Abstract

   Pyruvate metabolism defects lead to severe neuropathies such as the
   Leigh syndrome (LS) but the molecular mechanisms underlying neuronal
   cell death remain poorly understood. Here, we unravel a connection
   between pyruvate metabolism and the regulation of the epitranscriptome
   that plays an essential role during brain development. Using
   genetically engineered mouse model and primary neuronal cells, we
   identify the transcription factor E4F1 as a key coordinator of
   AcetylCoenzyme A (AcCoA) production by the pyruvate dehydrogenase
   complex (PDC) and its utilization as an essential co-factor by the
   Elongator complex to acetylate tRNAs at the wobble position uridine 34
   (U[34]). E4F1-mediated direct transcriptional regulation of Dlat and
   Elp3, two genes encoding key subunits of the PDC and of the Elongator
   complex, respectively, ensures proper translation fidelity and cell
   survival in the central nervous system (CNS) during mouse embryonic
   development. Furthermore, analysis of PDH-deficient cells highlight a
   crosstalk linking the PDC to ELP3 expression that is perturbed in LS
   patients.

   Subject terms: tRNAs, Neurodegeneration
     __________________________________________________________________

   The mechanisms by which pyruvate metabolism defects lead to
   neurological symptoms in Leigh syndrome patients remain unclear. Here,
   the authors show a link between pyruvate metabolism and translation
   fidelity that is important for brain development.

Introduction

   The Leigh syndrome (LS, OMIM#256000) is a severe inborn
   neurodegenerative encephalopathy affecting one in forty thousand
   newborns that stems from mitochondrial deficiencies in pyruvate
   metabolism or in the electron transport chain (ETC). Mutations in over
   ninety-five genes encoded by either the nuclear or the mitochondrial
   genomes have been linked to this inborn metabolic disorder, including
   genes encoding key components or regulators of the pyruvate
   dehydrogenase (PDH) complex (PDC). PDC is a mitochondrial multisubunit
   complex that converts pyruvate into AcetylCoenzymeA (AcCoA) to fuel the
   tri-carboxylic acid (TCA) cycle. Most LS patients display chronic
   lactate acidemia and bilateral necrotizing lesions of the basal ganglia
   and the brain stem. Affected children also develop a broad range of
   motor and neurological symptoms which lead to premature death during
   early childhood, including muscular atonia, growth and mental
   retardation, ataxia and vision loss^[112]1. No efficient cure has been
   identified so far for PDC-linked mitochondrial disorders although few
   patients have shown a partial improvement of their clinical symptoms
   under a ketogenic diet, a regimen that metabolically bypasses the PDC
   complex, or upon its pharmacological stimulation by dichloro-acetate
   (DCA)^[113]2. Paradoxically, despite most of the genetic alterations
   implicated in the LS have been identified, the molecular mechanisms
   underlying the clinical symptoms of these children, and in particular
   their neurological defects, remain poorly understood.

   Converging evidence suggests that deregulation of the multifunctional
   protein E4F1 is associated to PDH deficiency and to the LS. The E4F1
   protein was originally identified as a transcriptional regulator of the
   viral E4 gene that is hijacked by the oncoprotein E1A during adenoviral
   infection^[114]3. Since, E4F1 was also characterized as an atypical E3
   ligase of the tumor suppressor p53 that controls its transcriptional
   activities independently of degradation^[115]4. Beside its role in
   regulating the balance between cell proliferation and cell
   death^[116]4–[117]6, E4F1 was also identified as a major regulator of
   the PDC complex. Thus, E4F1 directly binds to the promoter of several
   genes encoding essential subunits or regulators of the PDC complex,
   including DLAT and DLD, the E2 and E3 core subunits of the PDC, the
   regulatory subunit of the PDH phosphatase complex PDPR, the pyruvate
   transporter of the inner mitochondrial membrane MPC1, and the SLC25A19
   protein involved in mitochondrial uptake of thiamine pyrophosphate, an
   essential co-factor of the PDC^[118]7–[119]10. A potential connection
   between E4F1 and the LS was identified through the clinical management
   of two siblings of an italian family who developed clinical symptoms
   reminiscent of those observed in LS patients, including muscular
   hypotonia, microcephaly, hyperlactacidemia and organic aciduria
   associated to impaired PDH activity in their skeletal muscles. Brain
   MRI-imaging revealed symmetric necrotic lesions in the subthalamic
   nuclei and in the substantia nigra that are common in LS
   patients^[120]11. Strikingly, extensive genetic analyzes of these
   patients failed to identify any of the pathogenic variants previously
   linked to LS but highlighted a homozygous E4F1^K144Q missense mutation.
   Consistent with a pivotal role of E4F1 in pyruvate metabolism and LS
   pathogenesis, knock-out (KO) mice lacking E4f1 in their skeletal
   muscles exhibit decreased PDH activity and display phenotypes that are
   reminiscent of some of the clinical symptoms of LS patients, including
   chronic lactate acidemia and exercice intolerance^[121]8. However, no
   major neuropathological abnormalities were observed in this animal
   model, suggesting that the neuropathy of LS patients results from
   tissue specific defects in the central nervous system (CNS). Here, we
   generated mice lacking E4F1 in the CNS to investigate its role during
   neuronal development. Detailed molecular characterization of this
   animal model during embryonic development and of E4F1-deficient primary
   fibroblasts demonstrated that in addition to its role in pyruvate
   metabolism, E4F1 is also a key transcriptional regulator of Elp3, a
   gene encoding the catalytic subunit of the Elongator complex. The
   Elongator complex is an acetyl-transferase which targets a subset of
   tRNAs in the acceptor stem loop at the wobble position uridine 34
   (U[34]) to promote a stepwise cascade of chemical modifications
   (methoxy-carbonyl-methylation and thiolation) to produce mature tRNAs
   and thereby ensure protein translation fidelity^[122]12–[123]14. Our
   data indicate that perturbation of E4F1 activities during embryonic
   development results in severe brain defects and microcephaly associated
   to impaired tRNAs U[34] modifications, translation defects and the
   induction of an integrated stress response (ISR) leading to neuronal
   cell death. Moreover, we identified a crosstalk between the PDC and
   ELP3 expression that is perturbed in cells isolated from LS patients.
   Altogether, our data highlight the importance of the links between
   pyruvate metabolism and the epitranscriptome to ensure proteome
   homeostasis in the CNS.

Results

E4F1 plays a critical role in neuronal development

   To evaluate the roles of E4f1 during brain development, we generated
   animals lacking E4F1 in their CNS by crossing E4f1 constitutive and
   conditional KO (cKO) mice with transgenic animals expressing the CRE
   recombinase under the control of the Nestin (Nes) promoter that is
   predominantly active in all neuronal and glial progenitors from day 11
   of embryonic development (E11) onwards, and with a reporter strain
   allowing Cre-mediated expression of the green fluorescent protein (GFP)
   from the Rosa26 locus (Supplementary Figure [124]S1A)^[125]15–[126]18.
   In these compound mutant animals (hereafter referred to as
   E4f1^(Nes)KO) and their control littermates (CTL), immunohistochemistry
   (IHC) analysis of the GFP reporter on head or whole embryo sagittal
   sections isolated at different developmental stages confirmed that
   Cre-mediated recombination was mainly restricted to the CNS and the
   spinal cord (Fig. [127]1A and Supplementary Figure [128]S1B). The
   efficiency and the specificity of Cre-mediated inactivation of E4f1 in
   the CNS was confirmed at E14.5 by RT-qPCR and immunoblotting,
   respectively (Fig. [129]1B, C). E4f1^(Nes)KO mice were represented at
   the expected mendelian ratio at E14.5, E16.5 and E18.5 and showed no
   obvious growth developmental abnormalities, but were absent at P1,
   indicating that E4F1 deficiency in the CNS leads to perinatal death
   (Fig. [130]1D and Supplementary Figure [131]S1C,D). Moreover,
   E4f1^(Nes)KO E18.5 embryos exhibited microcephaly, a developmental
   defect also observed in some LS patients^[132]1 including those
   harboring the homozygous E4F1^A430C mutation which encodes an
   hypomorpic allele of E4F1, as shown by our complementation assays in
   E4f1^cKO Mouse Embryonic Fibroblasts (MEFs) (Supplementary
   Figure [133]S1E–G).

Fig. 1. E4F1 plays a critical role in corticogenesis during mouse embryonic
development.

   [134]Fig. 1
   [135]Open in a new tab

   A Immunohistochemistry (IHC) analysis of GFP expression on brain
   sagittal sections prepared from E14.5, E16.5 and E18.5 E4f1^(Nes)KO and
   CTL littermates. Scale bars, 2.5 mm. Images are representative of
   experiments performed on n = 4 animals/group. B RT-qPCR analysis of
   E4f1 mRNA levels in the indicated tissues in E18.5 E4f1^(Nes)KO and CTL
   embryos (n = 6 for brain samples and n = 3 for other tissues). C
   Immunoblot analysis of E4F1 and ACTIN protein levels in the brain of
   E14.5 E4f1^(Nes)KO and CTL embryos. Histobars represent the
   quantification of n = 4 immunoblots performed on independent samples
   for each genotype. D Numbers of E4f1^(Nes)KO and CTL embryos identified
   at different developmental stages. Expected numbers were calculated
   based on a Mendelian distribution. E, F Analysis of brain structures in
   E14.5 and E16.5 E4f1^(Nes)KO embryos by High Resolution Episcopic
   Microscopy (HREM). E Histobars represent the relative volume of the
   neuroepithelium tissue and of brain ventricles of E4f1^(Nes)KO and CTL
   embryos at E14.5 (n = 3) and E16.5 (n = 3 for CTL and n = 5 for
   E4f1^(Nes)KO). F Representative HREM images showing the 3D volumic
   reconstruction of different areas of the brain (forebrain in green,
   midbrain in blue and hindbrain in red) and of ventricles (lateral
   ventricles in yellow, third ventricle in pink, mesencephalic vesicle in
   orange and fourth ventricle in red) of E16.5 E4f1^(Nes)KO and CTL
   littermates (n ≥ 4). Scale bars, 280 μm. Data are presented as mean +
   standard error of the mean (SEM) for Fig. 1B or mean + standard
   deviation (SD) for Fig. 1C, E from the indicated number of animals.
   Statistical significance was evaluated using unpaired two-sided
   Student’s t-test (ns, not significant).

