.7 macrophages treated with oxLDL | Downregulated in HMDMs upon oxLDL treatment | 74 |
| hsa–miR–24–3p | SCARB1 | Models of cells involved in cholesterol homeostasis: HepG2 hepatocytes and PMA–induced THP–1 macrophages incubated with differently labeled HDL | NI | 75 |
| hsa– and, respectively mmu–miR–185–5p/miR–96–5p/miR–223–3p | SCARB1 | Models of cells, involved in cholesterol homeostasis: in vitro HepG2, Bel–7402 and HL–7702 hepatocytes treated with Dil–HDL and PMA–induced THP–1 macrophages; in vivo high–fat diet fed ApoE–/– mice | Livers of the high–fat diet fed ApoE–/– mice: miR–185–5p and miR–96–5p levels reduced. THP–1 upon PMA stimulation: miR–185–5p, miR–96–5p and miR–223–3p levels increased | 76 |
| hsa– miR–223–3p and mmu– miR–223–3p | SCARB1 | Models of cells, involved in cholesterol homeostasis: in vitro J774 macrophages and Huh7 hepatocytes treated with LPDS, HCAECs treated with HDL–C; in vivo cocoa butter diet fed ApoE–/– mice | Downregulated in J774 and Huh7 upon LPDS treatment and upregulated upon LDL–C treatment, elevated levels in cocoa butter diet fed ApoE–/– mice | 77 |
| hsa– miR–758–5p | CD36 | Foam cells in atherosclerosis: THP–1 macrophages, treated with fluorescently labeled DiI–oxLDL | NI | 78 |
| hsa– miR–26a–5p | CD36 | NAFLD (steatohepatitis): HepG2 hepatocytes treated with PM2.5 liposoluble extracts | Downregulated upon PM2.5 liposoluble extracts treatment | 79 |
| hsa– miR–204–3p | CD36 | Foam cells in atherosclerosis: BMDMs, HMDMs, TEPMs and Raw 264.7 macrophages treated with oxLDL | hsa–miR–204–3p is localized in the nuclei of BMDMs, HMDMs and TEPMs. The levels in the cytoplasmic fractions remained unchanged upon oxLDL treatment | 74 |
| hsa–let–7g–5p and mmu– let–7g–5p | OLR1 | Proliferation and migration of vascular smooth muscle cells (neointimal hyperplasia) in atherosclerosis: in vitro HASMCs treated with oxLDL; in vivo high–fat diet fed C57BL/6J mice | Downregulated upon oxLDL treatment. Was also on a high–fat diet fed mice as well in serum of hypercholesterolemic human subjects compared with healthy controls | 80 |
| hsa–let–7g–5p and mmu– let–7g–5p | OLR1 | Neointimal hyperplasia in atherosclerosis: in vitro HASMCs treated with oxLDL; in vivo high–fat diet fed ApoE–/– mice | NI | 81 |
| hsa–miR–590–5p and mmu– let–7g–5p | OLR1 | Angiogenesis in atherosclerosis: in vitro HUVECs treated with oxLDL; in vivo Marigel plugs inserted in the mice subcutaneous space | Downregulated upon oxLDL treatment | 82 |
| hsa–miR–320a–3p | OLR1 | Endothelial dysfunction in atherosclerosis: HUVECs treated with oxLDL | Downregulated upon oxLDL treatment | 83 |
| hsa–miR–98 | OLR1 | Endothelial dysfunction in atherosclerosis: in vitro hypoxic injured HUVECs; plasma from patients with single–vessel, double–vessel, and multi–vessel coronary artery disease / healthy controls | Downregulated upon hypoxic injury in HUVECs and reduced in patients with occluded coronary artery disease | 84 |
| hsa–miR–24–3p | OLR1 | No particular pathology: two human hepatic cell lines with different genotypes for rs1050286 (HeLa A/G versus HepG2 A/A) | NI | 85 |
AGEs
Both exogenous and endogenous AGEs undoubtedly play an important role in AGE/RAGE signaling; however, it is not possible to consider their miRNA dysregulation because neither the AGEs themselves (there are no specific genes and therefore no mRNA transcripts for them), nor the transcripts of the genes of the enzymes that form them (since, as mentioned above, the reactions of AGE formation do not require enzymes), can be the subject of direct miRNA regulation. Note that in this regard, it would be more accurate to use the term “ligand/RAGE signaling” in our review, but the term “AGE/RAGE signaling” is generally accepted and implies, among other things, the involvement of non-glycated ligands, so we have chosen not to replace it.
