Nuclear and mitochondrial tRNA-lookalikes in the human genome

Abstract

We are interested in identifying and characterizing loci of the human genome that harbor sequences resembling known mitochondrial and nuclear tRNAs. To this end, we used the known nuclear and mitochondrial tRNA genes (the "tRNA-Reference" set) to search for "tRNA-lookalikes" and found many such loci at different levels of sequence conservation. We find that the large majority of these tRNA-lookalikes resemble mitochondrial tRNAs and exhibit a skewed over-representation in favor of some mitochondrial anticodons. Our analysis shows that the tRNA-lookalikes have infiltrated specific chromosomes and are preferentially located in close proximity to known nuclear tRNAs (z-score ≤ -2.54, P-value ≤ 0.00394). Examination of the transcriptional potential of these tRNA-lookalike loci using public transcript annotations revealed that more than 20% of the lookalikes are transcribed as part of either known protein-coding pre-mRNAs, known lncRNAs, or known non-protein-coding RNAs, while public RNA-seq data perfectly agreed with the endpoints of tRNA-lookalikes. Interestingly, we found that tRNA-lookalikes are significantly depleted in known genetic variations associated with human health and disease whereas the known tRNAs are enriched in such variations. Lastly, a manual comparative analysis of the cloverleaf structure of several of the transcribed tRNA-lookalikes revealed no disruptive mutations suggesting the possibility that these loci give rise to functioning tRNA molecules.

Keywords: human genome; mitochondrial tRNA; nuclear tRNA; tRNA; tRNA fragment.

Figure 1

Number of identified tRNA-lookalikes as a function of the allowed number of mismatches. In addition to locating all 632 tRNA-Reference entries (green bars) we found many other nuclear genomic loci that harbored tRNA-lookalikes. Of the tRNA-lookalikes that we have identified, only a few (dark red bars) are currently labeled as tRNAs in RepeatMasker (RM) whereas the rest are novel. See Methods for the two filtering criteria (minimum length and maximum allowed mismatches) that we used.

Figure 2

Identity of the best matching source template. Many of the discovered tRNA-lookalikes best resemble a nuclear tRNA source sequence (dark blue segments) with twice as many resembling a mitochondrial tRNA source (cyan segments).

Figure 3

Distribution of only the tRNA-lookalikes grouped by the coding amino acid. (A) Distribution based on the coded amino acid. (B) Distribution based on the actual anticodon sequence that is used. In each panel the distribution is shown as a function of the allowed mismatches. No tRNA-Reference entries were included in this chart.

Figure 4

Distribution and density of true tRNAs and lookalikes across the various chromosomes. (A) Distribution of only the true tRNA-Reference entries across the chromosomes. (B) Distribution of only the tRNA-lookalikes that resemble true tRNAs across chromosomes shows the speed of chromosomal penetration by the lookalikes as a function of the allowed number of mismatches. (C) Distribution of the total number of tRNA instances (=true tRNA-Reference entries + tRNA-lookalikes) across the chromosomes. In all three panels, the number shown in each cell is the number of tRNA instances for the corresponding combination of chromosome and number of mismatches whereas the color of each cell represents the density of the chromosome in tRNA instances per million bases (See Methods).

Figure 5

TRNA-lookalikes and annotated RNA transcripts. Distribution of the number of tRNA-lookalikes that are wholly present in known, annotated transcripts as a function of the number of allowed mismatches.

Figure 6

Examining the functional potential of tRNA-lookalikes. Examples of the mitochondrial tRNA-Reference entry for Methionine (A) and of two tRNA-lookalike hypothesized secondary structures shown in panels (B,C). The nucleotides shown in red correspond to sequence differences from the reference gene. The lookalike in (B) belongs to Category A, while the one in (C) belongs to Category C (See Methods). The lookalike shown in panel (B) is located on the forward (+) strand of chromosome 1, between positions 564952 and 565019 inclusive; the lookalike shown in panel (C) is located on the reverse (–) strand of chromosome 9, between positions 81357661 and 81357728 inclusive.

Figure 7

Examples from the compiled dataset. (A) Example of two entries, corresponding to a tRNA-lookalike and a tRNA-Reference gene (KnowntRNAgene) respectively from the database that we have generated. The entries correspond to an AspGTC tRNA. The shown columns, from left to right, contain a description of the genetic locus, the percentage identity and the number of mismatches (i.e., number of non-identical bases with the best tRNA-Reference entry) with the best-matched hit, the presence (or absence) in RepeatMasker (RM) or in NUMT, the genomic location (chromosome, strand, “from” and “to” coordinate, the best-matching hit (if a lookalike) or the name of the tRNA (if the entry is already in tRNA-Reference), the sequence of 50 nts up-stream of the hit and of 50 nts down-stream of it. (B) Clustal-W alignment for the tRNA and its lookalike and separately for each of its two flanking regions. Asterisks indicate identical bases and dashes indicate gaps. The anticodon of the shown tRNA is GTC and shown here boldfaced and underlined in context: […]ATCCCCGCCTGTCACGCGGGAGACC[…]. All sequences are shown in 5′ to 3′ orientation.