   Next, we performed high-resolution episcopic microscopy (HREM) and
   generated three-dimension (3D) volumic brain reconstructions of E14.5
   and E16.5 E4f1^(Nes)KO embryos and CTL littermates to evaluate the
   morphology of different regions of the CNS (Fig. [136]1E,F and
   Supplementary Figure [137]S2A,B). These analyzes revealed that the
   forebrain, midbrain and hindbrain neuroepithelium of E16.5 E4f1^(Nes)KO
   embryos presented large areas of degeneration. These defects were
   illustrated by patchy losses of neuroepithelium or as areas of
   cavitation leading to alterations in the structure of all regions of
   the brain. The space between the brain and the cranium was also
   enlarged in E4F1-deficient mutants when compared to control
   littermates. 3D reconstructions showed that the volume of the total and
   subdivided neural tissues (forebrain, midbrain, hindbrain) was smaller
   in E4f1^(Nes)KO E16.5 embryos, whereas the volume of the total and
   subdivided ventricles was bigger in E4F1-deficient animals than in
   control littermates. Minor brain lesions were detected in the
   forebrain, but not in the midbrain and hindbrain, at E14.5, suggesting
   that the brain defects resulting from E4F1 deficiency started at E14.5
   and worsened at later stages of embryonic development (Fig. [138]1E,F
   and supplementary Figure [139]S2A,B). Histological analyzes of
   hematoxylin and eosin (H&E)-stained sagittal sections of E14.5 and
   E16.5 embryos confirmed the structural lesions observed by HREM and
   showed that these brain defects started at E14.5 and increased at later
   developmental stages. The thickness of the cortical plate (CP), which
   is enriched in neurons, was reduced in E14.5 E4f1^(Nes)KO embryos. E4F1
   deficiency in the CNS resulted in decreased cellularity of all cortical
   layers (Supplementary Figure [140]S2C). No major infiltration by CD45
   positive immune cells was observed at E16.5 nor at E18.5 (Supplementary
   Figure [141]S2D). Moreover, IHC and immunoblot analyzes showed enhanced
   cleavage of the apoptotic marker caspase 3 (C-CASP3) at E16.5 and
   E18.5, but very mild increase at E14.5 (Fig. [142]2A-C). Protein levels
   of cleaved caspase 8, a regulator of the extrinsic apoptic pathway and
   of other forms of cell death, was not induced in the brain of E16.5
   E4f1^(Nes)KO embryos (Supplementary Figure [143]S2E). To identify the
   neural populations in which E4F1 plays an important role, we determined
   in coronal brain sections prepared from E14.5 embryos the number of
   apical and intermediate progenitors, as well as that of neurons,
   identified on the basis of SOX2, TBR2 and TBR1 expression,
   respectively. The number of TBR2 positive intermediate progenitors, and
   that of TBR1 neurons decreased in the brain of E4f1^(Nes)KO E14.5
   embryos, whereas the number of SOX2 positive apical progenitors
   remained comparable to that of control animals. We also evaluated the
   expression of C-CASP3 and of the proliferation marker Ki67
   (Fig. [144]2D). These in situ analyzes showed that E4F1 deficiency
   impaired the proportion of cycling progenitors, likely explaining the
   decreased cellularity observed in various regions of the
   neuroepithelium. Moreover, and consistently with our IHC analyzes, no
   C-CASP3 staining was observed at E14.5, suggesting that the
   proliferation defects of E4F1-deficient progenitors at E14.5 preceded
   the neuronal cell death detected at later stages of development. These
   data indicate that E4F1 plays a critical role in neuronal proliferation
   and cell survival during mid-corticogenesis and that perturbation of
   E4F1 functions in the CNS results in microcephaly and perinatal death.

Fig. 2. E4f1 inactivation impairs neuronal development during
mid-corticogenesis.

   [145]Fig. 2
   [146]Open in a new tab

   A Left panels: immunoblot analysis of uncleaved and cleaved CASPASE 3
   (CASP3 and C-CASP3, respectively) and ACTIN protein levels in the brain
   of E14.5, E16.5 and E18.5 E4f1^(Nes)KO and CTL embryos. Histobars
   represent the quantification of immunoblot analyzes of the ratio
   C-CASP3/CASP3 protein levels in the brain of E4f1^(Nes)KO and CTL
   embryos at the indicated developmental stage (n = 4 animals/group). B
   IHC analysis of C-CASP3 protein in brain sagittal sections prepared
   from E14.5, E16.5 and E18.5 E4f1^(Nes)KO and CTL embryos. Insets show
   images at higher magnification of the forebrain neuroepithelium. Images
   are representative of experiments performed on n = 5 animals/group for
   E14.5, n = 6 CTL animals and n = 7 E4f1^(Nes)KO animals for E16.5, and
   n = 4 CTL animals and n = 3 E4f1^(Nes)KO animals for E18.5. Scale bars,
   100 μm and 20 μm for the insets. Black stars point at C-CASP3 positive
   cells. Fb= forebrain; LV= lateral ventricule. C Histobars represent the
   number of C-CASP3 positive cells in different brain regions of
   E4f1^(Nes)KO and CTL embryos at the indicated developmental stage
   (n = 5 animals/group for E14.5, n = 6 CTL animals and n = 7
   E4f1^(Nes)KO animals for E16.5, and n = 4 CTL animals and n = 3
   E4f1^(Nes)KO animals for E18.5). D Upper panel: schematic
   representation of embryonic neurogenesis and of the dorso-lateral
   region of the medial plane used to determine the relative numbers of
   the different neuronal progenitors and their respective proliferation
   and survival rates by immunofluorescence (BioRender Created in
   BioRender. David, A. (2025) [147]https://BioRender.com/s82s985; NIAID
   Visual & Medical Arts. 30/08/2024. Pyramidal Neuron. NIAID NIH BIOART
   Source. bioart.niaid.nih.gov/bioart/424; NIAID Visual & Medical Arts.
   16/09/2024. Fibroblast. NIAID NIH BIOART Source.
   bioart.niaid.nih.gov/bioart/152; NIAID Visual & Medical Arts.
   29/08/2024. Outer Radial Glial Cell. NIAID NIH BIOART Source.
   bioart.niaid.nih.gov/bioart/400). The different protein markers used to
   distinguish apical and intermediate progenitors and neurons are
   indicated. Lower panel: immunofluorescence (IF) analysis of SOX2, TBR2,
   TBR1, C-CASP3 and Ki67 protein levels in brain coronal sections
   prepared from E14.5 E4f1^(Nes)KO and CTL embryos. Scale bars, 50 μm.
   Histobars represent the number of positive cells for the indicated
   protein marker expressed as % of the control (n = 4 for SOX2, TBR1,
   TBR2 analysis, and n = 3 for C-CASP3, Ki67 analysis). Data are
   presented as mean + standard deviation (SD) for Fig. 2A or mean +
   standard error of mean (SEM) for Fig. 2C, D from the indicated number
   of animals. Statistical significance was evaluated using unpaired
   two-sided Student’s t-test (ns, not significant).

Identification of an E4F1 core transcriptional program operating in the CNS

   To gain insights into the molecular mechanisms underlying impaired
   neurogenesis in E4F1-deficient animals, we compared the gene expression
   profile of total brains harvested from E4f1^(Nes)KO and CTL embryos at
   E14.5 using RNA sequencing (RNA-seq). Upon E4f1 inactivation in the
   CNS, 185 genes displayed a fold change superior to 2 with a false
   discovery rate (FDR) inferior to 0.05, among which 116 were
   significantly down-regulated and 69 were up-regulated. Among the top 35
   differentially expressed genes with the lowest false discovery rate
   ((FDR)-adjusted p-values) in E4F1-deficient brains were Wdr7, Kif1bp,
   Neurl4, Fktn, Dlat, Znhit6, Senp8, Tube1, Elp3, Leo1, Tti2, Dph5, Taz,
   Prbm1, Ddi2, Mrfap1, Taz, Dnajc19, Mrpl15, Nsun5, Eef1g and Rad52
   (Fig. [148]3A, B, Supplementary Figure [149]S3A and Supplementary
   Data [150]1). Interestingly, all these genes were previously identified
   as E4F1-direct target genes in mouse fibroblasts and embryonic stem
   (ES) cells as well as in human breast cancer cells by chromatin
   immunoprecipitation (ChIP) combined with next generation sequencing
   (ChIP-seq)^[151]7,[152]8,[153]10. Gene set enrichment analysis (GSEA)
   indicated that this E4F1-signature was significantly down-regulated in
   the brain of E4f1^(Nes)KO embryos, highlighting an E4F1 core
   transcriptional program operating in various cell types (Fig. [154]3C,
   Supplementary Figure [155]S3B and Supplementary Data [156]2). Impaired
   expression of a subset of genes composing this core E4F1 program was
   confirmed by RT-qPCR on mRNAs extracted from total brains or forebrains
   of E14.5 E4f1^(Nes)KO embryos (Fig. [157]3D and Supplementary
   Figure [158]S3C). Of note, GSEA and pathway enrichment analysis (PEA)
   also indicated that in addition to the deregulation of this E4F1 core
   program, an ISR was induced in E4F1-deficient brains (Fig. [159]3E,F
   and Supplementary Figure [160]S3D).