HMGB1 (gene HMGB1)
When vascular endothelial cells were treated with excess glucose, administration of an hsa–miR–126 mimetic significantly reduced inflammation and oxidative stress. It has been shown that the protective effect of hsa–miR–126 is precisely due to the suppression of HMGB1 (a proven direct target of miR–126) and NOX (as a consequence of HMGB1 suppression).67 Hsa–miR–126 has been proposed as a marker for the prediction of prediabetes and T2DM: a significant decrease of this miRNA was found in the blood of patients with T2DM.86 In a cell experiment with primary human endothelial progenitor cells, the expression of hsa–miR–126 decreased significantly when AGEs were added.87 Thus, hsa–miR–126 not only interferes with the genes of the AGE/RAGE axis but is also influenced by AGEs themselves. This leads to the hypothesis that a decreased HMGB1 level can be observed in diabetes due to increased AGEs.
In the livers of patients with hepatocellular carcinoma (which can develop as a result of NAFLD), miR–320a–3p was significantly reduced. Researchers found that hsa–miR–320a suppressed HMGB1 in the liver, preventing it from functioning as a DAMP. On the contrary, suppression of hsa–miR–320a–3p promoted the activation of the AGE/RAGE axis through increased expression of HMGB1. In addition, a negative correlation between hsa–miR–320a–3p and HMGB1 was found in liver tissue from hepatocellular carcinoma patients.68 It is known that the plasma level of hsa–miR–320a was reduced in women with MetS compared to non–MetS women with a normal body mass index and showed a significant negative correlation with body mass index, waist circumference, triglyceride level, glucose level, and the homeostasis model assessment of insulin resistance index.69 Another miRNA, hsa–miR–328–3p, significantly reduced HMGB1 expression, homeostasis model assessment of insulin resistance levels, and oxLDL-induced inflammation in a cell-based model of endothelial dysfunction in atherosclerosis.70
S100 (various genes listed in section 1.1)
No studies were found on the relationship between miRNA and S100 in the context of MetS and its comorbidities.
Beta–amyloid (gene APP)
The authors have chosen not to include this gene in this review for the following reasons:
Beta-amyloid as a RAGE ligand is not the APP gene product itself, but peptides (beta-amyloid fibrils) formed by proteolytic cleavage of the APP product; therefore, miRNA regulation of beta-amyloid cannot be directly considered88;
Beta-amyloid fibrils are primarily associated with neurodegenerative diseases, and their role in MetS is secondary.
RAGE (gene AGER)
Surprisingly, only one paper was found on miRNA dysregulation of the major receptor of the AGE/RAGE axis: hsa–miR–5591–5p directly targeted the 3′–UTR of AGER and repressed its expression. In vitro, hsa–miR–5591–5p was shown to promote AT stem cell survival and enhance the ability to repair diabetic wounds via the AGE/AGER axis.71 It is worth noting that the journal editors left a note on 01/20/2022 that the study is under investigation, but as of 11/22/2024, the article has not yet been retracted.
sRAGE
Researchers found that hsa–miR–21–3p was significantly increased in vascular epithelial cells at high glucose concentrations in the culture medium. It was established that hsa–miR–21–3p targets ADAM10, an enzyme that stimulates the formation of sRAGE. When the expression of hsa–miR–21–3p increased, ADAM10 was degraded to a greater extent, leading to a decrease in sRAGE levels and progression of atherosclerosis.72
AGE–R1 (gene DDOST) and AGE–R2 (gene PRKCSH)
No studies were found on the relationship between miRNA and these genes in the context of MetS and its comorbidities.
AGE–R3 (galectin–3, gene LGALS3)
In a study of mild cognitive impairment (MCI) in T2DM, the plasma levels of galectin-3 were found to be significantly elevated in the group of patients with MCI. On this basis, the authors suggested using galectin-3 as a prognostic marker for MCI in T2DM. In the same study, hsa–miR–128–3p was found to be significantly reduced in the peripheral blood mononuclear cells (PBMCs) of T2DM patients and negatively correlated with the expression level of LGALS3 in PBMCs. Hsa–miR–128–3p directly targets the 3′–UTR of LGALS3.73
Stabilins 1 and 2 (genes STAB1 and STAB2)
No studies were found on the relationship between miRNA and stabilins in the context of MetS and its comorbidities.
