MicroRNA-Regulated Protein–Protein Interaction Networks: How Could They Help in Searching for Pro-Longevity Targets?

Introduction

A lmost 800 longevity-associated genes (LAGs) have thus far been established in model organisms. ¹ One of the remarkable features of these genes is that they are highly evolutionarily conserved. Consequently, the majority of LAGs have orthologs in humans, suggesting their relevance to human aging and longevity as well. ^2,3 In support of this is the fact that many LAGs are also involved in the major human age-related diseases (ARDs), including atherosclerosis, cancer, type 2 diabetes, and Alzheimer's disease (AD), ⁴ which are the main causes of death in the elderly and thus the main limiting factor of human life span. If so, the genes/proteins involved in ARDs would also have an impact on longevity and vice versa.

Aging was suggested as the main risk factor for the development of ARDs. ⁵ We believe that ARDs not only are direct consequences of aging but actually represent its diverse manifestations, being an essential component of the “normal” aging process. ^1,6 This means that common molecular mechanisms stand behind both aging/longevity and ARDs. These mechanisms could be greatly related to the epigenetic control of gene expression, ^{6 –10} which includes DNA methylation, histone modifications, and a recently discovered mechanism of RNA interference (RNAi). The effector molecules in RNAi are microRNAs (miRNAs), small double-stranded RNAs that cause gene inactivation at the posttranscriptional level by binding to mRNAs, with their consequent degradation or repression of translation (translational silencing). The rapidly emerging (though still incomplete) data in the field indicate that changes in the patterns of epigenetic regulation or in its components appear to have a profound impact on both longevity and ARDs. ^{10 –14} For example, the possibility to extend life span by modulating gene expression through RNAi in Caenorhabditis elegans was shown recently. ¹⁵

The extremely high complexity of the subject requires appropriate analytic tools. During the last years, a network-based approach has been increasingly used for the analysis of aging, longevity, and ARDs. ^{2 –4,16 –25} In particular, the analysis of protein–protein interaction (PPI) networks showed numerous molecular links between different ARDs ²⁶ and between ARDs and longevity. ⁴ Apart from the contribution to a better understanding of the mechanisms of aging, the network approach could be useful for developing new research strategies in the field and searching for potential pro-longevity targets. One of the important conclusions from the network analysis is that multitarget interventions could be much more efficient for combating aging and ARDs than the currently dominant approach, which focuses on highly specific single-target drugs. ^4,27 In this work, we analyzed how the miRNA-regulated PPI network common for longevity and ARDs could help in searching for pro-longevity targets.

Materials and Methods

The construction of the networks was described in detail elsewhere. ^2,4,28 Briefly, given that the LAGs act in cooperative manner, we have recently constructed the PPI Human Longevity Network (HLN) ² (for updated HLN, see www.netage-project.org ²⁸ ). This network includes the human orthologs of established LAGs and their first-order interacting partners. Similarly, PPI networks were built for the major human ARDs. ^3,4 All of these networks are now available in the NetAge database. ²⁸ At the time of writing, NetAge hosts the human miRNA-regulated interactome, which integrates PPIs from the BioGRID database, release 2.0.5, ²⁹ with experimentally validated miRNA targets from TarBase v.5c. ³⁰ Associations between miRNAs and diseases were retrieved from HMDD, The Human MicroRNA Disease Database. ³¹ Transcription factor–miRNA regulation data were retrieved from the TransmiR database. ³² Mouse phenotypic data, including gene essentiality, were extracted from the Mouse Genome Informatics database (www.informatics.jax.org).

Network analysis, including simulations with randomly selected genes/proteins, was performed using YABNA (Yet Another Biological Networks Analyzer), a flexible set of tools developed in our laboratory. The graphical output of networks was generated using Cytoscape 2.6.0. ³³

All statistical calculations were performed using the Statistical Package for the Social Sciences Inc. (SPSS; Chicago, IL) software. For comparison of observed and expected values, the chi-squared test was used. The differences were considered significant at p value < 0.05. To compare the average values of connectivity between different groups of genes, the number of PPIs (k) for each gene (protein) was ln-transformed.