Fig. 3. Gene expression profile of E4F1-deficient neurons.

   [161]Fig. 3
   [162]Open in a new tab

   A,B RNA-seq analysis of brains prepared from E14.5, E4f1^(Nes)KO and
   CTL embryos (n = 3 for each genotype). A Heatmap representing the
   relative mRNA levels (fold change, log[2] ratio) of the differentially
   expressed genes (DEG) with the most statistically significant False
   Discovery Rate (likelihood ratio tests (LRT) - FDR-adjusted p-values).
   B Volcano plot depicting the significance (FDR) and magnitude of
   difference (fold change, log[2] ratio) of DEG identified in E14.5
   E4f1^(Nes)KO embryos. C Gene set enrichment analysis (GSEA) of RNA-seq
   data showing decreased expression of E4F1 target genes. The normalized
   enrichment score (NES) and false discovery rate (FDR) are indicated.
   P-values were corrected using Benjamini–Hochberg’ FDR. D RT-qPCR
   analysis of E4f1 mRNA levels and those of a subset of its direct target
   genes (Ndufs5, Dnajc19, Taz, Wdr7, Neurl4) in total RNAs prepared from
   the forebrain of E14.5 E4f1^(Nes)KO and CTL embryos (n = 5). Data are
   presented as mean + standard error of the mean (SEM) for the indicated
   number of animals. Statistical significance was evaluated using
   unpaired two-sided Student’s t-test. E List of the top 10 pathways
   identified by pathway enrichment analysis (PEA) as differentially
   modulated in E14.5 E4f1^(Nes)KO embryos. The size of the histobars
   represents the number of genes in the indicated category with an FDR ≤
   0,001. F GSEA showing the induction of the Integrated Stress Response
   (ISR) signature in E14.5 E4f1^(Nes)KO embryos. P-values were corrected
   using Benjamini–Hochberg’ FDR.

E4F1 regulates the PDC complex and pyruvate metabolism in the CNS

   Among the top three most deregulated genes identified by RNA-seq in the
   brain of E4f1^(Nes)KO embryos was Dlat, an E4F1-direct target gene
   encoding the E2 subunit of the PDC complex. Interestingly, RT-qPCR
   analysis of mRNAs prepared from E14.5 E4f1^(Nes)KO forebrains indicated
   that the role of E4F1 in the regulation of the PDC in neurons extended
   to other E4F1-direct target genes encoding several key subunits or
   regulators of the PDC, including Dld, Mpc1, Pdpr and Slc25a19
   (Fig. [163]4A). Furthermore, immunoblot analyzes confirmed that E4f1
   inactivation in the CNS impinged on DLAT and MPC1 protein levels,
   whereas those of PDHA1, the E1 subunit of the PDC complex, that were
   evaluated as a control, as well as DLD, remained unaffected
   (Fig. [164]4B). Consistent with this expression profile, PDH activity
   was significantly impaired in the brain of E4f1^(Nes)KO embryos and
   resulted in decreased amounts of AcCoA (Fig. [165]4C, D).
   Paradoxically, the abundance of histone H3 acetylation marks previously
   linked to nuclear PDC did not decrease in E4F1-deficient brains,
   suggesting that the pool of pyruvate-derived AcCoA poorly contributes
   to histone acetylation in the CNS (Supplementary
   Figure [166]S3E)^[167]19,[168]20. Furthermore, circulating lactate
   levels were significantly higher in E14.5 E4f1^(Nes)KO embryos than in
   CTL animals (Fig. [169]4E) and IHC analyzes indicated that the protein
   levels of the lactate transporter MCT4 was upregulated in the brain of
   E4F1-deficient embryos (Fig. [170]4F). These data show that E4f1
   inactivation in the CNS impacts on pyruvate oxidation by the PDC and
   favors the redirection of the glycolytic flux towards lactate
   production, a feature reminiscent of the metabolic reprogramming
   occurring in LS patients.

Fig. 4. E4F1 regulates pyruvate metabolism in the CNS.

   [171]Fig. 4
   [172]Open in a new tab

   A RT-qPCR analysis of E4f1 mRNA levels and those of genes encoding key
   subunits or regulators of the pyruvate dehydrogenase complex (PDC)
   (Dlat, Dld, Mpc1, Pdpr, Slc25a19 and Pdha1) in the forebrain of E14.5
   E4f1^(Nes)KO and CTL embryos (a minimum of n = 4 animals/group was used
   to perform these analysis). B Representative immunoblots showing E4F1,
   DLAT, MPC1, DLD, PDHA1, and ACTIN (loading control) protein levels in
   the brain of E14.5 E4f1^(Nes)KO and CTL embryos. Right panel: Histobars
   represent the quantification of immunoblots performed on n = 4
   independent samples/group. C Pyruvate Dehydrogenase (PDH) activity in
   protein extracts prepared from E18.5 E4f1^(Nes)KO and CTL embryos
   (n = 5 animals/group). D Relative AcetylCoenzyme A (AcCoA) levels in
   brain from E18.5 E4f1^(Nes)KO and CTL embryos measured by HR-MS (n = 4
   animals). E Blood lactate levels in E18.5 E4f1^(Nes)KO (n = 43) and CTL
   (n = 99) embryos. F IF analysis of MCT4 protein levels in brain
   sagittal sections prepared from E14.5 E4f1^(Nes)KO and CTL embryos
   (n = 5 animals/group). Fb= forebrain; Mb= middle brain; LV= lateral
   ventricule. Scale bars, 50 μm. Data are presented as mean + standard
   error of mean (SEM) for Fig. 4A, C or mean + standard deviation (SD)
   for Fig. 4B, D, E from the indicated number of animals. Statistical
   significance was evaluated using unpaired two-sided Student’s t-test
   (ns, not significant).

E4f1 controls the Elongator complex and U[34] codon-biased translation in the
CNS

   Strikingly, our RNA-seq datasets indicated that Elp3 mRNA levels were
   also strongly reduced in E4F1-deficient brains (Fig. [173]3A, B). Elp3
   encodes the catalytic subunit of the Elongator complex, an
   acetyl-transferase that was previously implicated in
   neurogenesis^[174]21. This prompted us to further investigate the links
   between E4F1 and the Elongator complex. Analysis of previous ChIP-seq
   experiments performed in murine embryonic fibroblasts (MEFs) and in ES
   cells suggested that Elp3 is a direct target gene of E4F1
   (Fig. [175]5A)^[176]7,[177]22. Quantitative ChIP (qChIP) experiments
   performed with a previously validated polyclonal anti-E4F1
   antibody^[178]7 confirmed that E4F1 binds to the promoter region of
   Elp3 on its canonical binding site in neuronal N2A cells
   (Fig. [179]5B). We then determined by RT-qPCR the mRNA levels of
   different subunits of the Elongator complex in the brain of
   E4f1^(Nes)KO embryos. Consistent with the role of E4F1 in the direct
   regulation of Elp3 transcription, Elp3 mRNA levels were reduced in
   E14.5 E4F1-deficient brains, whereas those of Elp1, Elp2 and Elp4,
   remained unchanged (Fig. [180]5C). At later developmental stages (E16.5
   and E18.5), the mRNA levels of all these core subunits of the Elongator
   complex were affected by E4F1 deficiency (Supplementary
   Figure [181]S4A). Immunoblots showed that E4f1 inactivation also
   impacted on ELP3 protein levels (Fig. [182]5D). Furthermore, and in
   line with previous reports showing that the stability of the catalytic
   subunit ELP3 and that of the scaffold ELP1 are
   interdependent^[183]21,[184]23, ELP1 protein levels also decreased in
   the brain of E14.5 E4f1^(Nes)KO embryos (Fig. [185]5D).

Fig. 5. E4f1 controls the expression and activity of the Elongator complex.

   [186]Fig. 5
   [187]Open in a new tab

   A E4F1 ChIP-seq read densities in transformed MEFs (tMEFs) and mouse ES
   cells (MESC) at the Elp3 locus. The sequence and the position of the
   E4F1 consensus motif in the Elp3 locus are indicated. Arrows indicate
   gene orientation. B Quantitative ChIP (qChIP) experiments on the Elp3
   promoter region using an anti-E4F1 polyclonal antibody (E4F1) or an
   irrelevant antibody (irr) and chromatin prepared from neuronal N2A
   cells (n = 3 independent experiments). Results are represented as the
   relative ratio between the mean value of immunoprecipitated chromatin
   (calculated as a percentage of the input) with the anti-E4F1 antibody
   and the one obtained with the control antibody. C RT-qPCR analysis of
   Elp1, Elp2, Elp3, and Elp4 mRNA levels in total RNAs prepared from the
   brain of E14.5 E4f1^(Nes)KO and CTL embryos (n = 5 animals/group). D
   Immunoblot analysis showing ELP1, ELP3 and ACTIN (loading control)
   protein levels in the brain of E14.5 E4f1^(Nes)KO and CTL embryos.
   Right panel: histobars represent the quantification of immunoblots
   performed on n = 4 and n = 5 independent samples for ELP1 and ELP3
   respectively. E Liquid-chromatography coupled to tandem mass
   spectrometry (LC-MS/MS) analysis of the 5-carbamoylmethyl uridine
   (ncm^5U[34]), 5-methoxy-carbonyl-methyl-2-thio uridine (mcm^5s^2U[34]),
   methyl-3-cytosine (m^3C) and methylguanosine (Gm) tRNAs modifications,
   normalized to the total amount of Uridine (U), Cytosine (C) or
   Guanosine (G) (n = 5 independent samples/group). Data are presented as
   mean + standard error of the mean (SEM) from the indicated number of
   samples. Statistical significance was evaluated using unpaired
   two-sided Student’s t-test (ns, not significant).