SR–AI (gene MSR1)
Hsa–miR–204–5p suppressed MSR1 by directly targeting its 3′–UTR. The expression level of hsa–miR–204–5p was significantly reduced in the foam cell model of atherosclerosis.74
SR–BI (gene SCARB1)
It was shown that hsa–miR–24–3p significantly reduced the uptake of HDL–C in HepG2 and THP1 cell cultures. The authors of the study demonstrated that this effect was achieved by inhibiting the SCARB1 receptor (proven to be a direct target of hsa–miR–24–3p) in hepatocytes and macrophages, which is responsible for selective HDL–C uptake. Thus, the hsa–miR–24–3p antagonist could play a protective role in atherosclerosis.75 Hsa–miR–185–5p, hsa–miR–96–5p, and hsa–miR–223–3p also led to a similar phenomenon: the addition of their mimetics to HepG2 cells resulted in significantly reduced uptake of labeled HDL–C. All three miRNAs were shown to directly target SCARB1 and, as in the study described above, contributed to the development of atherosclerosis by interfering with this gene (as they impeded cholesterol metabolism in hepatocytes). It is noteworthy that the authors of the study did not address the other side of the issue—namely, the effects of oxLDL uptake on the liver itself and the possible development of NAFLD.76 A similar study showed that hsa–miR–223–3p in Huh7 cells also regulated HDL–C uptake by suppressing SCARB1.77SCARB1 is highly expressed not only in the liver but also in endothelial cells: it was also shown that inhibition of hsa–miR–223–3p led to increased SCARB1 levels and HDL–C uptake in human coronary artery endothelial cells.
CD36 (gene CD36)
The miRNA mimetic hsa–miR–758–5p interfered with CD36 in a foam cell model of atherosclerosis: hsa–miR–758–5p negatively regulated macrophage binding of oxLDL and reduced cholesterol accumulation in macrophages, which is suggested by the researchers as a promising strategy for the treatment of atherosclerosis.78 In human monocyte-derived macrophages, hsa–miR–204–3p suppressed the expression of CD36, but according to the results of the luciferase reporter assay, there is no direct canonical interaction between hsa–miR–204–3p and CD36. The researchers demonstrated an alternative pathway: hsa–miR–204–3p was found to be translocated to the nucleus where, together with the Argonaute-2 protein, it did not repress translation as canonically assumed, but rather affected the nuclear transcription of the CD36 gene.74 Another miRNA, hsa–miR–26a, interfered with CD36 (its proven direct target) in a cellular model of steatosis and steatohepatitis, thereby suppressing the processes of cellular lipogenesis.79
OxLDL receptor 1 (gene OLR1)
The receptor for oxLDL (the protein is often referred to as lectin like oxidized low density lipoprotein receptor 1 (LOX-1) in studies) is associated with the development of atherosclerosis. When oxLDL was added to the primary cell culture of human aortic smooth muscle cells, the expression of OLR1 increased significantly in response to an increase in the concentration of its major ligand. At the same time, the addition of oxLDL suppressed the expression of hsa–let–7g, which directly targets OLR1. This negative relationship between oxLDL level and hsa–let–7g expression could explain the fact that oxLDL causes an increase in OLR1 expression in a cellular model of atherosclerosis.80 This was later also demonstrated in hyperlipidemic mice (ApoE–/–): intravenous administration of hsa–let–7g mimetics attenuated atherosclerotic lesion formation, which was accompanied by a significant reduction in Lox–1 production. Taken together, these studies suggest that hsa–let–7g has a protective and anti-atherosclerotic effect by suppressing OLR1.81
The hsa–miR–590–5p mimetic significantly reduced levels of OLR1 (its proven direct target) in a model of endothelial dysfunction in atherosclerosis, thereby preventing oxLDL-mediated angiogenesis.82 In a similar cell model of atherosclerosis, hsa–miR–320a–3p was also shown to interfere with OLR1 (its proven direct target), leading to increased cell viability.83 Decreased expression of hsa–miR–98 was detected in the human umbilical vein endothelial cell line (when cultured under hypoxic conditions), and the level of hsa–miR–98 was also decreased in the plasma of patients with single-, double-, and multivessel coronary artery disease. The hsa–miR–98 mimetic decreased OLR1 levels and thereby significantly increased the viability of endothelial cells in culture.84 No luciferase reporter was used in this study, so a direct link between hsa–miR–98 and OLR1 could not be established.
In addition, the strength of miRNA–mRNA interaction may depend on the presence of some polymorphisms. For example, hsa–miR–24–3p (OLR1 is a proven direct target) interfered differently with OLR1 in HeLa and HepG2 cell cultures, depending on the presence of the rs1050286 polymorphism: thus, OLR1 was significantly more strongly suppressed in the HeLa cell line (the hsa–miR–24–3p site in the OLR1 gene is heterozygous) than in the HepG2 cell culture (the binding site in the OLR1 gene is homozygous for the polymorphism).85
Thus, the influence of miRNAs on AGE receptors and their ligands in the context of MetS and its comorbidities seems to be poorly studied: There are virtually no data on miRNA regulation of the major AGE receptors, and data on scavenger receptors focus on the uptake of different cholesterol fractions and not on AGEs.
