Abstract

   TET-family members play a critical role in cell fate commitment.
   Indeed, TET3 is essential to postnatal development due to yet unknown
   reasons. To define TET3 function in cell differentiation, we have
   profiled the intestinal epithelium at single-cell level from wild-type
   and Tet3 knockout mice. We have found that Tet3 is mostly expressed in
   differentiated enterocytes. In the absence of TET3, enterocytes exhibit
   an aberrant differentiation trajectory and do not acquire a
   physiological cell identity due to an impairment in oxidative
   phosphorylation, specifically due to an ATP synthase assembly
   deficiency. Moreover, spatial metabolomics analysis has revealed that
   Tet3 knockout enterocytes exhibit an unphysiological metabolic profile
   when compared with their wild-type counterparts. In contrast, no
   metabolic differences have been observed between both genotypes in the
   stem cell compartment where Tet3 is mainly not expressed. Collectively,
   our findings suggest a mechanism by which TET3 regulates mitochondrial
   function and, thus, terminal cell differentiation at the metabolic
   level.

   Subject terms: Stem-cell differentiation, Mechanisms of disease, DNA
   methylation, Disease model, Energy metabolism
     __________________________________________________________________

   TET3 is essential to postnatal development due to unknown reasons. Here
   Mulet et al. demonstrate that TET3 is required for ATP synthase
   assembly in differentiated cells and that its loss leads to an
   unphysiological metabolic profile.

Introduction

   Methylation of the cytosine base at the fifth carbon position
   (5-methylcytosine; 5mC) is a key epigenetic modification of DNA that
   regulates development and disease^[58]1. In mammalian cells, 5mC can be
   oxidized into 5-hydroxymethylcytosine (5hmC)^[59]2. To date, the
   Ten-eleven-translocation gene family (Tet1, Tet2 and Tet3) are the only
   enzymes that have been described to convert 5mC into 5hmC through their
   α-ketoglutarate/Fe(II)-dependent dioxygenase activity^[60]2.

   TET enzymes seem to play an essential role in lineage specification
   across all three germ layers and their dysfunctions result in skewed or
   arrested differentiation^[61]3–[62]5. Loss-of-function mouse models
   from the three Tet family members have provided insight into their
   unique and overlapping functions. Tet1 knockout mice have been
   described to be viable and fertile^[63]6. Likewise, Tet2 knockout mice
   survive to an adult stage but develop lethal myeloid
   malignancies^[64]7. Tet1 and Tet2 double knockout mice are born at
   mendelian ratios but half of them die perinatally, suggesting that Tet1
   and Tet2 enhance survival but that their loss do not impair postnatal
   development^[65]8. Remarkably, Tet3 knockout mice develop to term but
   die perinatally due to unknown reasons, suggesting that Tet1 and Tet2
   cannot compensate for Tet3 critical functions at birth^[66]9.
   Altogether, the Tet paralogs exhibit functional differences that are
   not fully understood.

   TET3 loss leads to developmental defects in multiple cell types
   indicating that TET3 is a crucial factor for lineage commitment and
   terminal differentiation^[67]10. In adult homeostasis, the role of Tet3
   has been studied in a few tissues together with Tet2, as both paralogs
   are expressed in many differentiated cells^[68]5,[69]11. TET3 can
   rescue TET2 loss during hematopoietic differentiation to some lineages
   but not to others, suggesting that TET2 and TET3 share some but not all
   targets^[70]5. Mechanistically, TET3 has been described to be
   recruited, via interaction with transcription factors, to certain
   lineage—associated genomic loci to promote DNA demethylation and
   stimulate gene expression^[71]10,[72]12. It has also been hypothesized
   that an increase in DNA hydroxymethylation levels due to TET3 activity
   might enhance chromatin accessibility and binding of key
   lineage—specific transcription factors^[73]11. However, the precise
   molecular mechanism by which TET3 physiologically regulates adult stem
   cell differentiation remains unresolved.

   In this study, we have used the mouse intestinal epithelium as a model
   to investigate the role of TET3 during adult stem cell differentiation.
   Our data demonstrates that TET3 is required for ATP synthase assembly
   and, thus, is critical for mitochondrial function in cell
   differentiation. Overall, in the absence of TET3, enterocytes exhibit
   an unphysiological metabolic profile due to a deregulation in oxidative
   phosphorylation.

Results

Tet3 expression mainly correlates with enterocyte differentiation

   To better understand the precise physiological role of TET3 in stem
   cell-hierarchical tissues, we generated a Tet3 knockout mouse line
   through homologous recombination in which exon 5 was excised (Supp.
   Figure [74]1a). Knockout mice were born at predicted mendelian ratios
   but died on postnatal day (P) 1 showing no gross abnormalities (Supp.
   Figure [75]1b, c, d). 5hmC staining on whole embryo sections on
   embryonic day (E) 18.5 pointed to the intestinal epithelium as the
   tissue with the most significant reduction in 5hmC levels in the
   absence of TET3 (Supp. Figure [76]1e). Hence, we decided to focus on
   the intestinal epithelium for further studies assuming that the
   substantial decrease of 5hmC observed in this tissue might help us to
   pinpoint TET3 critical functions. A histological analysis of E18.5
   intestinal epithelium showed no differences between the wild-type,
   heterozygous and Tet3 knockout genotypes (Supp. Figure [77]2a). Of
   note, E18.5 small intestinal epithelium does not contain crypts, which
   develop around P5^[78]13. Next, we quantified RNA transcripts of the
   three Tet-family members in a cell-type specific manner with single
   molecule RNA fluorescent in situ hybridization (smFISH) using RNAscope
   probes in intestinal sections. To this end, Tet1, Tet2 and Tet3 RNAs
   were simultaneously detected together with Olfm4, Vil1, Muc2 or Chga
   RNAs, as markers of intestinal stem cells and their progeny:
   enterocytes, goblet and enteroendocrine cells, respectively
   (Fig. [79]1). Paneth cells have not emerged at this developmental stage
   yet^[80]14. In each section, only the continuous intestinal epithelial
   monolayer containing entire villi, in which the tip and the base of
   each villus could be unequivocally identified, were used for subsequent
   cell segmentation and quantification with HALO software (Supp.
   Figure [81]2b). In the wild-type, our analysis shows that Tet3 RNA copy
   number inversely correlates to Olfm4, Muc2 and Chga RNA copy number. In
   contrast, a positive correlation is clearly observed between Tet3 and
   Vil1 RNA copy numbers. These results imply that Tet3 expression
   increases during enterocyte maturation since Vil1 expression is
   enhanced in differentiated enterocytes^[82]15. A spatial plot depicting
   the location of individual intestinal epithelial cells and their Tet3
   copy numbers is displayed in Supp. Figure [83]2c and reveals a pattern
   in which Tet3 RNA copy numbers increase as the cells mature during
   their migration towards the villi tip. Likewise, a Tet2 positive
   correlation with Vil1 could also be observed (Supp. Figure [84]3).
   However, Tet1 was only detected at very low copy numbers in very few
   cells and did not show a correlation to any specific cell-type marker
   (Supp. Figure [85]4). In the Tet3 knockout intestinal epithelium, Tet1
   and Tet2 exhibited the same expression pattern as in the wild-type
   (Supp. Figure [86]2f, [87]3 and [88]4). Quantitative RT-PCR on bulk
   wild-type intestinal epithelial cells demonstrated that Tet2 and Tet3
   are expressed at similar levels (Supp. Figure [89]2d). The low level of
   Tet1 expression identified by in situ hybridization was almost
   undetectable by bulk quantitative RT-PCR in both genotypes. Loss of
   Tet3 expression was accompanied with an increase in Tet2 expression
   levels in the Tet3 knockout. Altogether, our findings suggest that TET2
   and TET3 might play a role in enterocyte differentiation.

Fig. 1. Enterocytes are the main cell-type expressing Tet3 in the intestinal
epithelium.

   [90]Fig. 1
   [91]Open in a new tab

   a, b, c and d Representative images of multiplex RNAscope in situ
   hybridization for Tet1, Tet2 and Tet3 plus one cell type—specific
   marker: a Olfm4 (intestinal stem cells), b Chga (enteroendocrine
   cells), c Muc2 (goblet cells) and d Vil1 (enterocytes) in wild-type
   intestinal epithelial sections. Scale bars represent from left to
   right, 100, 20 and 10 μm, respectively. a, b, c and d Co-expression of
   Tet3 with each cell type—specific marker at the single-cell level in
   the intestinal epithelium. The axes represent the number of RNA copies
   detected for each gene by RNAscope in situ hybridization and the color
   code indicates the number of cells for each combination. The total
   number of analyzed cells is shown at the top. Source data are provided
   as a Source Data file.