   Next, we explored the impact of E4F1-deficiency on the activity of the
   Elongator complex. In mammalian cells, the Elongator complex promotes
   the acetylation of a subset of tRNAs (tRNAs^Gln[UUG], tRNAs^Glu[UUC]
   and tRNA^Lys[UUU]) on the uridine located at position 34 of the
   anti-codon stem loop (U[34]). This acetylation reaction initiates a
   stepwise cascade that first leads to the attachment of a carboxy-methyl
   group (cm^5U[34]), followed by a methylation step involving the
   Alkylation repair homolog 8 complex (ALKBH8/TRM112) to generate
   5-methoxycarbonylmethyl uridine (mcm^5U[34]), and then the addition of
   a thiol group by the Cytosolic thiouridylase proteins 1/2 (CTU1/2)
   enzymes (mcm^5s^2U[34]). From ELP3-acetylated tRNAs, another branch of
   the pathway generates 5-carbamoylmethyl uridine (ncm^5U[34]) through a
   yet unknown enzyme (Supplementary
   Figure [188]S4B)^[189]13,[190]14,[191]24. Liquid-chromatography coupled
   to tandem mass spectrometry (LC-MS/MS) analysis of tRNAs purified from
   E14.5 E4f1^(Nes)KO and CTL brains showed that the relative abundance of
   ncm^5U[34] and mcm^5s^2U[34] decreased in E4F1-deficient neurons,
   whereas that of other tRNAs modifications not linked to the Elongator
   complex, including 3-methylcytosine (m^3C) and methylguanosine (Gm),
   remained unchanged (Fig. [192]5E). RT-qPCR analyzes showed that the
   total amount of tRNAs targeted by the Elongator complex,
   tRNAs^Gln[UUG], tRNAs^Glu[UUC] and tRNA^Lys[UUU], were comparable in
   the brain of E14.5 E4f1^(Nes)KO and CTL embryos, demonstrating that the
   decreased abundance of mcm^5s^2U[34] and ncm^5U[34] in E4F1-deficient
   neurons was not the consequence of an impaired production of these
   tRNAs (Supplementary Figure [193]S4C). Since the Elongator complex has
   also been involved in tubulin and histone acetylation, we evaluated the
   ratio of acetylated/total tubulin (α-TUB-Ac/TUB) and that of acetylated
   histone H3 on lysine 14/total Histone H3 (H3K14Ac/H3)^[194]25,[195]26.
   We failed to detect any significant decrease in α-TUB-Ac nor in H3K14Ac
   in the brain of E14.5 E4f1^(Nes)KO embryos, indicating that E4F1
   deficiency in the CNS, and the associated impaired activity of the
   Elongator complex, preferentially impacted on tRNAs acetylation
   (Supplementary Figure [196]S4D).

   Because changes in Elongator-dependent tRNAs U[34] modifications have
   been linked to codon-biased translation defects, we next performed
   polysome profiling experiments^[197]27. We isolated total cytosolic
   mRNAs, as well as mRNAs from pooled light polysomal (1 to 3 ribosomes)
   and heavy polysomal (more than 3 ribosomes) fractions prepared from the
   brain of E14.5 E4f1^(Nes)KO embryos and control littermates and then
   determined by next generation sequencing the abundance of these
   transcripts (Fig. [198]6A and Supplementary Data [199]3). We first
   focused on a group of genes exhibiting similar total transcript levels
   but differential mRNA abundance in either the light or the heavy
   fractions purified from E4F1-deficient brains. Bioinformatic analysis
   of codons content showed that these mRNAs are significantly enriched in
   two U[34]-codons (AAA, GAA) in comparison to transcripts of the whole
   transcriptome (Fig. [200]6B). Moreover, 12% (282 out of 2288) of the
   corresponding proteins, including KCNA3, contain at least one
   penta-hydrophilic motif previously linked to protein aggregation and
   degradation upon codon-dependent translation defects linked to the
   Elongator complex (Fig. [201]6C)^[202]28. We confirmed by
   immunoblotting and RTqPCR that E4F1-deficiency in the brain of E14.5
   embryos impacted on KCNA3 protein, but not mRNA, levels (Fig. [203]6D
   and Supplementary Figure [204]S4E). Altogether, these data support the
   notion that E4f1 inactivation in the CNS leads to perturbed activity of
   the Elongator complex and translation defects that are linked, at least
   in part, to impaired tRNAs U[34] modifications.

Fig. 6. E4F1 regulates U[34] codon-biased translation in the CNS.

   [205]Fig. 6
   [206]Open in a new tab

   A Polysome profiling analysis of brains of E14.5 E4f1^(Nes)KO and CTL
   embryos (n = 5 animals/group). Dot plots represent transcript levels
   (Log[2] fold change) between E4f1^(Nes)KO and CTL embryos in the pooled
   light (less then 3 ribosomes, left panel) or heavy (more than 3
   ribosomes, right panel) polysomal fractions, according to their
   abundance in the total cytosolic fraction. The lists indicate the 16
   genes which transcript abundance is the most deregulated (log[2]FC) in
   either the light or the heavy polysomal fractions prepared from
   E4F1-deficient brains but showing similar total transcript levels. B
   Codon enrichment in transcripts regulated at the post-transcriptional
   level in the brain of E14.5 E4f1^(Nes)KO and CTL embryos. U[34]-linked
   codons are indicated in green. Codon enrichment was assessed by χ^2
   tests. C Bioinformatic analysis showing that 12% of the proteins (282
   out of 2288) corresponding to genes regulated at the
   post-transcriptional level in the brain of E14.5 E4f1^(Nes)KO embryos
   contain at least one penta-hydrophilic amino-acid sequence previously
   linked to protein aggregation and degradation upon U[34]
   codon-dependent translation defects. Presence of the penta-hydrophilic
   sequence was determined by AME (Analysis of Motif Enrichment). D
   Immunoblot analysis showing KCNA3 and HSP90 (loading control) protein
   levels in the brain of E14.5 E4f1^(Nes)KO and CTL embryos. Immunoblot
   is representative of 3 independent experiments.

E4F1 deficiency triggers the ISR that leads to neuronal cell death

   Previous findings showed that genetic inactivation of Elp3 during
   corticogenesis leads to the induction of the ISR mediated in part by
   ATF4, a member of the Activating Transcription Factor (ATF) family of
   proteins^[207]21. Consistent with these findings, IHC analysis of brain
   sections prepared from E14.5 E4f1^(Nes)KO embryos confirmed the
   induction of ATF4 protein levels (Fig. [208]7A). Furthermore, the
   induction of this ATF4-driven transcriptional response correlated with
   the activation of its upstream regulator eIF2a, as evidenced by its
   increased phosphorylation. Co-staining experiments with antibodies
   recognizing glial (GFAP) or neuronal (NeuN) markers indicated that both
   cell types exhibited increased phospho-eIF2a levels (Fig. [209]7B). In
   agreement with these data, RNA-seq and RT-qPCR analyzes indicated that
   E4F1 deficiency in the CNS led to increased mRNA levels of ATF4-target
   genes implicated in the ISR, including Atf3, Atf5, Eif4ebp1, Asns,
   Sesn2, Slc6a9, Chop, Chac1, Trib3, Mthfd2, Slc3a2 and Slc7a5
   (Figs. [210]3, [211]7C and Supplementary Data [212]1). The ISR involves
   the derepression of integral ER membrane proteins which activate
   independent signaling pathways^[213]29. Similar to Elp3 cKO mice, the
   mRNA levels of Xbp1 and its spliced transcripts Xbp1-u and Xbp1-s, as
   well as those of Atf6, remained unchanged in the brain of E14.5
   E4f1^(Nes)KO embryos, supporting the notion that E4f1 deficiency
   triggered specifically an ATF4-driven response (Fig. [214]7C and
   Supplementary Figure [215]S4F). Furthermore, ten days upon genetic
   inactivation of E4f1 in vitro, ISR-related genes, including Asns,
   Eif4ebp, Sens2, Trib3 and Slc7a3, were also induced at the mRNA level
   in primary cortical cultures composed of neurons and glial cells but
   not in MEFs, indicating that the induction of the ISR triggered by E4F1
   deficiency occurred in a cell autonomous and in a cell type specific
   manner (Fig. [216]7D and Supplementary Figure [217]S4G). To evaluate
   whether the neuronal cell death observed at E16.5 and E18.5 was linked
   to the induction of the ISR, we daily administered ISRIB, a selective
   ISR inhibitor, to pregnant females from E10.5 onwards. The induction of
   the ISR-related genes Atf5, Eif4ebp, Chac1 and Trib3 in the brain of
   E14.5 E4f1^(Nes)KO embryos was impeded by ISRIB (Fig. [218]7E).
   Furthermore, immunoblot analysis of protein extracts prepared from
   ISRIB or mock-treated E16.5 embryos indicated that cleaved-caspase 3
   levels in the brain of E4F1-deficient embryos were reduced upon
   inhibition of eIF2a (Fig. [219]7F). However, administration of ISRIB
   was insufficient to rescue the neonatal lethality of E4f1^(Nes)KO mice
   (Fig. [220]7G). These data support a model in which perturbation of
   E4F1-mediated control of the Elongator complex impacts on proteome
   homeostasis leading to the induction of an ISR and ultimately to
   neuronal cell death.