Deductive “top–down” approach: Differentially expressed miRNAs in MetS comorbidities and their involvement in dysregulation of the AGE/RAGE axis
The inductive “bottom–up” approach presented in the previous section has proven to be rather insufficiently systemic: It remains unclear to what extent comorbid pathologies in MetS differ in how miRNAs modulate AGE/RAGE signaling. Finally, the literature search conducted revealed a lack of data on miRNA dysregulation of central genes of the AGE/RAGE axis, namely the ligands S100s and various receptors for AGEs.
For all these reasons, we decided to complement our review with a reverse deductive (or synthetic–systematic) “top–down” approach. Briefly, this approach can be described as follows: We collected human miRNA profiling data related to MetS comorbidities and derived differentially expressed miRNAs (DEMs) from these studies. We then performed an automated search for target genes of the collected DEMs in the miRTarBase database for experimentally proven miRNA–target interactions (MTIs) and compared the target genes of these DEMs with genes of the AGE/RAGE pathway (extended WikiPathways WP2324 pathway) to analyze the degree of miRNA dysregulation of this pathway in the context of MetS. In implementing this approach, we faced a number of important dilemmas that required compromises in either the scope of our analysis or its accuracy.
A search for human miRNA profiling studies was performed in PubMed and Gene Expression Omnibus. We selected studies based on the following parameters/aspects:
Pathology. MetS itself and the following comorbidities were considered: obesity, IR, atherosclerosis, NAFLD, periodontitis, and PCOS.
Localization of the tissue sample. Preference was given to tissues “corresponding” to the pathology: e.g., liver in NAFLD, subcutaneous white adipose tissue (sWAT) in obesity, etc. Two exceptions were made that are worth explaining in more detail: Obesity was also considered in visceral AT exosomes in addition to sWAT, as visceral AT is an active metabolic organ and can contribute significantly to systemic miRNA dysregulation via exosomes.89 MetS was considered both in plasma (which contains circulating miRNAs, and can therefore be considered a systemic reflection of miRNA dysregulation) and in PBMCs, as meta-inflammation is one of the key aspects of MetS.90 In the context of the above, we use the term “pathology–localization” in this review for the sake of clarity. This means that the effects of miRNA dysregulation in pathology are considered in a specific organ/tissue and not in pathology in general.
Method to study miRNA expression. We limited ourselves to data obtained exclusively with microarrays: This method has acceptable parameters for both accuracy and performance and, moreover, is commonly used in miRNA profiling practice.91
All retracted articles were excluded.
The year of publication was not taken into account.
As the analysis of the original articles has shown, even within the same method, there is no agreement on how to define “differential expression” (i.e., DEMs). Therefore, we included in our deductive analysis those miRNAs that were defined as DEMs by the authors themselves according to their own criteria. It is important to emphasize that the miRNAs considered in this section are not equivalent in their potential contribution to dysregulation of the AGE/RAGE pathway, as the strength of the interaction is influenced by the level of miRNA expression (and by other factors that we have not considered in this approach, such as the strength of miRNA binding to its target, steric accessibility of the landing site, etc.).
