Coding, or non-coding, that is the question

Abstract

The advent of high-throughput sequencing uncovered that our genome is pervasively transcribed into RNAs that are seemingly not translated into proteins. It was also found that non-coding RNA transcripts outnumber canonical protein-coding genes. This mindboggling discovery prompted a surge in non-coding RNA research that started unraveling the functional relevance of these new genetic units, shaking the classic definition of "gene". While the non-coding RNA revolution was still taking place, polysome/ribosome profiling and mass spectrometry analyses revealed that peptides can be translated from non-canonical open reading frames. Therefore, it is becoming evident that the coding vs non-coding dichotomy is way blurrier than anticipated. In this review, we focus on several examples in which the binary classification of coding vs non-coding genes is outdated, since the same bifunctional gene expresses both coding and non-coding products. We discuss the implications of this intricate usage of transcripts in terms of molecular mechanisms of gene expression and biological outputs, which are often concordant, but can also surprisingly be discordant. Finally, we discuss the methodological caveats that are associated with the study of bifunctional genes, and we highlight the opportunities and challenges of therapeutic exploitation of this intricacy towards the development of anticancer therapies.

Fig. 1. Bifunctional genomic loci express mRNAs and ncRNAs.

Genomic loci are defined bifunctional when they can be considered both coding (orange) and non-coding (blue), because they express both an mRNA and a ncRNA, through one of the following mechanisms. a The ncRNA is the product of alternative splicing, for example through the retention of introns or the choice of an alternative splice site. b In back-splicing, the joining of an upstream 3′ splice acceptor with a 5′ downstream splice donor leads to the production of a non-coding circRNA. c NATs are non-coding RNA molecules transcribed from the opposite DNA strand. According to the degree of overlap with sense coding mRNAs, NATs are defined as head-to-head (the overlap is in the 5′ region, left), tail-to-tail (the overlap is in the 3′ region, middle) or embedded (the overlap is complete, right). d Exons compose the mRNA that is translated into a protein, while introns compose a non-coding RNA such as a miRNA. The ncRNA is called “intragenic”, while the coding gene in which it resides is called “host gene”.

Fig. 2. Bifunctional mRNAs exert coding functions and non-coding functions.

In mRNA molecules (middle) three regions can be identified: the 5′UTR (red), the CDS (light blue) and the 3′UTR (green). (Top, orange) mRNAs are primarily protein-coding RNA molecules: they carry a primary ORF (the CDS), and they may also present a short uORF in the 5′UTR. (Bottom, blue) non-canonical non-coding functions have been attributed to mRNAs. The 5′UTR can exert non-coding functions in cis or in trans, by interacting with proteins. The CDS can be involved in non-coding RNA–protein or RNA–RNA interactions. The 3′UTR can exert non-coding functions within the concept of ceRNAs: due to MREs, 3′UTRs can sponge miRNAs, leading to the de-silencing of ceRNA partners that share MREs for the same miRNAs.

Fig. 3. Bifunctional ncRNAs exert non-coding functions and coding functions.

ncRNAs can be divided into two groups: housekeeping ncRNAs (upper panels) and regulatory ncRNAs (lower panels). Among housekeeping ncRNAs there are rRNAs, tRNAs, snRNAs and snoRNAs. Regulatory ncRNAs are further divided according to their length. Among short ncRNAs (< 200 nt) there are miRNAs (light blue). Long ncRNAs (lncRNAs, ≥ 500 nt) include lincRNAs (light green), pseudogenic RNAs (PGs, light gray), NATs (see Fig. 1c), and circRNAs (light purple). The primary function of miRNAs is non-coding (blue) and relates to post-transcriptional regulation of gene expression: mRNA degradation or translational repression mediated by the RISC complex. Among non-coding functions of lincRNAs there are: chromatin remodeling by epigenetic modification, transcription and splicing regulation, sponging of miRNAs and proteins, post-translational modification of proteins. Non-coding functions of PGs can be parental gene (PA)-related or unrelated. They include sponging of miRNAs and proteins, and mRNA degradation through endosiRNAs. S sense strand, AS antisense strand. circRNAs can be composed by only exons (EcircRNAs), both exons and introns (EIcircRNAs), or only introns (ciRNAs). They exert non-coding functions by acting as sponges for miRNAs and proteins, and by forming circRNPs that modulate signaling pathways. These groups of regulatory ncRNAs can all exert non-canonical coding functions as well, because they can be translated into ncPEPs, i.e., small peptides or proteins (orange).

Fig. 4. The complexity of expression and function of the coding and non-coding products derived from the same bifunctional gene.

The coding product is represented as a generic orange protein and the non-coding product as a generic blue ncRNA. a Possible scenarios for gene expression. The protein and the ncRNA are expressed together in the same context (left), or separately in two different contexts (right). b Possible in cis regulatory mechanisms. The protein and the ncRNA positively or negatively regulate the gene from which they originate, or each other. c Possible in trans regulatory mechanisms. The protein and the ncRNA positively or negatively regulate downstream effectors (other genes, RNAs or proteins) that can be distinct or the same for both.

Fig. 5. Concordant or discordant impact of the coding and non-coding products derived from the same bifunctional gene in cancer.

Graphical representation of the possible combinations of cancer-related functions of coding (orange) and non-coding (blue) partners. Upper left panel: both the coding and the non-coding products have tumor-suppressive properties. PTEN mRNA encodes the tumor suppressor PTEN protein and, due to its 3′UTR, it can sponge oncogenic miRNAs. Upper right panel: the coding product is a tumor suppressor, while the non-coding product is an oncogene. Zbtb7a pre-mRNA undergoes canonical splicing and generates a mature mRNA that encodes the Pokémon protein with tumor suppressor properties; the same pre-mRNA also undergoes back-splicing of exon 2, leading to the formation of oncogenic circPOK. Bottom left panel: the coding product is an oncogene, while the non-coding product is a tumor suppressor. AKT3 pre-mRNA is translated into AKT3 oncogenic protein, but exons 3–7 undergo back-splicing, leading to the production of a circRNA (hsa_circ_0017250) that exerts tumor-suppressive effects through its translation into the AKT3-174aa ncPEP. Bottom right panel: both the coding and the non-coding products have oncogenic properties. MCM7 pre-mRNA is spliced to produce a mature mRNA that is translated into the oncogenic MCM7 protein; in addition, intron 13 hosts oncogenic miR-106b~25 cluster. From a therapeutic point of view, the optimal approach is to enhance/restore tumor-suppressive activities (green “plus” symbol) and/or, on the other hand, abolish/inhibit oncogenic activities (red “minus” symbol). This is easier in the case of bifunctional genes whose products have concordant outputs, while it can be ineffective or even deleterious in the case of a discordant output.