https://orcid.org/0000-0002-1481-1061
Figures
Abstract
Pseudoexfoliation glaucoma is a severe form of secondary open angle glaucoma and is associated with activation of the TGF-β pathway by TGF-β1. MicroRNAs (miRNAs) are small non-coding RNA species that are involved in regulation of mRNA expression and translation. To investigate what glaucomatous changes occur in the trabecular meshwork and how these changes may be regulated by miRNAs, we performed a bioinformatics analysis resulting in a miRNA-mRNA interactome. Primary human trabecular meshwork cells originating from normal donors were treated with TGF-β1 at 5 ng/mL for 24h; total RNA was extracted followed by RNA-Seq and miRNA-Seq. For both mRNA and miRNA species, differential expression was determined using a bioinformatics pipeline consisting of FastQC, STAR, FeatureCounts, edgeR (for miRNA) and DESeq2 (for mRNA). Putative mRNA-miRNA interactions between differentially expressed mRNA and miRNA species were determined using interaction databases miRWalk, miRTarBase, TarBase and TargetScan. To classify mRNA species by function and pathway, gene enrichment was performed using Enrichr. The resulting miRNA-mRNA interactome consisted of 1202 interactions. Some highly connected microRNAs were hsa-let-7e-5p, hsa-miR-20a-5p, hsa-miR-122-5p, and hsa-miR-29c-3p. Most differentially expressed genes were indicated to be regulated by miRNAs. The sub-interactomes of genes involved in specific pseudoexfoliation glaucoma related enrichment terms such as oxidative stress, unfolded protein response, signal molecules and ECM remodelling were determined. This is the first study to present a genome-wide microRNA-mRNA regulatory network for human trabecular meshwork cells treated with TGF-β1 and may serve to generate unbiased hypotheses about regulatory functions and mRNA targets of miRNAs in pseudoexfoliation glaucoma and may help to develop miRNA-based therapeutics.
Citation: Roodnat AW, Doyle C, Callaghan B, Lester K, Henry M, Sheridan C, et al. (2025) Investigating the miRNA-mRNA interactome of human trabecular meshwork cells treated with TGF-β1 provides insights into the pathogenesis of pseudoexfoliation glaucoma. PLoS ONE 20(1): e0318125. https://doi.org/10.1371/journal.pone.0318125
Editor: Sharif Moradi, Royan Institute for Stem Cell Biology and Technology, ISLAMIC REPUBLIC OF IRAN
Received: October 1, 2024; Accepted: January 10, 2025; Published: January 30, 2025
Copyright: © 2025 Roodnat et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the manuscript and its Supporting Information files. Furthermore, the results can be reproduced by bioinformatics scripts which are available at https://github.com/Antauniau/TGFb1_HTM_interactome.
Funding: This work was supported by a Glaucoma UK (https://glaucoma.uk/) Royal College of Ophthalmologists Award 2018 to SA and CEW; BC and miRNA research was supported by Fight for Sight (UK; https://www.fightforsight.org.uk/), grant 5055-5056 to CEW. The funders had no say in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Pseudoexfoliation glaucoma (XFG) is a severe form of secondary open angle glaucoma which develops from pseudoexfoliation syndrome (XFS). The prevalence of XFS has been estimated to be 60 million people and in half this population pseudoexfoliation glaucoma (XFG) will eventually develop [1]. The aetiology of pseudoexfoliation glaucoma is not fully understood but it is associated with the production and release of exfoliation material (XFM) in the anterior segment of the eye which is deposited in the trabecular meshwork (TM) and other structures within the anterior segment [2]. Deposition of XFM in the TM is believed to contribute to the increased intraocular pressure (IOP) observed in XFG by blocking the outflow facility. Important components of XFM are amongst others LOXL1, fibrillin, latent TGF-β binding proteins and elastin [3]. Conventional glaucoma therapies lower IOP pharmacologically or surgically [4] but cellular and molecular changes in the TM in XFG remained unchecked. Understanding the mechanisms underlying the regulation of gene and protein expression in the TM in XFG would provide a foundation for developing gene-based therapies for this clinically aggressive glaucoma [5,6]. The pathogenesis of XFS and XFG is multifactorial and is associated with single nucleotide polymorphisms of LOXL1 [7,8] and other genes [2] but also with ageing [1], UV exposure [1], folate intake [9], hyperhomocysteinaemia [10], oxidative stress [11], dysfunctional autophagy [12] and dysregulated retinoic acid signalling [13]. Furthermore, XFG is associated with increased concentration of TGF-β1, both in active form in the aqueous humour and in latent form which is bound to the extracellular matrix [14]. TGF-β1 is an extracellular signal molecule involved in several cellular processes such as growth, extracellular matrix remodelling, oxidative stress, cell migration and apoptosis and is associated with wound-healing and fibrosis [15]. Treatment of normal TM cells with TGF-β1 induces changes in biological processes that show many similarities with changes observed in the TM in XFG such as production of XFM components, activation of the unfolded protein response, changes in the antioxidant system and fibrotic ECM remodelling [16].
MicroRNAs (miRNAs) are small noncoding RNA species that bind to specific mRNA sequences, usually in the 3’ untranslated region, and by doing so regulate gene expression by increased mRNA degradation or by reducing translation of the target mRNA. MiRNA regulation of gene expression is not very gene-specific, but a miRNA rather regulates a group (module) of genes and expression of one gene may be regulated by several miRNAs simultaneously [17]. Several miRNAs have been found to be differentially expressed in ocular tissues from XFG patients compared to normal tissue [2,18,19]. Furthermore, when normal TM cells are treated with TGF-β1 this induces changes in the expression of miRNAs [20] and there is a significant overlap with changes in miRNA expression observed in XFG. Previously several specific miRNA-gene interactions have been studied in ocular tissues [21–23] but these studies only reveal a small fraction of a vast interactive miRNA-gene network at the genome level. A representation of this genome-wide network in which all putative miRNA-gene interactions are shown for a certain condition is called a miRNA-gene interactome. Such genome-wide interactomes have been reported for other conditions such as osteoarthritis [24–26], ventricular hypertrophy [27] and cancer [28–31] but not for the trabecular meshwork. Therefore, the main goal of this work was to construct a miRNA::mRNA interactome for TM cells treated with TGF-β1 using our previously published miRNA [20] and mRNA [16] datasets, to determine how differentially expressed genes and altered biological processes may be regulated by miRNAs on a genome-wide level to investigate the role of miRNAs in XFG pathology. This interactome may help to understand the pathophysiology of XFG and could be used to identify interesting candidates for miRNA-based therapeutics.
Materials and methods
mRNA and miRNA sequencing
Primary normal human trabecular meshwork (TM) tissue was harvested from donor eyes originating from four different donors (n = 4) obtained from the Liverpool Research Eye Bank in the period 01/06/2018–01/01/2019.
Medical history for the donor eyes was unknown, however, no donors had previous ocular surgery or a known glaucoma diagnosis. Donor eyes were excluded if the maximum post-mortem time exceeded 48 h. More detailed donor information can be found in our previous publications [16,20] and the associated Short Read Archive BioProjects PRJNA1099874 and PRJNA882267 as mentioned in those publications. TM cells from four different donors were cultured followed by either TGF-β1 treatment or no treatment (i.e. controls) resulting in a paired experiment. The RNA concentration was measured using the NanoDrop 2000 (Thermofisher Scientific, Horsham, UK) and was higher than 50 ng/uL for miRNA and higher than 20 ng/uL for mRNA for all samples. RNA quality was determined using the Bioanalyzer 2100 (Agilent Technologies, Stockport, UK) and measured RIN was higher than 8 for miRNA and higher than 7 for mRNA. RNA extracted from all eight cell cultures was subjected to both regular RNA sequencing and small RNA sequencing. More details concerning the sequencing process can be found in our previous publications [16,20]. Subsequent data processing resulted in differentially expressed mRNA species and associated functional enrichment [16] and differentially expressed miRNA species [20] as reported previously by our group.
miRNA-mRNA interaction database
A combined miRNA-mRNA interaction database (or more loosely miRNA-gene interaction database) was created by combining multiple online available databases into one merged database in order to align gene and miRNA symbols, accession codes and the interaction evidence presented by the databases (e.g. experimental versus predicted interactions). Both predicted miRNA::mRNA interactions from TargetScan 8.0 (accessed 01/10/2022) [32] and miRWalk 2.0 (accessed 30/09/2022) [33] together with experimentally confirmed interactions from Tarbase v8 (accessed 25/01/2021) [34] and miRTarBase 8.0 (accessed 20/01/2021) [35] were added to establish a combined database of miRNA::mRNA interactions using an R script. In addition, the combined database mirDIP (accessed 30/10/2022) [36] was called from an R script and results were added to this ‘in-house’ combined miRNA::mRNA interaction database. From each database, interactions were selected using lists of expressed mRNAs and miRNAs obtained from our RNA-Seq and miRNA-Seq experiments as described above. Next it was attempted to estimate a probability of interaction in the form of a p-value using the information available in each online interaction database. In miRTarBase, the strength of evidence was used (strong: p = 0.01, weaker: p = 0.1); in TargetScan, the p-value was estimated using the context-score value. In TarBase, p = 0.01 was assigned to interactions obtained by Western Blot, qPCR or luciferase assays and p = 0.1 to other interactions. Furthermore, all TarBase interactions obtained via indirect evidence were filtered out and excluded. In miRWalk, binding probability was used as a measure of p-value (so p = 1 –binding probability). In mirDIP only interactions with an interaction score of at least ‘high’ were utilised and the ‘confidence score’ was mapped to a p-value. After combining the interactions from these databases, the final p-value was determined by taking the lowest (i.e. best) value of individual p-values per database. The rationale for this method is that if database A provides strong evidence (e.g. p = 0.01) and another database B provides weaker evidence (e.g. p = 0.1) then the information from database A is most likely to be correct and hence p = 0.01. After combining, the total miRNA::mRNA interaction database for human TM cells consisted of 3,711,158 interactions between 2757 unique miRNAs and 13784 unique genes.
Creation of a miRNA-mRNA interactome
Using sets of differentially expressed (FDR<0.05, |log2FC|>0.26) miRNAs and mRNAs together with the combined miRNA-mRNA database, a set of miRNA-mRNA interactions was created as depicted in Fig 1. This set of interactions was subjected to the following filters:
- Differential expression of the regulating miRNA should be in the opposite direction to the differential expression of the regulated mRNA so if the expression of a regulating miRNA is increased then the expression of the regulated mRNA should be decreased and vice versa (Inverse Expression Correlation).
- If a gene is indicated to be regulated by multiple miRNA species, then the miRNAs that have an expression level that is at least 8 orders lower than the expression level of the miRNA with the highest expression level (as expressed in log2CPM) are filtered out. This is a novel step intended to weed out less important miRNA regulators. The value 8 (equivalent to a 256x lower expression) was chosen based on the observation that the range of expression of a set of miRNAs regulating a particular gene can be in the order of 12 (when expressed as log2CPM) so 8 seems to be a fair threshold value.