Studies describing the tissue type, patient sample, and number of miRNAs analyzed, as well as the number of DEMs identified by the authors, are listed in Table 2.89,90,92-106
Table 2miRNA profiling studies included in the top–down approach
| Disease | Biological sample | Case group and control group | Micro array size | Number of DEMs | References |
|---|
| Metabolic syndrome | PBMCs isolated immediately after blood collection | Subjects undergoing venipuncture: 20 MetS and 10 healthy control subjects | 1146 human miRNAs (detected 688) | 52 up– and 29 downregulated | 90,92 |
| Metabolic syndrome | Blood plasma | Women undergoing venipuncture under 8–12 h fasting conditions: 10 MetS, 8 healthy non–obese | 1773 human miRNAs (detected NI) | 8 up- and 13 downregulated | 93,94 |
| Obesity | Adipocyte–exosomes from visceral AT (greater omentum) | Adolescent women undergoing bariatric surgery or unrelated abdominal procedures: 7 obese and 5 lean | 1773 human miRNAs (detected NI) | 16 up- and 28 downregulated | 89,95 |
| Obesity | Abdominal subcutaneous white AT | Women undergoing: 1: elective surgical procedures: 6 non–obese and 13 obese without T2DM; 2: fat biopsy by needle aspiration: 26 non–obese and 30 obese | 1: 723 human miRNAs (detected NI); 2: 678 human miRNAs (detected NI) | 1: 5 up- and 4 downregulated; 2: 2 up- and 18 downregulated, 1 miRNA is common (hsa–miR–139–5p downregulated) | 1: 96,97; 2: 98,99 |
| Insulin resistance | Abdominal subcutaneous white AT | Women undergoing fat biopsy by needle aspiration: 8 obese insulin–resistant and 21 obese insulin–sensitive | 847 human miRNAs (detected 205) | 4 up- and 10 downregulated | 100,101 |
| NAFLD | Liver tissue | Subjects undergoing a laparoscopic gastric bypass or sleeve gastrectomy: 15 NAFLD and 15 non–NAFLD (histology proven) | 1438 human miRNAs (detected between 299 and 389) | 39 up- and 1 downregulated | 102 |
| Atherosclerosis | Vascular samples (plaques) | Subjects undergoing endarterectomy, or an abdominal aortic bypass, or coronary artery bypass: 12 atherosclerotic plaques from peripheral arteries and 6 non–atherosclerotic left internal thoracic arteries (each biospecimen from different subject) | 866 human miRNAs (detected NI) | 42 up– and 33 downregulated | 103 |
| PCOS | Ovarian theca interna tissue | Women undergoing laparoscopy and/or ovarian wedge resection: 10 PCOS and 8 non–PCOS with normal insulin sensitivity | 1700 miRNAs annotated in miRBase 16.0 (detected NI) | 7 up– and 3 downregulated (targeting insulin action related and androgen producing related genes) | 104 |
| Periodontitis | Gingival tissue | Subjects with severe periodontitis: 10 normal–weight and 10 obese (without T2DM) | 1773 human miRNAs (detected NI) | 13 up– and 22 downregulated | 105,106 |
There are two main approaches to determine the MTI: bioinformatic prediction and experimental confirmation. Bioinformatic methods are based on nucleotide sequence and/or machine learning; experimental methods are based on next-generation sequencing, microarrays, polymerase chain reaction, western blot, luciferase reporter, and other methods.107 Confirmed miRNA–target interactions are collected in specialized databases (miRTarBase, DIANA–TarBase, and miRecords). The authors acknowledge and draw the reader’s attention to the fact that experimental methods of MTI verification are inevitably subject to the so-called streetlight effect (or drunkard’s search), i.e., they limit the list of MTIs to those that, for one reason or another, have previously been of interest to researchers (appeared on a microarray or were selected for polymerase chain reaction/blot/luciferase reporters, etc.). Despite this bias, experimental methods were favored in this review as they better and more accurately reflect the current state of knowledge. MTIs were searched in miRTarBase v. 9.0 (this database is not only large, but the data in it are also regularly revised).107 miRTarBase v. 9.0 contains both low-throughput (more accurate) and high-throughput (less accurate) experimental methods for MTI detection, which the reader should consider when evaluating the results presented.
Originally, the same list of genes considered in the previous section (ligands and receptors only) was intended for this section. However, these genes have a relatively low number of experimentally proven MTIs, which limits the applicability of such an approach for them (as you can see in Supplementary Table 1 on the “Limited Analysis” list). In this context, it was decided to extend the list of signaling genes to all genes of the AGE/RAGE pathway. Genes involved in AGE/RAGE signaling were exported from WP2324 (https://www.wikipathways.org/pathways/WP2324.html ) and supplemented with the following genes: CD36, PRKCSH, STAB1, STAB2, OLR1, and SCARB1 (other receptors for AGEs,23HMGB1, S100A1, S100A4, S100A6, S100A7, S100A8, S100A9, S100A11, S100A12, S100A13, S100A14, S100B, S100P (ligands for the RAGE receptor),25,108ADAM10 (metalloprotease that determines the formation of sRAGE109), SOD2 (superoxide dismutase, fulfills similar functions as SOD1, which is already listed in WP2324110), NOX1, CYBB, NOX3, NOX4, NOX5 (in WP2324 signaling they are listed under the general name “NOX” but are not listed as specific genes themselves; evidence for their common involvement can be found at the link111).
