Evolutionarily conserved inhibitory uORFs sensitize Hox mRNA translation to start codon selection stringency

Abstract

Translation start site selection in eukaryotes is influenced by context nucleotides flanking the AUG codon and by levels of the eukaryotic translation initiation factors eIF1 and eIF5. In a search of mammalian genes, we identified five homeobox (Hox) gene paralogs initiated by AUG codons in conserved suboptimal context as well as 13 Hox genes that contain evolutionarily conserved upstream open reading frames (uORFs) that initiate at AUG codons in poor sequence context. An analysis of published cap analysis of gene expression sequencing (CAGE-seq) data and generated CAGE-seq data for messenger RNAs (mRNAs) from mouse somites revealed that the 5' leaders of Hox mRNAs of interest contain conserved uORFs, are generally much shorter than reported, and lack previously proposed internal ribosome entry site elements. We show that the conserved uORFs inhibit Hox reporter expression and that altering the stringency of start codon selection by overexpressing eIF1 or eIF5 modulates the expression of Hox reporters. We also show that modifying ribosome homeostasis by depleting a large ribosomal subunit protein or treating cells with sublethal concentrations of puromycin leads to lower stringency of start codon selection. Thus, altering global translation can confer gene-specific effects through altered start codon selection stringency.

Keywords: Hox genes; eIF1; eIF5; start codon selection stringency; uORF.

Fig. 1.

A schematic representation of the role of eIF1 and eIF5 in regulating stringency of start codon selection in eukaryotes. (Left) mRNA with the mORF AUG start codon in suboptimal context (mAUG*). (Right) mRNA with inhibitory uORF initiated by an AUG codon in suboptimal context (uAUG*). High levels of eIF1 confer high stringency of start codon selection leading to 1) repressed expression of genes initiated by mAUG* codons and 2) increased leaky scanning of uAUG* codons resulting in the derepression of mORF translation. Low stringency of start codon selection in cells with high levels of eIF5 has the opposite effects, stimulating translation of ORFs starting with mAUG* and repressing the translation of the mORF on mRNAs with inhibitory uORFs initiating at uAUG* codons.

Fig. 2.

Hox genes with conserved poor context mAUG start codon or conserved uORF initiated by suboptimal uAUG codon. (A) The initiation context for the mAUG start codon of five Hox genes in which the mAUG is the main or only start codon. The suboptimal context of human eIF1 and the consensus optimal (Kozak) context are also shown. (B) A cartoon representation, approximately to scale, of the seven Hox leaders containing uORFs (light blue) conserved from mammals to fish. The suboptimal context of the three uAUG start codons of human eIF5 are also shown. The initiating AUG codon is shown in green; critical consensus optimal nucleotides at position −3 and +4 are shown in brown; nonoptimal residues at positions 3 and +4 are shown in red; consensus residues in positions −6, −5, −4, −2, and −1 are shown in cyan; nonconsensus residues in the same positions are shown in black. (C) Screenshot from the GWIPS-viz (https://gwips.ucc.ie/) browser (31) of global aggregate ribosome footprint profiling (riboseq) coverage in the leaders of mouse Hoxa1, Hoxa9, and Hoxa11. Reads mapping to the uORFs are boxed in blue and represented in the cartoons by with blue rectangles. The mORFs are depicted by yellow rectangles. The first nucleotide of each sequence matches the 5′ end of the mRNA as determined by CAGE-seq of somites (Fig. 3 and SI Appendix, Figs. S1 and S2).

Fig. 3.

The CAGE-seq analysis of mouse somites reveals relatively short leaders on the Hoxa1, Hoxa9, and Hoxa11 mRNAs. Combined reads of three replicates of CAGE-seq analysis mapping the transcription start sites in mouse somites for (A) Hoxa1, (B) Hoxa9, and (C) Hoxa11. For Hoxa9 and Hoxa11, the schematic at Top shows proposed “IRES” elements (red box) (red box, 14) in the transcript drawn to scale. The schematics at Bottom display the National Center for Biotechnology Information (NCBI) annotated transcripts, shown to scale; exons are shown as rectangles, introns as lines; the mORF coding region is shown as a black rectangle, with the start and stop codons indicated by inverted green or red triangles, respectively. Chromosomal coordinates (GRCm38/mm10) of the mapping data and 1-kb scale bar are shown above and below, respectively, each figure. Complementary qRT-PCR analyses reveal that the Hoxa9 CDS is at least 200-fold more abundant in total somite mRNAs that the putative IRES (SI Appendix, Fig. S7).

Fig. 4.

The uORFs of Hoxa1, Hoxa9, and Hoxa11 confer translation regulation following global changes in the stringency of start codon selection. (A) U2OS cells were transfected with luciferase reporters containing the leaders of Hoxa1, Hoxa9, and Hoxa11 or mutant versions in which the uAUG codon was mutated to a noninitiating AAA codon. The relative activity of Renilla luciferase reporter to firefly luciferase transfection control is plotted. The percent activity of wild-type versus mutant reporter is indicated on Right. (B) U2OS cells were cotransfected with luciferase reporters described in A, a reporter initiated by an AUG codon in poor EIF1 context (residues −6 to +4) or a reporter containing the leader of EIF5, and with plasmids overexpressing eIF1 (blue), eIF5 (orange), or with empty vector (gray). Activities of the test reporter, shown schematically on Left, are presented as the percentage of the control reporter, shown schematically on Right. The control reporter for eIF1 is initiated by an AUG codon in optimal context; for other controls, the uAUG codons are mutated to noninitiating AAA codons. For Hox reporters, test Renilla luciferase activities were normalized to firefly luciferase activity expressed from the same plasmid. For EIF1 and EIF5 reporters, test firefly activity was normalized to activity from cotransfected Renilla n.s., not significant, *P < 0.05, **P < 0.01, ***P < 0.001 (Student’s two-tailed t test; n = 8, assayed in duplicate).

Fig. 5.

Inhibition of global translation reduces stringency of start codon selection. A comparison of luciferase activities of U2OS cells transfected with luciferase reporters initiated by AUG codons in poor EIF1 context (residues −6 to +4), shown schematically on Left, or optimal context, shown schematically on Right. (A) Cells cotransfected with an empty vector or a plasmid expressing an shRNA targeting RPL11. (B) Untreated cells or cells treated with 750 ng/mL puromycin. Firefly activity was normalized to Renilla activity from a cotransfected plasmid initiating from an AUG codon in optimal context. Error bars denote SD. ***P < 0.001 (Student’s two-tailed t test; n = 8, assayed in duplicate). No statistically significant changes in reporter mRNA levels were detected in these experiments (SI Appendix, Fig. S10B).