Fig. 7. E4F1 deficiency triggers the ISR that leads to neuronal cell death.

   [221]Fig. 7
   [222]Open in a new tab

   A IHC analysis of ATF4 protein levels in brain sagittal sections
   prepared from E14.5 E4f1^(Nes)KO and CTL embryos. Data are
   representative of n = 4 independent experiments. Scale bars, 100 μm. B
   Immunofluorescence (IF) analysis of phospho-eIF2a (p-eIF2a), NeuN and
   GFAP protein levels in brain sagittal sections prepared from E14.5
   E4f1^(Nes)KO and CTL embryos. Data are representative of n = 4
   independent experiments. Scale bars, 50 μm. C Relative mRNA levels of
   genes related to the ISR response in the forebrain of E14.5
   E4f1^(Nes)KO and CTL embryos determined by RT-qPCR (n = 5 independent
   samples/group). D Relative mRNA levels of genes related to the ISR
   response in populations of E4f1^cKO and CTL primary neurons and glial
   cells determined by RT-qPCR, 10 days after E4f1 inactivation in vitro
   (n = 6 independent populations of cells/group). E Relative mRNA levels
   of genes related to the ISR response in the brain of mock or ISRIB
   -treated E14.5 E4f1^(Nes)KO and CTL embryos determined by RT-qPCR
   (n = 5 independent samples/group). F Upper panel: immunoblot analysis
   showing Cleaved-Caspase-3 (C-CASP3) and TUBULIN (loading control)
   protein levels in the brain of mock or ISRIB -treated E16.5
   E4f1^(Nes)KO and CTL embryos. Lower panel: histobars represent the
   quantification of immunoblots performed on n = 5 independent
   samples/group. G Numbers of ISRIB-treated E4f1^(Nes)KO and CTL embryos
   identified at birth (P1). Expected numbers were calculated based on a
   normal mendelian distribution. Data are presented as mean + standard
   error of the mean (SEM) for Fig. 7C or mean + standard deviation (SD)
   for Fig. 7D–F from the indicated number of samples. Statistical
   significance was evaluated using unpaired two-sided Student’s t-test
   (ns, not significant).

E4F1 coordinates pyruvate-derived AcCoA production by the PDC and tRNAs U[34]
acetylation

   The key role of E4F1 in pyruvate metabolism and in the regulation of
   the Elongator complex prompted us to evaluate whether E4F1 coordinates,
   at the transcriptional level, the production of pyruvate-derived AcCoA
   by the PDC and its utilization by the Elongator complex to promote
   tRNAs acetylation and ensure translation fidelity. To challenge this
   hypothesis, we first evaluated Dlat and Elp3 expression upon acute E4f1
   inactivation in MEFs. E4F1-deficient MEFs exhibited impaired Dlat and
   Elp3 mRNA and protein levels (Fig. [223]8A, B). Moreover, and
   consistent with our findings in E4f1^(Nes)KO embryos, these cells also
   exhibited decreased amounts of mcm^5s^2U[34] and ncm^5U[34]
   (Fig. [224]8C). Next, we performed stable isotope tracing experiments
   to evaluate whether pyruvate-derived AcCoA produced by the PDC serves
   as a co-factor and a carbon source for the Elongator complex to
   acetylate tRNAs at U[34]. MEFs were incubated in presence of ^13C
   uniformally labeled pyruvate ([U-^13C]pyruvate) for 6 h and
   ^13C-enrichment was measured by LC-MS/MS in the ionized fragments of
   ncm^5U[34] and mcm^5s^2U[34] that were predicted to incorporate the two
   ^13C-labeled carbons of the acetyl-group provided by pyruvate-derived
   ^13C-AcCoA into the side chain of the uridine ring upon acetylation by
   the Elongator complex (Fig. [225]8D). The m + 2 isotopologues of
   ncm^5U[34] and mcm^5s^2U[34] were significantly detected in MEFs
   incubated with^13C-pyruvate for 6 h, a process affected upon
   pre-incubation with UK-5099, a pharmacological inhibitor of the
   pyruvate mitochondrial transporter MPC1 that indirectly blocks PDH
   activity (Fig. [226]8E and Supplementary Figure [227]S5A). In the same
   experimental set-up, ^13C-acetate, another alternative source of AcCoA,
   contributed to a much lesser extent to U[34] tRNAs modifications,
   suggesting that pyruvate and acetate are channeled differently in MEFs
   to fuel specific acetylation reactions (Supplementary Figure [228]S5B).
   These data demonstrate that pyruvate-derived AcCoA produced by the PDC
   is an important co-factor for Elongator-mediated acetylation reactions
   targeting U[34] tRNAs.

Fig. 8. E4F1 coordinates AcCoA production by the PDC and its utilization by
the Elongator complex to ensure U[34] tRNAs acetylation.

   [229]Fig. 8
   [230]Open in a new tab

   A RT-qPCR analysis of E4f1, Dlat and Elp3 mRNA levels in E4f1^cKO and
   CTL MEFs (n = 4 independent populations of MEFs/group), 7 days after
   activation of the Cre recombinase. B Left panel: immunoblot analysis of
   E4F1, DLAT, ELP3 and TUBULIN protein levels in the same cells than in
   A. Right panel: histobars represent the quantification of immunoblots
   performed on independent samples (n = 4 populations/group). C LC-MS/MS
   analysis of ncm^5U[34], mcm^5s^2U[34] and Um in tRNAs purified from the
   same cells than in A and normalized to the total amount of Uridine (U)
   (n = 4 independent populations of MEFs/group). D Schematic
   representation of pyruvate contribution to U[34] tRNAs modifications
   when cells are incubated with uniformly labeled [U-^13C]-pyruvate
   (NIAID Visual & Medical Arts. 29/08/2024. Mitochondria. NIAID NIH
   BIOART Source. bioart.niaid.nih.gov/bioart/352). E Stable isotope
   tracing experiments in MEFs incubated with DMSO or the MPC1 inhibitor
   UK-5099 and cultured in the presence of [U-^13C]-pyruvate.
   ^13C-enrichment in ncm^5U[34] and mcm^5s^2U[34] was determined by
   LC-MS/MS in two independent experiments performed in triplicates. Data
   are represented as the relative ratio between the ^13C-labeled m + 2
   isotopologue and the corresponding unlabeled U[34] modification. F
   Immunoblot analysis of PDHA1, ELP3 and TUBULIN protein levels in MEFs
   transduced with control or Pdha1 shRNAs lentiviruses (shPdha1 #1 and
   #2). Right panel: histobars represent the quantification of immunoblots
   performed on n = 4 independent samples. G Immunoblot analysis of ELP3
   and ACTIN protein levels in human skin fibroblasts isolated from Leigh
   Sydrome (LS) patients harboring mutations in the indicated genes or
   from age-matched healthy donors. Immunoblot was performed in 2
   independent experiments. H Immunoblot analysis of PDHA1, ELP3 and ACTIN
   protein levels in human skin fibroblasts isolated from LS patients
   harboring mutations in the PDHA1 gene transducted with an empty control
   or a WT-PDHA1 encoding retrovirus. Right panel: histobars represent the
   quantification of immunoblots performed on cells isolated from n = 3 LS
   patients. Data are presented as mean + standard error of the mean (SEM)
   for Fig. 8A, C or mean + standard deviation (SD) for Fig. 8B, E, F and
   H from the indicated number of samples. Statistical significance was
   evaluated using unpaired two-sided Student’s t-test (ns, not
   significant).

PDH deficiency impacts on ELP3 expression

   To further understand the relationship between pyruvate metabolism and
   the Elongator complex, we depleted PDHA1 in MEFs using two independent
   short hairpin RNAs (shRNAs). Strikingly, Pdha1 knock-down (KD) cells
   displayed decreased ELP3 protein, but not mRNA, levels (Fig. [231]8F
   and Supplementary Figure [232]S6A). Comparable decrease of ELP3 protein
   levels were obtained upon shRNA-mediated depletion of Dlat or
   pharmacological inhibition of PDC activity by β-Fluoropyruvate,
   unraveling a crosstalk between the PDC and ELP3 expression
   (Supplementary Figure [233]S6B-C). Furthermore, skin fibroblasts
   isolated from LS patients harboring mutations in the PDHA1 or the E4F1
   genes, but not in those with mutations in SURF1 or NDUFS6 displayed
   reduced ELP3 protein levels (Fig. [234]8G). Since the LS is a mendelian
   disease, we next evaluated the effect of ectopic wild-type (WT) PDHA1
   on endogenous ELP3 protein levels in these patient-derived
   PDHA1-mutated cells. Ectopic expression of WT-PDHA1 increased ELP3
   protein levels in these PDHA1-mutated LS fibroblasts, confirming the
   notion that ELP3 and PDC expression are tightly coordinated
   (Fig. [235]8H).