miRNA dysregulation of the common AGE/RAGE axis: Comparison of different pathology–localizations in MetS comorbidities
Using the described approach, 304 unique DEMs were collected, of which 41 DEMs were found in more than one pathology–localization. Of these recurrent DEMs, 25 showed similar dynamics of expression changes in all pathology–localizations (increased or decreased everywhere), while the remaining 16 showed inconsistent (opposite) dynamics. The following observation was also very interesting: When comparing the lists of DEMs, it was found that the localization shaped the similarity of the altered miRNA profile much more than the pathology itself. Thus, obesity was considered in two localizations: in exosomes derived from visceral AT (44 DEMs) and in sWAT (combined) (28 DEMs),89,96,98 a comparison of these lists did not reveal a single common miRNA. However, the miRNA profile in sWAT was considered in another study, but then in IR,100 and comparison of the profiles in obesity (28 DEMs) and IR (14 DEMs) revealed five common DEMs with complete concordance of expression dynamics (in both pathologies hsa–miR–143–3p, hsa–miR–145–5p, hsa–miR–26a–5p, hsa–miR–378a–3p, and hsa–miR–652–3p were downregulated in sWAT).
AGE/RAGE signaling was found to be significantly dysregulated by DEMs in each MetS comorbidity, as shown in Figure 2, which shows how many MTIs were found in each of the pathologies, as well as the three most “significant” (with the largest number of MTIs with AGE/RAGE signaling genes) miRNAs. The greatest dysregulation was observed in atherosclerosis in plaque samples (64 of 75 DEMs targeted AGE/RAGE pathway genes with 284 experimentally proven MTIs), MetS in PBMCs (64 of 81 DEMs with 203 MTIs), and obesity in sWAT (27 of 28 DEMs with 147 MTIs). The results obtained are somewhat consistent with the findings of the above “bottom–up” approach (see Table 1), as a significant part of the information on miRNA dysregulation of AGE/RAGE genes that we found in research articles was specifically described in the context of atherosclerosis.
The full results of the analysis (all DEMs in the pathology–localizations retrieved from the studies and all their MTIs with AGE/RAGE signaling genes) are listed in Supplementary Table 1. Supplementary Table 1 comprises six sheets (all contents are listed on their corresponding sheet “Description”), it is not necessary for the understanding of this article: all important results of the analysis are presented graphically and in tables in the article itself. The supplementary table is added merely for the transparency of the publication and may be useful for those readers who wish to reproduce a similar method or check the validity of the conclusions drawn in this article.
Interesting data were also obtained from the DEMs themselves, for example: hsa–miR–34a–5p, which has 15 known MTIs involving 15 AGE/RAGE axis genes (AKT1, CASP3, CASP8, CASP9, CYBB, CYCS, HMGB1, MAP2K1, MAPK3, MMP2, MMP7, NFKB1, S100P, SRC, and STAT1), was elevated in PBMCs in MetS as well as in atherosclerotic plaques. It is well known that monocyte–macrophage cells transform into foamy macrophages in the presence of excess lipids and make an undeniable contribution to atherogenesis.112 Indeed, hsa–miR–34a–5p has already been shown to promote atherosclerosis development,113 but its effects in the context of the AGE/RAGE axis have not yet been considered.
As mentioned above, expression of hsa–miR–145–5p and hsa–miR–143–3p was significantly reduced in sWAT in both obesity and IR – both miRNAs have 11 MTIs with a total of 19 AGE/RAGE axis genes (genes common to both miRNAs are in bold): AKT1, CD36, EGFR, HIF1A, HRAS, IRS1, MAPK1, MMP13, MMP14, MMP2, MMP9, ROCK1, SMAD2, SMAD3, SP1, STAT1, STAT3, STAT5A, and TIRAP, suggesting an important role of AGE/RAGE signaling in the development of these pathologies.
A striking finding was that hsa–miR–92a–3p (13 MTIs with the genes ADAM10, DIAPH1, EZR, HIF1A, HMGB1, MAPK8, MAPK9, NFKB1, PRKCA, SCARB1, SP1, STAT3, and TIRAP) was simultaneously reduced in obesity (sWAT), atherosclerosis, and PCOS. The development of PCOS is associated with obesity and the progression of CVD, including atherosclerosis: Women with PCOS have higher intima–media thickness (one of the diagnostic criteria for atherosclerosis) compared to the control group.114 Thus, reduced hsa–miR–92a–3p expression could represent a link between the development of atherosclerosis in obesity and PCOS.