   5hmC globally accumulates as the intestinal epithelial cells migrate
   upwards to the lumen (Supp. Figure [92]2g). Interestingly, the Tet3
   knockout intestinal epithelium exhibited a reduction in the intensity
   and number of 5hmC—positive cells whereas the 5hmC levels on lamina
   propria and interstitial space were unaffected, as shown by
   immunofluorescence (Supp. Figure [93]2g). Liquid chromatography-tandem
   mass spectrometry (LC-MS/MS) quantified a 50% reduction in global 5hmC
   levels in the Tet3 knockout intestinal epithelium (Supp.
   Figure [94]2e). This result is in line with the intestinal epithelium
   being mainly composed of enterocytes and Tet3 being mainly expressed in
   this cell-type (Fig. [95]1). Tet2 is also expressed in enterocytes
   which explains the remaining 5hmC levels in the absence of TET3.
   Collectively, our results demonstrate that Tet2 and Tet3 are the
   Tet-family members responsible for 5hmC deposition in enterocytes.

TET3 is essential to acquire a physiological enterocyte signature

   Next, we investigated the specific role of TET3 on enterocyte cell fate
   commitment. To this end, we used the single cell 10X Genomics platform
   to perform single-cell RNA sequencing (scRNA-seq) on 12,000 individual
   EpCam—positive epithelial cells sorted from the small intestine of
   wild-type and Tet3 knockout E18.5 embryos (Supp. Figure [96]5a, b). Our
   analysis resolved seventeen clusters that were associated with distinct
   cell identities namely, stem/transit-amplifying cells (clusters 9 and
   11), goblet cells (cluster 16), immature proximal and distal
   enterocytes (clusters 4, 6, 7, 10, 12, 14 and 15), proximal enterocytes
   (clusters 1, 2, 8 and 13) and distal enterocytes (clusters 0, 3 and 5),
   based on known markers from adult intestinal epithelium^[97]16 since
   the information available for newborns is limited (Fig. [98]2a). The
   top 100 genes defining each cluster are shown in Supp. Data [99]1 and
   the 3 key genes determining the identity of each cluster are depicted
   in Supp Fig. [100]5e. Both wild-type and Tet3 knockout goblet cells are
   represented by a single distinct cluster (cluster 16) that contains a
   similar percentage of wild-type and knockout cells (Supp.
   Table [101]1). Likewise, no major differences in gene expression and
   number of cells of each genotype are observed in clusters 9 and 11 that
   correspond to stem/transit-amplifying cells (Supp. Table [102]1). In
   contrast, enterocytes segregate in different clusters depending on
   their genotype, in which clusters 0, 1, 5 and 7 are mainly composed of
   wild-type cells whereas clusters 2, 3, 8 and 13 are composed almost
   entirely of Tet3 depleted cells (Fig. [103]2b, c and Supp.
   Table [104]1). A trajectory analysis confirmed a bifurcating trajectory
   from stem/transit-amplifying cells into the secretory and the
   absorptive lineage and a totally different enterocyte differentiation
   trajectory between the wild-type and the Tet3 knockout cells
   (Fig. [105]2d). Indeed, the wild-type cells progress from the
   stem/transit-amplifying cells into enterocytes (clusters 0, 1, 5 and
   7). Instead, the majority of the Tet3 knockout cells follow an aberrant
   trajectory and accumulate in clusters 2, 3, 8 and 13. These results are
   consistent with the positive correlation observed between Tet3 and Vil1
   in enterocytes (Fig. [106]1) and indicate that most enterocytes do not
   reach a physiological cell identity in the absence of TET3.

Fig. 2. Enterocytes acquire an aberrant transcriptional signature in the
absence of TET3.

   [107]Fig. 2
   [108]Open in a new tab

   a Cluster color-coded scRNA-seq t-SNE visualization of 5933 E18.5
   wild-type and 5983 E18.5 Tet3 knockout small intestinal epithelial
   cells. Clusters are labelled with numbers and cell identities are
   delineated with dashed lines. b t-SNE visualizations of (upper panel)
   only wild-type cells n = 5933 cells and (lower panel) only wild-type
   specific clusters n = 3816 cells. c t-SNE visualizations of (upper
   panel) only Tet3 knockout cells n = 5983 cells and (lower panel) only
   Tet3 knockout specific clusters n = 3211 cells. d Cluster color-coded
   (upper panel) or genotype color-coded (lower panel) cell
   differentiation trajectory containing both wild-type (n = 5933) and
   Tet3 knockout (n = 5983) cells for each genotype. Black and white
   arrows in lower panel show differentiation into the physiological and
   aberrant absorptive lineage, respectively. Dashed lines delimit
   wild-type and Tet3 knockout—specific clusters. e, h Top 8 significantly
   enriched ontology terms identified using g:Profiler in specific
   wild-type clusters (n = 3816 cells) or in specific Tet3 knockout
   clusters (n = 3211 cells), respectively. f, i Expression levels of
   genes representative of the mitochondrial electron transport pathway or
   lipid metabolism pathway in specific wild-type clusters (n = 3816
   cells) or in specific Tet3 knockout clusters (n = 3211 cells),
   respectively. Median as well as the first quartile and the third
   quartile are depicted in green and yellow, respectively. g GSEA plot
   showing oxidative phosphorylation as a negatively enriched pathway in
   Tet3 knockout epithelial cells. j Differential expression and
   statistical significance between wild-type (n = 3816 cells) and Tet3
   knockout (n = 3211 cells) specific clusters for mitochondrial subunits
   showing p-values < 0,05. Fisher’s one-tailed test with g:SCS multiple
   testing correction (e, h) and Wilcoxon rank-sum test with
   Benjamini-Hochberg correction (f, i and j) was used.***p-value < 0.001.
   Source data are provided as a Source Data file.

   A gene ontology analysis was performed to identify the main biological
   functions enriched only in wild-type (clusters 0, 1, 5 and 7) or only
   in aberrant (clusters 2, 3, 8 and 13) enterocytes. The analysis
   revealed a significant enrichment of oxidative phosphorylation in the
   wild-type (Fig. [109]2e, f) but of fatty acid metabolism and sterol
   biosynthesis in the Tet3 knockout enterocytes (Fig. [110]2h, i),
   suggesting that energy metabolism is dysregulated in the absence of
   TET3. Gene set enrichment analysis between wild-type (clusters 0, 1, 5
   and 7) and Tet3 knockout (clusters 2, 3, 8 and 13) specific enterocyte
   clusters confirmed the oxidative phosphorylation pathway as negatively
   enriched in the absence of TET3 (Fig. [111]2g and Supp. Figure [112]5c,
   d). Indeed, a volcano plot of all the components of the mitochondrial
   electron transport chain showed that a large number of complex I and
   F[1]F[0]ATP synthase (also known as complex V) subunits were
   significantly downregulated in the Tet3 knockout enterocytes
   (Fig. [113]2j and Supp. Figure [114]5f, g). Overall, our results
   demonstrate that the loss of TET3 leads to enterocytes with an aberrant
   metabolic transcriptional signature.

Mitochondrial morphology is abnormal in the absence of TET3

   The gene expression analysis suggests a dysregulation in oxidative
   phosphorylation in the absence of TET3 (Fig. [115]2g). Next, we sought
   to ascertain the impact of TET3 loss on mitochondrial morphology as it
   associates with mitochondrial function^[116]17. In the intestinal
   epithelium, while the intestinal stem cells rely largely on glycolysis
   to fulfill their energetic demands, enterocyte dependence on oxidative
   phosphorylation increases with differentiation^[117]18. Mitochondrial
   elongation has been suggested to correlate with oxidative
   metabolism^[118]19 and, accordingly, enterocytes exhibit more elongated
   mitochondria than intestinal stem cells^[119]20. Hence, electron
   microscopy analysis was employed to study the mitochondria size in the
   villi tip and in the intervilli region in both wild-type and Tet3
   knockout (Fig. [120]3a). As previously reported, the wild-type
   mitochondria are significantly larger in the differentiated cells
   located at the villus tip than at the stem cell compartment
   (Fig. [121]3b). In contrast, the mitochondrial area does not increase
   in the Tet3 knockout villi tip, pointing to a mitochondrial dysfunction
   during intestinal epithelial cell differentiation in the absence of
   TET3 (Fig. [122]3b). Interestingly, there is no difference in
   mitochondrial area between the wild-type and the Tet3 knockout
   mitochondria located in the intervilli (Fig. [123]3b). These findings
   suggest that TET3 loss has no impact on the stem cell compartment and
   correlate with the results obtained by single cell gene expression
   profiling (Fig. [124]2d). In contrast, mitochondrial size is
   significantly larger in the wild-type in comparison to the Tet3
   knockout at the villi tip (Fig. [125]3b), indicating that mitochondria
   does not properly develop upon TET3 loss. Overall, mitochondria show an
   abnormal morphology in differentiated cells in the absence of TET3,
   which points to an alteration in mitochondrial fate during intestinal
   epithelial differentiation.

Fig. 3. Mitochondrial morphology remains immature in the absence of TET3.