Discussion

   The LS is a severe pediatric neurodegenerative disease linked to either
   PDC or ETC deficiencies. The prognosis for LS patients is very poor and
   current treatments provide limited improvement of their clinical
   symptoms. The paucity of animal models recapitulating the genetic
   alterations found in LS patients has limited studies aiming at
   understanding the molecular defects underlying their clinical symptoms
   and the development of efficient treatments. Moreover, because the
   relative contribution of cell autonomous versus systemic metabolic
   defects to the various clinical manifestations associated with LS
   remain elusive, the development of genetically engineered mouse models
   allowing tissue specific inactivation of genes related to the LS is of
   major interest. A brain specific Pdha1-cKO mouse model was previously
   generated in which PDH deficiency led to the expected metabolic
   reprogramming including impaired glucose-derived flux into the TCA
   cycle and increased lactate production. These mice also exhibited
   defective synthesis of neuro-transmitters linked to the TCA cycle
   including glutamate that was associated to neuronal excitability
   defects, cerebral hypotrophy and epilepsy which led to post-natal death
   around P25. In the same animal model, acetate and propionate were found
   to be efficient anaplerotic sources for the TCA which compensate, at
   least in part, for the lack of glucose oxidation by the
   PDC^[236]30,[237]31. Interestingly, combining a ketogenic diet with the
   systemic administration of propionate mitigated the growth defects and
   the neuronal phenotypes of these mice but only delayed the post-natal
   death of these animals by a few weeks. Hence, a better understanding of
   the molecular defects associated to PDH deficiency is still needed to
   develop more efficient therapies. The detailed molecular
   characterization of mice in which we inactivated E4f1, a gene encoding
   a key transcriptional regulator of the PDC, in both neuronal and glial
   progenitors during early embryonic development uncovered an unexpected
   link between the PDC and the control of translation fidelity. Despite
   these mice display a reduction but not a complete inhibition of PDH
   activity, all E4F1-deficient animals died within hours after birth,
   likely because of reduced maternal-pup interactions and inability to
   feed. A comparable neonatal death was described in other mouse models
   developing embryonic neuronal defects, such as those resulting from the
   genetic inactivation of the splicing factor Tra2b or Tsc1, a component
   of the tuberous sclerosis complex^[238]32,[239]33. Noteworthy, and
   consistent with our previous findings that E4F1 is a key regulator of
   the cell cycle and of the intra-S checkpoint^[240]5,[241]7,[242]10,
   E4F1-deficient animals died earlier than Pdha1 brain-specific cKOs,
   strongly suggesting that E4F1 functions in the CNS extend beyond the
   regulation of the PDC. Using E4f1-brain specific cKO mice, we unraveled
   a network involving E4F1, the PDC and ELP3 that links pyruvate
   metabolism to translation fidelity. We don’t exclude the possibility
   that the deregulation of other E4F1-target genes, including Nsun5 (a
   methyl-transferase targeting ribosomal RNAs) or Eef1g (a translation
   elongation factor), also contributes to the translation defects
   observed in E4F1-deficient animals. Further investigations will be
   required to evaluate whether these different components of the E4F1
   core transcriptional program synergistically contribute to the
   translation of the same subset or of different pools of mRNAs.
   Nevertheless, our data indicate that these mistranslated transcripts
   display a codon content bias towards U[34] codons and that most of the
   corresponding proteins contain the hydrophilic motif previously
   associated with protein aggregation occurring upon perturbations of
   U[34]-enzymes. Altogether, these results support the notion that
   impaired activity of the Elongator complex plays a pivotal role in the
   translation defects observed in the CNS of E4F1-deficient embryos.
   Finally, our data also highlighted a crosstalk between the PDC and
   ELP3, independent of E4F1, that is perturbed in cells isolated from
   PDH-deficient patients, suggesting that impairement of protein
   translation fidelity is a common feature of cells exhibiting impaired
   PDH.

   How metabolic perturbations lead to changes in gene expression is a
   field of active research. Many studies highlighted the role of
   different metabolites, including AcCoA and S-Adenosyl-Methionine (SAM),
   in the regulation of the epigenome. By serving as essential co-factors
   of acetyltransferases and methyltransferases, these metabolites
   directly contribute to DNA or histone modifications including
   acetylation or methylation, respectively^[243]34–[244]36. Although it
   is known for decades that some of these acetyl- or methyl- transferases
   also target RNAs, including mRNAs, tRNAs, or ribosomal RNAs, the
   importance of their metabolic co-factors in regulating the
   epitranscriptome remains poorly understood^[245]37. Previous work
   showed that serine metabolism in mitochondria provides methyl donors to
   produce the taurinomethyluridine base at the wobble position of some
   mitochondrial tRNAs and control mitochondrial translation of
   OXPHOS-related transcripts encoded by the mitochondrial genome^[246]38.
   In our study, we validated an innovative stable isotope tracing
   approach to track the metabolic fate of pyruvate-derived carbons into
   U[34] tRNAs modifications and demonstrated that AcCoA produced by the
   PDC directly contributes to tRNAs modifications linked to the Elongator
   complex. Strikingly, in the same experimental set-up, ^13C-labeled
   acetate contributed much less to tRNAs U[34] acetylation, suggesting
   that these metabolic substrates are channeled differently to fuel
   different pools of AcCoA that are utilized for specific acetylation
   reactions. Our data support a model where E4F1 regulates a
   transcriptional program coordinating AcCoA production by the PDC and
   its utilization by the Elongator complex to promote tRNAs U[34]
   modifications. Perturbation of this program impaired protein
   translation fidelity and associated with the induction of an ISR that
   ultimately led to neuronal cell death. Interestingly, despite
   E4F1-mediated control of Elp3 occurred in different cell types
   including MEFs, the induction of the ISR was only observed in the CNS
   in vivo and in neuronal cells in vitro. The mechanisms driving this
   cell-type specific and cell autonomous effect remain to be identified.

   It is noteworthy that some antibiotics, including tetracyclines, have
   been shown to improve the fitness of nutrient-deprived cells isolated
   from patients suffering mitochondrial diseases with ETC deficiencies.
   Strikingly, tetracycline analogs, such as doxycycline, selectively
   attenuated mitochondrial translation in these cells, and when
   administrated to Ndufs4-mutant animals, improved their neuronal defects
   and prolonged significantly their lifespan. Although these antibiotics
   induced a clear ATF4-driven cellular response, their protective effects
   were rather attributed to the induction of a mitohormetic response
   which restored a proper redox balance and thereby improved cell
   survival^[247]39. These results suggest that modulating mitochondrial
   translation, a process amenable to pharmacological intervention, plays
   a pivotal role in protecting cells exhibiting ETC deficiency. In E4F1
   brain-specific cKO embryos, in which we demonstrated impaired PDH
   activity, pharmacological inhibition of eIF2a mitigated the induction
   of the ISR and diminished neuronal cell death, suggesting that either
   the intensity or the duration of the ATF4-response may lead to
   different outputs in these neuronal cells. Interestingly, previous
   studies also suggested that inhibiting mTORC1 activity, activating a
   hypoxic response or modulating NAD metabolism, can partly rescue
   phenotypes associated to respiratory chain
   deficiencies^[248]40–[249]42. The efficacy of these approaches in
   PDH-deficient-related mitochondrial diseases has not been investigated
   yet but our data pave the way for further studies evaluating whether
   the interplay between the PDC and the Elongator complex also represents
   a new avenue for therapeutic intervention.

Methods

Study Approval

   All mouse procedures were approved by the ethic committee for animal
   warefare of the region Languedoc Roussillon (Comité d’Ethique en
   Expérimentation Animal Languedoc-Roussillon), an accredited institution
   of the French Minister for Education, Research and Innovation
   (agreement number #18030-201812111250227). For human fibroblasts,
   informed consent was obtained from all investigated subjects according
   to the Declaration of Helsinki principles and was approved by Research
   Ethics committees (agreement numbers DC 2009-939 from the “Comité
   d’Ethique de L’hopital Bicêtre Assistance Publique Hopitaux de Paris”
   and [250]GGP15041 from the “Comitato Etico Regione Lombardia”).

Animal housing

   The following strains were used in this study: E4f1^tm1.1Llca,
   E4f1^tm1Pisc, Tg(Nes-cre)^1Kln,
   Gt(ROSA)26Sor^tm1(cre/ERT2)Tyj^[251]15–[252]18. Mice were interbred and
   maintained on a mix 129 Sv/J; C57Bl/6 J background and were housed in a
   pathogen free barrier facility (room temperature 22 °C; relative
   humidity 55%, and a 12-h-light–dark cycle). Animal housing and
   euthanasia were performed in accordance with the 3 R rules. Mice were
   maintained under chow diet (A03, Safe) containing 22 kcal% protein,
   65 kcal% carbohydrate and 13 kcal% fat. Mice were genotyped by PCR on
   tail genomic DNA using Red-N extract kit (Sigma, XNAT), with primers
   indicated in Supplementary Table [253]1. Pregnant females were treated
   with ISRIB (2.5 mg/kg/day, Sigma, SML0843) by intraperitoneal
   injections from E10.5 onwards. All mouse strains are available upon
   request.

Patient material

   Skin fibroblasts were isolated from LS patients or aged-match control
   individuals (between 1 month and 5 years of age) recruited at Paris
   Kremlin Bicêtre hospital or at the Foundation IRCSS Institute of
   Neurology Carlo Besta (Milan, Italy). The sex of the patients was not
   considered as a variable in this study although all experiments were
   conducted on fibroblasts isolated from both male and female Leigh
   syndrome patients and control individuals.

