If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Louvain Drug Research Institute (LDRI), Université catholique de Louvain (UCLouvain), Brussels, BelgiumPole of Pharmacology and Therapeutics, Institut de Recherche Expérimentale et Clinique (IREC), Université catholique de Louvain (UCLouvain), Brussels, Belgium
Despite being initially regarded as a metabolic waste product, lactate is now considered to serve as a primary fuel for the tricarboxylic acid cycle in cancer cells. At the core of lactate metabolism, lactate dehydrogenases (LDHs) catalyze the interconversion of lactate to pyruvate and as such represent promising targets in cancer therapy. However, direct inhibition of the LDH active site is challenging from physicochemical and selectivity standpoints. However, LDHs are obligate tetramers. Thus, targeting the LDH tetrameric interface has emerged as an appealing strategy. In this work, we examine a dimeric construct of truncated human LDH to search for new druggable sites. We report the identification and characterization of a new cluster of interactions in the LDH tetrameric interface. Using nanoscale differential scanning fluorimetry, chemical denaturation, and mass photometry, we identified several residues (E62, D65, L71, and F72) essential for LDH tetrameric stability. Moreover, we report a family of peptide ligands based on this cluster of interactions. We next demonstrated these ligands to destabilize tetrameric LDHs through binding to this new tetrameric interface using nanoscale differential scanning fluorimetry, NMR water–ligand observed via gradient spectroscopy, and microscale thermophoresis. Altogether, this work provides new insights on the LDH tetrameric interface as well as valuable pharmacological tools for the development of LDH tetramer disruptors.
). The elevated glycolytic flux in cancer cells has two origins: adaptation to hypoxia (anaerobic glycolysis) and adaptation to high proliferation rates (aerobic glycolysis, also known as the “Warburg effect”) (
). This higher glycolytic flux provides cancer cells with the energy and biomass essential for the sustainment of their anabolic growth. At the end of the glycolytic pathway stands the reduction of pyruvate to lactate catalyzed by the lactate dehydrogenase (LDH) family.
While lactate has long been considered as a mere byproduct of glycolysis, it is now regarded as a potential purpose of accelerated glycolysis in cancer, in the light of the numerous benefits it provides to tumor growth (
). Lactate metabolism further establishes a metabolic symbiosis between oxidative cancer cells that use lactate preferentially to glucose as a fuel and glycolytic cancer cells that rapidly convert glucose to lactate (
LDHs are key enzymes at the core of this adaptive metabolism as they catalyze the terminal reaction of lactate biosynthesis with the interconversion of pyruvate and NADH to lactate and NAD+. LDHs function as obligate tetramers constituted by the homoassociation or heteroassociation of two isoenzymes, LDH-H (lactate dehydrogenase heart isozyme; encoded by the LDHB gene) and LDH-M (lactate dehydrogenase muscle isozyme; encoded by the LDHA gene) (
). The two LDH homotetramers, lactate dehydrogenase heart isozyme homotetramer (LDH-1) (LDH-H4) and lactate dehydrogenase muscle isozyme homotetramer (LDH-5) (LDH-M4), are the most extensively studied forms of LDH and constitute appealing targets for cancer therapy (
). However, more recent reports about the implications of LDH-1 in cancer pathogenesis shed light on LDH-1 inhibition. First, LDH-1 was reported to interact with lysosomal vesicular ATPase, thus regulating autophagy, and is essential for metabolic reprogramming through p53 and Ras mutations (
). In fact, LDHs are usually recognized as poorly druggable targets, and different reasons can account for this. First, LDH active-site inhibitors face a challenge in achieving selectivity over other dehydrogenases, notably because of a common NAD-binding domain. For instance, gossypol derivatives, which are among the first LDH inhibitors reported (
). Second, the LDH catalytic site presents nonoptimal physicochemical properties with high solvent exposure and hydrophilicity, leading to challenging absorption, distribution, metabolization, and excretion properties for most LDH active-site inhibitors (
). Finally, an inherent difficulty in achieving therapeutic LDH inhibition stems from its high intracellular concentration; LDHs are indeed highly concentrated in cancer cells, with protein concentrations reported in the micromolar range (
). This high cellular concentration often hampers the observation of cell-based inhibition below that micromolar threshold, even for the more potent nanomolar inhibitors reaching micromolar concentrations in tumors (
These different challenges to LDH inhibition called for developing new strategies to target this enzymatic family of high therapeutic potential. To this end, tool compounds able to target the LDH oligomeric interface instead of its active site have been recently developed (
). Targeting a protein oligomeric state is still an underexplored strategy that can provide several benefits over active-site targeting and could thus overcome the existing difficulties encountered with LDH orthosteric inhibitors. Targeting LDH self-assembly could indeed lead to the identification of new and potentially more druggable allosteric sites (
). Noteworthy, as LDH subunits can form homotetramers and heterotetramers, the tetrameric interface is shared between the two different isoenzymes. Targeting LDH tetrameric interface can thus yield to molecules disrupting both LDH-1 and LDH-5, which is in line with the current pan-LDH inhibition strategy (
). This model allows to study the LDH tetrameric interface and previously led to the identification of a first allosteric site involving the N-terminal residues. The present study reports on the existence of a second allosteric site, based on which we developed a new family of peptidic ligands functioning as LDH destabilizers.
