A structural element within the 5'UTR of β-catenin mRNA modulates its translation under hypoxia

Abstract

Tight regulation of translation initiation is crucial for cellular adaptation to environmental changes. Stress conditions like hypoxia trigger translational reprogramming of mRNAs encoding proteins essential for stress recovery and cell survival. Recent studies highlight alternative translation initiation pathways based on specific motifs in mRNA 5' untranslated regions (5'UTRs). Notably, β-catenin is of particular interest since maintaining its translation promotes cancer cell persistence and plasticity. β-Catenin, an oncogenic protein, plays a key role in Wnt signalling. Besides dysregulation of the β-catenin/Wnt pathway, chemotherapy-induced hypoxia leads to abnormal nuclear β-catenin accumulation, modulating gene expression linked to cancer progression and metastasis. However, the mechanism sustaining β-catenin translation in stressed cells remains elusive. To explore how β-catenin mRNA evades global translational blockade in hypoxic cancer cells, we analysed its 5'UTR and identified a translation regulatory element in cellulo. We discovered a GC-rich three-way junction (TWJ) structure within the β-catenin 5'UTR enhancing its hypoxia-driven translation. A polypurine region within the TWJ anchors eIF4B, eIF4A, and eIF4G2. Importantly, the TWJ makes β-catenin mRNA translation eIF4A-dependent and sensitive to silvestrol, a selective eIF4A inhibitor and promising anticancer agent. This study elucidates the 5'UTR-driven β-catenin mechanism under hypoxia, paving the way to inhibit its translation in cancer.

Graphical Abstract

Figure 1.

β-Catenin 5′UTR exhibits the presence of two structurally distinct domains. Secondary structure of β-catenin 5′UTR based on SHAPE, DMS, and CMCT chemical probing. The reactivities are shown as average reactivity from three independent experiments. A representation of reactivities is assigned to each probe as color code depending on a range of values as shown in the figure legend on the right. Based on these reactivities, two domains—here delimited by dashed grey boxed —were identified. RNA helices are indicated by H1-6 while start codon is highlighted by a cyan blue square. GC distribution diagram is represented by plotting the average GC-content value of 30-nucleotide windows each 15 nucleotides. The β-catenin untranslated regions (UTRs) and the ORF are represented below the GC distribution diagram.

Figure 2.

The GC-rich element of β-catenin 5′UTR enhances mRNA translation under hypoxia. (A) Schematic representation of the CTNNB1 variants and the control transcript (HBB) used in this study. (B) TE of CTNNB1 variant transcripts measured 4 h after mRNA transfection in HeLa cells, performed in triplicate with Renilla luciferase as a cotransfection control. An unpaired t-test was performed comparing translation levels to that of CTNNB1_1. (C) TE of CTNNB1 variant transcripts measured 4 h after mRNA transfection in HeLa cells cultured under normoxic and/or hypoxic conditions, in triplicate with Renilla luciferase as a cotransfection control. Unpaired t-test was performed to compare β-catenin mRNA translation in normoxia versus hypoxia. (D) Stability of CTNNB1 mRNA variants 4 h after their transfection in HeLa cells cultured under normoxic and severe hypoxic conditions. (E) Fold change in endogenous CTNNB1 mRNA level in HeLa and SW480 cells exposed to various hypoxic condition compared to standard oxygen levels. (F) Cap-dependence of β-catenin mRNA translation was assessed using a hypoxic and/or nHCTS. The CTNNB1_1 mRNA construct was used either m7G-capped (m7G) or A-capped (A). ns (no significant), *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001.

Figure 3.

The enrichment of eIFs involved in 43S PIC recruitment in the hypoxic LS48S IC correlates with the presence of the CTNNB1 GC-rich element. (A) Semiquantitative mass spectrometry analysis of normoxic and hypoxic HeLa LS48S IC, indicating the abundance of each eIFs based on the normalized spectrum counts (NSCs). The NSCs are presented as heatmaps using a greyscale gradient, where white indicates low abundance and black denotes high abundance. The list includes eIFs, as well as proteins related to translation initiation—such as ABCE1, PDCD4, PKR, and RACK1—and to mRNA maturation, such as eIF4A3. Stars indicate the values of the coefficients of variation calculated for each NSC. (B) Fold change and adjusted p-value from the comparison between normoxic and hypoxic FL-CTNNB1 LS48S IC protein composition is represented in a volcano plot. Significantly enriched proteins are represented as grey dots, with eIFs specifically highlighted in red (normoxic complex) and green (hypoxic complex). (C) Volcano plot comparing protein composition of the hypoxic ΔGC-CTNNB1 LS48S IC versus FL-CTNNB1 LS48S IC. Red and green dots represent the proteins significantly more abundant in the hypoxic complex bound to ΔGC-CTNNB1 mRNA or to FL-CTNNB1 mRNA, respectively. For both volcano plot analyses, the threshold was set to Log2FoldChange cutoff of ± 1 and -log10(adjp) ≥ 1.3 (P < 0.05).

Figure 4.

The GC-rich element binds eIF4A1 and eIF4B in vitro through H2 internal loops and H3 stem–loop. (A) EMSA experiments revealed the binding of human eIF4B to the GC-rich element. RNAs used in this study are schematized on the right and named RNA 1, 2, and 3: GC-rich element, AU-rich element and ORF (102 nts CTNNB1 ORF). EMSAs were performed with a fixed concentration of radiolabelled RNA ± 0.5 μM eIF4A1 and/or ± 0.5 μM eIF4B. (B) Footprinting of eIF4A1 and eIF4B on the purine track of the GC-rich element. In the RNA experimental condition, SHAPE reactivities were determined by comparing the electropherograms of BzCN-modified RNA with those of unmodified RNA. For protein conditions, electropherograms of BzCN-modified RNA with or without protein incubation were compared. Footprinting experiments were done in triplicates and only one replicate of each condition is represented.

Figure 5.

Silvestrol inhibits β-catenin translation exclusively in presence of the GC-rich element. (A) mRNA transfection experiments were performed on HeLa cells in normoxic or hypoxic conditions using Renilla luciferase as cotransfection control. Cells were treated either with 0.005% DMSO (v/v) or 25 nM silvestrol prior to transfection. (B) Same experiment was made in SW480 and U87MG cells for the CTNNB1_1 construct. An unpaired t-test comparing translation of one mRNA treated with DMSO to its level under silvestrol treatment. Experiments were performed in triplicates. ns (no significant), *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001.

Figure 6.

Schematic representation of β-catenin mRNA translation regulation under distinct oxygen condition. Under normoxia (left), β-catenin mRNA translation initiation proceeds via the canonical cap-dependent mechanism. Under hypoxia (right), the TWJ element within the β-catenin 5′ UTR anchors a 43S-recruiting complex, in which eIF4G2 connects to the cap via eIF3d, thus enabling eIF4E-independent translation.