- Both participants in a miRNA::mRNA interaction should be significantly differentially expressed when corrected for multiple observations, i.e. the false discovery rate (FDR) should be <0.05 and the absolute value of the log2 of fold-change |log2FC| >0.26.
Shown is the bioinformatics flow that has been used to create the miRNA-gene interactome in which blue boxes indicate data objects, grey boxes indicate processing steps and green boxes indicate external databases. Human trabecular meshwork cells are sequenced to determine both mRNA (genes) and miRNA species. Furthermore, an interaction database is compiled using publicly available interaction databases. This results in a full interaction network which can be utilised to extract subsets of the complete network for certain enrichment themes. Such a subset can be presented as heatmap, cytoscape network or table.
This FDR can be calculated in two ways: the simplest approach is to only use miRNA and mRNA species of which the FDR<0.05; this is the standard interactome similar to other studies [24].
Instead of correcting for multiple hypothesis testing by calculating the FDR for mRNA and miRNA separately, which would also correct for non-interacting miRNA-mRNA combinations, another option is to calculate the p-value per interaction based on the p-values of the regulating miRNA and regulated mRNA and then determine the FDR for the set of interactions. Assuming it is certain that a miRNA will interact with a certain mRNA and we require that both participants should be differentially expressed due to TGF-β1 treatment then the p-value for a certain interaction reflects the probability that either miRNA or mRNA is not differentially expressed and can be calculated as follows:
(1)
So, for example, if the p-value of a regulating miRNA is 0.04 and that of the regulated mRNA is 0.05 then the probability that at least one of these is in fact not differentially expressed is approximately 0.09. Using this method, an enhanced list of interactions and their p-values was obtained. Using this set, the (Benjamini-Hochberg) FDR was calculated per interaction and only interactions with FDR<0.05 were selected.
In summary the total set of interactions consists of standard interactions in which both miRNA and regulated gene are differentially expressed when corrected for multiple testing, so FDR<0.05 for both gene and miRNA, and the enhanced set of interactions in which the interaction is significant (FDR<0.05) when corrected for multiple testing. Because the number of interactions is different from the number of genes or miRNAs, the correction for multiple testing will be different.
At this point, we have arrived at a ‘Boolean’ interactome which shows that there may or not may be an interaction but there is no information about the interaction strength. To get some sense of the regulatory strength of an interaction, an interaction score S was estimated that contains miRNA molarity (in the form of CPM), Pearson correlation in expression between miRNA and mRNA and the evidence presented in the combined interaction database as follows:
(2)
in which evidence.exper = 1 if the considered interaction has been reported to be confirmed experimentally in a database, not just predicted. Parameter ‘standard.iact’ represents if the interaction originates from the standard rather than the enhanced interactome, in which case the value will be 1, otherwise 0. Parameter ‘abs(cor)’ represents the absolute value of the Pearson correlation between expression of miR and mRNA for this interaction and ‘mir.molarity’ is the molarity of the miRNA expressed as log2CPM.
The theoretical maximum value of this interaction score is about 7 for those interactions that (1) have been proven experimentally (2) both gene and miRNA have FDR<0.05 (3) correlation = -1 (4) high miRNA molarity, in practice limited to 15 or so. It should be noted that the Pearson correlation per sample could be calculated in our analysis because both RNA-Seq and miRNA-Seq were performed on the same samples. More details and considerations with respect to the use of these filters can be found in the discussion. The code to reproduce this analysis has been made available on github, see section ‘Data Availability’.
Enrichment analysis of interactions
To investigate regulation of subsets of mRNAs (genes) by miRNAs, functional enrichment analysis of differentially expressed genes was performed as described in [16]. In short, Enrichr was used to arrive at a set of enrichment terms from KEGG, Reactome and GO biological processes and molecular functions which were subsequently clustered based on gene overlap. Using these enrichment results and the gene-miRNA interactome it can be determined what miRNAs may play a role in a certain enrichment cluster by determining what cluster-genes are part of the interactome and by what miRNA species they are regulated. This analysis results in a matrix of enrichment clusters versus differentially expressed miRNA species. Another level of detail can be obtained by analysing gene-miRNA interactions per enrichment cluster resulting in an interaction matrix of genes versus miRNAs for one particular enrichment cluster of interest. To determine what differentially expressed genes in our dataset are transcription factors, gene list GO:0000981 called “DNA-binding transcription factor activity, RNA polymerase II-specific” was used. In addition to subsets that have been obtained after enrichment, it is also interesting to investigate what genes may be regulated by a set of miRNA species (such as the miR-29 family).
The code to reproduce this analysis has been made available on github, see section ‘Data Availability’.
Determining miRNA regulation of biological processes
To estimate what biological processes a certain differentially expressed miRNA may be regulating, a matrix of miRNAs versus functional enrichment clusters was created in which each matrix entry represents the number of genes regulated by a certain miRNA for a certain functional enrichment cluster divided by the total number of miRNA-regulated genes in that cluster so each matrix entry must be between 0.0 (this miRNA does not regulate any genes in this enrichment cluster) and 1.0 (all regulated genes in this cluster are also regulated by this miRNA).
Analysis of miRNA—transcription factor loops
By combining interactions in which a miRNA is regulating a transcription factor (TF) and interactions in which a TF regulates the expression of a miRNA, feedback loops may be detected. To determine what genes may be transcription factors, GO gene list “RNA polymerase II cis-regulatory region sequence-specific DNA binding” (GO:0000978) was used. To determine which miRNAs may be regulated by these TFs, TF–miRNA interactions in which TFs regulate miRNA expression were obtained from miRNet (accessed 29/04/2023) [37]. By combining both types of interactions, it becomes possible to detect double negative miRNA—| TF—| miRNA feedback loops. In such a loop, an upregulated TF should have a repressive effect on miRNA transcription and this repression will cause higher expression of the miRNA-controlled TF.
Ethics statement
The study was conducted in accordance with the Declaration of Helsinki and ethical approval was obtained from the local ethics review board of the University of Liverpool (RETH000833) in addition to national ethical approval from the UK National Research Ethics Service Committee North West—Haydock (IRAS ID 119105; REC reference 13/NW/0061, extended by IRAS ID 239185; REC reference 18/NW/0136). Written consent from relatives of the eye donors was obtained by the Liverpool Research Eye Bank.
Results
Differentially expressed mRNA and miRNA species in response to TGF-β1 stimulation
As reported earlier by our group, 689 mRNA species were found to be differentially expressed with a false discovery rate (FDR) < 0.05 in primary human TM cells after treatment with TGF-β1 (5ng/mL) for 24 hrs when compared to untreated controls [16]. Functional annotation of the set of differentially expressed genes set resulted in 219 enrichment terms originating from KEGG, Reactome, Gene Ontology (GO) biological processes and molecular functions. Using gene set overlap analysis several enrichment terms could be clustered into 53 functional clusters [16]. Furthermore, in the same samples, 107 miRNAs were found to be differentially expressed with p<0.05 [20] of which 52 miRNAs were differentially expressed when corrected for multiple testing i.e. FDR<0.05.
miRNA-mRNA interactome for TGF-β1 stimulated human TM cells
By combining miRNA-gene interactions from miRTarBase, TarBase, TargetScan, miRWalk and mirDIP followed by selecting only those genes and miRNAs expressed in our TM experimental datasets [16,20], a combined database of miRNA-gene interactions was created. Using this combined interaction database together with the list of 689 differentially expressed genes and 52 differentially expressed miRNAs, a set of possibly interacting miRNA-gene pairs was determined. This set was further purified by adding the constraint of opposite fold-change: if expression of the regulated gene is increased, expression of the regulating miRNA should be decreased and vice versa. Because both mRNA and miRNA species were determined for the same sample, also the correlation of expression of mRNA and miRNA over samples could be determined. This correlation was used as part of an interaction score but not as a filter. The procedure resulted in a bipartite network consisting of 52 miRNAs and 466 genes connected by 1202 likely interacting miRNA-gene pairs, see Fig 2 and S2 Table. Due to the constraint of inverse expression, the network consists of (at least) two unconnected subnetworks, one for up-regulated genes regulated by down-regulated miRs and one for down-regulated genes regulated by up-regulated miRNAs. In addition to this standard interactome which only consists of genes and miRs for which the false discovery rate FDR<0.05 also an enhanced interactome was determined in which the p-value per interaction was determined and correction for multiple testing was performed on the set of interactions rather than the sets of genes or miRs followed by a selection of those interactions for which FDR<0.05. This procedure resulted in 2151 additional interactions between 912 genes and 85 miRNAs. Of these interactions, 1054 originate from experimental database evidence and 1097 from predicting database evidence. The total (standard + enhanced) interactome consists of 3353 interactions.
The node colour indicates if the RNA species (miRNA or mRNA) is upregulated (orange) or downregulated (blue) and the colour intensity has been mapped to log2FC. The node size has been mapped to log2CPM. Edge width has been mapped to magnitude of correlation.
MiRNAs with the highest number of interactions (hub-miRs) were hsa-let-7e-5p (113 interactions), hsa-miR-335-5p (100 interactions), hsa-miR-15a-5p (96 interactions), hsa-miR-20a-5p (96 interactions), hsa-miR-218-5p (96 interactions), hsa-miR-181a-5p (94 interactions), hsa-miR-29b-3p (90 interactions), hsa-miR-122-5p (87 interactions), hsa-miR-19b-3p (87 interactions), hsa-miR-30b-5p (86 interactions), hsa-miR-181b-5p (84 interactions), hsa-miR-182-5p (81 interactions), hsa-miR-125a-3p (80 interactions), hsa-miR-26a-5p (79 interactions), hsa-miR-29c-3p (77 interactions), hsa-miR-195-5p (74 interactions), hsa-miR-204-5p (74 interactions), hsa-miR-29a-3p (74 interactions), hsa-miR-19a-3p (63 interactions) and hsa-miR-744-5p (63 interactions). The average number of interactions per miRNA was 39.0 and each gene was on average regulated by 2.95 miRNAs.
miRNA regulation of biological processes altered by TGF-β1 in human TM cells
A matrix of miRNAs versus enrichment clusters was created as described in the Methods section, see Fig 3 and S1 Table. It appears that a miRNA cluster consisting of hsa-miR-335-5p, hsa-miR-29a-3p, hsa-miR-29b-3p, hsa-miR-29c-3p, hsa-miR-378a-3p, hsa-miR-15a-5p, hsa-hsa-miR-195-5p, hsa-miR-101-3p, hsa-miR-146b-5p and hsa-miR-20a-5p co-regulate a cluster of functional themes consisting of amongst others “eye morphogenesis”, “regulation of cell migration”, “phagosome” and “cellular response to hypoxia”. Another cluster of functional themes consisting of “epithelial to mesenchymal transition”, “unfolded protein response”, “fibroblast proliferation; G1/S”, “ECM remodelling” and several others was co-regulated by almost all differentially expressed miRNAs except hsa-let-7e-5p and hsa-miR-451a. Other miRNAs that seem to co-regulate similar processes are hsa-miR-204-5p and hsa-miR-218-5p; hsa-miR-19a-3p and hsa-miR-19b-3p; hsa-miR-432-5p and hsa-miR-708-5p and members of the miR-29 family. In the next sections the specific gene-miRNA interactions per functional theme will be investigated for several clusters believed to be of specific interest to XFG.