Results and discussion
In silico mapping of the LDH-1 tetrameric interface identifies a new cluster of interactions
LDH quaternary state is a “dimer of dimers” (Fig. 1A) (
). According to X-ray structures, three different subunit orientations could account for LDH dimeric conformation. We have indeed previously found that LDH N-terminal domain truncation leads to dimers (LDH-Htr) (Fig. 1A) (
), only the association of dimers A to C and B to D in a tetramer can explain the role of this N-terminal domain in the stabilization of the tetrameric state (Fig. 1A). Based on this hypothesis, we first mapped the interactions made by one subunit with an LDH dimer (A–C or B–D) using the Molecular Operating Environment (MOE; Chemical Computing Group [ChemComp]) software (
). Mapping these contact points highlighted two clusters: A (A1 and A2) and B (B1 and B2) (Fig. 1B). Clusters A1 and A2 corresponded to the LDH N-terminal tetramerization domain (Fig. 1C) and to its related tetramerization site, respectively, which were previously investigated (
). Clusters B1 and B2 matched with a 22-amino acid α-helix and its interacting site, which to our knowledge were never reported before. Interestingly, the sequence corresponding to cluster B1 was highly conserved among vertebrates (Fig. S1). Overall, clusters A1 and B1 corresponded to continuous epitopes interacting with discontinuous oligomerization sites A2 and B2.
LDH-Htr behaves as a weak tetramer through cluster B
Next, we aimed to confirm this interaction model and the anticipated symmetry axis of LDH dimers using our model of dimeric LDH (LDH-Htr) (
). According to the interaction map, LDH-Htr lacks cluster A1 but still possesses clusters A2, B1, and B2. We thus reasoned that LDH-Htr might still be able to self-interact at high concentrations via cluster B. Comparison between LDH-Htr and LDH-1 elution profiles by size-exclusion chromatography indeed suggested that LDH-Htr could be in an equilibrium between tetramers and dimers (
). Consistently, the evaluation of LDH-Htr self-interaction by microscale thermophoresis (MST) revealed that the dimeric protein interacts with itself at high concentrations (Fig. 2A; Kd = 1.25 μM [0.96–1.62 μM]). According to our model, this interaction can only be the result of LDH-Htr forming dimers of dimers through cluster B. Monitoring this interaction using MST hence provided valuable information on the overall potency of cluster B. We then evaluated whether LDH-Htr self-association could stabilize the protein complex, as the oligomerization of a protein often results in its stabilization (
). LDH-Htr denaturation profile was evaluated using nanoscale differential scanning fluorimetry (nanoDSF) and revealed that the protein exhibited a concentration-dependent destabilization (Fig. 2B) and conformational change (Fig. 2C). We also evaluated LDH-Htr oligomeric state using mass photometry (MP). MP is a recent technique that allows for single-molecule detection and mass measurement in solution based on light scattering (
). MP analysis of LDH-Htr revealed an equilibrium between dimers and tetramers in solution (Fig. 2D). Altogether, these results demonstrated that the truncation of the LDH N-terminal domain does not entirely prevent the protein from forming tetramers. This ability of the truncated dimers to interact and form weak tetramers validated our in silico model and provided valuable information about this new tetrameric interface.
Identification of peptide ligands of the LDH tetrameric interface
We then set out to further characterize the continuous epitope B1 in order to identify peptides targeting the LDH tetrameric interface. As discussed previously, cluster B1 corresponds to a 22-amino acid peptide folding into a long and “kinked” α-helix ended by a short loop (Fig. 3). We thus decided to study the interaction between “cluster B1”-derived peptide (named LP-22, LEDKLKGEMMDLQHGSLFLQTP) and the LDH-H tetrameric interface. To that end, we performed a set of biophysical evaluation using NMR water–ligand observed via gradient spectroscopy (WaterLOGSY), MST, and nanoDSF experiments.