   [126]Fig. 3
   [127]Open in a new tab

   a Representative transmission electron microscopy images from E18.5
   wild-type and Tet3 knockout mitochondria located at intervilli and
   villi tips are shown. Semi-thin sections images with boxed areas
   indicating the analyzed villi regions are included. Scale bars: 500 nm.
   b Quantitative mitochondrial area comparison between wild-type
   intervilli, Tet3 knockout intervilli, wild-type villi tip and Tet3
   knockout villi tip (n = 45, 53, 155, and 150 mitochondria,
   respectively). Center line, median; box limits, upper and lower
   quartile; whiskers, 1.5x interquartile range; points, all measures.
   Student t test, two-tailed. Source data are provided as a Source Data
   file.

TET3 regulates mitochondrial F[1]F[0]-ATP synthase assembly

   Tet3 knockout enterocytes showed a reduction in the expression of a
   large number of mitochondrial complex I and F[1]F[0]-ATP synthase
   subunits, thus we next attempt to functionally validate these results.
   However, isolating wild-type (clusters 0, 1, 5 and 7) and Tet3 knockout
   (clusters 2, 3, 8 and 13) differentiated enterocytes is technically not
   feasible. Cultured fibroblasts obtained from patient skin biopsies have
   been previously used in the diagnostic evaluation of patients with a
   suspected mitochondrial disorder related to nuclear-encoded
   genes^[128]21. Therefore, we decided to assess mitochondrial complex I
   and ATP synthase activities in E18.5 tail tip derived-fibroblasts that,
   like enterocytes, also exhibit a terminally differentiated phenotype
   and a significant reduction in 5hmC levels (Supp. Figure [129]6a, b).
   Quantitative gene expression analysis in Tet3 knockout fibroblasts
   revealed a decrease in the expression of several complex I and ATP
   synthase mitochondrial subunits at similar levels to those observed in
   Tet3 knockout enterocytes, suggesting that the results obtained in
   fibroblasts could be extrapolated to enterocytes (Supp.
   Figure [130]6c). Differences in complex I assembly into supercomplexes
   were not observed despite the reduction in gene expression of many of
   its subunits, as shown by blue native gel electrophoresis (BNGE)
   followed by either in gel activity or western blotting for NDUFA9
   (Fig. [131]4a). Next, we assessed the activity of mitochondrial complex
   I that is known to transition between active and deactive
   forms^[132]22. NADH:ubiquinone activity informs on the ability of
   complex I to extract electrons from NADH and transfer them to
   ubiquinone, allowing to determine the active form. On the other hand,
   NADH:Fe(CN)[6] oxidoreductase activity informs on the available complex
   I regardless if it is in active or deactive form. For the purpose of
   estimating both the amount of complex I and the proportion of active
   complex I, we measured NADH:ubiquinone and NADH:Fe(CN)[6]
   oxidoreductase activities spectrophotometrically in isolated wild-type
   and Tet3 knockout mitochondria (Fig. [133]4b). Our results show a
   significant decrease in complex I NADH:Fe(CN)[6] but not in
   NADH:ubiquinone oxidoreductase activity. Altogether, our findings
   demonstrate that the activity of complex I is unaffected upon TET3 loss
   because the reduction in the total amount of complex I observed in the
   absence of TET3 is compensated by a higher percentage of complex I in
   its active form. Self-association of F[1]F[0]-ATP synthase monomers
   into rows of dimers is a fundamental biological process involved in
   energy efficiency^[134]23. In mammals, F[1]F[0]-ATP synthase
   dimerization and oligomerization is mediated by specific subunits
   located at the F[1]F[0]-ATP synthase membrane domain. Strikingly, most
   of the F[1]F[0]-ATP synthase subunits that were downregulated in both
   Tet3 knockout enterocytes and fibroblasts were located in the membrane
   domain, as shown in the bovine F[1]F[0]-ATP synthase structure that is
   usually employed as a surrogate for mammalian homologs^[135]24
   (Fig. [136]4c). Hence, we used clear native gel electrophoresis to
   determine the oligomerization state of the ATP synthase upon TET3 loss.
   In gel ATPase activity and western blotting for the ATP synthase β
   subunit in isolated mitochondria revealed a significant decrease in the
   dimer/monomer ratio and a total loss of oligomers in the Tet3 knockout
   fibroblasts (Fig. [137]4d, e). Indeed, F[1]F[0]-ATP synthase showed a
   60% reduced activity upon TET3 loss, as measured by spectrophotometry
   (Fig. [138]4f). Next, we performed a rescue experiment in which
   overexpressing only the Atp5e subunit in Tet3 knockout fibroblasts was
   almost able to recover the dimer/monomer and the oligomer/monomer ratio
   as well as the F[1]F[0]-ATP synthase activity to wild-type levels
   (Supp. Figures [139]6d, e, f). The overexpression of all the other
   affected subunits might be required to completely recover the
   phenotype. However, this rescue experiment demonstrates that a 1.5-fold
   reduction in Atp5e transcript levels is sufficient to impair ATP
   synthase self-assembly and activity. Mitochondrial cristae formation is
   inseparably linked to the formation of ATP synthase rows of dimers and
   a reduced dimerization ability has been described to result in widened
   cristae^[140]25. Accordingly, a larger cristae area was observed in
   mitochondria from Tet3 knockout fibroblasts in comparison with their
   wild-type counterparts, as shown by electron microscopy (Supp.
   Figure [141]6g). In addition, we have also investigated mitochondrial
   morphology in heart that is considered the best tissue to detect
   mitochondrial functional aberrations^[142]21. Tet3 knockout
   cardiomyocytes exhibit reduced 5hmC levels as well as smaller
   mitochondria, fewer developed cristae and wider cristae than their
   wild-type homologs (Supp. Figure [143]6h, k). Furthermore, we evaluated
   whether an ATP synthase activity reduction might contribute to
   mitochondrial hyperpolarization. The inner mitochondrial membrane
   potential was significantly increased in the absence of TET3
   (Fig. [144]4g and Supp. Figures [145]6i, j), as previously described
   upon ATP synthase assembly impairment^[146]26. The mitochondrial mass,
   the citrate synthase activity and the cellular respiration status
   remained unaffected (Fig. [147]4g–i). Overall, our data demonstrate
   that TET3 loss-of-function leads to a reduction in the F[1]F[0]-ATP
   synthase self-assembly and activity.

Fig. 4. TET3 regulates mitochondrial F[1]F[0]-ATP synthase assembly.

   [148]Fig. 4
   [149]Open in a new tab

   a Blue-Native gel electrophoresis followed by in gel complex I activity
   or western blot with anti-NDUFA9 in isolated mitochondria from
   wild-type and Tet3 knockout fibroblasts. SDHA was used as loading
   control. I, II, III and IV refer to the different mitochondrial
   complexes associated into supercomplexes. Subscript number refers to
   the units of each complex. n = 2 biologically independent replicates,
   each sample corresponds to a pool of two independent biological
   samples. b Mitochondrial complex I NADH: Fe(CN)6 oxidoreductase (left
   panel) and NADH:ubiquinone (middle panel) activities
   spectrophotometrically measured in isolated mitochondria from wild-type
   and Tet3 knockout fibroblasts. Percentage of complex I activation in
   isolated mitochondria (right panel). n = 3 biologically independent
   samples. c Bovine ATP synthase dimer state1:state3 model (PDB ID: 7AJD)
   in which the subunits showing lower expression levels in the absence of
   TET3 have been color-coded. d Clear native gel electrophoresis followed
   by in gel ATPase activity or blue native gel electrophoreses followed
   by western blot with anti-ATPβ from isolated mitochondria in wild-type
   and Tet3 knockout fibroblasts. ATP synthase O, oligomer; D, dimer; M,
   monomer. n = 2 biologically independent replicates, each sample
   corresponds to a pool of two independent biological samples. e
   Dimer/monomer ratio quantification detected by in gel activity in (d).
   n = 2 biologically independent replicates, each sample correspond to a
   pool of two independent biological samples. f F[1]F[0]ATP synthase
   activity measured spectrophotometrically in mitochondria isolated from
   wild-type and Tet3 knockout fibroblasts. n = 2 biological replicates. g
   Mitochondrial membrane potential and mitochondrial mass quantification
   in wild-type and Tet3 knockout fibroblasts, as measured by confocal
   imaging. n = 3 biologically independent samples. a.u., arbitrary units.
   h Citrate synthase activity measured spectrophotometrically in
   wild-type and Tet3 knockout fibroblasts. n = 3 biological replicates. i
   Oxygen consumption rate (OCR)-derived quantification of basal
   respiration in wild-type and Tet3 knockout fibroblasts. n = 6
   biologically independent replicates. Data are presented as mean values
   +/− SD. Student’s t test, two-tailed. Source data are provided as a
   Source Data file.

