Letter
https://doi.org/10.1038/s41586-018-0243-7
Codon-specific translation reprogramming promotes resistance to targeted therapy Francesca Rapino1,2, Sylvain Delaunay1,2, Florian Rambow3,4, Zhaoli Zhou1,2, Lars Tharun5, Pascal De Tullio6, Olga Sin7,8,9, Kateryna Shostak2,10, Sebastian Schmitz1,2, Jolanda Piepers11, Bart Ghesquière12, Latifa Karim2,13, Benoit Charloteaux2,13, Diane Jamart1,2, Alexandra Florin5, Charles Lambert2, Andrée Rorive14, Guy Jerusalem14, Eleonora Leucci3,4, Michael Dewaele3,4, Marc Vooijs11, Sebastian A. Leidel7,8,9, Michel Georges2,13, Marianne Voz2, Bernard Peers2, Reinhard Büttner5, Jean-Christophe Marine3,4, Alain Chariot2,10,15 & Pierre Close1,2,15*
Reprogramming of mRNA translation has a key role in cancer development and drug resistance 1. However, the molecular mechanisms that are involved in this process remain poorly understood. Wobble tRNA modifications are required for specific codon decoding during translation2,3. Here we show, in humans, that the enzymes that catalyse modifications of wobble uridine 34 (U34) tRNA (U34 enzymes) are key players of the protein synthesis rewiring that is induced by the transformation driven by the BRAFV600E oncogene and by resistance to targeted therapy in melanoma. We show that BRAFV600E-expressing melanoma cells are dependent on U34 enzymes for survival, and that concurrent inhibition of MAPK signalling and ELP3 or CTU1 and/or CTU2 synergizes to kill melanoma cells. Activation of the PI3K signalling pathway, one of the most common mechanisms of acquired resistance to MAPK therapeutic agents, markedly increases the expression of U34 enzymes. Mechanistically, U34 enzymes promote glycolysis in melanoma cells through the direct, codon-dependent, regulation of the translation of HIF1A mRNA and the maintenance of high levels of HIF1α protein. Therefore, the acquired resistance to anti-BRAF therapy is associated with high levels of U34 enzymes and HIF1α. Together, these results demonstrate that U34 enzymes promote the survival and resistance to therapy of melanoma cells by regulating specific mRNA translation. Elongator proteins Elp1 and Elp3 and their partners in U34 tRNA modification, the cytosolic thiouridylases subunits 1 and 2 (Ctu1 and Ctu2) were found to be upregulated in BRAFV600E;tp53−/− melanoma lesions4 compared to normal adjacent skin in zebrafish that expressed human BRAF(V600) (Extended Data Fig. 1a, b). Note that about 50% of human melanoma carry BRAF mutations, and BRAFV600E is the most prevalent mutation, which results in constitutively overactive MAPK– MEK–ERK signalling5. Notably, genetic inactivation of elp3 using CRISPR–Cas9 impaired the development of BRAFV600E melanoma in zebrafish (Fig. 1a and Extended Data Fig. 1c). ELP1, ELP3 and CTU2 protein levels were also significantly increased in tumour biopsies of human melanoma, in short-term melanoma cultures (MM lines) and established cells lines compared to human primary melanocytes or fibroblasts (Fig. 1b, c and Extended Data Fig. 1d, e). This upregulation was independent of the mutational status of the sample. Short hairpin RNA (shRNA)-mediated knockdown of ELP3 or CTU1/CTU2 markedly compromised the viability of human BRAFV600E melanoma cells, but not of normal human melanocytes or two independent BRAFWT melanoma cultures (Fig. 1d and Extended Data Fig. 1f, g). Moreover, stable expression of human oncogenic BRAF(V600E) in B16 mouse