Strikingly, WaterLOGSY experiments showed that LP-22 undergoes a saturation transfer with dimeric LDH-Htr, but not with the tetrameric LDH-1, thus demonstrating that it interacts at the LDH tetrameric interface (Fig. 4A). MST further validated this interaction with an estimated Kd of 156 μM with LDH-Htr (Fig. 4B). Thermal shift experiments using nanoDSF revealed a stabilization (ΔTm = 2.8 °C at 500 μM) (Fig. 4D) of dimeric truncated LDH-Htr with LP-22, consistent with an interaction occurring at the exposed oligomeric interface and with the stabilization that usually follows a ligand–protein interaction because of an increase in its Gibbs free energy of unfolding (
). On the opposite, LP-22 destabilized tetrameric LDH in a concentration-dependent manner (Fig. 4, E and F) with an EC50 = 47 μM (32–68 μM). These results are coherent with the observation that ligands interacting at oligomeric interfaces often induce protein thermal destabilization (
). Interestingly, a recent report of selective LDH-5 inhibitors disclosed an inhibitor interacting at LDH-5 tetrameric interface near cluster B. This inhibitor was highly selective toward LDH-5, which is coherent with the lower stability of LDH-5 tetrameric complex (
), LDH destabilization was also dependent on protein concentration, as an increasing amount of subunit reverted the effect (Fig. S2). Overall, these results consistently demonstrated that LP-22 interacts at the LDH tetrameric interface and destabilizes the tetrameric enzyme.
Biophysical and computational experiments identify LP-22 essential binding region
We next compared LP-22 WaterLOGSY and 1H-NMR spectra. Because WaterLOGSY is a ligand-based NMR spectroscopy that relies on protein–ligand saturation transfer, LP-22 residues that do not interact with the protein will be absent of the WaterLOGSY spectrum (
). A careful comparison between LP-22 1H and WaterLOGSY spectra highlighted 1H chemical shift regions characteristic to lysine, glutamate, aspartate, and leucine aliphatic regions that were not undergoing saturation transfer (Fig. 5, A and B). Calculation of the contribution of each residue to the peptide-binding free energy further suggested that N-terminal LP-22 residues LEDKLK, a region rich in those particular amino acids, did not account for much of LP-22 binding energy (Fig. 5C). Removal of these six N-terminal residues led to GP-16 (GEMMDLQHGSLFLQTP), a peptide that had a similar WaterLOGSY spectrum (Fig. 5A), thus suggesting that the two peptides were interacting very similarly. MST indicated that the interaction with dimeric LDH was slightly weakened, with a Kd = 240 μM (Fig. S3). Consistently, nanoDSF established that GP-16 could still destabilize tetrameric LDH in a concentration-dependent manner (EC50 = 262 μM [142–383 μM]) (Fig. 5D).
Probing GP-16 and LDH-H tetrameric interface hot spots
Computational and biophysical data suggested that the GP-16 sequence represents an essential binding region of the LDH tetrameric interface. To verify this hypothesis, we probed the contribution of each residue of cluster B1 to the stability of LDH-H oligomeric state. To that end, we performed an alanine scanning of the LDH-1 sequence corresponding to GP-16. We thus designed, produced, and purified the 16 corresponding LDH-H recombinant alanine variants and evaluated these mutants for their thermal and chemical stability (Fig. 6, A and B and Table 1). These variants were also evaluated by MP (Table 1 and Fig. 6C and Fig. S4).
Table 1Oligomeric state and stability of LDH-H variants
Reported values are means ± SEM for melting temperatures and EC50 (n = 6) and means ± SD for the molecular weight (the values presented here were obtained with MP from one measurement and were repeated at least 3 times with similar results). The reported molecular weights correspond to the main oligomeric state of the protein.