The colour intensity of this heatmap corresponds to the number of mRNAs (genes) regulated by a certain miRNA divided by the total number of miRNA-regulated mRNAs for a certain biological process and is therefore a measure of how strongly a certain biological process is regulated by a certain miRNA. Only miRNAs of the standard interactome have been shown (FDR<0.05).
miRNA regulation of TGF-β1 induced expression of exfoliation material genes
Our analysis showed that the expression of 10 exfoliation material (XFM) genes may be regulated by 22 miRNAs via 24 interactions in the standard interactome and 12 additional interactions in the enhanced interactome, see Fig 4 and Table 1. Expression of fibrillin 1 (FBN1), ADAM19, elastin (ELN) and myosin light chain 6 (MYL6) was regulated by the miR-29 family (hsa-miR-29a-3p, hsa-miR-29b-3p, hsa-miR-29c-3p). In addition, ELN was also regulated by hsa-miR-195-5p and possibly hsa-miR-29c-5p (enhanced interactome) and ADAM19 was also regulated by hsa-miR-335-5p and hsa-miR-30b-5p. FBN1 was regulated by the miR-29 family members and also hsa-miR-19b-3p, hsa-miR-660-5p, hsa-miR-1-3p and hsa-miR-340-5p of which hsa-miR-29c-3p had the strongest correlation (-0.93) and highest interaction score (S = 5.5) and substantial interaction evidence and miR molarity. Versican (VCAN) was regulated by hsa-miR-101-3p, hsa-miR-335-3p, hsa-miR-218-5p of which hsa-miR-101-3p had the highest interaction score (S = 5.3) due to good correlation combined with strong database evidence and miRNA molarity. Vitronectin (VTN) may be regulated by hsa-miR-1275 (enhanced interactome, predicted database interaction), IRAG1 may be regulated by hsa-miR-1275, hsa-miR-139-5p, hsa-miR-183-5p and hsa-miR-1294 of which hsa-miR-1275 showed the best correlation and all interactions originate from predicting databases. Both lysyl oxidase like 1 (LOXL1) and Latent Transforming Growth Factor Beta Binding Protein 2 (LTBP2) may be regulated by hsa-miR-335-5p. LDHA was regulated by hsa-miR-4683, hsa-miR-378a-3p and hsa-miR-502-3p of which hsa-miR-378a-3p showed the highest interaction score (S = 4.0).
Heatmap of the interactome in which the heatmap colour maps to the interaction score which is calculated from database evidence, miRNA expression and miRNA-gene correlation using Eq (2). The annotation bars on the left represent miRNA fold-change as log2FC (in red/blue) and log2CPM (grey-scale) and the annotation bar at the top represents log2FC of genes (in red/blue).
miRNA regulation of TGF-β1 induced extracellular matrix remodelling
This relatively large sub-interactome contains 250 interactions between 60 miRNAs and 72 genes all involved in extracellular matrix remodelling. The standard interactome showed 149 interactions between 44 genes and 33 miRs while the enhanced interactome showed 101 interactions between 50 genes and 52 miRs, see Fig 5 and Table 2. As shown in Fig 5 a large cluster consists of miR-29 family members miR-29a-3p, miR-29b-3p and miR-29c-3p which contribute to the regulation of several collagen species COL4A1, COL4A2, COL4A4, COL5A1, COL5A2 (and additional species such as COL1A1 regulated by hsa-miR-218-5p, hsa-miR-29b-3p), metallopeptidases ADAM12, ADAM19, MMP15 and MMP24, and amine oxidases LOX and LOXL2. In addition, SPARC, P3H1, fibrillin 1 (FBN1), elastin (ELN) and SERPINH1 were found to be members of this cluster.
Heatmap of the interactome in which the heatmap colour maps to interaction score which is calculated from database evidence, miRNA expression and miRNA-gene correlation using Eq (2). The annotation bars on the left represent miRNA fold-change as log2FC (in red/blue) and log2CPM (grey-scale) and the annotation bar at the top represents log2FC of genes (in red/blue).
Prolyl 4-Hydroxylase Subunit P4HA1 was shown to be regulated by hsa-miR-335-5p, hsa-miR-452-3p, hsa-miR-30b-5p while P4HA2 was regulated by hsa-miR-30b-5p. Peroxidasin (PXDN), involved in ROS-dependent crosslinking of basement membrane component COL4, was shown to be regulated by hsa-miR-19a-3p, hsa-miR-19b-3p, hsa-miR-29b-3p, hsa-miR-29c-3p. FURIN, an endoprotease of the TGF-β1 precursor, was shown to be upregulated by downregulation of hsa-miR-15a-5p, hsa-miR-195-5p, hsa-miR-20a-5p. SMOC1 was indicated to be mainly regulated by hsa-miR-19a-3p, hsa-miR-19b-3p, hsa-miR-204-5p and hsa-miR-20a-5p.
Cell adhesion molecule NCAM1 was indicated to be regulated by hsa-miR-125a-3p, hsa-miR-122-5p, hsa-miR-210-3p of which hsa-miR-122-5p showed the highest interaction score (S = 4.6) due to good correlation (r = -0.7) and strong interaction database evidence. ICAM1 was found to be regulated by hsa-miR-146a-5p. In the integrin family, ITGAV was found to be regulated by hsa-miR-146b-5p, hsa-miR-340-5p and hsa-miR-218-5p of which the latter has the highest interaction score. ITGA2 was regulated by hsa-miR-19a-3p, hsa-miR-19b-3p, hsa-miR-20a-5p, hsa-miR-335-5p, hsa-miR-378a-3p, hsa-miR-15a-5p and hsa-miR-195-5p of which hsa-miR-15a-5p and hsa-miR-195-5p show the strongest correlation. ITGB3 was regulated by hsa-miR-30b-5p and hsa-miR-204-5p while ITGB5 was shown to be regulated by hsa-miR-335-5p (experimental database evidence) and by hsa-miR-335-3p (predicted database evidence). From the enhanced interactome, ITGB1 was indicated to be regulated by hsa-miR-20a-5p, hsa-miR-218-5p and hsa-miR-29b-3p and ITGB8 was regulated by mostly hsa-miR-181b-5p, hsa-miR-181a-5p, hsa-miR-432-5p, hsa-miR-708-5p and hsa-let-7e-5p.
miRNA regulation of TGF-β1 induced expression of signal molecules and downstream pathways
The interactome analysis resulted in a total of 188 interactions between 63 miRs and 50 genes functionally enriched in signal molecules and their pathways. The total subset of interactions consisted of 83 standard interactions between 29 genes and 29 miRs and 105 additional interactions between 41 genes and 55 miRs from the enhanced interactome, see Fig 6 and Table 3. Fig 6 shows the interactions resulting from the standard interactome method. Starting with the canonical TGF-β pathway, receptor protein gene TGFBR1 was regulated by a large number of miRNAs namely hsa-miR-101-3p, hsa-miR-204-5p, hsa-miR-20a-5p and hsa-miR-335-5p. When considering the correlation in expression and molarity the key regulators are likely hsa-miR-101-3p, hsa-miR-204-5p and hsa-miR-20a-5p. Gremlin1 (GREM1) was regulated by hsa-miR-204-5p, hsa-miR-218-5p and perhaps hsa-miR-335-5p. Of these miRNAs, hsa-miR-204-5p has the highest correlation and is the only interaction that originates from an experimental database. LEFTY2 was indicated to be regulated by hsa-miR-502-3p (only predicted database evidence). When considering standard interactions, SMAD7 was regulated by hsa-miR-15a-5p, hsa-miR-195-5p, hsa-miR-20a-5p and hsa-miR-99a-5p. Of these miRNAs, both hsa-miR-15a-5p and hsa-miR-20a-5p had high interaction scores. SMAD3 appeared in the enhanced interactome (not shown in Fig 6), and a key regulator was hsa-miR-708-5p (S = 4.0). Another miRNA that may regulate SMAD3 is hsa-miR-744-5p. In the standard interactome, CCN2 (also called connective tissue growth factor: CTGF) was regulated by hsa-miR-19a-3p, hsa-miR-19b-3p and hsa-miR-218-5p. In addition, the enhanced interactome showed regulation by hsa-miR-18a-5p and hsa-miR-26a-5p. Of these miRNAs, hsa-miR-218-5p had the highest interaction score (S = 4.9) due to good correlation (r = -0.6) and high molarity. Moving to genes involved in the Wnt pathway, WNT2B was indicated to be regulated by hsa-miR-122-5p (only predicted database evidence). Furthermore, WNT5B was regulated by hsa-miR-660-5p and WNT7B was regulated by hsa-miR-335-5p, hsa-miR-19a-3p, hsa-miR-19b-3p, hsa-miR-378a-5p and hsa-miR-505-5p which were all experimentally verified interactions. DKK1 was regulated by hsa-miR-335-5p and hsa-miR-1-3p. Frizzled class receptor 6 (FZD6) was regulated by hsa-miR-101-3p, hsa-miR-15a-5p, hsa-miR-19b-3p, hsa-miR-20a-5p and, from the enhanced interactome, also hsa-miR-21-5p. Of these miRNAs the interaction score was strongest for hsa-miR-21-5p (S = 5.6). AXIN2 was regulated by hsa-miR-1275 with high interaction confidence and strong correlation (r = -0.95) and S = 2.9.
Heatmap of the interactome in which the heatmap colour maps to interaction score which is calculated from database evidence, miRNA expression and miRNA-gene correlation using Eq (2). The annotation bars on the left represent miRNA fold-change as log2FC (in red/blue) and log2CPM (grey-scale) and the annotation bar at the top represents log2FC of genes (in red/blue).