Cell culture

   MEFs were isolated from E13.5 E4f1^-/flox; Rosa26-CRE^ER embryos and
   cultured for no more than 3 passages in DMEM complemented with 10% of
   heat-inactivated fetal bovine serum (FBS, Fisher Scientific). In vitro
   E4f1 inactivation in these primary cells was obtained upon incubation
   with 4-hydroxy-tamoxifen (1 μM, Sigma, H7904) overnight (E4f1^cKO). For
   stable isotope tracing experiments, MEFs were cultured for the last
   24 h in serum-free media (custom DMEM/F12 containing 5.5 mM Glucose;
   1 mM Glutamine; 0.2 mM Branched Chain Amino Acids and a mix of growth
   factors (ATCC, #PCS201040)), in presence of the MPC-1 inhibitor UK5099
   (50μM, Sigma, PZ0160) for 12 h and with 1 mM of universally labeled
   ^13C-pyruvate (Sigma, 490717) or ^13C-Acetate (Sigma, 282014) for the
   last 6 hours. Skin fibroblasts isolated from LS patients or control
   individuals were cultured for no more than 8 passages in DMEM
   complemented with 10% of heat-inactivated fetal bovine serum (FBS,
   Fisher Scientific). The N2A murine neuroblastoma cell line (provided by
   Nguyen Lab) was cultured in DMEM/F12 complemented with 10% of
   heat-inactivated fetal bovine serum (FBS, Fisher Scientific). Primary
   cortical cultures (composed of neurons and glial cells) were prepared
   from the brain of P1 E4f1^flox/flox; Rosa26-CRE^ER newborns and
   cultured in BrainPhys medium (Stem Cell Technologies, 05790)
   supplemented with 2% B27 (Gibco, 17504044) and 0.25% Glutamax (Gibco,
   35050037)^[254]43. Low passage populations of primary skin fibroblasts
   isolated from LS patients and age-matched control individuals (between
   P6 and P10), or MEFs of interest (P1–P3) can be provided upon request.

RNA extraction, RNA-seq and quantitative PCR

   Total RNAs were isolated from various organs or from cells using TRIzol
   Reagent (Invitrogen, 15596018). For RNA-seq, mRNA libraries were
   generated following the manufacturer recommendations (mRNA hyper prep
   from ROCHE). Bar-coded libraries were pooled and sequenced on the
   Novaseq6000 200 cycles Illumina sequencer, with a sequencing depth
   corresponding to a minimum of 40 million of 2x100bp paired-end reads
   per sample after demultiplexing. Quantitative real time PCR (RT-qPCR)
   was performed on a LightCycler 480 (Roche). cDNAs were synthesized from
   500 ng of total RNA using the SuperScript™ III Reverse Transcriptase
   (Invitrogen, 18080044). The relative mRNA copy number was calculated
   using Ct values and was normalized to 18S and tubulin β5 mRNA levels.
   Primers used for RT-qPCR analyzes are listed in the Supplementary
   Data [255]3. For RNA-sequencing, STAR v2.5.3a was used to align reads
   on the reference genome mm10. Gene expression was determined with RSEM
   1.2.28 on RefSeq catalog, prior to normalization and differential
   analysis with DESeq2 bioconductor package using likelihood ratio tests
   (LRT). Multiple hypothesis adjusted p-values were calculated with the
   Benjamini–Hochberg procedure to control FDR. GSEA and PEA were
   performed with fgsea bioconductor package on Hallmark gene sets from
   MSigDB collections, completed with some custom gene sets. As for
   differential expression analysis, p-values were corrected using
   Benjamini–Hochberg’ FDR.

Polysome profiling

   Entire brain from E14.5 embryos were lysed in polysome lysis buffer (7%
   sucrose, 50 mM Tris-HCl pH 7.5, 5 mM MgCl[2], 25 mM KCl, 1% NP40,
   10 U/mL RNAseOUT (Invitrogen, 10777019) supplemented with protease
   inhibitors and 20 μg/mL Emetin (Sigma, E2375)) by hard shaking with
   1.4 mm ceramic spheres (Lysing matrix D MPBio, 1169130-CF) in a
   FastPrep instrument (MPBio). 30 μg of total RNA were loaded on 15–50%
   sucrose gradient and ultracentrifuged at 2,10,000×g for 2.5 h at 4 °C
   in a SW41 rotor (Beckman Coulter). Approximately 10% of the cytosolic
   RNAs were kept as an input. The different fractions were separated
   through a live optical density (OD) 254 nm UV spectrometer and
   collected with an ISCO (Lincoln, NE) density gradient fractionation
   system. The absorbance at 254 nm was measured continuously as a
   function of gradient depth. Total cytosolic mRNAs, as well as mRNAs
   associated with 1 to 3 ribosomes (Light polysomal fraction) and mRNAs
   associated with more than 3 ribosomes (Heavy polysomal fraction) were
   extracted using TRIzol LS (Invitrogen, 10296028) according to the
   manufacturer instructions. For sequencing, mRNA libraries were
   generated following the manufacturer recommendations (mRNA hyper prep
   from ROCHE). Bar-coded libraries were pooled and sequenced on a
   Novaseq6000 200 cycles Illumina sequencer, with a sequencing depth
   corresponding to a minimum of 80 million of 2 × 100 bp paired-end reads
   per sample after demultiplexing.

   Polysome profiling datasets were pre-processed using the nf-core
   RNA-seq pipeline v3.11.0^[256]44. Initial quality control of raw reads
   was conducted with FastQC v0.11.9. Trim Galore v0.6.7 was used for
   reads trimming, with parameters set to a minimum read length of 25 bp,
   a quality threshold of Phred score 30, and a stringency of 10.
   Additionally, 1 bp was clipped from both 5’ ends of R1 and R2 reads.
   Reads were aligned to the mouse reference genome (GRCm38) using STAR
   v2.7.9a. Transcript-level quantification was performed using Salmon
   v1.10.1, and gene-level summarization was carried out using tximport
   v1.32.0. Post-alignment quality was assessed with RSeQC v3.0.1 to
   ensure data integrity and alignment accuracy.

   For statistical analysis, datasets from polysome fractions (light,
   heavy, and total) were analyzed independently using R v4.4.0.
   Differential expression (DE) analysis for each fraction was conducted
   with edgeR v4.2.1, comparing gene expression between E4f1^(Nes)KO and
   CTL animals. Normalized count data were fitted to a generalized linear
   model (GLM) according to E4f1 status as a cofactor (KO or WT). After
   estimating the signal dispersion, the abundance of transcripts in a
   given fraction was considered significantly different in likelihood
   ratio tests when exhibiting an absolute fold-change ≥ 1.2 and a
   FDR-adjusted p-value < 0.2. To integrate results from the different
   fractions, the corresponding genes were categorized as controlled at
   the post-transcriptional level (differential abundance of their
   transcripts in either the heavy and/or the light polysomal fractions,
   but not in the total fraction), at the transcriptional level
   (differential abundance of their transcripts in the total fraction but
   not in the heavy nor in the light polysomal fractions), and at both the
   transcriptional and post-trancriptional (Mix) levels (differential
   abundance of their transcripts in the total fraction and in the heavy
   and/or in the light polysomal fractions). Codon usage bias was assessed
   by comparing codon frequencies in transcripts controlled at the
   post-trancriptional level in E4F1-deficient cells to those in
   transcripts from the entire GRCm38 transcriptome using χ^2 tests. In
   the corresponding proteins, enrichment of the specific hydrophilic
   motif linked to U[34]-dependent translational defects was conducted
   using the AME (Analysis of Motif Enrichment)^[257]45 tool from the MEME
   suite v1.12.0, using the occurrence of this motif in the full GRCm38
   proteome as a reference.

Measurement of PDH activity and blood lactate levels

   PDH activity was measured using Pyruvate Dehydrogenase Activity Assay
   Kit (MAK183-1KT, Sigma) as recommended by the manufacturer. Lactate
   levels were measured in blood using a lactometer (EKF Diagnostics).

AcetylCoEnzyme A (AcCoA) quantification

   Brains from E14.5 embryos were lysed in 10% trichloroacetic acid (TCA)
   diluted in HPLC-grade water (Fisher Scientific) precooled at 4 °C.
   Samples were centrifuged and supernatants were loaded onto SPE columns
   (Oasis HLB 3cc; Waters WAT094226). Columns were washed with HPLC-grade
   water, then eluted with HPLC-grade MeOH with 25 mM ammonium acetate.
   Samples were dried with unheated speed vac, resuspended in HPLC-grade
   water, transfered onto HPLC glass vials and placed in the autosampler
   at 4 °C. LC–high resolution mass spectrometry quantification was
   performed using a Vanquish chromatographic system coupled with a
   Q-Exactive Plus mass spectrometer (Thermo Fisher Scientific) fitted
   with an electrospray source. Chromatographic separation was performed
   on an InfinityLab Poroshell HPH-C18 column (2.7 μm, 3.0 × 100 mm;
   Agilent) maintained at 30 °C.

   Identification was determined by extracting the accurate mass of
   acyl-CoAs with a mass accuracy of 5 ppm and comparing the retention
   time to standards injected in the same sequence (mix of 5 μM acyl-CoA).
   Peak areas of acyl-CoAs were divided by peak area of the internal
   standard used, MES, and by the weight of the corresponding brain. Data
   treatment was done with Xcalibur Freestyle Software (Thermo Fisher
   Scientific).