TET3 loss-of-function leads to hypermethylation of metabolic-related pathways

   TET3 can modulate gene expression through catalytic and non-catalytic
   activities^[150]27,[151]28. To assess whether the downregulation of ATP
   synthase subunits upon TET3 loss was due to TET3—mediated DNA
   demethylation, we examined global DNA methylation patterns in wild-type
   and Tet3 knockout intestinal epithelium by whole-genome bisulfite
   sequencing (WGBS). By mapping DNA methylation at base-pair resolution,
   we observed an increase in methylation levels in all analyzed genomic
   regions in the absence of TET3 (Fig. [152]5a, b).

Fig. 5. TET3 deletion leads to hypermethylation in the intestinal epithelium.

   [153]Fig. 5
   [154]Open in a new tab

   a, b WGBS data represented as 5mC average levels over a different gene
   regions and b gene body regions for all annotated genes in wild-type
   and Tet3 knockout intestinal epithelium. The methylation level has been
   calculated after dividing each region in 20 bins. n = 2 biological
   replicates. c Distribution of hyper and hypo DMRs across different gene
   regions. n = 2 biological replicates. d Distribution of hyper and hypo
   DMRs across different chromosomes. TE, repeat elements. n = 2
   biological replicates. e Heatmap of H3K27ac and H3K4me1 at all
   enhancers identified in P0 intestinal cells. Data from ENCODE. f Line
   plot of H3K27ac and H3K4me1 enrichment at all enhancers. g Heatmap of
   H3K27ac, H3K4me1 and DNA methylation from WGBS in wild-type and Tet3
   knockout intestinal epithelium at hypermethylated enhancers. h DNA
   methylation levels at hypermethylated enhancers in Tet3 knockout
   intestinal epithelial cells. i Percentage of hyperDMRs associated to
   enhancers in P0 intestinal cells. j, k Top 20 gene ontology terms
   enriched in all Tet3 knockout hyperDMRs (j) or only in Tet3 knockout
   hyperDMRs located at the promoter (k). Rich factor indicates the ratio
   of the number of hypermethylated genes in the pathway to the total
   number of genes in that pathway.

   To identify genomic regions displaying changes in 5mC levels, we
   searched the DNA methylome profiles for differentially methylated
   regions (DMRs). Consistent with TET3 catalytic role in DNA
   demethylation, the Tet3 knockout epithelium showed a larger number of
   hypermethylated DMRs (hyperDMRs) than hypomethylated DMRs, and these
   hypermethylated DMRs were equally distributed across all chromosomes
   and gene regions (Fig. [155]5c, d). Ansari and collaborators have
   previously reported that a double Tet2 and Tet3 deletion in the
   intestinal epithelium induces profound epigenetic changes at putative
   enhancer regions^[156]29. To determine if the hyperDMRs identified in
   the Tet3 knockout intestinal epithelium were enriched at enhancer
   regions, we used available ChIPseq data for the enhancer hallmarks
   H3K4me1 and H3K27ac from P0 small intestine to call enhancers^[157]30.
   We identified 16,431 enhancers and only 1497 enhancers (6,7%)
   overlapped with the hyperDMRs identified in the Tet3 knockout
   intestinal epithelium (Fig. [158]5e–i). These regions were associated
   to 543 genes (Supp. Data [159]2). Our results suggest that TET3 alone
   does not seem to play a major role in intestinal epithelial enhancer
   demethylation, and that the enhancer hypermethylation observed upon
   combined loss of TET2 and TET3 in prior studies^[160]29 might be
   largely due to Tet2 deletion, which is known to be enriched at
   enhancers in many cell types. Next, all hyperDMRs as well the hyperDMRs
   located only at promoters were subjected to pathway enrichment analysis
   revealing an association with metabolism-related terms (Fig. [161]5j,
   k), as also observed at the transcriptomic level (Fig. [162]2e, h).
   Interestingly, genes encoding for ATP synthase membrane domain subunits
   that were downregulated upon TET3 loss were not associated to hyperDMR,
   excluding TET3 demethylation activity as the mechanism underlying their
   downregulation in gene expression.

TET3 loss-of-function impairs mitochondrial metabolic function

   Our results in cultured fibroblasts demonstrating that TET3
   loss-of-function leads to an ATP synthase assembly deficiency
   (Fig. [163]4) cannot be validated in the intestinal epithelium due to
   the inability to sort mature enterocytes. Then, we sought to assess the
   metabolic profile of differentiated enterocytes since an ATP synthase
   deficiency will certainly have an impact on cellular metabolism. To
   this end, an untargeted spatial metabolomic analysis was performed
   using matrix-assisted laser desorption/ionization mass spectrometry
   imaging (MALDI-MSI) to determine the relative ion intensities and
   spatial location of metabolites in the intestinal epithelium^[164]31.
   In total, 52 metabolites were annotated (Supp. Table [165]2). The base
   and the tip of the villi, where the proliferative and the
   differentiated compartments are respectively located, were separately
   analyzed in both the wild-type and the Tet3 knockout intestinal
   epithelium (Fig. [166]6a). Three metabolites from the tricarboxylic
   acid (TCA) cycle were detected, showing non-significant differences in
   the intestinal stem cells but a significant increase in Tet3 knockout
   differentiated enterocytes (Fig. [167]6b). Unfortunately, only the
   metabolite 3-phosphoglyceric acid could be determined from the
   glycolysis pathway. Again, significant higher levels were observed at
   the villus tip but not at the base in the absence of TET3
   (Fig. [168]6c). The same pattern was followed by nicotinamide adenine
   dinucleotide (NADH) and the NAD+ precursor, nicotinamide ribotide
   (Fig. [169]6d). The redox couple glutathione (GSH) and oxidized
   glutathione (GSSG), which play a relevant role as non-enzymatic
   antioxidants, exhibited a significant increase upon TET3 loss only at
   the tip of the villi but not at the base (Fig. [170]6e). The levels of
   coenzyme A, that is required for fatty acid synthesis and for pyruvate
   oxidation in the TCA cycle, were also augmented only in the Tet3
   knockout enterocytes (Fig. [171]6f). Several metabolites involved in
   hexosamine pathway and carbohydrate biosynthesis were also higher upon
   TET3 loss-of-function (Fig. [172]6g, h). Finally, pyrimidine and purine
   mono- and di-nucleotides were significantly increased in the Tet3
   knockout villi tips but cytidine-, guanosine-, and
   uridine-5’-triphosphate showed no differences between both genotypes
   (Fig. [173]6i). Interestingly, ATP levels are significantly increased
   upon TET3 loss. Collectively, our results demonstrate that the Tet3
   knockout intestinal epithelium exhibits a metabolic profile that
   differs from its wild-type counterpart mainly at the villi tip where
   the differentiated enterocytes are located. Indeed, 39 metabolites were
   found exhibiting significant differences between the wild-type and the
   Tet3 knockout only at the villi tip. On the other hand, only 3
   metabolites showed significant differences only at the base between
   both genotypes.

Fig. 6. TET3 loss-of-function leads to an abnormal metabolic profile.

   [174]Fig. 6
   [175]Open in a new tab

   a Hematoxylin and eosin—stained sections as well as MS-MALDI outlined
   sections for wild-type and Tet3 knockout. Insets in the intestinal
   section images show the location of the magnified villus in which the
   measured regions are indicated. Red and green refer to base and tip,
   respectively. Scale bars, 200 and 50 µm b–i MALDI-MS relative
   quantification of metabolites. Center line refers to the median; box
   limits, upper and lower quartiles; whiskers 1.5x interquartile range;
   points, outliers. n = 4 independent measurements from four different
   histological sections. *p-value < 0,05, **p-value < 0,01 and
   ***p-value < 0,001. Student’s t test, two-tailed. Source data are
   provided as a Source Data file.