Molecular weight (kDa)
Ratio 350/330 nm
155 ± 17
0.953 ± 0.012
74.5 ± 0.1
88 ± 13
57.5 ± 0.1
142 ± 10
0.735 ± 0.018
69.8 ± 0.1
77 ± 18
58.9 ± 0.1
141 ± 14
0.891 ± 0.012
71.6 ± 0.1
149 ± 18
0.845 ± 0.010
68.7 ± 0.1
143 ± 12
0.521 ± 0.012
56.4 ± 0.1
137 ± 23
0.630 ± 0.011
61.1 ± 0.1
134 ± 16
0.893 ± 0.010
73.1 ± 0.1
154 ± 15
0.619 ± 0.016
67.2 ± 0.2
153 ± 18
0.722 ± 0.015
73.5 ± 0.1
144 ± 13
0.580 ± 0.009
66.0 ± 0.1
148 ± 12
67.1 ± 0.1
70 ± 17
53.9 ± 0.1
137 ± 10
0.348 ± 0.008
62.0 ± 0.1
143 ± 10
0.636 ± 0.020
71.3 ± 0.1
145 ± 10
0.439 ± 0.012
64.9 ± 0.1
142 ± 17
0.752 ± 0.011
72.2 ± 0.1
a Reported values are means ± SEM for melting temperatures and EC50 (n = 6) and means ± SD for the molecular weight (the values presented here were obtained with MP from one measurement and were repeated at least 3 times with similar results). The reported molecular weights correspond to the main oligomeric state of the protein.
Among the different single-point alanine mutations, three of them significantly impacted the LDH-1 oligomeric state, with variants E62A and F72A behaving mainly as dimers in solution and variant L71A behaving as a mixture of tetramers and dimers according to MP results (Fig. 6C). Consistently, nanoDSF experiments showed that LDH-HF72A and LDH-HE62A exhibited denaturation patterns comparable to dimeric LDH-Htr, with a lower initial 350/330 nm ratio and a red shift instead of the blue shifts usually observed for tetrameric LDH variants (Fig. 6A). Interestingly, LDH-HL71A exhibited a lower initial ratio and a red shift as well but a Tm 10 °C higher than LDH-Htr. This mixed profile is coherent with the mixture of dimers and tetramers that appear to be present in solution with this variant. Comparison of the tryptophan fluorescence spectra of these variants with LDH-1 showed decays in fluorescence intensity characteristic of the dimeric forms of LDHs for variants E62A and F72A but not L71A (Fig. 7A) (
Mutations of L73 and D65, two other hot spots previously suggested by in silico analysis, resulted in tetrameric variants displaying a significant reduction of stability as assessed by both thermal and chemical denaturation methods (Fig. 5 and Table 1). In line with the expected reduction of tetrameric stability, dilution experiments of the D65A variant resulted in concentration-dependent destabilization of the protein and in the apparition of a second unfolding event (Fig. 7, B and C). Interestingly, variants L66A and L73A showed similar stabilities by nanoDSF, with Tm of 61.1 and 62.0 °C, respectively. However, chemical denaturation experiments demonstrated a striking difference in stability between these two mutants, with EC50 of 0.630 and 0.348 M, respectively (Fig. 7, D and E and Table 1). MP experiments validated the presence of an equilibrium between dimers and tetramers for L73A but not for L66A (Fig. S4). In accordance with our in silico calculation and available crystallographic data (
), these results confirm that mutation L73A reduces LDH-1 oligomeric stability. In contrast, mutation L66A appears to impact the stability of the protein differently, for instance, by disturbing its hydrophobic core. These results further highlight the importance of exploiting orthogonal methods when assessing the impact of a mutation on protein stability. Other mutations resulted in tetrameric proteins displaying a medium to low variation of their chemical and thermal stability compared with wildtype LDH-1, highlighting the lesser importance of these residues for the oligomeric state of the protein (Table 1).
Overall, the different stabilities of the mutants coherently matched with our in silico prediction of the ΔG of interaction and highlighted new molecular determinants of the LDH tetrameric interface (Fig. 5C). Cluster B1 hot spots are constituted by the two negatively charged amino acids, E62 and D65, and by the three consecutive hydrophobic residues, L71, F72, and L73. Based on the crystallographic structure (protein data bank ID: 1I0Z), E62 and D65 are involved in a hydrogen bond network with water and neighboring residues R170, K246, A252, and W251. L71, F72, and L73 perform hydrophobic interactions between each other and with residues L166, A169, P183, A252, and L255 (Fig. 8). Interestingly, cluster B1 is constituted of both polar and apolar hot spots, which contrast with the purely lipophilic hot spots that we had previously identified in the LDH tetramerization arm (
). Targeting protein self-association is an emerging concept in drug design that can bring several advantages over classical orthosteric inhibition. First, targeting the LDH oligomeric interface could unravel new allosteric sites, potentially leading to compounds displaying improved drug-like features compared with LDH active-site inhibitors. Second, molecules interacting at a protein homomeric interface can lead to its destabilization and degradation (
), providing compounds with a substoichiometric effect. In this study, we report the identification and characterization of a new LDH tetrameric interface and its essential residues, using a combination of MP, nanoDSF, and chemical stability experiments. These results are in accordance with previous computational studies that suggested that this region is involved in LDH tetramerization (
). Furthermore, we report the identification of a family of peptidic ligands that target the tetrameric interface of LDH, destabilize the tetrameric LDH, and stabilize the dimeric LDH-Htr. Altogether, this work provides a structural characterization of the molecular determinant of the LDH tetrameric interface as well as valuable pharmacological tools for the future development of compounds targeting the LDH oligomeric state.