Protein extraction and western blotting

   Total protein extracts were prepared from total brains isolated from
   E14.5, E16.5 or E18.5 E4f1^(Nes)KO and CTL embryos lysed in Laemmli
   buffer (80 mM Tris pH = 6,8, 2% SDS, 12% sucrose, 2% β-mercaptoethanol,
   bromophenol blue) and immunoblotting was performed using the following
   antibodies: anti- E4F1 (1/1000)^[258]46, C-CASP3 (Cell Signaling,
   9661S, lot 43, 1/1000), DLAT (Santa Cruz, sc-271534, lot C1813, 1/500),
   MPC1 (Sigma, HPA045119, lot E105936, 1/1000), DLD (Santa Cruz,
   sc-365977, lot F1611, 1/1000), PDHE1 (Life Technologies, 459400, lot
   J4161, 1/500), ELP3 (Abcam, ab190907, lot GR3180892-3, 1/500), ELP1
   (Abcam, ab62498, lot GR3369590, 1/500), TUBULIN (Sigma, T6199, lot
   115M4796V, 1/1000), ACTIN (Sigma, A3854, lot 100M4782, 1/7000),
   Acetylated-TUBULIN (Santa Cruz, sc23950, lot E2218, 1/500), Histone
   H3K9Ac (Cell Signaling, 9649S, lot 13, 1/1000), Histone H3K23Ac (Abcam,
   ab61234, lot GR82698-4, 1/1000), Histone H3K27Ac (Abcam, ab4729, lot
   17, 1/1000), Histone H3K14Ac (Cell Signaling, 7627, 1/1000), total
   Histone H3 (Cell Signaling, 4499S, lot 9, 1/1000), HSP90 (Cell
   Signaling, 13998S, lot 2, 1/1000), KCNA3 (Diagomics, AF6702, lot 23,
   1/500) and Cleaved-Caspase 8 (Proteintech, 66093-1, lot 10023946,
   1/1000).

LC-MS detection of tRNA modifications

   Total RNAs were isolated from brains of E14.5 E4f1^(Nes)KO and CTL
   embryos or from MEFs using TRIzol Reagent (Invitrogen, 15596018). tRNAs
   were purified on an agarose gel. 400 ng of tRNAs were digested with
   0.001U of nuclease P1 (Sigma, N8630) in 25 mM of ammonium acetate pH
   5.3 for 2 hours at 42 °C. Samples were then treated with 0.001U of
   alkaline phosphatase (Sigma, P4252) in 250 mM ammonium acetate pH 5.3
   at 37 °C for 2 h. The nucleoside solution was diluted twice and
   filtered with 0.22 µm filters (Millex®-GV, Millipore, SLGVR04NL). 5 µL
   of each sample was injected in triplicate in a LC-MS/MS. For stable
   isotope labeling experiments performed in MEFs using ^13C-sodium
   pyruvate (Sigma, 490717) or ^13C-sodium acetate (Sigma, 282014), 4 ug
   of total RNA were digested with 0.002 U of nuclease P1 and alkaline
   phosphatase by following the same procedure. Nucleosides were then
   separated using Nexera LC-40 systems (Shimadzu) and Synergi™ Fusion-RP
   C18 column (4 µm particle size, 250 mm × 2 mm, 80 Å) (Phenomenex,
   00G-4424-B0). Mobile phase was composed of 5 mM ammonium acetate pH5.3
   (solvent A) and acetonitrile (solvent B). The 30 minutes elution
   gradient started with 100% phase A followed by a linear gradient until
   reaching 8% solvent B at 13 min. Solvent B was further increased to 40%
   over the next 10 min. After 2 min, solvent B was decreased back to 0%
   at 25.5 minutes. Initial conditions were recovered by rinsing the
   LC-system with 100% solvent A for additional 4.5 min. The flow rate was
   set at 0.4 mL/min and the column temperature at 35 °C. Analysis was
   performed using a triple quadrupole 8060NX (Shimadzu corporation,
   Kyoto, Japan) in positive ion mode. Peak areas were measured using
   Skyline 21.2 software (MacCoss Lab Software; [259]https://skyline.ms)
   and the ratio of modified nucleosides on non-modified nucleosides was
   calculated.

Quantitative chromatin immunoprecipitation (qChIP) assays

   For qChIP assays in cells, N2A cells were incubated with 1%
   formaldehyde / 1% paraformaldehyde for 5 min. Nuclei were isolated and
   chromatin was extracted and sonicated (epishear, Active Motif). qChIPs
   on the Elp3 locus were carried out using a control irrelevant IgG or an
   affinity-purified rabbit anti-E4F1 polyclonal
   antibody^[260]7,[261]22,[262]46. Primers used for qChIP assays are
   listed in the supplementary table [263]1.

Lentivirus production

   Lentiviral and retroviral particles encoding Flag-tagged human
   WT-PDHA1, the PDHA1^C91A mutant, human E4F1^WT or the E4F1^K144Q
   mutant, and shRNAs targeting mouse Pdha1 (TRCN0000041914,
   TRCN0000041914, Sigma) or mouse Dlat (TRCN0000041608 and
   TRCN0000041609, Sigma) were produced in HEK293T packaging cells. 72 h
   after transfection, viral supernatants were harvested and added on MEFs
   overnight in presence of polybrene (5 μg/mL, Sigma). Selection of
   transduced cells was performed 48 h after transduction with puromycin
   (2 μg/mL, Invitrogen, J67236.8EQ) or hygromycin (50 μg/mL, Invitrogen,
   10687010).

Immunohistochemistry, immunofluorescence and histological analyzes

   The brain of E14.5, E16.5, E18.5 E4f1^(Nes)KO and CTL embryos were
   dissected in 0.1 M phosphate-buffered saline pH7.4 (PBS) and were fixed
   at 4 °C in 4% paraformaldehyde (PFA) for one hour. Fixed samples were
   cryoprotected overnight in 20% sucrose in PBS at 4 °C, embedded in OCT
   Compound (VWR International, 361603E) and sectioned (14 µm) onto slides
   (SuperFrost Plus, VWR International, 631-0108) using a cryostat. Entire
   head samples were fixed in 4% neutral-buffered formalin (24 h),
   decalcificated in 10% EDTA for 3 days and paraffin embedded. FFPE
   tissues were sectioned (4μm sections) and processed for
   immunohistochemistry (IHC) or hematoxylin and eosin (H/E) staining. For
   GFP (Invitrogen, A10262,, lot 2260907, 1/100), Cleaved-Caspase 3 (Cell
   Signaling, 9661S, lot 43, 1/100) and CD45 (Biosciences, 14-051-82, lot
   4316815, 1/500) stainings, IHC was performed on a VENTANA Discovery
   Ultra automated staining instrument (Ventana Medical Systems) according
   to the manufacturer’s instructions. The number of Cleaved-Caspase3
   positive cells were quantified on whole-slides digitized sections
   (Panoramic 250 Flash II/3DHISTECH whole-slide digital microscope)
   scanned at a 0.22 µm/pixel resolution, using Definiens Developer XD
   software (Definiens, Munich, Germany). IF stainings were processed
   manually. Briefly, brain sections were pretreated with target retrieval
   solution (DAKO, 20 min, 95 °C) prior incubation with the following
   primary antibodies overnight at 4 °C: anti- MCT4 (Proteintech,
   22787-1-AP, lot 00049309, 1/200), ATF4 (Cell Signaling, 11815, lot 6,
   1/200), phospho-eIF2a (Cliniscience, [264]P04387, lot [265]P04387,
   1/500), Ki67 (abcam, Ab15580, lot GR342706-1, 1/500), SOX2 (Santa Cruz,
   Sc17320, lot #H0516, 1/500), TBR2 (eBiosciences, 14-4875-82, lot
   2504948, 1/250), TBR1 (Abcam, ab31940, lot GR3182037-1, 1/500), GFAP
   (Avès Labs, AB2313547, lot 4937976, 1/200) and NeuN (Biolegend, 834501,
   lot B352650, 1/500). After washing, sections were incubated for 1 h at
   room temperature with either anti-rabbit, anti-rat or anti-goat
   secondary antibodies coupled to Alexa-488, Alexa-647 or Alexa-555 (Life
   Technologies). Nuclei were counterstained with DAPI (Sigma, D9542) and
   slides were washed in PBST and coverslipped using Mowiol mounting
   solution. For images analyzes, sections were visualized on a Thunder
   epifluorescent microscope (Nikon).

High-Resolution Episcopic Microscopy (HREM)

   E14.5 embryos and E16.5 heads were fixed in Bouin’s fixative for 24 h,
   then washed and stored in 70% EtOH. Fixed samples were gradually
   dehydrated in an increasing series of ethanol concentrations and were
   embedded in a methacrylate resin (JB-4 kit, Polysciences, Warrington,
   PA) containing eosin and acridin orange as contrasts agents^[266]47.
   The resin blocks were sectioned on the Histo3D system to generate data
   by repeated removal of 7 µm thick sections^[267]48. Resulting HREM data
   with a voxel size of 8 × 8 × 7 µm^3 (E14.5 whole embryos) or
   6.5 × 6.5 × 7 µm^3 (E16.5 heads) were generated from approximately 1000
   or 800 aligned images, respectively. All HREM images were converted
   into a volume dataset and segmented using Avizo 9.4.0 software
   (Thermofisher Scientific, France) to create 2- and 3-dimensional (3D)
   images, according to the Allen brain atlas.

Statistical analysis

   In vivo studies were performed on a sufficient number of animals per
   genotype with a minimum of 5 animals per experimental group. The in
   vitro data were obtained from 3 to 5 independent experiments as
   indicated in the figure legends. Statistical significance was evaluated
   using unpaired bilateral Student’s t-test. P-values inferior or equal
   to 0.05 were considered statistically significant.

Reporting summary

   Further information on research design is available in the [268]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [269]Supplementary Information^ (5.6MB, pdf)
   [270]41467_2024_55444_MOESM2_ESM.pdf^ (117.6KB, pdf)

   Description of Additional Supplementary Files
   [271]Supplementary Data 1^ (2.2MB, xlsx)
   [272]Supplementary Data 2^ (9.5KB, xlsx)
   [273]Supplementary Data 3^ (4.5MB, xlsx)
   [274]Reporting Summary^ (142.5KB, pdf)

Source data

   [275]Source Data^ (32.5MB, xlsx)

Acknowledgements