Ribosome Profiling Uncovers Selective mRNA Translation Associated with eIF2 Phosphorylation in Erythrocyte Progenitors

To keep circulating erythrocytes between narrow boundaries, the erythroid progenitor compartment has a large expansion capacity, that is tightly regulated by environmental factors. Erythropoietin (Epo), stem cell factor (SCF) and glucocorticoids (dexamethasone; Dex) can induce a large increase of erythroblasts both in vivo and in vitro. We previously observed that this cell increase depends on control of selective mRNA translation. Translation is regulated by the interplay of translation factors and transcript-specific domains in the 5'- or 3'UTR of the mRNA, including secondary structures, protein binding sequences and upstream open reading frames (uORFs). Two translation factors are rate limiting; the cap-binding eukaryotic Initiation Factor 4E (eIF4E), and the initiator tRNA-binding eIF2. Upon heme deficiency, ER stress, or lack of amino acids, eIF2 is phosphorylated by heme regulated kinase (HRI), PKR-like endoplasmic reticulum kinase (PERK), or GCN2, respectively. Phosphorylation reduces the availability of eIF2, which leads to global inhibition of translation. However, uORF containing transcripts are preferentially translated. Until recently uORFs could be predicted from the primary structure, but it was unknown whether they were translated, nor how translation affected protein synthesis. Ribosome footprinting is a novel technique that reveals at nucleotide reslution which mRNA sequences are translated. The aim of our study is to identify transcripts that are particularly sensitive to lack of iron or ER stress due to phosphorylation of eIF2, and to establish the role of uORFs in the regulation of translation.

Phosphorylation of eIF2α was induced in erythroblasts by 2.5μg/ml tunicamycin (Tm) applied for 90 minutes. Cells were exposed for 7 minutes to Harringtonin (Ht) to detect start codons, or for 5 minutes to cycloheximide (CHX) to stall elongating ribosomes. Ribosome protected fragments were isolated, sequenced, and aligned to the genome adapted from a previously published strategy (E. de Klerk, 2015 NAR 43:4408). In parallel, we performed RNA-seq of these cells. Together the data give information on (i) start codon usage through the analysis of ribosome footprints in presence of Ht, and (ii) ribosome loading through comparison of ribosome footprints in presence of CHX with transcript abundance (RNAseq). Translation of the protein coding ORF of Atf4, and its targets Ddit3 (Chop) and Ppp1R15A (Gadd34) was upregulated by Tm, as expected. The ribosome protects a sequence that extends 12nt upstream of the exit(E) position of the translated codon. We detected a satisfactory footprint periodicity, but even at high RNaseI conditions a majority of footprints extends 13 nt upstream of the codon at the E position.

Overall, we observed a general repression of mRNA translation with few hypersensitive transcripts, among which the RNA binding proteins Csde1 and Pabpc1. Both Csde1 and Pabpc1 are abundantly expressed in erythroblasts. Notably, Csde1 translation was also reduced in erythroblasts from DBA patients (Horos, 2014 Blood 119:262), and we recently found that Csde1 and Pabpc1 form a complex and act in concert.

When we examined the role of uORFs in the effects of tunicamycin on mRNA translation, we observed that uORFs contribute to the regulation of some, but not all transcripts. When uORFs contribute, they start at both AUG and CUG start codons.

In conclusion, we obtained insight into the changes in translation control induced by eIF2 phosphorylation. The results are important to understand recovery of peripheral erythrocytes after blood loss (blood donation), but may also give insight into impaired erythropoiesis in anemias such as β-thalassemia, in which oxidative stress plays a role.