Chemicals and peptides
All reagents were purchased from different chemical suppliers and used without further purification. Peptides were purchased from Genecust (https://www.genecust.com). Structure conformity and purity grade (>95%) were assessed by analytical HPLC analysis and MS. Peptide GP-16 was amidated and acetylated, respectively, at its C and N termini, and LP-22 was only amidated at his C terminus.
Production and purification of human LDH proteins
The hLDHB nucleotidic sequences used to produce full-length, truncated, and variant LDH-H proteins inserted in a pET-28a expression vector were ordered from Genecust. NdeI and Bpu1102I restriction sites were used for sequence insertion and allowed for an N-terminal 6-His tag addition. Protein production and purification were performed following a previously described procedure (
). Recombinant plasmids were then transformed in host bacterium Escherichia coli Rosetta (DE3). Transformants were cultured in terrific broth medium supplemented with 50 μg/ml kanamycin and 34 μg/ml chloramphenicol at 37 °C until an absorbance of 0.8 was reached. LDH expression was induced by the addition of 1 mM IPTG at 20 °C for 20 h. Then, cells were collected by centrifugation at 5000 rpm (rotor 11150; Sigma), 4 °C for 25 min. Pellets were suspended in a lysis buffer (Tris–HCl 50 mM pH 8.5, MgCl2 10 mM, NaCl 300 mM, imidazole 5 mM, and glycerol 10%), and then disrupted by sonication, followed by centrifugation at 4 °C, 10,000 rpm (rotor 12165-H; Sigma) for 30 min. The insoluble fraction was discarded, and 1 μl of β-mercaptoethanol was added per milliliter of soluble fraction. Purification of recombinant proteins was performed using 1 ml His-trap Fast-Flow-crude columns (GE Healthcare) according to the manufacturer's instructions. Finally, protein concentrations were measured using the Bradford method with the Protein Assay Kit (Bio-Rad), and sample homogeneity was assessed by SDS-PAGE with Coomassie brilliant blue as a staining agent.
Human LDH-H (full length and truncated)-6-His proteins for 1D NMR were expressed and purified from E. coli, as described previously. All experiments were performed on an Ascend Avance III 600 MHz system equipped with a broadband cryoprobe (Bruker) following a previously described procedure (
For WaterLOGSY NMR studies, samples were prepared in 10% heavy water containing 50 mM sodium phosphate buffer, pH 7.6, and 100 mM NaCl. The concentration of LDHs was ranging from 15 to 20 μM of monomer. Ligand binding was detected using a WaterLOGSY ephogsygpno.2 avance-version sequence with a 1 s mixing time, 4096 scans were collected at 277 K to yield a 16 K points free induction decay. Water signal suppression was achieved using an excitation-sculpting scheme, and a 50 ms spinlock was used to suppress protein background signals.
In silico evaluation
Calculation of the free binding energy and mapping of the interaction at LDH interface was performed using the MOE software with the LDH-1 crystallographic structure (Protein Data Bank entry: 1I0Z) (
). The truncated dimeric version of LDH-1 (LDH-Htr) was generated from the tetrameric complex by removing LDH-H subunits B and D as well as the N-terminal domain of subunits A and C. Minimization was performed before free energy calculation and interaction mapping.
NanoDSF was performed following previously described procedures (
Solutions of proteins (LDH-H, LDH-Htr or variants), stored in a 50 mM sodium phosphate, 100 mM NaCl, and 20% glycerol, pH 7.6, were evaluated on a Tycho NT.6 device (NanoTemper Technologies) using concentrations ranging from 25 to 65 μM. According to standard manufacturer's procedures, samples were poured into capillaries and heated up to 95 °C in 3 min, while following fluorescence emission at 330 and 350 nm. Melting temperatures were extracted from the derivative of the 350/330 nm fluorescence ratios upon increasing temperature.