   Next, we analyzed the metabolic profile of tail tip—derived fibroblasts
   since this cell-type is used in clinical settings for the diagnosis of
   mitochondrial disorders^[176]21. Here, we could use LC-MS/MS with
   reference standards in order to absolutely quantify selected
   metabolites because spatial information is not required. We targeted
   our analysis on central carbon metabolism based on the molecules
   identified in the MALDI-MS analysis shown above. In the absence of
   TET3, we observed an increase in NADH as well as in NAD+ and its
   precursor NAAD (Supp. Figure [177]7a, c), a result similar to that
   obtained in differentiated enterocytes in which both NADH and
   nicotinamide ribotide were elevated upon TET3 loss. NADPH is required
   for fatty acid synthesis and for maintaining the steady-state level of
   GSH. NADPH is slightly but not significantly increased in the absence
   of TET3 whereas NADP+ is significantly elevated (Supp. Figure [178]7a).
   As described in Tet3 knockout enterocytes, an increment in the levels
   of both GSH and GSSG was observed (Supp. Figure [179]7a). FAD was also
   increased in the absence of TET3 (Supp. Figure [180]7b). Next, we
   observed a significant increment in the levels of almost all
   metabolites from glycolysis (Supp. Figure [181]7d) and a slight but
   significant rise in lactic acid was measured in Tet3 knockout
   fibroblasts (Supp. Figure [182]7f). Furthermore, all the metabolites of
   the pentose phosphate pathway (PPP), which plays a pivotal role in the
   production of cellular NADPH and nucleotide biosynthesis, were elevated
   (Supp. Figure [183]7d). In fact, both pyrimidine and purine nucleotides
   concentrations were increased in the absence of TET3 (Supp.
   Figure [184]7e), thus correlating with the results described in Tet3
   knockout enterocytes. The metabolites involved in hexosamine pathway
   and tricarboxylic acid (TCA) cycle were also higher upon loss of TET3
   function (Supp. Figure [185]7d), as in the Tet3 knockout enterocytes.
   The levels of acetyl-CoA that is required for fatty acid synthesis were
   also augmented in the Tet3 knockout fibroblasts (Supp. Figure [186]7d),
   in line with the elevation in coenzyme A levels detected on Tet3
   knockout enterocytes. In addition, higher levels of several
   glycerophospholipids, which are the major structural lipids in
   eukaryotic membranes^[187]32, such as glycerophosphoetanolamine,
   glycerophosphoserine, glycerophosphoglycerol and
   glycerophosphoinositol, were also detected (Supp. Figure [188]7g).
   Moreover, acetoacetyl CoA that is essential for cholesterol synthesis
   also exhibited a significant increase in Tet3—deleted cells (Supp.
   Figure [189]7h). GDP glucose/mannose, a sugar donor for carbohydrate
   biosynthesis, displayed an increment in the absence of TET3 as well
   (Supp. Figure [190]7i), as in Tet3 knockout enterocytes. Some of the
   enzymes involved in these metabolic pathways show substrate promiscuity
   and produce side products that might be toxic^[191]33. A significant
   2-fold increase of 4-phosphoerythronate, the resulting side product
   generated by glyceraldehyde 3-phosphate dehydrogenase acting on
   erythrose-4-phosphate instead of on glyceradehyde-3-phosphate, was
   measured in Tet3 knockout fibroblasts. Likewise, 2-hydroxyglutarate
   that arises from the noncanonical activity of several dehydrogenases
   also accumulates in Tet3-deleted fibroblasts as well as methyl-malonic
   acid that is a byproduct of propionate metabolism. The accumulation of
   these side products correlates with the observed increase in the
   glycolytic and TCA flux (Supp. Figure [192]7j). Phosphocreatine, which
   is considered an energy buffer^[193]34 as well as ATP were increased in
   Tet3 knockout fibroblasts (Supp. Figures. [194]7e, k). Nicely, ATP was
   also significantly increased in Tet3 knockout enterocytes
   (Fig. [195]6i). Overall, our analysis indicates that Tet3 knockout
   fibroblasts and Tet3 knockout enterocytes located at the villi tip show
   a very similar metabolic profile enriched in anabolic pathways that are
   reminiscent of undifferentiated cells^[196]35. Therefore, our study
   suggests that the ATP synthase deficiency observed in fibroblasts can
   be extrapolated to mature enterocytes since both differentiated cell
   types show a similar downregulation in genes coding for F[1]F[0]-ATP
   synthase membrane domain subunits, an immature mitochondrial morphology
   and a comparable metabolic profile. Overall, our results demonstrate
   that TET3 impairs mitochondrial function and compromises terminal
   differentiation at the metabolic level.

Discussion

   In this study, we have demonstrated that enterocytes require TET3 to
   acquire a physiological cell identity. Tet3 is almost not expressed at
   the stem cell compartment, thus both wild-type and Tet3 knockout
   intestinal stem cells show an identical transcriptional profile and
   mitochondrial morphology. Upon differentiation, Tet3 expression
   increases in the enterocytes as they migrate towards the tip of the
   villi. In the absence of TET3, genes coding for critical ATP synthase
   subunits exhibit reduced gene expression levels leading to an abnormal
   mitochondrial morphology and an aberrant metabolic signature
   specifically in enterocytes located at the tip of the villi, where Tet3
   is expressed at its highest level, but not at the base where it is
   almost not expressed (Fig. [197]3, Fig. [198]6 and Fig. [199]7). These
   results are in line with enterocytes—dependence on oxidative
   phosphorylation increasing as they mature^[200]18. Tet3 knockout
   fibroblasts and Tet3 knockout enterocytes exhibit a similar reduction
   in the expression of genes coding for ATP synthase subunits and a very
   similar metabolic profile. Therefore, we consider that the ATP synthase
   impairment that we have identified in Tet3 knockout fibroblasts can be
   extrapolated to enterocytes (Fig. [201]4). A model summarizing our
   findings has been depicted in Fig. [202]7.

Fig. 7. TET3 regulates terminal cell differentiation at the metabolic level.

   [203]Fig. 7
   [204]Open in a new tab

   Tet3 is almost not expressed at the stem cell compartment, thus both
   wild-type and Tet3 knockout intestinal stem cells show an identical
   transcriptional profile and mitochondrial morphology. Upon
   differentiation, Tet3 expression increases in the enterocytes as they
   migrate towards the tip of the villi. In the absence of TET3, genes
   coding for critical ATP synthase subunits exhibit reduced gene
   expression levels leading to an abnormal mitochondrial morphology and
   an aberrant metabolic signature specifically in enterocytes located at
   the tip of the villi, where Tet3 is expressed at its highest level, but
   not at the base where it is almost not expressed (Fig. 3, Fig. 6 and
   Fig. 7). Finally, a biochemical analysis in fibroblasts identified an
   ATP synthase deficiency assembly in the absence of TET3. As Tet3
   knockout fibroblasts exhibit: i) a reduction in the expression levels
   of the same critical ATP synthase subunits affected in Tet3 knockout
   enterocytes and; ii) a metabolic profile very similar to Tet3 knockout
   enterocytes, we have extrapolated the ATP synthase impairment observed
   in fibroblasts to enterocytes.

   Two recent reports using an intestinal epithelium—specific Tet3 or a
   double Tet2 and Tet3 conditional knockout mouse model concluded that
   TET3 is essential for proper epithelial differentiation but the
   underlying molecular mechanism was not defined^[205]29,[206]36. It has
   also been proposed that TET3 ablation predisposes the intestinal
   epithelium to inflammation through microbiome alterations^[207]29.
   Likewise, another study has predicted TET enzymes to be risk genes in
   ulcerative colitis patients^[208]37. ATP synthase deficiency has been
   described to cause mitochondrial hyperpolarization and, in turn, an
   increment in the production of reactive oxygen species (ROS)^[209]38.
   Thus, our results suggest that the production of reactive species due
   to an ATP synthase assembly defect might be the underlying mechanism
   leading to the inflammation previously reported in the intestine upon
   TET3 loss. Furthermore, a study in the cardiac tissue showed that a
   cardiac-specific deletion of both Tet2 and Tet3 in mice leads to
   ventricular non-compaction cardiomyopathy with embryonic
   lethality^[210]11. Our findings might help to explain this cardiac
   phenotype since patients exhibiting ATP synthase deficiency due to
   single mutations in other ATP synthase subunits or assembling factors
   such as ATP5F1D or TMEM70 display cardiomyopathies^[211]39,[212]40.
   Even more, we found that Tet3 knockout cardiomyocytes exhibit smaller
   mitochondria with fewer developed cristae than their wild-type
   homologs, supporting a mitochondrial dysfunction as the mechanism
   underlying the cardiomyopathy observed in the absence of Tet2 and Tet3.
   Recently, a human Mendelian disorder caused by the disruption of TET3
   has been reported^[213]41. However, an insight into the disease
   mechanism was not described. We propose that the pathology observed in
   TET3 deficient patients might be due to a compromised F[1]F[0]ATP
   synthase dimerization, as none of the described TET3 mutations induce a
   complete TET3 loss-of-function^[214]41, and point to Tet3 as a new
   candidate gene to be considered for F[1]F[0]-ATP synthase deficiency
   diagnosis.