However, some significant DEMs related to dysregulation of the AGE/RAGE axis showed contradictory expression dynamics in different pathology–localizations. For example, let–7b–5p, which has 11 MTIs with 11 AGE/RAGE axis genes (DIAPH1, HIF1A, HMGB1, HRAS, MAPK1, MSN, NFKBIA, SOD2, SP1, STAT1, and STAT5A), was elevated in the liver in NAFLD but decreased in sWAT in IR. Glucose intolerance is recognized as an important factor in the pathogenesis of NAFLD,115 so the presence of common DEMs between these pathologies was expected (however, there was only one such DEM – the let–7b–5p mentioned above). Nonetheless, the different directions of the changes in their expression leave many questions unanswered: it is not clear what consequences this has both in the broadest sense and in the context of AGE/RAGE signaling.
miRNA dysregulation of AGE/RAGE pathway receptors and their ligands in pathology–localizations with MetS comorbidities
As written above, there is particularly little information in the literature on miRNA dysregulation of genes of greatest interest for this topic: AGE/RAGE ligands and receptors. Therefore, we further focused on the interpretation of the data obtained specifically for these genes. The results of the analysis are shown graphically in Figure 3. Accordingly, the HMGB1 ligand and the DDOST receptor were the most miRNA-dysregulated genes in MetS comorbidities.
For other ligands (except HMGB1 and OLR1) and receptors of the AGE/RAGE axis, the data were rather sparse. We would like to emphasize again that the picture shown does not reflect the actual extent of miRNA dysregulation, but the part studied so far, as the genes of interest have a different number of known experimentally proven MTIs. This is reflected in Table 3,107 which clearly shows that many AGE/RAGE axis genes have very few experimentally confirmed MTIs, namely S100A1, A4, A6, A7, A8, A9, A12, A14, S100B, S100P, as well as AGER itself, and the anti-receptors PRKCSH, LGALS3, and the scavenger receptor MSR1 have fewer than 10 entries in miRTarBase v. 9.0. And there are no entries at all for the genes S100A13, STAB1, and STAB2. Based on this information, it was possible to calculate the “fraction of miRNA dysregulation” for the genes of interest in the MetS comorbidities. Thus, for the ligands S100A6, S100A12, S100A14, S100P, AGER itself, and for PRKSCH, of all known MTIs, ≥50% were associated with MetS, confirming the importance of their miRNA dysregulation in the context of this syndrome.
Table 3Fraction of miRNA-dysregulated major genes of the AGE/RAGE axis in MetS comorbidities (The number of experimentally proven MTIs was calculated from the miRTarBase v. 9.0 database107)
| Gene | All mirTarBase v. 9.0 entries (MTIs) | Entries for DEMs in MetS comorbidities (MTIs) | Fraction of MTIs for MetS comorbidities in all mirTarBase v. 9.0 MTIs |
|---|
| HMGB1 | 179 | 30 | 17% |
| S100A1 | 4 | 1 | 25% |
| S100A4 | 1 | 0 | 0% |
| S100A6 | 1 | 1 | 100% |
| S100A7 | 1 | 0 | 0% |
| S100A8 | 3 | 1 | 33% |
| S100A9 | 7 | 1 | 14% |
| S100A11 | 24 | 3 | 13% |
| S100A12 | 2 | 2 | 100% |
| S100A13 | ND | ND | ND |
| S100A14 | 2 | 2 | 100% |
| S100B | 3 | 1 | 33% |
| S100P | 4 | 2 | 50% |
| AGER | 3 | 2 | 67% |
| ADAM10 | 15 | 5 | 33% |
| DDOST | 16 | 7 | 44% |
| PRKCSH | 6 | 3 | 50% |
| LGALS3 | 3 | 1 | 33% |
| CD36 | 21 | 2 | 10% |
| STAB1 | ND | ND | ND |
| STAB2 | ND | ND | ND |
| MSR1 | 7 | 1 | 14% |
| SCARB1 | 14 | 3 | 21% |
| OLR1 | 81 | 6 | 7% |
Finally, it remained to compare the results of the bottom–up and top–down approaches and to find out whether there were common miRNAs between them. It turned out that a significant proportion matched: of the 17 miRNAs for which studies were found in the first approach (see Table 1), eight were also found in the alternative second approach.
hsa–miR–126–3p – was downregulated in the cell model of atherosclerosis and suppressed HMGB167; was downregulated in PBMCs in MetS and sWAT in obesity; has MTIs with AKT1, IRS1, MMP7, NFKBIA, and ROCK1.
hsa–miR–328–3p – was downregulated in the cellular model of atherosclerosis and suppressed HMGB170; was upregulated in PBMCs in MetS; has MTIs with EZR, HMGB1, and DIAPH1.
hsa–miR–21–3p – was upregulated in a cellular model of atherosclerosis and suppressed ADAM1072; was upregulated in plaques in atherosclerosis; has MTIs with CYCS, CDC42, STAT3, RHOA, EGFR, and CASP8.
hsa–miR–758–5p – suppressed CD36 in a cell model of foamy macrophages in atherosclerosis78; was downregulated in visceral AT exosomes in obesity; has no MTIs with AGE/RAGE signaling genes.
