Upstream open reading frames (upORFs) are key regulatory elements located in the 5′untranslated (UTR) region of coding transcripts. UpORFs result from the presence of an upstream translation initiation site (uTIS) located within the 5′UTR and associated with an in-frame stop codon (uStop) located within the 5′UTR or the coding sequence (CDS). Different types of upORFs can be distinguished according to the position of the uStop with respect to the CDS (Figure 1). More precisely, when the uStop (i) is located within the 5′UTR, this results in a fully upstream ORF (uORF), (ii) is located within the CDS and is distinct from the main stop codon of the CDS, the uTIS is at the origin of an overlapping uORF (uoORF), and (iii) is the main stop codon of the CDS, this leads to an elongated CDS (eCDS). Approximately, half of the human transcripts naturally contain upORFs in their 5′UTR (1, 2) and these upORFs can contribute to modulate the production of the main protein encoded by the CDS by disturbing the translation initiation step and then the recognition of the main TIS by the ribosomes (3–5). The functional effect of a given upORF is highly variable and could be influenced by elements including the number of upORFs in the 5′UTR, their length, and the nucleotide context of the upORF as extensively discussed previously (6).

Often, the presence of upORF in general, and uoORF in particular, leads to a decrease of the expression of the main transcript (1). That could happen via the alteration of the translation mechanism (i.e., ribosome dissociation and ribosome stalling) or via transcript degradation by the non-sense-mediated decay process that recognizes the uStop as a premature stop codon (1, 7, 8). Nevertheless, under some conditions (i.e., hypoxia or cell stress), the presence of upORFs in a given transcript could be associated with an increase of the translation efficiency (9, 10). Indeed, upORF can modulate the activity of coexisting internal ribosome entry site (IRES) located on the same 5′UTR (11), thus regulating the IRES-dependent translation initiation in a context dependent manner. For instance, Chen et al., showed that the increase of fibroblst growth factor 9 (FGF9) protein levels under hypoxia happens via an IRES-dependent translation, regulated by the presence of a small upORF upstream to the IRES (12). In normal conditions (i.e., normaxia), FGF9 is present in low levels in human cells, thanks to the upORF-mediated translation inhibition of the CDS. Under hypoxia conditions, ribosomes probably switch from the upORF to IRES, thus activating the IRES-dependent translation and leading to efficient translation of FGF9 (12). That explains the increase of FGF9 under hypoxia conditions in cancer cells. In addition, upORFs could be translated into small-encoded peptides (SEPs) and play a regulatory role in health and disease contexts (13, 14).

High-throughput genomic studies have identified an increasingly number of single nucleotide variations (SNVs) located in 5′UTR and possibly altering upORFs by creating new ones or deleting/modifying existing ones suggesting that this kind of variants has been underestimated (15). Many of these variants have been characterized as disease causing by creating upORF and, thus, altering the production of the canonical protein, but surprisingly this has still not been investigated systematically. In fact, among the ∼4,000 disease-associated 5′UTR variants reported in different databases, the most deleterious ones are those creating or deleting uTIS or uStop, responsible of the creation or the disruption of upORFs (15). Whiffin et al. have recently shown that, among all the SNVs reported in the genome aggregation (GnomAD) database to locate in 5′UTR of 18,593 canonical transcripts, on an average of 30 SNVs per gene are variations creating a uAUG canonical initiation codon (15). They also showed that only 39 uAUG-creating and four stop-removing extremely rare variants were reported in Human Gene Mutation Database (HGMD) or likely pathogenic in ClinVar (15). Very interestingly, among these rare variants, nine uAUG-creating variants are located in genes implicated in cystic fibrosis, familial hypercholesterolemia, and hematologic diseases (16–22). Moreover, recent studies have also shown that upORF could be initiated by non-AUG codons and be disease causing (23, 24). Given the diversity of the functional implication of existing upORFs in the regulation of protein expression, the possible functional impacts of upORF-altering variants, hereafter called upSNVs, on protein expression could be highly variable. Up to very recently (25, 26), this type of genetic variants was not easily predicted by available bioinformatics tools. In addition, their functional characterization requires dedicated experimental strategies that have not yet been harmonized in order to demonstrate how they could affect gene expression and how the resulting dysregulations could lead to disease. Nevertheless, a first step in the assessment of the effect of upSNVs on the protein levels can be obtained using in vitro functional assays in which the 5′UTR and CDS of a given transcript are cloned in expression vectors followed by the expression of the produced vectors in human cells, both in the wild-type and upSNV contexts (27). upSNV-associated protein levels could then be evaluated by Western blot in comparison to the wild-type construct. Luciferase assays have also been widely used to study upSNVs. These assays are based on the cloning of the entire promoter of a given transcript before the coding sequence of a luciferase and the evaluation of the promoter activity in wild-type and mutant contexts in vitro by measuring the obtained luciferase luminescence normalized to a control vector. Additional methods used to characterize small ORFs and their potential translation into SEPs has been recently reviewed in (28). Altogether, upSNVs are still a neglected class of non-coding variations, and are often called as Variants of Unknown Significance when they are identified in routine clinical diagnosis, contributing then to medical wandering. In this work, with the aim of putting new light on upSNVs, we first provide a general overview of such type of variants known to cause rare cardiovascular disorders (CVDs). Then, we explore their potential role as candidates for explaining association signals detected in the context of genome-wide association studies (GWASs) for common complex CVDs.