Results and Discussion

Expression of many CGS genes is miRNA regulated

As a step toward constructing an integrative PPI–epigenetic network, we included miRNAs that have experimentally validated targets in CGS. The resultant miRNA-regulated PPI network is shown in Fig. 1. Thirty one miRNAs were found to target 77 genes in CGS, which in turn directly interact with other 360 CGS genes. That is, two thirds of the CGS genes could be miRNA regulated, directly or indirectly through PPIs of their target genes. Of note, all of these miRNAs were reported to have an altered expression in at least one ARD. ^10,31 The observed involvement of corresponding genes in ARDs could to a great extent be the result of an altered miRNA expression.

OPEN IN VIEWER

FIG. 1.

The microRNA (miRNA)-regulated Common Gene Signature (CGS) network. Presented is a graphical representation of the gene overlap between the human longevity network and the networks of major age-related diseases. Node size is proportional with the number of protein–protein interaction (PPIs) in the interactome. Nodes with black borders represent miRNA targets. LAGs, Longevity-associated genes.

When the miRNAs and their targets were “extracted” from CGS, they still formed a continuous network in which the genes are connected via either PPIs or miRNAs (Fig. 2). This is quite an interesting result, in view of the relatively small number of nodes. As seen in Fig. 2, the expression of several miRNAs could be also regulated by genes from the network. ³² For example, miR-21 and PTEN might form a typical feedback loop. Other cases include more complicated relationships. For example, SP1 was reported to influence the expression of miR-106A, which may target SP1 via its neighboring partner RB1. A similar “triangle” loop is observed for CCND1, miR-17, and NCOA3.

OPEN IN VIEWER

FIG. 2.

The network formed by microRNAs (miRNAs) and their targets in Common Gene Signature (CGS). With only two exceptions, all target genes in CGS are interconnected via either protein–protein interaction (PPIs) or miRNAs connections. Shown in brackets for each miRNA is the total number of targets in the interactome. LAGs, Longevity-associated genes.

Because the list of miRNA validated targets is far from being complete, this network includes a limited number of genes from the CGS. Nevertheless, the constructed network is important in view of possible pro-longevity interventions. Indeed, miRNAs (or small interfering RNAs [siRNAs]) represent a promising tool for controlled modification of gene expression, and the important point is that this could be done postdevelopmentally ¹⁵ or even later in life, including the postreproductive period. This approach appears to be much more reasonable, allowing for avoiding hardly predictable and potentially undesirable effects of gene inactivation on growth and development. Then, the essential CGS genes, and particularly those with multiple connections, could be the subject of RNAi. With this in mind, a group of 35 genes from the miRNA-regulated CGS network (Fig. 2), which are neighboring partners of LAGs, could be of particular interest. First, they are essential. Second, most of them were found in association with at least one ARD. This shows that they have features of antagonistic pleiotropy, and therefore could be relevant to aging and longevity. Consequently, these genes could be considered new candidates for pro-longevity interventions with the potential of being modulated through their miRNAs (siRNAs) later in life.

Concluding Remarks

It is reasonable to suggest that a well-balanced activity of CGS nodes including both miRNAs and genes/proteins would assure a favorable healthy status for the whole organism. Besides, the network organization offers the potential for self-compensation, which is very similar to the concept of self-stabilization, a highly desirable property of fault-tolerance systems in distributed computing. ³⁶ A distributed system that is self-stabilizing will end up in a correct state, after a finite number of execution steps, no matter what state it is initialized with and no matter what execution steps it will take. In biological systems, when the network reaches a certain degree of misbalance and self-compensation is not efficient enough, pathological conditions will manifest. It is quite obvious that concomitantly correcting all of the changes in the network is unrealistic. Moreover, it is unnecessary. Instead, we suggest that the restoration of an essential part of the network (“the network fingerprint”) could be sufficient for stabilizing and bringing it to a self-compensatory state.

As a first approximation, the network consisting of miRNAs and their validated targets (see Fig. 2) could be used as such “fingerprint” for CGS. Although the “fingerprint” expression profile may significantly differ among various tissues, it could provide useful information regarding the functional state of the CGS and suggest possible ways for corrections.