hsa–miR–26a–5p – was downregulated in a cell model of NAFLD and suppressed CD3679; was downregulated in sWAT in obesity and in IR; has MTIs with NOS2 and PRKCD.
hsa–miR–128–3p – suppressed LGALS3; was downregulated in PBMCs in diabetics with MCI73; was upregulated in plaques in atherosclerosis and in PBMCs in MetS; has MTIs with EGFR, CASP3, MAP2K1, MAPK14, SP1, SMAD2, LGALS3, and IRS1.
hsa–miR–185–5p – suppressed SCARB1 in cellular models of cholesterol uptake in atherosclerosis; was downregulated in the liver of diabetic mice fed a high-fat diet76; was downregulated in sWAT in obesity; has MTIs with AKT1, CDC42, DIAPH1, RHOA, SCARB1, and SOD2.
let–7g–5p – suppressed OLR1 in cell models of neointimal hyperplasia in atherosclerosis,80,81 was downregulated in this model80; was upregulated in ovarian theca tissue in PCOS; has MTIs with CASP3, HMGB1, SMAD2, and SOD2.
The data obtained allow us to consider these miRNAs as promising for closer applied studies, as their role as DEMs in MetS comorbidities has been demonstrated, as well as the fact that these miRNAs directly target key genes in AGE/RAGE signaling.
The limitations and further perspectives of the deductive “top–down” approach
In this review, we analyzed studies related to miRNA dysregulation of the major genes involved in the AGE/RAGE axis in MetS comorbidities. First, we performed a classic descriptive review of the relevant research studies. As this section was based on individual publications and conducted “bottom–up”—from the AGE/RAGE axis genes of interest to the miRNAs targeting them in models of MetS comorbidities—we labeled this approach “inductive”. The main conclusion that can be drawn from the inductive approach is that the information available in the literature on miRNA dysregulation of the AGE/RAGE axis is very limited. First, most information is devoted to the HMGB1 ligand and selected scavenger receptors (namely OLR1 and SCARB1), whereas miRNA dysregulation of AGER (including its soluble isoforms), other receptors for AGEs, and all S100 calgranulins remains virtually unexplored. Second, most studies were performed in atherosclerosis models (14 of the 19 articles), which inevitably gives a biased picture in the context of the whole MetS. Third, the studies focus primarily on miRNA regulation of cholesterol uptake processes (as a significant proportion of AGE receptors primarily take up various lipoproteins) rather than on AGEs themselves. Theoretically, the data obtained can be extrapolated to AGE signaling, but this has yet to be confirmed in practice.
Thus, the inductive approach proved to be insufficiently systemic and did not allow us to draw a holistic picture of miRNA dysregulation of the AGE/RAGE axis in MetS. This prompted us to test an alternative approach—”top–down”, i.e., in the opposite direction: from miRNAs significant for MetS comorbidities to genes of interest. This approach has been termed “deductive” or, alternatively, “synthetic” in the context of this work, as we merged data from different MetS comorbidities into a single system. The deductive approach has given us promising insights, but it is necessary to point out the hardly avoidable shortcomings of the deductive synthetic approach. The results we have obtained have a number of limitations:
The parameters used to determine the “DEM” in miRNA profiling studies vary, and therefore, the possible effects of these regulatory interactions differ as well.
The initial amount of miRNA contained in the microarray (i.e., the pool from which the DEM could be determined) varies between studies, which might affect the occurrence of false negatives in subsequent analysis.
The degree of acquired knowledge about the respective miRNA—and accordingly the number of its known MTI entries in databases—varies, which may, in turn, influence the occurrence of false negatives.
The inductive approach presents much less data, but their quality is higher, as almost all articles include the luciferase reporter method (the gold standard for determining the MTI), while the databases (such as miRTarBase) also include “less strongly evidenced” MTIs (e.g., proven by next-generation sequencing).
However, despite the limitations, the synthetic approach seems promising as it allows disparate data to be summarized in a formalized representation that is more accessible for analysis and understanding. We believe that the innovative synthetic approach we propose is the major advantage of our article. It is universal, as it can be applied to other signaling or molecular mechanisms in all pathologies or conditions (in which miRNA profiling has been performed). In the future, the approach can be improved by adding a layer of meta-analysis of data from different miRNA profiling studies (with standardization of “DEM” criteria between different studies) and/or introducing bioinformatic predictions instead of accessing databases of experimentally proven MTIs. The latter will significantly broaden the scope of the analysis and include miRNAs that have not been studied so far.