   TET catalytic loss-of-function strongly associates to tumorigenesis
   since most human cancers exhibit a dramatic loss of 5hmC when compared
   to their normal tissue counterparts^[215]42. Although TET mutations can
   be found in some tumor types such as myeloid malignancies, gliomas and
   a few others, most human cancers do not contain mutations on the TET
   genes or their co-factors, and yet, almost every tumor analyzed to date
   exhibits a clear loss of 5hmC^[216]42. However, it is important to keep
   in mind that TET enzymes can be enzymatically inhibited due, for
   instance, to hypoxia conditions or the presence of oncometabolites.
   Loss-of-function mutations in TET2 are associated with diverse myeloid
   and lymphoid leukemia in humans, however, these mutations per se are
   not sufficient to induce tumorigenesis and a second unknown hit is
   required. Tet2 knockout mice develop lethal myeloid malignancies at
   around 1-year of age^[217]7. Interestingly, a conditional double Tet2
   andTet3 knockout mouse model showed a very aggressive leukemia between
   3 and 7-weeks of age but the molecular mechanism underlying this
   oncogenic transformation was not defined^[218]43. An et al. suggested
   that TET2-depletion leads to an impairment in double-stranded break
   repair mechanism^[219]43. Our findings demonstrate that TET3 loss
   establishes a metabolic program in which oxidative capacity is reduced
   whereas glycolysis-dependent anabolic pathways are enhanced due to
   mitochondria with an immature phenotype. This metabolic signature
   resembles the metabolic profile observed in undifferentiated
   cells^[220]35 and suggests that TET3-depleted cells might be primed to
   malignancy transformation. Indeed, findings on TET loss in mutant mice
   suggest that malignant transformation occurs due to the maintenance of
   a stemness gene expression program^[221]44. Mitochondrial
   hyperpolarization due to compromised F[1]F[0]-ATP synthase dimerization
   and activity can generate an excess of reactive oxygen species and
   induce oxidative damage^[222]26. Hence, we hypothesize that a
   deficiency in double-stranded break repair mechanisms combined with an
   increase in oxidative stress might lead to genomic instability and
   oncogenic transformation in cells that, in addition, display a
   metabolic profile closely related to undifferentiated cells. This
   scenario might explain the extremely fast development of leukemias
   observed in the conditional double Tet2 andTet3 knockout mouse model.
   Collectively, our analysis has revealed the molecular mechanism by
   which TET3 function regulates terminal differentiation at the metabolic
   level in enterocytes and fibroblasts. Future work will need to assess
   whether our findings can be extrapolated to other differentiated
   cell-types expressing Tet3.

Methods

Generation of Tet3 knockout mice

   A targeting vector introducing loxP sites around Tet3 exon 5 was
   constructed and electroporated into (C57BL6/J x C3H/HeJ) F1 embryonic
   stem cells (ESCs). Deletion of the floxed region results in a
   frame-shift that introduces a premature stop codon in exon 6,
   triggering nonsense-mediated decay of the mutant transcript. The
   construct also contained an FRT-flanked selection cassette coding for
   neomycin and thymidine kinase. Individual neomycin-resistant ESC clones
   were screened for recombination in the proper genomic locus. Next,
   FIAU-negative selection was used to select for clones that had excised
   the thymidine kinase cassette after transient transfection of an
   Flp-coding plasmid. Then, a Cre-coding plasmid was transiently
   transfected to excise the loxP-flanked exon 5 and single ESC clones
   were genotyped to confirm the excision. Tet3 knockout ESCs were used to
   generate chimeric mice through aggregation with morula stage (C57BL6/J
   x C3H/HeJ) F1 embryos. The generated male chimeric mice were crossed to
   CD1 females to screen for germ line transmission of the Tet3 knockout
   allele. Finally, Tet3 heterozygous females were backcrossed with pure
   C57BL6/JRccHsd males at least 5 generations to obtain an incipient
   congenic line (N5 or >95% C57BL6/JRccHsd). Genotyping primers are shown
   in Supp. Table [223]3. Mice were time-mated and the presence of a
   vaginal plug was considered as embryonic day (E) 0.5. Experiments were
   performed on E18.5 embryos obtained by caesarian section, as the
   gestation length of this particular mouse line is 19 days. Both male
   and female Tet3 knockout mice exhibit a perinatal lethal phenotype,
   thus sex was not considered in the study design for not being relevant
   in this phenotype. Animals were housed in a controlled environment with
   a 12 h light/dark cycle in open cages at 20-24 °C and 55–65% humidity,
   with access ad libitum to a standard chow diet and water. All animal
   experiments were performed in accordance with the animal care
   guidelines of the European Union (2010/63/EU) prior approval of the
   Spanish National Research Council Ethics Committee and the Valencian
   Regional Government (Approval reference: 2016/VSC/PEA/00187).

RNAscope in situ hybridization and quantification analysis

   The first 2 cm of E18.5 mouse small intestines that include the whole
   duodenum and partial jejunum from wild-type and Tet3 knockout embryos
   were dissected, fixed in formalin and shipped to Charles River
   Laboratories Evreux. Then, the samples were embedded in paraffin,
   sectioned (3–5 µm thick), and stained with RNAscope probes (ACD). All
   images were stained with TET1 (probe 455228-C1), TET2 (probe
   511598-C4), TET3 (probe 505498-C3) and one of the following cell-type
   specific probes: OLFM4 (311838-C2), MUC2 (315458-C2), CHGA (447858-C2)
   or VIL1 (463308-C2). The probe 321818 and the probe 321838 were used as
   positive and negative controls, respectively. The staining was
   performed on a Leica BOND instrument. All samples were scanned at 40x
   on an Akoya PhenoImager (Akoya Biosciences) and the generated qptiff
   files were transferred to PharmaServices team at Indica Labs for
   analysis. The HALO® v3.6.4 (Indica Labs) quantitative image analysis
   software and the HALO® FISH v3.2.3 algorithm was employed to detect and
   count the fluorescent RNAscope probes after cell segmentation.

Immunofluorescence, confocal microscopy and image analysis

   E18.5 mouse small intestines were dissected, fixed in 4%
   paraformaldehyde, embedded in paraffin and sectioned (5 µm). Double
   stainings were performed first for 5hmC detection using the Tyramide
   Superboost kit (ThermoFisher Scientific) followed by stripping and
   second staining for 5mC. Antigen retrieval and antibody stripping was
   performed by incubating sections 25 min in sodium citrate buffer
   10 mM pH 6 in a water bath at 95 °C. Antibodies are listed in Supp.
   Table [224]4. Images were acquired with a SP8 confocal microscope
   (Leica Microsystems) or a LSM980 with Airyscan2 (Zeiss) and processed
   using ImageJ v.1.6.0. Hematoxylin and eosin (H&E) staining was
   performed as previously reported^[225]45. Whole embryo and H&E
   intestine section images were obtained with an Aperio VERSA Scanner
   (Leica Microsystems).

5hmC quantification

   DNA was extracted from epithelial cells sorted from the whole E18.5
   small intestine using the Qiamp DNA Mini kit following the
   manufacturer’s protocol. DNA was hydrolysed for 6 h at 37 °C using DNA
   degradase plus (Zymo Research). Mass spectrometry quantification was
   performed as previously described^[226]46 in the Babraham Institute
   mass spectrometry facility using three individual E18.5 wild-type and
   three individual E18.5 Tet3 knockout embryos.

RT - quantitative PCR

   RNA was extracted from villi isolated from the whole E18.5 small
   intestine using the Qiamp RNeasy micro kit, including the DNAse
   treatment step, following the manufacturer’s protocol. Quantitative PCR
   was performed in technical triplicates on a Quant Studio 5 Real-Time
   PCR System (Applied Biosystems) using Power up Sybr Green
   (Thermofisher) and a primer concentration of 375 nM. Primers are listed
   in Supp. Table [227]3. Actb and Gapdh were used as endogenous controls
   for gene expression analysis. Fold enrichment normalized to wild-type
   samples was calculated using the 2^−ΔΔCt method^[228]47.

Intestine dissociation and cell sorting

   A single cell suspension from whole mouse small intestinal epithelium
   was obtained following a previous reported protocol^[229]48. Briefly,
   the small intestine was dissected from E18.5 embryos, opened
   longitudinally, and placed on ice-cold PBS containing 30 mM EDTA and
   1.5 mM DTT for 20 min on ice. Next, the intestine was transferred to a
   tube containing 30 mM EDTA in PBS, incubated at 37 °C for 10 min and,
   then, vigorously shaken for 30 s to release the intestinal epithelium
   from the lamina propria. After removing the intact intestinal muscle,
   the villi suspension was used for mass spectrometry quantification and
   whole genome bisulfite sequencing or dissociated into single cell for
   fluorescence activated cell sorting. To generate a single cell
   suspension, the villi were incubated at 37 °C for 14 min in 10 ml HBSS
   (with Ca^2+ and Mg^2+) containing 8 mg of dispase II. During the
   incubation, the tube was vigorously shaken for 30 s every 2 min to
   dissociate the villi. Finally, 10% fetal bovine serum and 0.5 mg DNase
   I were added to the cellular suspension that was sequentially passed
   through 100 µm and 40 µm filters for obtaining single cells. The single
   cell suspension was incubated with CD31-PE, CD45-PE, Ter119-PE,
   EpCAM-APC780 fluorescent-labelled antibodies for 30 min. DAPI was used
   for staining dead cells. A BD FACS Aria Fusion (BD Biosciences) was
   used for sorting the live intestinal epithelial population
   (DAPI—negative, CD31-PE—negative, CD45-PE—negative, Ter119-PE—negative
   and EpCAM-APC780—positive cells). For all fluorescent channels,
   positive and negative cells were gated on the basis of fluorescent
   minus one control. Antibodies are listed in Supp. Table [230]4. For
   scRNA-seq, all steps were performed under RNase-free conditions and
   0.5U/µl of RiboLock RNase inhibitor (ThermoFisher Scientific) was added
   to all buffers. Cell sorting was performed in the University of
   Valencia Flow Citometry Unit.