The expression of interleukin IL11 was regulated by hsa-miR-204-5p and hsa-miR-502-3p of which hsa-miR-204-5p showed highest interaction score (S = 4.1) due to strong correlation (-0.67) and strong database evidence. Cytokine LIF was regulated by hsa-miR-19a-3p, hsa-miR-19b-3p, hsa-miR-20a-5p (best correlation) and miR-29c-3p. Furthermore, the enhanced interactome showed regulation by hsa-miR-18a-5p and hsa-miR-26a-5p so both the 17–92 cluster and the miR-29 family seemed to be involved in regulating LIF. Growth factor EPGN may be regulated by hsa-miR-335-3p. The enhanced interactome showed regulation of growth factor FGF2 by hsa-miR-101-3p, hsa-miR-15a-5p, hsa-miR-19b-3p and hsa-miR-218-5p. These interactions all originated from experimental databases. EGFR ligand HBEGF was regulated by hsa-miR-29a-3p, hsa-miR-29b-3p, hsa-miR-29c-3p and hsa-miR-335-5p and hsa-miR-29c-3p showed the strongest correlation in expression. VEGFA was regulated by a large number of miRNAs of which hsa-miR-378-3p and hsa-miR-101-3p showed the highest interaction score (S = 5.5) but hsa-miR-15-5p, hsa-miR-195-5p, hsa-miR-20a-5p and hsa-miR-29c-3p also showed good correlation. VEGFC was regulated by hsa-miR-15-5p and hsa-miR-195-5p as well and in addition by hsa-miR-218-5p. Angiotensinogen (AGT) was indicated to be regulated by hsa-miR-181b-5p (predicted interaction, not experimental).
miRNAs contribute to the regulation of oxidative stress in response to TGF-β1 in human TM cells
The interactome analysis indicated that miRNAs contribute to the regulation of mRNA expression of several genes that are associated with oxidative stress, antioxidant activity and oxidative activity. In total 66 interactions between 23 genes and 37 miRs were found of which 25 interactions were between 11 genes and 19 miRs in the standard interactome and 41 interactions between 18 genes and 29 miRs in the enhanced interactome, see Fig 7 and Table 4 below. NADPH Oxidase 4 (NOX4) was regulated by hsa-miR-99a-5p and showed a high interaction score (S = 4.9) and good correlation of -0.91 suggesting that this miRNA may be an important regulator. Superoxide dismutase 2 (SOD2) was regulated by amongst others hsa-miR-24-2-5p and hsa-let-7e-5p. Both interactions show up in both experimental and predicting miRNA databases. Superoxide 3 (SOD3) may be regulated by hsa-miR-744-5p (this interaction was only supported by predicted database evidence). Of the glutathione peroxidases, GPX1 was downregulated by hsa-miR-125-3p (high-confidence prediction) and hsa-miR-3065. GPX7 was downregulated by all 3 members of the miR-29 family: hsa-miR-29a-3p, hsa-miR-29b-3p, hsa-miR-29c-3p (interactions both experimentally verified and predicted database interaction). GPX3 was upregulated due to downregulation of hsa-miR-30b-5p (predicted by miRWalk and Tarbase) while GPX8 was upregulated due to downregulation of hsa-miR-340-5p with high interaction score (S = 4.3) and strong correlation (r = -0.75). GSR was regulated by hsa-let-7e-5p and hsa-miR-183-5p in which hsa-let-7e-5p had the highest score and molarity, but correlation was not very strong which may indicate that there are other non-miRNA regulators that play a role. PTGS2 (COX2) was regulated by hsa-miR-204-5p, hsa-miR-335-5p, hsa-miR-26a-5p, 101-3p and hsa-miR-146b-5p of which hsa-miR-26a-5p showed the strongest interaction score (S = 5.2). In the enhanced interactome, ERO1A was regulated by hsa-miR-551a, hsa-miR-18a-5p and hsa-miR-26a-1-3p with both hsa-miR-18a-5p and hsa-miR-26a-1-3p appearing as possible major regulators; judged by their interaction scores.
Heatmap of the interactome in which the heatmap colour maps to interaction score which is calculated from database evidence, miRNA expression and miRNA-gene correlation using Eq (2). The annotation bars on the left represent miRNA fold-change as log2FC (in red/blue) and log2CPM (grey-scale) and the annotation bar at the top represents log2FC of genes (in red/blue).
miRNA regulation of TGF-β1 induced unfolded protein response
The interactome showed 83 interactions between 32 miRNAs and 22 genes involved in the unfolded protein response. The standard interactome analysis resulted in 35 interactions between 10 genes and 16 miRNAs and the enhanced interactome yielded 48 interactions between 17 genes and 27 miRNAs, see S1 Fig and Table 5 below. The enhanced interactome indicated that HSPA5 (also known as BiP or GRP78) was regulated by hsa-miR-29a-3p, hsa-miR-30b-5p, and hsa-miR-335-5p and all these interactions originated from experimental interaction databases. Of these miRNAs, hsa-miR-30b-5p showed the highest interaction score (S = 4.0). EIF2AK3 (also known as PERK) was regulated by hsa-miR-204-5p, hsa-miR-335-5p, hsa-miR-26a-5p of which hsa-miR-204-5p showed the strongest correlation (r = -0.94) and hsa-miR-26a-5p the highest interaction score (S = 5.4). Both these interactions originated from experimental interaction databases. HSP90B1, also known as GRP94, was indicated to be controlled by hsa-miR-15a-5p, hsa-miR-218-5p, hsa-miR-101-3p and hsa-miR-335-5p. Heat shock protein DNAJB11 was regulated by three members of the miR-29 family: hsa-miR-29a-3p, hsa-miR-29b-3p and hsa-miR-29c-3p. UPR-related transcription factor ATF3 was indicated to be regulated by hsa-miR-181a-2-3p (only predicted evidence). FKBP14 is involved in collagen protein folding and was indicated to be regulated by hsa-miR-101-3p, hsa-miR-335-5p, hsa-miR-660-5p, hsa-miR-146b-5p, hsa-miR-20a-5p, hsa-miR-29b-3p and hsa-miR-29c-3p. HYOU1 was indicated to be regulated by hsa-miR-15a-5p (r = -0.88) and hsa-miR-146b-5p (r = -0.43). HERPUD1 was regulated by hsa-miR-335-5p and hsa-miR-340-5p. Within the cluster of UPR-genes, hsa-miR-335-5p seemed to be a major ‘hub’ miRNA and controlled 10 UPR genes out of 22 including key UPR genes HSPA5 (BiP) and EIF2AK3 (PERK). This also followed from the miRNA-enrichment matrix analysis described earlier.
miRNA regulation of TGF-β1 induced changes in transcription factor expression
This subset contains 175 interactions between 53 transcription factors (and some other DNA-binding proteins) and 52 miRs, namely 63 interactions between 18 genes and 30 miRs in the standard interactome and 112 interactions between 46 genes 43 miRs in the enhanced interactome, see S2 Fig and Table 6 below. MiRNAs with the highest numbers of interactions were hsa-miR-129-5p (11), hsa-let-7e-5p (10), hsa-miR-122-5p (9), hsa-miR-495-3p (9), hsa-miR-181a-5p (8), hsa-miR-182-5p (8), hsa-miR-125a-3p (6). It is interesting to see that miRNA hsa-miR-122-5p is highly ranked in this subset because this was the miRNA with the lowest FDR and high fold change (FDR = 9e-15, log2FC = 3.9) in our miRNA dataset. Forkhead box D1 (FOXD1) was indicated to be regulated by hsa-miR-99a-5p, hsa-miR-30b-5p, hsa-miR-378a-3p of which the latter showed the highest interaction score. The enhanced interactome showed that FOXO1 may be downregulated due to upregulation of hsa-miR-183-5p and hsa-miR-223-3p. Downregulated histone deacetylase 5 HDAC5, a co-repressor of RARA, was indicated to be regulated by hsa-miR-146a-5p and hsa-miR-495-3p (predicted database evidence only). Epithelial-mesenchymal transition (EMT) associated transcription factor ZEB1 was indicated to be regulated by hsa-miR-3065-5p, hsa-miR-183-5p, hsa-miR-223-3p in the standard interactome of which hsa-miR-183-5p showed the strongest correlation. ZEB1 was also indicated to be controlled by hsa-miR-200c-3p but the FDR of this interaction was just above the limit of FDR = 0.05 (0.054) and was therefore pruned. ZEB1 paralog ZEB2 was regulated by a different set of miRNAs: hsa-miR-122-5p, hsa-miR-125a-3p, hsa-miR-129-5p, hsa-miR-181a-5p, hsa-miR-182-5p (all in the enhanced interactome).
miRNA–transcription factor loops
Using the method described earlier, two double negative miRNA—| TF—| miRNA loops were identified: hsa-miR-15a-5p forms a double negative feedback loop with E2F7 and hsa-miR-183-5p forms a double-negative feedback loop with ZEB1. See S3 Table for TF—| miRNA interactions and S2 Fig for the corresponding miRNA—| TF interactions.
Discussion
Our analysis of miRNA-mRNA interactions provides a bioinformatics-based hypothesis-generating prediction of part of the regulatory mechanisms causing differential gene expression in normal human TM tissue when treated with TGF-β1. Normal human TM tissue treated with TGF-β1 shows overlap with features of XFG pathogenesis and may therefore serve as a simplified model for the TM in this condition [16]. As such the miRNA-gene interactome in this work may also partially model the regulation of changes in gene expression observed in XFG and may be a first step in explaining the regulatory function of differentially expressed miRNA species in XFG when compared to normal TM tissue. Many studies have been performed to determine what miRNA species are differentially expressed in XFG and other forms of glaucoma while other studies focussed on the regulatory targets for one species of miRNA [2,19]. To our knowledge, this is the first study that combines genome-wide miRNA and mRNA (gene) expression into one genome-wide interactome for normal TM cells treated with TGF-β1.
It is interesting to compare those miRNAs with the highest number of miRNA-gene interactions (hub-miRs) found in this study with miRNAs previously associated with XFG or glaucoma. MiRNA hsa-miR-20a-5p was previously reported to be downregulated in senescent TM cells [38]. hsa-miR-181a-5p was reported to be upregulated in TM cells subjected to mechanical stretch [39]. Changes in hsa-miR-15a-5p were observed in senescent TM cells [38,39]. hsa-miR-19b-3p was not previously associated with glaucoma to our knowledge. hsa-miR-30b-5p has not been reported in the context of glaucoma before but some close relatives have such as hsa-miR-30c-5p in plasma of XFG patients [40]. Upregulation of hsa-miR-122-5p was reported in the aqueous humour [41] and plasma [40] of XFG patients. MiRNA hsa-miR-181b-5p was not previously associated with glaucoma but the other strand hsa-miR-181b-3p has been reported in AH of glaucoma patients [42]. hsa-miR-26a-5p was previously observed in plasma of XFG patients [40]. hsa-miR-218-5p has not been associated with glaucoma before but has been reported in other fibrotic conditions [43].