Two complementary strategies were adopted to identify rare uAUG-creating variants in CVD genes. First, we selected variants from Supplementary Table 2 of (15) reporting upSNVs from ClinVar and HGMD. Then, we looked for additional variants in ClinVar and HGMD that were not reported in Whiffin et al., and scanned research articles in PubMed using the following keywords: “upstream ORF” and “cardio-vascular.”

To investigate whether some association signals detected in GWAS for CVDs could be explained by upSNVs, we deployed MORFEE¹ on the 1,000 Genome reference dataset (phase 3-v20130502) in order to identify all the common (allele frequency > 1%) predicted upSNVs in 5′UTR regions. In a second step, we checked whether these predicted upSNVs could be in linkage disequilibrium (LD) with lead SNVs identified in GWAS studies for coronary artery disease (CAD), stroke, venous thrombosis (VT), platelets, and lipid traits. LD information was retrieved from the European populations genetic database available through the LDlink web-based tool² and from which we considered two SNVs to be in LD when the absolute value of their pairwise D’ was greater than 0.7. For CAD, GWAS loci and lead SNVs were selected from Matsui et al. (29) and Hartiala et al. (30), while Malik et al. (31) and Lindström et al. (32) were used to identify GWAS loci and corresponding lead SNVs for stroke and VT, respectively. For platelets and lipids traits, we selected all the SNVs reported in the Geospatial Resource for Agriculture Species and Pests (GRASP) server³ (33) as of September 2021 to associate at p < 5.10^–8 with any of their related quantitative traits, including mean platelet volume, platelet count, platelet aggregation or platelets’ response to medication for platelet therapy, and high-density lipoprotein (HDL)-/low-density lipoprotein (LDL)/total cholesterol, triglycerides for lipids. Finally, this selection strategy led to a list of 749 CVD traits associated loci scrutinized for harboring common upSNVs.

While there is increasingly awareness of the impact of rare upSNVs in rare Mendelian disorders, there has been so far little initiative to investigate the possible role of such variants in the susceptibility to common diseases and their quantitative risk factors. From a list of ∼700 loci identified in GWAS for CVD traits, we only identified a very minor proportion of loci (5: FGF21, FRDM5, PEAR1, PSORS1C1, and SLC18A1) where the GWAS signal could be partially explained by upSNVs. We focused here on CVDs but similar investigations merit to be conducted for other human diseases. Our results were based on in silico observations (bioinformatics predictions coupled to LD analyses) and deserve to be further investigated through fine-mapping association analysis and experimental molecular characterization. Several molecular techniques (gene reporter assays, toeprinting, polysome profiling, among others) are available to evaluate the effect of upSNVs on the translation machinery and/or protein expression. Here, we would like to highlight the recent advances in the antisense oligonucleotides (ASOs) strategy targeting upORFs, as it also offers therapeutic perspectives in the context of rare diseases. ASOs are very efficient molecular tools designed to modulate gene expression through Watson–Crick base pairing with specific motifs on target transcripts (65, 66). Initially, ASOs were used to downregulate gene expression or to modify RNA splicing. Recently, ASOs have been proposed to ameliorate gene expression by directly targeting uAUG (67). Liang et al., have shown that this technique depends on many factors on the RNA and on the chemical structure of the used ASOs (67). However, targeting upORF using ASOs seems to be a very innovative and efficient genetic tool to assess in vitro the functional impact of upSNVs on protein levels. Beyond their in vitro utility, effective ASOs capable of restoring protein levels could be used as a therapeutic approach to treat rare diseases caused by upSNVs. ASOs have indeed demonstrated great potential for treating rare diseases (68–71) due to coding or splice mutations. The antisense field has remarkably progressed over the last few years with the approval of several antisense drugs and with the development of even more potent compounds (72), opening promising perspectives to treat upORF-altering variants.

In this analytic review, we focused on SNVs known, or predicted, to create upORFs. We did not discuss molecular tools that are available to determine whether these upORFs could be at the origin of functional small micropeptides that could have specific physiological roles. This topic has recently been addressed in an independent review (28). Finally, we only examined in this work SNVs that could create uAUG resulting upORFs, the most known class of variants among those that affect non-canonical ORFs. Ribosome profiling data have shown the presence of small ORFs (sORFs) in coding transcripts outside the 5′UTR but also in non-coding RNAs (73, 74). Some of these sORFs have been shown to be translated into small encoded peptides and/or to have a regulatory role on gene expression (75, 76). Thus, one can easily speculate that genetic alterations in such sORFs could also have functional consequences and be involved in human diseases. The next steps would then be to characterize the spectrum of SNVs creating or deleting TIS or Stop in non-coding transcripts.

CM and DA developed and applied the MORFEE bioinformatics tool. OS and D-AT designed the study, conducted the systematic review, and drafted the manuscript. OS and CP performed in silico annotations of the predicted upORFs. AG and ME completed the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.