Single cell RNA-seq

   Epithelial cells sorted from the whole E18.5 small intestine were fixed
   in 90% ice-cold methanol and stored at −80 °C until analysis. Tet3
   knockout and wild-type samples correspond to a pool of intestinal
   epithelial cells from two embryos each. 6000 cells per sample were
   analyzed. For analysis, cells were resuspended in 3x saline sodium
   citrate buffer containing 0.04% BSA, 1% SUPERase·In RNAs Inhibitor
   (Invitrogen) and 40 mM DTT^[231]49. Single-indexed libraries were
   prepared using the Chromium Single Cell 3’ v3/v3.1 chemistry (10x
   Genomics) following the manufacturer’s instructions. The generated
   libraries were sequenced on an Illumina Novaseq (Illumina) system using
   the following transcript read lengths: 28 bp for cell barcode and UMI,
   8 bp for sample index and 91 bp for insert. 100,000 reads per cell were
   sequenced. scRNA-seq analysis was performed at Princess Margaret
   Genomics Centre (Toronto, Canada).

Single cell RNA-seq data analysis

   For scRNA-seq data processing, the alevin pipeline integrated with the
   salmon software (version salmon-0.14.1) was used, performing cell
   barcode (CB) detection, read mapping, unique molecular identifier (UMI)
   deduplication, gene count estimation and cell barcode
   whitelisting^[232]50. Reads were aligned and annotated to the mouse
   GRCm38 reference genome, transcriptome and annotation, downloaded from
   GENOME website (version 21). Alevin’s forceCells option was set to 6000
   for both wild-type and Tet3 knockout populations to specify the number
   of CBs to consider for whitelisting. Standard procedures for filtering,
   variable gene selection, dimension reduction and clustering were
   performed using the Scanpy toolkit (version 1.7.1)^[233]51. Cells with
   fewer than 200 detected genes were removed from the analysis as well as
   cells with a mitochondrial versus all gene fraction greater than 0.15.
   Genes detected in more than 3 cells were considered as expressed, while
   genes with more than 100,000 total counts were excluded. The data
   matrix was total-count normalized to 10,000 reads per cell and
   stabilized by computing the natural logarithm of one plus the counts
   for each cell. Highly variable genes were filtered for feature
   selection, remaining a data matrix of 11871 cells (5933 WT and 5983
   Tet3 knockout) and 1350 genes. Each gene was scaled to unit variance
   and values exceeding standard deviation 10 were clipped. For
   dimensional reduction and clustering, the t-Distributed Stochastic
   Neighbor Embedding (t-SNE)^[234]52 and the Louvain graph-clustering
   method^[235]21 were used. The ranking for the characterizing genes of
   each cluster was computed using the Wilcoxon test with
   Benjamini-Hochberg correction method. For trajectory inference, a
   partition-based graph abstraction (PAGA) map preserving the global
   topology of data was performed^[236]51 and single-cell graphs using
   ‘fa’ (ForceAtlas2)^[237]53 layout and PAGA initialization were made,
   one for the Louvain cluster groups, and another for the wild-type and
   Tet3 knockout groups. Gene ontology analysis was performed using the
   g:Profiler web server with default parameters^[238]54 using the top 100
   gene markers for wild-type or Tet3 knockout specific clusters. Gene Set
   Enrichment Analysis (GSEA) was performed with GSEA v.4.1.0 desktop
   software using the GSEA Preranked tool with
   SIGN(log2FC)*(-log10(p-value)) as ranking parameter for all detected
   genes, where SIGN is the sign operator and FC the Fold Change, and
   c2.cp.kegg.v7.4.symbols.gmt as enrichment gene set^[239]55.

Transmission Electron Microscopy

   Small intestines were obtained from Tet3 knockout and wild-type
   18.5 d.p.c. embryos by caesarian section. The first two centimeters of
   the small intestine that include the entire duodenum and jejunum were
   immediately fixed overnight at 4 °C by immersion in 0.1 M phosphate
   buffer pH 7.4, containing 4% paraformaldehyde and 2% glutaraldehyde,
   followed by postfixation with 1% osmium tetroxide- 0,8% potassium
   ferrocyanide mixture in water for 1 hour at 4 °C. Samples were then
   dehydrated with ethanol and propylene oxide and embedded in Epon resin.
   Semi-thin sections were obtained from the resin blocks and stained with
   toluidine blue to identify regions of interest in intestine and heart.
   Ultra-thin sections were stained with lead citrate and examined in a
   HT780 Hitachi electron microscope. Sample preparation and imaging were
   generated in the University of Valencia Electron Microscopy Service.
   Images were analyzed with ImageJ v.1.6.0.

Mitochondria isolation

   Isolation of mitochondria from E18.5 tail tip—derived fibroblasts was
   performed from 1–3 150 mm plates, according to the differential
   centrifugation method. Briefly, cells were resuspended in medium A
   (0.32 M sucrose, 1 mM EDTA, 10 mM Tris-HCl, pH 7.4) and homogenized in
   a Teflon potter-type tissue homogenizer. The mixture was centrifuged at
   1000 g for 5 minutes at 4 °C and the generated supernatant was
   subsequently centrifuged at 12000 g for 12 minutes at 4 °C. Finally,
   the pellet was resuspended in medium B (25 mM sucrose, 75 mM sorbitol,
   100 mM KCl, 0,05 mM EDTA, 5 mM MgCl[2], 10 mM Tris-HCl pH 7,4 and 10 mM
   H[3]PO[4], final pH=7,4) and stored at −80 °C.

Mitochondria solubilization

   Solubilization of mitochondrial membranes was carried out with
   digitonin in order to visualize both respiratory supercomplexes and
   free form complexes. The ratio of grams of detergent: grams of
   mitochondrial protein used was 4:1, except in gel where F[1]F[0]ATP
   synthase was visualized, in which the ratio was lowered to 2.5:1.
   100 μg of purified mitochondria resuspended at 10 μg/μl (50 mM NaCl
   buffer, 50 mM imidazole and 5 mM aminocaproic acid) and containing
   digitonin at the desired concentration were incubated 10 minutes on
   ice. The insoluble portion was removed by centrifugation at 13000 g for
   30 minutes at 4 °C. The pellet obtained was discarded and the
   supernatant was mixed with 4x loading buffer (5% Coomassie Blue-G250 in
   1 M aminocaproic acid) for gel loading.

Blue Native-PAGE and Clear Native-PAGE

   3–13% gradient polyacrylamide gels were prepared in house with a
   gradient former. Clear native (CN) gels were essentially the same as
   blue native (BN) gels, but 0.01% of digitonin was added to all gel
   solutions and Ponceau red buffer (Ponceau red, glycerol) was used
   instead of the normal loading buffer. The amount of sample loaded in
   each well was that obtained from solubilization of 100 μg of
   mitochondria. Cathode buffer A (tricine 50 mM, bis-tris 15 mM, pH 7.0,
   Coomassie Blue G-250 0.02%) and cathode buffer B (tricine 50 mM,
   bis-tris 15 mM, pH 7.0, Coomassie Blue G-250 0.002%) were used for
   electrophoresis. Electrophoresis was performed in a cold chamber. The
   first half hour was run at 90 volts with cathode buffer A. After that
   time, the cathode buffer was exchanged for cathode buffer B.
   Electrophoresis continued for approximately one more hour at 300 volts,
   until the dye began to run out of the bottom of the gel. In CN gels the
   whole electrophoresis was performed in cathode buffer B, with no
   changes.

In-gel complex I activity

   Measurement of NADH dehydrogenase activity of complex I was determined
   on the same gel after BN-PAGE electrophoresis. The gel was incubated in
   0.1 M Tris-HCl, pH 7.4, 0.14 mM NADH and 1 mg/ml NitroBlue tetrazolium
   solution at room temperature. Visualization was achieved by the purple
   precipitate produced after reduction of NitroBlue tetrazolium by the
   NADH dehydrogenase activity of complex I.

In-gel F[1]F[0]ATP synthase activity

   Measurement of NADH dehydrogenase activity of F[1]F[0]ATP synthase was
   determined on the same gel after CN-PAGE electrophoresis. The gel was
   incubated in 270 mM glycine, 35 mMTris, pH 8.4 for 3 hours, then ATP
   hydrolysis was assayed in 270 mM glycine, 35 mMTris, 4 mM ATP, 14
   mMMgSO4, 0.2%Pb(NO3)2, pH 8.4. The reaction was stopped after 5–10 min
   by 30 min incubation in 50% methanol/50% water after the formation of
   phosphate lead precipitates. The gel was then scanned.