Downregulation of mir-29 family members hsa-miR-29a-3p, hsa-miR-29b-3p, hsa-miR-29c-3p was previously observed in iridial and ciliary body tissue specimens obtained from XFG patients when compared to controls [44] and in lamina cibrosa (LC) tissue from POAG patients [45]. MiRNA hsa-miR-204-5p was reported to be downregulated in TM tissue subjected to oxidative stress and senescence [18] and in the glaucomatous retina [46] but upregulated in plasma of XFG patients [40]. MiRNA hsa-miR-18a-5p was reported to be downregulated due to stress-induced senescence [38] and altered in plasma of XFG patients [40]. MiRNA hsa-miR-195-5p and hsa-miR-139-5p were reported to be altered due to stress-induced senescence in TM tissue [38].
Previous findings indicate that glaucoma induces differential expression of a large number of miRNAs when compared to normal controls [19] and our results show that a large fraction of those differentially expressed miRNAs also appear in normal TM tissue when treated with TGF-β1. To investigate what the regulatory effect of these differentially expressed miRNA species on gene expression may be, we constructed an interactome describing putative regulatory relationships between differentially expressed miRNAs and oppositely differentially expressed genes.
Coupling functional enrichment of differentially expressed genes to their putative regulating miRNAs indicates that miRNAs appear to be involved in the regulation of all altered biological processes in TM tissue when treated with TGF-β1. Individual miRNA species usually regulate a wide range of genes and biological processes, and each individual biological process or pathway is regulated by several miRNAs. Therefore, it does not usually seem to be possible to pinpoint one key miRNA which is mostly responsible for a change in a biological process. Rather, it seems that groups of genes (modules) or biological processes are co-regulated by the combined effect of multiple co-regulating miRNAs [47]. For example, one previously studied case of co-regulation by multiple miRNAs is the miR-15/107 group [48], which indeed also showed co-regulation of multiple biological processes by members hsa-miR-15a-5p and hsa-miR-195-5p in our results. Another example is the regulation of ECM remodelling by members of the miR-29 family and ECM remodelling [49].
Our results showed upregulation of several XFM components due to downregulation of miR-29 members hsa-miR-29a-3p, hsa-miR-29b-3p and hsa-miR-29c-3p. The miR-29 family plays a key role in regulation of extracellular matrix components [49] and XFM also contains several ECM-related proteins. Regulation of LOXL1 by hsa-miR-335-5p has not been reported in glaucoma so far but has previously been reported in breast cancer tissue using microarrays so this may be a downstream regulatory effect [50]. Regulation of ADAM19 by hsa-miR-335-5p as a result of TGF-β1 treatment has been reported in HK-2 cells [51]. Regulation of elastin (ELN) by hsa-miR-195-5p has not been reported before in the context of XFG but was experimentally demonstrated in aortic aneurysmal disease [52]. Regulation of versican (VCAN) by hsa-miR-218-5p, hsa-miR-101-3p was identified in our results but has not been reported in XFG or other forms of glaucoma to our knowledge. Overall, the miR-29 family and hsa-miR-335-5p have the highest number of interactions in the XFM functional cluster of genes.
Oxidative stress is associated with XFG [11]. Our results show that in the trabecular meshwork, several key genes involved in oxidative stress and antioxidant activity may be regulated by miRNAs in response to TGF-β1 treatment. In agreement with our findings, regulation of NOX4 by hsa-miR-99a-5p was also reported in human endothelial cells [53]. Also, several glutathione peroxidases are regulated by miRNAs. A bioinformatics study on this topic was performed by [54] which suggested regulation of GPX3 by hsa-miR-30b-5p and GPX1 by hsa-miR-125a-3p which agrees with our results. GPX7 has been reported to be regulated by hsa-miR-29b [55] which is in agreement with our findings. Regulation of peroxidasin PXDN by hsa-miR-29a has been experimentally confirmed in T-cells [56]. PTGS2 (also known as COX2) was experimentally shown to be regulated by hsa-miR-26a-5p [57].
Treatment of TM cells with TGF-β1 causes differential expression of several extracellular signal molecules and their related downstream genes [16,58]. Our interactome analysis indicated that TGF-β1 induced differentially expressed miRNAs contribute to this differential gene expression. TGFBR1 (also known as ALK5) regulation by hsa-miR-101-3p was demonstrated previously [59]. TGFBR1 regulation by hsa-miR-204-5p was also reported in [60] and regulation by hsa-miR-20a-5p was demonstrated previously as well [61] in endothelial cells in agreement with our results. SMAD3 has been shown to be downregulated by upregulation of hsa-miR-708-5p [62] which agrees with our findings. SMAD7 was demonstrated to be directly regulated by hsa-miR-15b-5p [63] which supports regulation of SMAD7 by hsa-miR-15a-5p in our results as these miRNAs share the same seed sequence. SMAD7 expression is known to be upregulated in response to TGF-β pathway activation via SMAD3 [64] so it appears that SMAD7 expression is regulated by both transcription factors and miRNAs in response to TGF-β1 treatment which may point to the existence of a feed-forward motif. CCN2 (also known as connective tissue growth factor CTGF) regulation by hsa-miR-218-5p was demonstrated experimentally [65] which agrees with our results. CCN2 was also shown to be controlled by hsa-miR-19 members [66] and hsa-miR-18a-5p [21] which also agrees with our results. Vascular Endothelial Growth Factor A VEGFA was reported to be regulated by many miRNAs including hsa-miR-101-3p [67] and by members of the miR-29 family [68] which is in keeping with our findings. The Wnt/β-catenin pathway is associated with glaucoma, IOP regulation and EMT [69–71]. Several key genes of the Wnt/β-catenin pathway were reported to be regulated by miRNAs and several key regulators also appear in our results such as hsa-miR-335-5p [72] and hsa-miR-122-5p [73]. Interleukin IL11 is essential for development of kidney and heart fibrosis and is involved in EMT [74,75]. Regulation of IL11 by hsa-miR-204-5p was demonstrated previously in TM tissue [76] and other tissue types [77] and also appears in our interactome.
Activation of the unfolded protein response has been implicated in glaucoma [78]. Several key UPR genes have been reported to be regulated by miRNAs [79].
Our results indicate that EIF2AK3 (also known as PERK) is regulated by several miRNAs including hsa-miR-204-5p which agrees with the literature [80]. Regulation of HSPA5 (also known as BiP, GRP78) by miR-30a was reported before [81] and may in part support our finding of regulation of HSPA5 by hsa-miR-30b-5p. According to our interactome a key miRNA involved in UPR regulation appears to be hsa-miR-335-5p but this has not been explicitly mentioned in literature as yet.
Remodelling of the extracellular matrix is associated with both TM [82] and LC [83] in glaucoma. Regulation of components, enzymes and other proteins of the ECM by miRNAs has been reported extensively in literature [19,23]. Upregulation of collagen species due to downregulation of miRNA-29 family members hsa-miR-29a-3p, hsa-miR-29b-3p and hsa-miR-29c-3p has been reported in several studies [49] and agrees with our findings. In addition, hsa-miR-146b-5p and hsa-miR-218-5p may play a role in regulating these genes. In keeping with our findings, regulation of SPARC by miR-29 family members was reported before in vitro in TM cells in response to TGF-β2 treatment [84].
Transcription factor ZEB1 is associated with fibrosis [85–87] and, in agreement with our interactome, has been reported to be regulated by hsa-miR-183-5p and hsa-miR-223-3p with strong experimental evidence [35]. Related transcription factor ZEB2 is also associated with fibrosis and fibroblast to myofibroblast transdifferentiation [88]. There is experimental evidence that ZEB2 may be regulated by hsa-miR-181a-5p [89] and furthermore several predictive databases propose regulation by hsa-miR-122-5p but without experimental evidence to date. Kruppel Like Factor 13 KLF13 is an antifibrotic transcription factor and is downregulated by TGF-β1 treatment in our study [90] and the interactome only showed predicted interactions of KLF13. KLF15 is also antifibrotic and downregulated and inhibits expression of pro-fibrotic cytokine CCN2 (CTGF) [91]. We demonstrated that KLF15 was downregulated due to TGF-β1 and has previously been reported to be regulated by hsa-miR-181a-5p in six different publications according to Tarbase [34]. Transcription factor FOXO1 is involved in several stress responses [92] and is regulated by hsa-miR-183-5p [93] and hsa-miR-223-3p [94] with strong experimental evidence which is in agreement with our interactome. Transcription factor E2F7 is involved in cell cycle control. There is experimental PAR-CLIP based evidence from at least six different sources that E2F7 can be regulated by hsa-miR-15a-5p, see interaction MIRT502570 in miRTarBase [35].
Two double-negative feedback motifs between miRNAs and transcription factors were identified in our analysis. The double-negative feedback loop consisting of hsa-miR-183-5p and ZEB1 has been reported before [86] while the feedback loop consisting of hsa-miR-15a-5p and E2F7 may be a novel finding. Depending on the strength of regulation, a double-negative feedback loop, which is effectively a positive feedback loop, can bring the cell into a different stable state [95]. In the context of glaucoma, XFG and fibrosis, especially the process of EMT is relevant in which ZEB1 plays an important role [87]. Apart from hsa-miR-183-5p, also the miR-200 family has been reported to regulate ZEB1 [96] with a double-negative feedback motif. This interaction did not show up in our results because of the higher FDR of hsa-miR-200c-3p although the p-value of this miR is p = 0.03.
A large proportion of the interactions found in our analysis originates from experiment-based databases and is therefore already validated to some extent albeit usually not in trabecular meshwork cells. Several publications do cover part of our interactome and therefore may serve as validation. In agreement with our interactome, regulation of ELN, COL5A2, FSTL1 and COL1A1 was demonstrated experimentally using a miR-29b mimic and antagomir in primary human skin fibroblasts and other models in the context of treatment of keloid formation and fibroplasia using miR-29b mimic Remlarsen [97]. Further validation of our interactome is provided by another experiment in which human trabecular meshwork cells were transfected with hsa-miR-29b-3p [23] which demonstrated regulation of ADAM12, COL5A1, COL1A1 and SPARC by this miRNA. Another publication that may serve as validation of our interactome is provided by a similar in vitro experiment involving hsa-miR-204-5p which demonstrated regulation of IL11, SERPINE1, CXCL3, AP1S2, BIRC2, SERP1, SERPINE1, PLAUR by this miRNA in TM tissue [76].
Our in silico analysis combines miRNA-Seq and mRNA-Seq (gene) data of the same samples together with putative interactions obtained from experimental and predicting miRNA-mRNA interaction databases to arrive at an interactome that describes how genes in TM cells may be regulated by miRNAs in response to TGF-β1 treatment. Many interactions in the interactome have already been described in other publications which shows the predictive value of our interactome. Despite this fact, there are certain limitations to this study.
The experimental interaction information available in interaction databases is usually not determined in TM tissue while it is known that miRNA regulatory strength of genes has cell type and cell context specific characteristics [47,98].
