**4. Conclusions**

Studies strongly suggest that the tRNA cloverleaf structure unfolded prior to the appearance of a fully functional ribosomal core, making it one of the most ancient RNAs of the RNA world [70, 97] or even the oldest [98]. Though the "RNA-world" hypothesis is well accepted, the successive events leading to the emergence of different partners playing a role in translation and the involvement of tRNAs in this evolution are highly controversial coveted field [99]. However, some hypotheses as the "tRNA core" [88] strongly suggest that tRNAs would be at the origin of the primitive genetic material and gave rise to mRNA and rRNA, as well as the conformational structure of the first proto-ribozymes. The base module being a pleiofunctional RNA that can adopt the cloverleaf structure is found today in various sequences without direct link with translation. One may conclude that "one should not change a winning secondary structure." In a precellular context, a molecule with ss-tRNA characteristics (small ORF associated with cloverleaf structure) would be advantageous. Putatively, ss-tRNA-like molecules cumulating both tRNA and mRNA functions would have been the first molecules on Earth to support nonrandom protein synthesis.

The antiquity of ss-tRNAs can be discussed, and it is very likely that the TAR10 (and especially TAG) triplets played very early a critical role in the tertiary folding of some tRNAs. Their implication in translation termination would be an exaptation where firstly, they were part of a structural signal. Origin of ATR49 triplets is less clear perhaps tracing to the first endosymbiosis. Hence it would be apomorphic (derived character). Analyzes by taxa and tRNA species suggest a nonhomogeneous evolution. At the beginning of the RNA/protein world, it has quickly become essential to start peptide synthesis at particular codons and one cannot exclude that ATR49 was an ancestral state which would have not been retained as intergenic spaces increased. Analyzes of known tRNAs of α-proteobacteria and cyanobacteria could suggest that in organelles, ATR49 triplets would have been selected with genome reduction. Organelle genomes may be under increased pressure for size reduction with resulting overlaps (see, [100]). However, several features strongly suggest that overlapping genes are not a direct mechanism to substantially reduce genome size. Gene overlaps allow mtDNA genome compaction while avoiding the loss of tRNA genes [53]. Nevertheless, overlaps may allow a more efficient control in the regulation of gene expression, the regulatory pathways are simplified, and the number of proteins (and genes) required decreases [100]. Among others, short antiparallel overlaps may be involved in antisense regulatory mechanisms. Consequently, genomes with compact sizes enable putatively less flexible but more efficient physiologies.

The selection of tRNAs had to be done mainly on two seemingly opposite criteria, stability and plasticity, making it a kind of Swiss army knife of the RNA world. This explains that beyond their central role in protein synthesis, tRNAs have many other crucial functions. To date, it can be hypothesized that ss-tRNAs might regulate gene expression, stress responses, and metabolic processes. Indeed, *in silico* analyzes allowed to speculate that several overlapping sequences may code simultaneously for mRNAs and tRNAs in most of the metazoan mt-genomes. These overlaps can have a variable (sometimes large) number of nts; however, when annotating their genomes, several authors voluntarily underestimated the number and the size of overlaps, speculating that there would be upstream abbreviated stop codons or downstream alternative start codons but most often without any direct demonstration so far. However, the high number of possible overlaps on the same strand in which the first in-frame complete stop codon or standard start codon are located at specific positions in the sequences of *trn* genes (TAR10 and ATR49, respectively) strongly suggest an exclusive relationship between obtaining tRNAs and translation of mRNAs and/or the development of repair system to keep the two genes functional due in some cases to co-evolution during several hundred MY. We can therefore speculate that ss-*trn* genes could allow true tRNA punctuation and initiation. Noted that ss-tRNAs seem to be hybrid molecules which would contain three essential coding or decoding informations in the form of nt triplets (i.e., anticodon and stop/ start codons) which are all at least in part integrated into stem or loop; moreover, after the ATR49, nt triplets play the role of internal sense codons. To date, it is unclear what biochemical mechanism would allow to choose between different alternate cleavage sites, leading to the complete tRNA rather than to the mRNA or *vice versa*, but reduced/expanded proteins can be functional, and various processes including editing suggest this also for incomplete tRNAs. Hence, despite lacking experimental evidence, TAR10 and ATR49 triplets have probable roles, including regulation. Future analyzes of the processed bicistronic transcripts (tRNA/ protein-encoding or the contrary) are required. Moreover, even if mt-*trn* genes are most often expressed at very low levels [53], only direct sequencing of tRNAs can validate transcription, epitranscriptomic maturation and can pinpoint nt modifications including post-transcriptionally edited positions. Purified native, or even synthetic, tRNAs should also be tested for their *in vitro* activity to confirm the functionality of aberrant transcripts. Similar experiments must be made on the flanking mRNAs and their products. If as we think, ss-tRNAs could play regulatory roles, initially experiments should compare stress and nonstress conditions.

Here, the bias for metazoan mtDNA does not allow for a complete picture of variation in the entire eukaryotic world, and protist mt-genomes should also be considered. Special attention should also be paid to noncanonical base pairings potentially formed by UAR10 and AUR49 nts, in perspective with tRNA structure and V-R length. Accounting for TAR10 and ATR49 triplet presences in the algorithms predicting tRNAs could improve mt-genome annotations, reducing numbers of false positives and negatives, and more accurately determine tRNA termini while accounting tRNA species, taxa, and genomic systems.

MtDNA plays a central role in apoptosis, aging, and cancer [13]. Moreover, mt-diseases are among the most common inherited metabolic and neurological disorders [101]. In addition, as new functions and new mechanisms of action of tRNAs are continuously discovered [1] and as ss-*trn* genes could affect the cellular dynamic during normal and stress conditions leading to pathologies, potential subtleties of action and regulation of these genes and products should be more thoroughly investigated.