Solutions of proteins (LDH-1, LDH-5, or LDH-Htr) with peptides were evaluated on a Tycho NT.6 device. Evaluations were performed in a 50 mM sodium phosphate and 100 mM NaCl (pH 7.6) buffer. According to standard manufacturer's procedures, samples were poured into capillaries and heated up to 95 °C in 3 min, while following fluorescence emission at 330 and 350 nm. Melting temperatures were extracted from the derivative of the 350/330 nm fluorescence ratios upon increasing temperature.
MST measurements were performed on a NanoTemper Monolith NT.115 instrument (NanoTemper Technologies) using red dye–N-Hydroxysuccinimide fluorescent labeling. LDH-Htr purified to homogeneity was labeled with the Monolith red dye–N-Hydroxysuccinimide second-generation labeling dye (NanoTemper Technologies), according to the supplied protocol. Measurements were performed in 50 mM sodium phosphate, pH 7.6, and 100 mM NaCl containing 0.01% Tween-20 in standard-treated capillaries (NanoTemper Technologies). The final concentration of proteins in the assay was 100 nM. Ligands were titrated in 1:1 dilutions following manufacturer's recommendations. Experiments were performed in triplicates using 40% light-emitting diode power, high MST power, laser on time 20 s and laser off time 3 s. Peptides were evaluated for their thermophoretic pattern, and Kds were extracted from raw data at a 1.5 s MST on time following the manufacturer's instructions.
Protein landing was recorded using a Refeyn OneMP (Refeyn Ltd) MP system by adding 1 μl of the protein stock solution (1 μM) directly into a 16 μl drop of filtered PBS solution. Movies were acquired for 60 s (6.000 frames) with the AcquireMP (version 2.1.1; Refeyn Ltd) software using standard settings. Data were analyzed using default settings on DiscoverMP (version 2.1.1; Refeyn Ltd) (
Full tryptophan fluorescence spectra were recorded using an excitation wavelength of 286 nm and recording the emission spectra from 320 to 400 nm at room temperature. The raw fluorescence of each experiment was subtracted to a corresponding control experiment without the protein. Experiments were performed in a 50 mM sodium phosphate and 100 mM NaCl, pH 7.6, buffer. For dissociation in subunits, increasing amounts of guanidinium hydrochloride ranging from 0.3 to 2 M were put in contact with the studied proteins (1.3 μM), and fluorescence spectra were recorded afterward.
All quantitative data are expressed as means ± SEM. Error bars were sometimes smaller than symbols. n refers to the total number of replicates. Data were analyzed using the GraphPad 7.0 software (Prism).
All data are contained within the article and the supporting information.
The authors declare the following competing financial interest(s): L. T., L. B., P. S., and R. F. are inventors of European Patent Application EP19172347.7, LDH inhibitor polypeptides for use in treatment of cancer; L. T., M. L., P. S., and R. F. are inventors of European Patent Application EP21154636.1, polypeptide inhibitors of LDH activity for use in cancer therapy. The authors declare no other conflicts of interest.
The authors thank Dr Jamal El Bakali, Severine Ravez, Vincent Stroobant, and Didier Colau for fruitful discussions. Funding sources had no role in study design, data collection, and interpretation, or the decision to submit the work for publication.
P. S. and R. F. are co-senior authors. L. T., P. S., and R. F. designed the research. L. T., M. L., K. K., L. B., and E. Y. performed research experiments. L. T., M. L., N. J., J. M., and J. W. analyzed the data. L. T., K. K., and M. L. set up the experimental methods. L. T., P. S., and R. F. wrote the article. All the authors approved the article content.
Funding and additional information
This work was supported by European Research Council independent researcher starting grant 243188 (TUMETABO), the Belgian Science Policy Office (Belspo; Interuniversity Attraction Pole Grant P7/03 ), the French Community of Belgium ( ARC 14/19-58 ), the Belgian Fonds National de la Recherche Scientifique (FRS—FNRS; grants 3.4575.11 , 2.5025.12 , and 31288018 ), the Belgian Fondation contre le Cancer (Fundamental Research Grant FAF-F/2018/1282 ), a VIB grant (to J. M.), the Research Foundation—Flanders— Fonds de la Recherche Scientifique Excellence of Science project ( 30829584 to J. M.), and the Belgian Télévie (grants 7.4611.11 , 7.4548.13 , 7.4507.14 , 7.4647.20 , and 7.8504.18 ). P. S. is a research director of the FRS—FNRS. L. T. was a PhD fellow of the FRS—FNRS and is currently a postdoctoral researcher of the Belgian Fondation contre le Cancer.