Our initial interactome was Boolean by nature, i.e. an interaction between miRNA is reported to be either present or not present, but there was no indication of the strength of interaction. In an attempt to predict the strength of interaction beyond just a Boolean yes/no we developed the interaction score which attempts to give an indication of the strength of regulatory action of a certain miRNA-gene pair. One parameter in this score is the molarity of the miRNA (in the form of log2CPM), from miRNA-target reaction equations one would expect that the magnitude of differential mRNA expression due to miRNA regulation would depend on the (local) concentration of the regulating miRNA and the binding strength between both species [99,100].
Another parameter which may be a measure of miRNA regulatory strength is correlation between regulating miRNA and regulated mRNA may be calculated [24] and indeed if the gene is regulated by one dominant regulatory miRNA this interaction is likely to show the best (closest to -1.0) correlation but by nature of miRNA regulation, a gene is usually regulated by the combined effort of multiple miRNAs [98]. If these miRNAs have similar regulatory strength and are uncorrelated amongst themselves, then this can result in weaker (i.e. closer to 0.0) correlation values.
Besides miRNAs, also other regulators act on mRNA expression such as lncRNAs and transcription factors which have not been taken into account in this analysis. It is possible that a gene is regulated in one direction by one type of regulator and in the opposite direction by a miRNA. If this is the case, then this miRNA regulation may be missed.
The local concentration of miRNA and mRNA species may differ from what would be expected from the number of reads obtained from RNA-Seq because both miRNA-Seq and mRNA-Seq are bulk sequencing procedures which show the averaged expression of multiple cell-types while the TM is known to consist of several cell-types [101]. This unknown compartmentalisation may therefore lead to false positives if the supposedly interacting miRNA and mRNA in reality reside in different compartments or cell types. To partially solve this problem, it would be better to use single-cell sequencing.
MiRNAs can exert their regulatory role as part of several different regulatory motifs [102], an overview is shown in Fig 8; motif A shows the basic regulatory motif. From this motif, it follows that when the miR expression increases, the target expression should decrease which is the rationale behind applying the constraint of opposite expression as we have done. This simple motif may be valid in case there is one dominant regulating miR or group of miRs in concert such as several interactions involving members of the miR-29 family. A target mRNA may be regulated by miRs of approximately equal regulatory strength (motif E) which may be partially missed by our method if the regulating miRs have unequal signs of expression (some up, some down). Other motifs that will be missed by our method are (B) negative feed-back loop and both incoherent feed-forward loops (C, D) because in these cases the sign of expression of miR and target may be the same. Motifs that will usually be detected by our method are (F) double-negative feedback-loop, and both coherent feed-forward loops (G, H). To complicate things further, a certain regulated target mRNA may be part of multiple different motifs leading to a very complex network indeed of which the interaction parameters are often unknown or inaccurate.
In which miR = miRNA, gene = mRNA, TF = transcription factor. (A) Basic miRNA-gene interaction (B) negative feedback loop (C) incoherent feedforward loop with TF as dual regulator of both miRNA and regulated gene (D) incoherent feedforward loop with miRNA as dual regulator of both TF and regulated gene (E) regulation of one gene by multiple miRNAs (F) double-negative feedback loop (G) coherent feedforward loop with TF as dual regulator (H) coherent feedforward loop with miRNA as dual regulator. Adapted from [102].
Rather than trying to accurately model the complete interaction network it is also possible to approximate the network by determining dominant regulators (both miRNAs and transcription factors) that regulate groups of co-regulated mRNA species. What mRNA species (genes) are co-regulated can be determined by using correlation-based techniques such as weighted gene correlation network analysis (WGCNA) [103]. To establish by what common miRNA regulators the co-regulated genes are controlled, our Boolean interactome may be employed. A similar methodology has been applied in the context of idiopathic pulmonary fibrosis [103]. Unfortunately, WGCNA requires at least 15 samples [104] so this method could not be applied to our experiment.
Conclusions
Pseudoexfoliation glaucoma (XFG) shows changes in both mRNA and miRNA expression. When normal HTM tissue is treated with TGF-β1 there is significant overlap in differential mRNA and miRNA expression in this simple model and XFG, but it is often unclear what the regulatory effect of differentially expressed miRNA may be. In this work we present a bioinformatics analysis that provides a hypothesis-generating prediction of the regulatory functions of differentially expressed miRNAs causing differential gene expression in normal HTM tissue when treated with TGF-β1. To our knowledge this is the first study that combines genome-wide miRNA and mRNA expression into one genome-wide interactome for normal TM cells treated with TGF-β1. This interactome may help to understand the pathophysiology of XFG and could be used to identify potential candidates for miRNA-based therapeutics.
Supporting information
S1 Fig. miRNA-gene interactome of genes involved in the Unfolded Protein Response (UPR) for human trabecular meshwork cells stimulated by TGF-β1.
Heatmap of the interactome in which the heatmap colour maps to interaction score which is calculated from database evidence, miRNA expression and miRNA-gene correlation using Eq (2).
https://doi.org/10.1371/journal.pone.0318125.s001
(TIF)
S2 Fig. miRNA-gene interactome of transcription factors for human trabecular meshwork cells stimulated by TGF-β1.
Heatmap of the interactome in which the heatmap colour maps to interaction score which is calculated from database evidence, miRNA expression and miRNA-gene correlation using Eq (2).
https://doi.org/10.1371/journal.pone.0318125.s002
(TIF)
S1 Table. miRNA regulation of biological processes altered by TGF-β1 in HTM cells.
This table contains altered biological processes, the genes within these processes and what miRNAs may be regulating these genes.
https://doi.org/10.1371/journal.pone.0318125.s003
(XLSX)
S2 Table. miRNA-gene interactome for TGF-β1 stimulated HTM cells.
This file contains all miRNA-gene interactions.
https://doi.org/10.1371/journal.pone.0318125.s004
(XLSX)
S3 Table. miRNA-TF loops for TGF-β1 stimulated HTM cells.
This file contains miRNA-TF loops.
https://doi.org/10.1371/journal.pone.0318125.s005
(XLSX)
References
- 1.
Aboobakar IF, Johnson WM, Stamer WD, Hauser MA, Allingham RR. Major review: Exfoliation syndrome; advances in disease genetics, molecular biology, and epidemiology. Experimental Eye Research. Academic Press; 2017. pp. 88–103. https://doi.org/10.1016/j.exer.2016.11.011 pmid:27845061
- 2. Schlötzer-Schrehardt U, Khor CC. Pseudoexfoliation syndrome and glaucoma: From genes to disease mechanisms. Current Opinion in Ophthalmology. Lippincott Williams and Wilkins; 2021. pp. 118–128. pmid:33332884
- 3. Zenkel M, Schlötzer-Schrehardt U. The composition of exfoliation material and the cells involved in its production. J Glaucoma. 2014;23: S12–S14. pmid:25275897
- 4. Jonas JB, Aung T, Bourne RR, Bron AM, Ritch R, Panda-Jonas S. Glaucoma. Lancet. 2017;390: 2183–2193. pmid:28577860
- 5. Vesti E, Kivela T. Exfoliation syndrome and exfoliation glaucoma. Prog Retin Eye Res. 2000;19: 345–368. pmid:10749381
- 6. Grødum K, Heijl A, Bengtsson B. Risk of glaucoma in ocular hypertension with and without pseudoexfoliation. Ophthalmology. 2005;112: 386–390. pmid:15745763
- 7. Thorleifsson G, Magnusson KP, Sulem P, Walters GB, Gudbjartsson DF, Stefansson H, et al. Common Sequence Variants in the LOXL1 Gene Confer Susceptibility to Exfoliation Glaucoma. Science (1979). 2007;317: 1397–1400. pmid:17690259
- 8. Schlötzer-Schrehardt U. Molecular Biology of Exfoliation Syndrome. J Glaucoma. 2018;27: S32–S37. pmid:29419649
- 9. Kang JH, Loomis SJ, Wiggs JL, Willett WC, Pasquale LR. A prospective study of folate, vitamin B₆, and vitamin B₁₂ intake in relation to exfoliation glaucoma or suspected exfoliation glaucoma. JAMA Ophthalmol. 2014;132: 549–559. pmid:24699833
- 10. Roedl JB, Bleich à S, Reulbach U, Rejdak R, H Naumann GO, Kruse FE, et al. Vitamin deficiency and hyperhomocysteinemia in pseudoexfoliation glaucoma. J Neural Transm. 2007;114: 571–575. pmid:17238009
- 11. Tanito M, Kaidzu S, Takai Y, Ohira A. Status of systemic oxidative stresses in patients with primary open-angle glaucoma and pseudoexfoliation syndrome. PLoS One. 2012/11/26. 2012;7: e49680–e49680. pmid:23189153
