Translation initiation downstream from annotated start codons in human mRNAs coevolves with the Kozak context

Abstract

Eukaryotic translation initiation involves preinitiation ribosomal complex 5 ′-to-3 ′ directional probing of mRNA for codons suitable for starting protein synthesis. The recognition of codons as starts depends on the codon identity and on its immediate nucleotide context known as Kozak context. When the context is weak (i.e., nonoptimal), leaky scanning takes place during which a fraction of ribosomes continues the mRNA probing. We explored the relationship between the context of AUG codons annotated as starts of protein-coding sequences and the next AUG codon occurrence. We found that AUG codons downstream from weak starts occur in the same frame more frequently than downstream from strong starts. We suggest that evolutionary selection on in-frame AUGs downstream from weak start codons is driven by the advantage of the reduction of wasteful out-of-frame product synthesis and also by the advantage of producing multiple proteoforms from certain mRNAs. We confirmed translation initiation downstream from weak start codons using ribosome profiling data. We also tested translation of alternative start codons in 10 specific human genes using reporter constructs. In all tested cases, initiation at downstream start codons was more productive than at the annotated ones. In most cases, optimization of Kozak context did not completely abolish downstream initiation, and in the specific example of CMPK1 mRNA, the optimized start remained unproductive. Collectively, our work reveals previously uncharacterized forces shaping the evolution of protein-coding genes and points to the plurality of translation initiation and the existence of sequence features influencing start codon selection, other than Kozak context.

Determinants of genome-wide distribution and evolution of uORFs in eukaryotes

“…Interestingly, NTEs showed lower O/E ratios than both oORFs and nORFs in 457 out of 478 species (Supplementary Fig. S4 ), suggesting that novel NTEs were selected against as they might alter protein functions 21 .…”

“…For an AUG triplet in the 5′ UTR (defined as “uAUG” hereafter), it can function as the start codon of a uORF that has a stop codon either preceding the start codon of the downstream CDS (nonoverlapping uORF, nORF) or residing in the body of the downstream CDS (out-of-frame overlapping uORF, oORF) 4 , 11 – 18 . Less frequently, an uAUG can function as the start codon of an ORF whose stop codon overlaps with the stop codon of the downstream CDS (N-terminal extension, NTE) 4 , 19 – 21 . The advent of ribosome profiling 22 – 26 , a method that determines the ribosome occupancy on mRNAs at the codon level, has enabled the genome-wide characterization of uORFs and NTEs that showed evidence of translation in various species with high sensitivity and accuracy 12 , 16 , 17 , 27 – 36 .…”

Combinatorial analysis of translation dynamics reveals eIF2 dependence of translation initiation at near-cognate codons

“…In the case of genes with an initial AUG codon in a weak Kozak context, the 40S complex continues to scan beyond this point in a process known as leaky scanning, and translation is initiated at the next downstream AUG codon ( 55 ). Reporter construct analysis has recently validated that 10 human genes are translated from an AUG initiation codon downstream of the canonical AUG initiation codon ( 56 ), and TISCA identified 4 of the 10 truncated ORFs ( Supplementary Figure S4A ). Both canonical and truncation AUG initiation codons were identified for the AASDHPPT gene, whereas only the truncation AUG initiation codon was identified for the other three genes ( CMPK1 , ISL2 and LIMK1 ), suggesting that the translation initiation efficiency of the canonical AUG initiation codon for these genes is low ( Supplementary Figure S4A ).…”

Differential processing and localization of human Nocturnin controls metabolism of mRNA and nicotinamide adenine dinucleotide cofactors

“…For NOCT, initiation at M1 would produce the full length 431 amino acid protein. However, the sequence context of NOCT M1 (5′-CCCGGCAUGU, Kozak consensus match in bold) differs from the optimal context, raising the possibility for leaky scanning and initiation at the downstream M67 (5′-UGUUCCAUGG) (30).…”