melanoma cells increased the sensitivity of these cells to both vemurafenib (that is, BRAF inhibition) and depletion of Elp3 and Ctu2 (Extended Data Fig. 1h, i). Collectively, these data indicate that U34 enzymes are upregulated in melanoma and that BRAFV600E melanoma cells are dependent on high expression levels of these enzymes. Notably, expression of catalytic-domain-deficient ELP3 mutants6 in ELP3depleted cells failed to rescue BRAFV600E cell viability, indicating that the catalytic activity of elongator proteins is required for the survival of BRAFV600E melanoma cells (Extended Data Fig. 1j, k). Quantitative proteomic analysis after stable isotope labelling by amino acids in cell culture indicated that the abundance of proteins enriched in codons that require U34 tRNA modification for decoding (that is, the AAA, GAA and CAA codons) was significantly increased in BRAF(V600E)-expressing mouse cells (such as Glb1 or Cfdp1; Extended Data Fig. 1l, m and Supplementary Table 1; the U34 enrichment value is 8.82% of the fourth quartile of the normal distribution of the frequency of sensitive codons in all of the genome; χ2 test, P = 4.06 × 10−34). According to gene set enrichment analysis and STRING (search tool for the retrieval of interacting genes/ proteins; https://string-db.org) interactome analyses, BRAF(V600E)expressing mouse cells are markedly involved in glycolytic and hypoxiaresponse pathways (gene set enrichment analysis, P = 1.35 × 10−14 and STRING, P = 1.2 × 10 −23; Extended Data Fig. 1l–n and Supplementary Table 2). In accordance with this in silico prediction, expression levels of hypoxia-induced factor 1α (HIF1α), which induces the glycolysis gene expression program7, was enhanced in BRAFV600Eexpressing cells and clinical specimens of melanoma compared to BRAFWT samples (Extended Data Fig. 2a, b). Concomitantly, increased expression of HIF1α-target genes, increased glucose uptake and lactate production (Fig. 1e and Extended Data Fig. 2c–f) were observed in primary human BRAFV600E melanoma lines. Tracer experiments using nuclear magnetic resonance, Seahorse and metabolomics confirmed the prevalence of the HIF1α-dependent metabolism in BRAFV600Eexpressing melanoma cells, as evidenced by high levels of lactate production and a reduced dependency on the tricarboxylic acid cycle (Fig. 1e, f and Extended Data Fig. 2g, h). Although short-term BRAFV600E-expressing melanoma cultures exhibited increased glucose uptake compared to BRAFWT cultures, previous studies have established that the metabolic profile of melanoma cells is independent of their mutational status8,9. Accordingly, both short-term cultures of MM011 and MM047 carry the NRASQ61* mutation, but only one of these lines (that is, MM047, but not MM011) exhibited a high HIF1α metabolism and was sensitive to silencing
1 Laboratory of Cancer Signaling, University of Liège, Liège, Belgium. 2GIGA-institute, University of Liège, Liège, Belgium. 3Laboratory for Molecular Cancer Biology, Department of Oncology, KU Leuven, Leuven, Belgium. 4Center for Cancer Biology, VIB, Leuven, Belgium. 5Institute for Pathology, University Hospital Cologne, Cologne, Germany. 6Centre for Interdisciplinary Research on Medicines (CIRM), University of Liège, Liège, Belgium. 7Max Planck Research Group for RNA Biology, Max Planck Institute for Molecular Biomedicine, Münster, Germany. 8Faculty of Medicine, University of Münster, Münster, Germany. 9Cells-in-Motion Cluster of Excellence, University of Münster, Münster, Germany. 10Laboratory of Medical Chemistry, University of Liège, Liège, Belgium. 11 Department of Radiotherapy (MAASTRO)/GROW, School for Developmental Biology and Oncology, Maastricht University, Maastricht, The Netherlands. 12Metabolomics Core Facility, Center for Cancer Biology, VIB, Leuven, Belgium. 13Unit of Animal Genomics, University of Liège, Liège, Belgium. 14Medical Oncology, CHU Sart Tilman Liège, University of Liège, Liège, Belgium. 15WELBIO, University of Liège, Liège, Belgium. *e-mail:
[email protected]
N A t U r e | www.nature.com/nature
© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
RESEARCH Letter
30
Gapdh
20
elp3wt
10
elp3mut
20
CTU1
0.2
CTU2 Adjacent tissue
P = 7 × 10–5
Control shRNA ELP3 shRNA CTU1 shRNA
P=4×
10–6
f
MM074 BRAF(V600E)
P = 3 × 10–10
MM029 BRAF(V600E)
5 #2 MM117 MM163 MM074 MM029
Melanocytes
BRAF(V600E) melanoma
P = 0.003 P = 4.07 × 10–5 P = 2.31 × 10–5 P = 0.005 P = 0.008 P = 0.002 P = 0.0006 P = 0.001
10
0
BRAF(WT) melanoma
Lactate
20
MM117 BRAF(WT) MM074 BRAF(V600E) MM029 BRAF(V600E) 0
2
4 6 Time (h)
MM117 BRAF(WT)
P = 2 × 10–9
10
#1
GAPDH
Melanoma
15
0
Lactate (AU)
0.4
20
Melanoma
BRAF(WT) BRAF(V600E)
ELP3
Fractional labelling
Relative nuclear fragmentation (%)
15 Time (weeks)
Melanocytes ELP1
CTU2
0.6
0
0 10 25
e
P = 0.0102
0.8
c
P = 9.2 × 10–43 –20 P = 4.7 × 10–12 P = 1.3 × 10
M 1 M 17 M 16 3 M M 07 4 M M 02 9
Elp3(mut)
ELP1 ELP3
#1 #2
40
1.0
M
Elp3(wt)
0.9
0.6 P = 0.0002
0.6
0.4 P = 0.02
0.3
0.2
0 Alanine
Fractional labelling
Tumour appearance (%)
d
b
elp3wt elp3mut
50
Relative expression
a
P = 0.001
Citrate 0.6
0.2
0.2 α-KG
0.4
0.4
P = 0.02
0
P = 0.001
0.6
0.6 0.4
0
P = 0.0001
0
P = 0.003 P = 0.002
0.2
Malate
0
P = 0.006 P = 0.002
Aspartate
8
Fig. 1 | U34 enzymes promote survival of melanoma that are dependent on HIF1α metabolism. a, Tumour appearance in control (n = 84) or Elp3 mutant (n = 44) Tg(mitfa:BRAFV600E);tp53−/− zebrafish. GehanBreslow–Wilcoxon test. b, Immunohistochemistry analysis of samples from patients with melanoma. n = 10 biologically independent samples, two-sided t-test, the mid line indicates the mean, whiskers are minimum to maximum. c, Protein levels of U34 enzymes in human cells. n = 2 independent experiments. d, Death of primary short-term melanoma
by fluorescence-activated cell sorting. n = 3 independent experiments, two-sided t-test, data are mean + s.d. e, Nuclear magnetic resonance (NMR) quantification of lactate in the medium. AU, arbitrary units normalized to protein concentration. n = 3 independent experiments, two-sided t-test, data are mean + s.d. f, Liquid chromatography–tandem mass spectometry quantification of 13C-glucose fractional labelling of tricarboxylic acid intermediates. α-KG, α-ketoglutarate. n = 4 independent experiments, two-sided t-test, data are mean + s.d.
of U34 enzymes (Extended Data Fig. 2i–m). This indicates that the dependence of melanoma cells on U34 enzymes relies mostly on their metabolic, rather than mutational, status. Wobble U34 tRNA modification is required to decode AAA, GAA and CAA codons in mRNAs during translation10–12. Using ribosome profiling, we found that ELP3 depletion did not globally affect ribosome footprinting density. Rather, subsets of mRNAs that were particularly rich in U34 codons (top 1%, n = 137; top 5%, n = 683; top 10%, n = 1,366; top 20%, n = 2,732) collectively presented an increase in ribosome footprint occupancy with respect to other mRNAs upon ELP3 depletion (P