- 12. Want A, Gillespie SR, Wang Z, Gordon R, Iomini C, Ritch R, et al. Autophagy and Mitochondrial Dysfunction in Tenon Fibroblasts from Exfoliation Glaucoma Patients. PLoS One. 2016;11. pmid:27391778
- 13. Zenkel M, Hoja U, Gießl A, Berner D, Hohberger B, Weller JM, et al. Dysregulated Retinoic Acid Signaling in the Pathogenesis of Pseudoexfoliation Syndrome. Int J Mol Sci. 2022;23: 5977. pmid:35682657
- 14. Schlötzer-Schrehardt U, Zenkel M, Küchle M, Sakai LY, Naumann GOH. Role of transforming growth factor-β1 and its latent form binding protein in pseudoexfoliation syndrome. Exp Eye Res. 2001;73: 765–780. pmid:11846508
- 15. Tzavlaki K, Moustakas A. TGF-β Signaling. Biomolecules. 2020;10. pmid:32210029
- 16. Roodnat AW, Callaghan B, Doyle C, Henry M, Goljanek-Whysall K, Simpson DA, et al. Genome-Wide RNA Sequencing of Human Trabecular Meshwork Cells Treated with TGF-β1: Relevance to Pseudoexfoliation Glaucoma. Biomolecules. 2022;12: 1693. pmid:36421707
- 17. Bartel DP. Metazoan MicroRNAs. Cell. 2018;173: 20–51. pmid:29570994
- 18. Gonzalez P, Li G, Qiu J, Wu J, Luna C. Role of MicroRNAs in the Trabecular Meshwork. Journal of Ocular Pharmacology and Therapeutics. 2014;30: 128. pmid:24383444
- 19. Greene KM, Stamer WD, Liu Y. The role of microRNAs in glaucoma. Exp Eye Res. 2022;215. pmid:34968473
- 20. Doyle C, Callaghan B, Roodnat AW, Armstrong L, Lester K, Simpson DA, et al. The TGFβ Induced MicroRNAome of the Trabecular Meshwork. Cells 2024, Vol 13, Page 1060. 2024;13: 1060. pmid:38920689
- 21. Knox J, Bou-Gharios G, Hamill KJ, Willoughby CE. MiR-18a-5p Targets Connective Tissue Growth Factor Expression and Inhibits Transforming Growth Factor β2-Induced Trabecular Meshwork Cell Contractility. Genes (Basel). 2022;13. pmid:36011411
- 22. Luna C, Li G, Qiu J, Epstein DL, Gonzalez P. MicroRNA-24 Regulates the Processing of Latent TGFβ1 During Cyclic Mechanical Stress in Human Trabecular Meshwork Cells Through Direct Targeting of FURIN. J Cell Physiol. 2011;226: 1407. pmid:20945401
- 23. Luna C, Li G, Qiu J, Epstein DL, Gonzalez P. Role of miR-29b on the regulation of the extracellular matrix in human trabecular meshwork cells under chronic oxidative stress. Mol Vis. 2009;15: 2488. Available: /pmc/articles/PMC2786891/ pmid:19956414
- 24. Coutinho De Almeida R, Ramos YFM, Mahfouz A, Den Hollander W, Lakenberg N, Houtman E, et al. RNA sequencing data integration reveals an miRNA interactome of osteoarthritis cartilage. Ann Rheum Dis. 2019;78: 270–277. pmid:30504444
- 25. Iliopoulos D, Malizos KN, Oikonomou P, Tsezou A. Integrative MicroRNA and Proteomic Approaches Identify Novel Osteoarthritis Genes and Their Collaborative Metabolic and Inflammatory Networks. PLoS One. 2008;3. pmid:19011694
- 26. Ross AK, Coutinho de Almeida R, Ramos YFM, Li J, Meulenbelt I, Guilak F. The miRNA-mRNA interactome of murine induced pluripotent stem cell-derived chondrocytes in response to inflammatory cytokines. FASEB J. 2020;34: 11546–11561. pmid:32767602
- 27. Chouvarine P, Photiadis J, Cesnjevar R, Scheewe J, Bauer UMM, Pickardt T, et al. RNA expression profiles and regulatory networks in human right ventricular hypertrophy due to high pressure load. iScience. 2021;24. pmid:33786422
- 28. Angius A, Uva P, Pira G, Muroni MR, Sotgiu G, Saderi L, et al. Integrated Analysis of miRNA and mRNA Endorses a Twenty miRNAs Signature for Colorectal Carcinoma. Int J Mol Sci. 2019;20. pmid:31434359
- 29. Winkler I, Bitter C, Winkler S, Weichenhan D, Thavamani A, Hengstler JG, et al. Identification of Pparγ-modulated miRNA hubs that target the fibrotic tumor microenvironment. Proc Natl Acad Sci U S A. 2020;117: 454–463. pmid:31871210
- 30. Lou W, Ding B, Xu L, Fan W. Construction of Potential Glioblastoma Multiforme-Related miRNA-mRNA Regulatory Network. Front Mol Neurosci. 2019;12. pmid:30971889
- 31. Ye Y, Li SL, Wang SY. Construction and analysis of mRNA, miRNA, lncRNA, and TF regulatory networks reveal the key genes associated with prostate cancer. PLoS One. 2018;13. pmid:30138363
- 32. Agarwal V, Bell GW, Nam JW, Bartel DP. Predicting effective microRNA target sites in mammalian mRNAs. Elife. 2015;4. pmid:26267216
- 33. Sticht C, De La Torre C, Parveen A, Gretz N. miRWalk: An online resource for prediction of microRNA binding sites. PLoS One. 2018;13: e0206239. pmid:30335862
- 34. Karagkouni D, Paraskevopoulou MD, Chatzopoulos S, Vlachos IS, Tastsoglou S, Kanellos I, et al. DIANA-TarBase v8: a decade-long collection of experimentally supported miRNA–gene interactions. Nucleic Acids Res. 2018;46: D239. pmid:29156006
- 35. Huang HY, Lin YCD, Li J, Huang KY, Shrestha S, Hong HC, et al. miRTarBase 2020: updates to the experimentally validated microRNA–target interaction database. Nucleic Acids Res. 2020;48: D148. pmid:31647101
- 36. Tokar T, Pastrello C, Rossos AEM, Abovsky M, Hauschild AC, Tsay M, et al. mirDIP 4.1-integrative database of human microRNA target predictions. Nucleic Acids Res. 2018;46: D360–D370. pmid:29194489
- 37. Chang L, Zhou G, Soufan O, Xia J. miRNet 2.0: network-based visual analytics for miRNA functional analysis and systems biology. Nucleic Acids Res. 2020;48: W244–W251. pmid:32484539
- 38. Li G, Luna C, Qiu J, Epstein DL, Gonzalez P. Alterations in microRNA Expression in Stress-induced Cellular Senescence. Mech Ageing Dev. 2009;130: 731. pmid:19782699
- 39. Youngblood H, Cai J, Drewry MD, Helwa I, Hu E, Liu S, et al. Expression of mRNAs, miRNAs, and lncRNAs in Human Trabecular Meshwork Cells Upon Mechanical Stretch. Invest Ophthalmol Vis Sci. 2020;61. pmid:32392310
- 40. Rao A, Chakraborty M, Roy A, Sahay P, Pradhan A, Raj N. Differential miRNA Expression: Signature for Glaucoma in Pseudoexfoliation. Clin Ophthalmol. 2020;14: 3025. pmid:33116354
- 41. Drewry MD, Challa P, Kuchtey JG, Navarro I, Helwa I, Hu Y, et al. Differentially expressed microRNAs in the aqueous humor of patients with exfoliation glaucoma or primary open-angle glaucoma. Hum Mol Genet. 2018;27: 1263–1275. pmid:29401312
- 42. Hindle AG, Thoonen R, Jasien J V., Grange RMH, Amin K, Wise Jet al. Identification of Candidate miRNA Biomarkers for Glaucoma. Invest Ophthalmol Vis Sci. 2019;60: 134. pmid:30629727
- 43. Sun B, Zhao C, Mao Y. MiR-218-5p Mediates Myocardial Fibrosis after Myocardial Infarction by Targeting CX43. Curr Pharm Des. 2021;27: 4504–4512. pmid:34587879
- 44. Zenkel M, Mößner A, Kruse FE, Schlötzer-Schrehardt U. Dysregulated Expression Of Fibrogenic microRNAs In Eyes With Pseudoexfoliation Syndrome. Invest Ophthalmol Vis Sci. 2012;53: 3853–3853.
- 45. Lopez NN, Rangan R, Clark AF, Tovar-Vidales T. Mirna expression in glaucomatous and tgfβ2 treated lamina cribrosa cells. Int J Mol Sci. 2021;22. pmid:34201109
- 46. Jayaram H, Cepurna WO, Johnson EC, Morrison JC. MicroRNA Expression in the Glaucomatous Retina. Invest Ophthalmol Vis Sci. 2015;56: 7971–7982. pmid:26720444
- 47. Wang P, Zhou Y, Richards AM. Effective tools for RNA-derived therapeutics: siRNA interference or miRNA mimicry. Theranostics. 2021;11: 8771–8796. pmid:34522211
- 48. Finnerty JR, Wang WX, Hébert SS, Wilfred BR, Mao G, Nelson PT. The miR-15/107 group of microRNA genes: evolutionary biology, cellular functions, and roles in human diseases. J Mol Biol. 2010;402: 491. pmid:20678503
- 49. Smyth A, Callaghan B, Willoughby CE, O’Brien C. The Role of miR-29 Family in TGF-β Driven Fibrosis in Glaucomatous Optic Neuropathy. Int J Mol Sci. 2022;23. pmid:36142127
- 50. Tavazoie SF, Alarcón C, Oskarsson T, Padua D, Wang Q, Bos PD, et al. Endogenous human microRNAs that suppress breast cancer metastasis. Nature. 2008;451: 147–152. pmid:18185580
- 51. Qiu Z, Zhong Z, Zhang Y, Tan H, Deng B, Meng G. Human umbilical cord mesenchymal stem cell-derived exosomal miR-335-5p attenuates the inflammation and tubular epithelial-myofibroblast transdifferentiation of renal tubular epithelial cells by reducing ADAM19 protein levels. Stem Cell Res Ther. 2022;13. pmid:35902972
- 52. Zampetaki A, Attia R, Mayr U, Gomes RSM, Phinikaridou A, Yin X, et al. Role of miR-195 in aortic aneurysmal disease. Circ Res. 2014;115: 857–866. pmid:25201911
- 53. Kriegel AJ, Baker MA, Liu Y, Liu P, Cowley AW, Liang M. Endogenous microRNAs in human microvascular endothelial cells regulate mRNAs encoded by hypertension-related genes. Hypertension. 2015;66: 793–799. pmid:26283043
- 54. Matoušková P, Hanousková B, Skálová L. MicroRNAs as Potential Regulators of Glutathione Peroxidases Expression and Their Role in Obesity and Related Pathologies. Int J Mol Sci. 2018;19. pmid:29662007
- 55. Zhou Y, Wu H, Wang F, Xu L, Yan Y, Tong X, et al. GPX7 Is Targeted by miR-29b and GPX7 Knockdown Enhances Ferroptosis Induced by Erastin in Glioma. Front Oncol. 2022;11. pmid:35127512
- 56. Oliveira LH, Schiavinato JL, Fráguas MS, Lucena-Araujo AR, Haddad R, Araújo AG, et al. Potential roles of microRNA-29a in the molecular pathophysiology of T-cell acute lymphoblastic leukemia. Cancer Sci. 2015;106: 1264. pmid:26251039
- 57. Cai B, Qu X, Kan D, Luo Y. miR-26a-5p suppresses nasopharyngeal carcinoma progression by inhibiting PTGS2 expression. Cell Cycle. 2022;21: 618–629. pmid:35073820
- 58. Chakraborty M, Sahay P, Rao A. Primary human trabecular meshwork model for pseudoexfoliation. Cells. 2021;10. pmid:34943956
- 59. Wang Y, Jia RZ, Diao S, He J, Jia L. miRNA-101 Targets TGF-βR1 to Retard the Progression of Oral Squamous Cell Carcinoma. Oncol Res. 2020;28: 203–212. pmid:31831099
- 60. Wang FE, Zhang C, Maminishkis A, Dong L, Zhi C, Li R, et al. MicroRNA-204/211 alters epithelial physiology. The FASEB Journal. 2010;24: 1552. pmid:20056717
- 61. Correia ACP, Moonen JRAJ, Brinker MGL, Krenning G. FGF2 inhibits endothelial-mesenchymal transition through microRNA-20a-mediated repression of canonical TGF-β signaling. J Cell Sci. 2016;129: 569–579. pmid:26729221
- 62. Hao C, Yang S, Xu W, Shen JK, Ye S, Liu X, et al. MiR-708 promotes steroid-induced osteonecrosis of femoral head, suppresses osteogenic differentiation by targeting SMAD3. Sci Rep. 2016;6. pmid:26932538
- 63. Tijsen AJ, Van Der Made I, Van Den Hoogenhof MM, Wijnen WJ, Van Deel ED, De Groot NE, et al. The microRNA-15 family inhibits the TGFβ-pathway in the heart. Cardiovasc Res. 2014;104: 61–71. pmid:25103110
- 64. Yan X, Liu Z, Chen Y. Regulation of TGF-β signaling by Smad7. Acta Biochim Biophys Sin (Shanghai). 2009;41: 263. pmid:19352540
- 65. Lun W, Wu X, Deng Q, Zhi F. MiR-218 regulates epithelial–mesenchymal transition and angiogenesis in colorectal cancer via targeting CTGF. Cancer Cell Int. 2018;18: 83. pmid:29977158
- 66. van Almen GC, Verhesen W, van Leeuwen REW, van de Vrie M, Eurlings C, Schellings MWM, et al. MicroRNA-18 and microRNA-19 regulate CTGF and TSP-1 expression in age-related heart failure. Aging Cell. 2011;10: 769–779. pmid:21501375
- 67. Zhang J, Han C, Zhu H, Song K, Wu T. miR-101 inhibits cholangiocarcinoma angiogenesis through targeting vascular endothelial growth factor (VEGF). Am J Pathol. 2013;182: 1629–1639. pmid:23608225
- 68. Tang W, Guo J, Gu R, Lei B, Ding X, Ma J, et al. MicroRNA-29b-3p inhibits cell proliferation and angiogenesis by targeting VEGFA and PDGFB in retinal microvascular endothelial cells. Mol Vis. 2020;26: 64. Available: /pmc/articles/PMC7043642/ pmid:32165827
- 69. Mao W, Cameron Millar J, Wang WH, Silverman SM, Liu Y, Wordinger RJ, et al. Existence of the Canonical Wnt Signaling Pathway in the Human Trabecular Meshwork. Invest Ophthalmol Vis Sci. 2012;53: 7043. pmid:22956608
- 70. Yuan Y, Schlötzer-Schrehardt U, Ritch R, Call M, Chu FB, Dong F, et al. Transient expression of Wnt5a elicits ocular features of pseudoexfoliation syndrome in mice. PLoS One. 2019;14: e0212569–e0212569. pmid:30840655
- 71. Li X, Yamagata K, Nishita M, Endo M, Arfian N, Rikitake Y, et al. Activation of Wnt5a-Ror2 signaling associated with epithelial-to-mesenchymal transition of tubular epithelial cells during renal fibrosis. Genes to Cells. 2013;18: 608–619. pmid:23755735
- 72. Sun Q, Liu S, Feng J, Kang Y, Zhou Y, Guo S. Current Status of MicroRNAs that Target the Wnt Signaling Pathway in Regulation of Osteogenesis and Bone Metabolism: A Review. Med Sci Monit. 2021;27: e929510–1. pmid:33828067
- 73. Peng Y, Zhang X, Feng X, Fan X, Jin Z. The crosstalk between microRNAs and the Wnt/β-catenin signaling pathway in cancer. Oncotarget. 2017;8: 14089. pmid:27793042
- 74. Henderson NC, Rieder F, Wynn TA. Fibrosis: from mechanisms to medicines. Nature 2020 587:7835. 2020;587: 555–566. pmid:33239795
- 75. Schafer S, Viswanathan S, Widjaja AA, Lim WW, Moreno-Moral A, DeLaughter DM, et al. IL-11 is a crucial determinant of cardiovascular fibrosis. Nature 2017 552:7683. 2017;552: 110–115. pmid:29160304
- 76. Li G, Luna C, Qiu J, Epstein DL, Gonzalez P. Role of miR-204 in the Regulation of Apoptosis, Endoplasmic Reticulum Stress Response, and Inflammation in Human Trabecular Meshwork Cells. Invest Ophthalmol Vis Sci. 2011;52: 2999. pmid:21282569
- 77. Pollari S, Leivonen SK, Perälä M, Fey V, Käkönen SM, Kallioniemi O. Identification of MicroRNAs Inhibiting TGF-β-Induced IL-11 Production in Bone Metastatic Breast Cancer Cells. PLoS One. 2012;7. pmid:22629385
- 78. Peters JC, Bhattacharya S, Clark AF, Zode GS. Increased Endoplasmic Reticulum Stress in Human Glaucomatous Trabecular Meshwork Cells and Tissues. Invest Ophthalmol Vis Sci. 2015;56: 3860–3868. pmid:26066753
- 79. Bartoszewska S, Kochan K, Madanecki P, Piotrowski A, Ochocka R, Collawn JF, et al. Regulation of the unfolded protein response by microRNAs. Cell Mol Biol Lett. 2013;18: 555. pmid:24092331
- 80. Xu G, Chen J, Jing G, Grayson TB, Shalev A. miR-204 Targets PERK and Regulates UPR Signaling and β-Cell Apoptosis. Mol Endocrinol. 2016;30: 917–924. pmid:27384111
- 81. Chien CY, Chen YC, Lee CH, Wu JR, Huang TW, Huang RY, et al. Dysregulation of the miR-30a/BiP axis by cigarette smoking accelerates oral cancer progression. Cancer Cell Int. 2021;21. pmid:34717640
- 82. Acott TS, Kelley MJ. Extracellular Matrix in the Trabecular Meshwork. Exp Eye Res. 2008;86: 543. pmid:18313051
- 83. Smyth A, Irnaten M, Powell SK, O’Brien CJ. Elevated Expression of ADAM 12 and 19 in Human Glaucoma Lamina Cribrosa Cells. Invest Ophthalmol Vis Sci. 2021;62: 2770. Available: https://iovs.arvojournals.org/article.aspx?articleid=2775759
- 84. Villarreal G, Oh DJ, Kang MH, Rhee DJ. Coordinated regulation of extracellular matrix synthesis by the microRNA-29 family in the trabecular meshwork. Invest Ophthalmol Vis Sci. 2011;52: 3391–3397. pmid:21330653
- 85. Lee JG, Jung E, Gestoso K, Heur M. ZEB1 Mediates Fibrosis in Corneal Endothelial Mesenchymal Transition Through SP1 and SP3. Invest Ophthalmol Vis Sci. 2020;61. pmid:32721022
- 86. Li XL, Hara T, Choi Y, Subramanian M, Francis P, Bilke S, et al. A p21-ZEB1 complex inhibits epithelial-mesenchymal transition through the microRNA 183-96-182 cluster. Mol Cell Biol. 2014;34: 533–550. pmid:24277930
- 87. Wellner U, Schubert J, Burk UC, Schmalhofer O, Zhu F, Sonntag A, et al. The EMT-activator ZEB1 promotes tumorigenicity by repressing stemness-inhibiting microRNAs. Nat Cell Biol. 2009;11: 1487–1495. pmid:19935649
- 88. Jahan F, Landry NM, Rattan SG, Dixon IMC, Wigle JT. The Functional Role of Zinc Finger E Box-Binding Homeobox 2 (Zeb2) in Promoting Cardiac Fibroblast Activation. Int J Mol Sci. 2018;19. pmid:30336567
- 89. Helwak A, Kudla G, Dudnakova T, Tollervey D. Mapping the human miRNA interactome by CLASH reveals frequent noncanonical binding. Cell. 2013;153: 654–665. pmid:23622248
- 90. Yang S, Xiang J, Ma C, Yang G, Wang X, Liu H, et al. Sp1-like protein KLF13 acts as a negative feedback regulator of TGF-β signaling and fibrosis. Cell Rep. 2023;42. pmid:37029927
- 91. Wang B, Haldar SM, Lu Y, Ibrahim OA, Fisch S, Gray S, et al. The Kruppel-like factor KLF15 inhibits connective tissue growth factor (CTGF) expression in cardiac fibroblasts. J Mol Cell Cardiol. 2008;45: 193–197. pmid:18586263
- 92. Rodriguez-Colman MJ, Dansen TB, Burgering BMT. FOXO transcription factors as mediators of stress adaptation. Nat Rev Mol Cell Biol. 2024;25: 46–64. pmid:37710009
- 93. Myatt SS, Wang J, Monteiro LJ, Christian M, Ho KK, Fusi L, et al. Definition of microRNAs that repress expression of the tumor suppressor gene FOXO1 in endometrial cancer. Cancer Res. 2010;70: 367–377. pmid:20028871
- 94. Wu L, Li H, Jia CY, Cheng W, Yu M, Peng M, et al. MicroRNA-223 regulates FOXO1 expression and cell proliferation. FEBS Lett. 2012;586: 1038–1043. pmid:22569260
- 95. Johnston RJ, Chang S, Etchberger JF, Ortiz CO, Hobert O. MicroRNAs acting in a double-negative feedback loop to control a neuronal cell fate decision. Proc Natl Acad Sci U S A. 2005;102: 12449–12454. pmid:16099833
- 96. Burk U, Schubert J, Wellner U, Schmalhofer O, Vincan E, Spaderna S, et al. A reciprocal repression between ZEB1 and members of the miR-200 family promotes EMT and invasion in cancer cells. EMBO Rep. 2008;9: 582–589. pmid:18483486
- 97. Gallant-Behm CL, Piper J, Lynch JM, Seto AG, Hong SJ, Mustoe TA, et al. A MicroRNA-29 Mimic (Remlarsen) Represses Extracellular Matrix Expression and Fibroplasia in the Skin. Journal of Investigative Dermatology. 2019;139: 1073–1081. pmid:30472058
- 98. Bartel DP. MicroRNA Target Recognition and Regulatory Functions. Cell. 2009;136: 215. pmid:19167326
- 99. Carancini G, Carletti M, Spaletta G. Modeling and Simulation of a miRNA Regulatory Network of the PTEN Gene. Mathematics 2021, Vol 9, Page 1803. 2021;9: 1803.
- 100. Ragan C, Zuker M, Ragan MA. Quantitative prediction of miRNA-mRNA interaction based on equilibrium concentrations. PLoS Comput Biol. 2011;7. pmid:21390282
- 101. Patel G, Fury W, Yang H, Gomez-Caraballo M, Bai Y, Yang T, et al. Molecular taxonomy of human ocular outflow tissues defined by single-cell transcriptomics. Proc Natl Acad Sci U S A. 2020;117: 12856–12867. pmid:32439707
- 102. Cora D, Re A, Caselle M, Bussolino F. MicroRNA-mediated regulatory circuits: outlook and perspectives. Phys Biol. 2017;14. pmid:28586314
- 103. McDonough JE, Kaminski N, Thienpont B, Hogg JC, Vanaudenaerde BM, Wuyts WA. Original article: Gene correlation network analysis to identify regulatory factors in idiopathic pulmonary fibrosis. Thorax. 2019;74: 132. pmid:30366970
- 104.
WGCNA package: Frequently Asked Questions. [cited 23 May 2023]. Available: https://horvath.genetics.ucla.edu/html/CoexpressionNetwork/Rpackages/WGCNA/faq.html