Spectrophotometric activities

   Measurement of complex I activity involved a two-step measurement, as
   described previously^[240]56. First step was performed in 25 μg of 3x
   frozen-thawed mitochondria solubilized in Mg^2+-containing C1/C2
   buffer, DecylCoQ 130 μM and Antimycin A 1 μM. Absorbance measurement at
   37 °C, 340 nm started after addition of NADH 100 μM and lasted 2–4 min.
   Then, rotenone 1 μM was added and absorbance was again measured for
   2–4 min. Next, 1 mM Fe(CN)[6] was added and measurement was performed
   at 420 nm at 37° for 2–4 min. Later, Diphenyleneiodonium chloride (DPI)
   was added and absorbance was again captured in the same conditions.
   F[1]F[0]ATP synthase activity was measured as a readout of its ATPase
   activity. As previously reported^[241]57, oligomycin-sensitive ATPase
   activity was calculated after measuring changes in absorbance at 340 nm
   driven by the pyruvate kinase reaction coupled to ADP phosphorylation
   by F[1]F[0]ATP synthase.

Analysis of mitochondrial membrane potential

   For mitochondrial membrane potential imaging, E18.5 tail tip—derived
   wild-type and Tet3 knockout fibroblasts were seeded onto 96-well plates
   (PerkinElmer) and mitochondrial inhibitors were added at Krebs Ringer
   Phosphate Glucose buffer (KRPG) supplemented with NaCl 145 mM;
   Na[2]HPO[4] 5,7 mM; KCl 4,86 mM; CaCl[2] 0,54 mM; MgSO[4] 1,22 mM;
   glucose 20 mM; pH 7,35. Next, cells were loaded with 10 nM TMRM
   (Sigma-Aldrich) and 1 µM cyclosporine-H (Sigma-Aldrich) in the Operetta
   CLS microscope (30 min at 37 °C in a 5 % CO[2] atmosphere) and confocal
   images were acquired at 40X, 1.4 NA objective (PerkinElmer).
   Mitochondrial uncoupler carbonyl cyanide
   p-tri-fluoromethoxyphenylhydrazone (FCCP, 10 µM, 15 min, Sigma-Aldrich)
   was added as a control of mitochondrial depolarization. Finally, images
   were analyzed using Harmony software (PerkinElmer).

Citrate synthase activities

   E18.5 tail tip—derived fibroblasts from wild-type and Tet3 knockout
   embryos were snap-frozen. After three cycles of freeze/thawing, to
   ensure cellular disruption, citrate synthase activities were
   determined. Citrate synthase activity was measured in the presence of
   93 mM Tris-HCl, 0.1% (vol/vol), Triton X-100, 0.2 mM acetyl-CoA and
   0.2 mM DTNB; the reaction was started with 0.2 mM oxaloacetate, and the
   absorbance was recorded at 412 nm (30 °C)
   (e = 13.6 mM^−1·cm^−1)^[242]58. The protein concentration of the
   samples was quantified by the BCA protein assay kit (Thermofisher)
   following the manufacturer’s instructions, using BSA as a standard.

Oxygen consumption measurement

   Oxygen consumption assays were performed as previously
   reported^[243]59. Briefly, oxygen consumption rate (OCR) was measured
   using an XF96 Extracellular Flux Analyzer (Seahorse Bioscience). E18.5
   tail tip—derived fibroblasts were plated 1 day before the experiment
   and preincubated with unbuffered DMEM 1 h at 37 °C in an incubator
   without CO[2] regulation. Successive injections of unbuffered DMEM,
   5 μg/mL oligomycin, 300 nM carbonyl cyanide 4-(trifluoromethoxy)
   phenylhydrazone (FCCP) and 1 μM rotenone plus 1 μM antimycin A were
   programmed. Calculations were performed following the manufacturer’s
   instructions.

Whole Genome Bisulfite Sequencing (WGBS) analysis

   5mC levels and genomic location was determined using whole genome
   bisulfite sequencing on Tet3 knockout and wild-type E18.5 entire small
   intestinal epithelium. Two lines of each genotype were analyzed.
   Briefly, genomic DNA spiked with lambda DNA was fragmented to
   200–400 bp. The DNA fragments were bisulfite treated using EZ DNA
   Methylation Gold kit (Zymo Research) to generate single strand DNA.
   Methylation sequencing adapters were ligated, followed by double strand
   DNA synthesis. Libraries were sequenced on an Illumina NovaSeq X Plus
   Series (PE150). Quality of the raw sequence reads was assessed with
   FastQC (fastqc_v0.11.8). Bismark software (version 0.24.0)^[244]60 was
   used to perform alignments of bisulfite-treated reads to the reference
   genome. The reference genome was firstly transformed into a
   bisulfite-converted version and then indexed using bowtie2^[245]61.
   Sequence reads that produce a unique best alignment from the two
   alignment processes (original top and bottom strand) were then compared
   to the normal genomic sequence and the methylation state of all
   cytosine positions in the read was inferred. The same reads that
   aligned to the same regions of genome were regarded as duplicated ones.
   The sequences were divided into 10 kb bins to calculate their
   methylation level. The sum of methylated and unmethylated read counts
   in each window was calculated. Differentially methylated regions (DMRs)
   were identified using the DSS software^[246]62. According to the
   distribution of DMRs through the genome, genes were related to DMRs
   when their gene body region (from TSS to TES) or promoter region
   (upstream 2 kb from the TSS) have an overlap with the DMRs. Gene
   Ontology (GO) enrichment analysis of genes related to DMRs was
   implemented by the GOseq R package^[247]63, in which gene length bias
   was corrected. GO terms with corrected p-value less than 0.05 were
   considered significantly enriched by DMR-related genes KOBAS
   software^[248]64 was used to test the statistical enrichment of DMR-
   related genes in KEGG pathways. The WGBS and subsequent analysis were
   performed by Novogene. To identify putative enhancers on
   hypermethylated DMRs, H3K27ac and H3K4me1 ChIPseq data performed on
   murine intestinal cells at postnatal day 0 (P0) were downloaded from
   ENCODE (experiments: ENCSR642VYW and ENCSR159RVN) and used to identify
   enhancers in P0 intestinal cells (ENCFF660XDP.bed and ENCFF349IDR.bw
   for H3K27ac; ENCFF523POG.bed and ENCFF559NPS.bw for H3K4me1). Bedtools
   intersect was used to identify overlapping H3K27ac and H3K4me1 peaks
   with default parameters. Line plots and heatmaps were generated using
   deepTools.

MALDI-MSI measurement

   All chemicals were purchased from Merck, if not otherwise mentioned.
   The first 2 cm of E18.5 mouse small intestines that include the whole
   duodenum and partial jejunum from wild-type and Tet3 knockout embryos
   were dissected and fresh frozen in liquid nitrogen. Then, the samples
   were embedded in 7.5%HPMC-2.5%PVP^[249]65 and cryo-sectioned for
   MALDI-MS imaging as described elsewhere^[250]66. Briefly, tissue
   sections were collected at 20 µm thickness and thaw mounted on
   Intellislides purchased from Bruker Daltonics. Slides were put in
   plastic mailers, vacuum sealed and stored at −80 °C until use. For
   MALDI MSI, a DAN-HCl matrix solution was prepared: 10 mg
   1,5-diaminonaphtalene (DAN), in 5 ml of solvent: 60% acetonitrile (ACN)
   in milliQ water (VWR) and 60 mM HCl. The matrix was sprayed using the
   HTX-M5 sprayer (HTX Technologies) with the following parameters: nozzle
   temperature 65 °C, passes 6, flow rate 0.07 ml/min, velocity
   1200 mm/min, track spacing 2 mm, pattern CC, pressure 10 psi, drying
   time 10 s, and nozzle height 40 mm. After spraying, red phosphorus
   suspended in acetone was spotted on the slide for mass calibration
   purposes. MALDI MSI data was acquired on a timsTOF fleX system (Bruker
   Daltonics) at 5 µm in negative ion mode using 1,5-diaminonaphtalene
   (DAN) as matrix. MALDI MSI data was acquired on a tims-TOF fleX system
   (Bruker Daltonics) equipped with a smartbeam 3D 10 kHz laser and
   microGRID, controlled by TimsControl v4.1 and flexImaging v7.2 software
   (Bruker Daltonics). Data was acquired in negative ion mode, m/z range
   of 80–1300, with 25 laser shots per pixel, 10 kHz laser frequency,
   laser spot size 5 µm^2 and raster 5 µm. The tune parameters were the
   following: MALDI Plate Offset 50 V, Deflection 1 Delta −70 V, Funnel 1
   RF 180 Vpp, isCID Energy 0 eV, Funnel 2 RF 180 Vpp, Multipole RF 180
   Vpp, Collision Energy 3 eV, Collision RF 800 Vpp, Quadrupole Ion Energy
   5 eV and Low Mass 70 m/z, Focus Pre TOF Transfer Time 80 µs and Pre
   Pulse Storage 6 µs, and Detection set to Focus Mode. After MALDI
   measurement, the matrix was removed by dipping the slide for 2 min in
   100% methanol. An automated staining apparatus, SunTissuePrep
   (SunChrome) was used for hematoxylin and eosin staining. A coverslip
   was mounted on top using Eukitt as quick-hardening mounting medium.
   Data was imported into SCiLS Lab MVS, Version 2024b Pro, from Bruker
   Daltonics, using the standard import parameters. High resolution scan
   of the H&E stained slide was imported as “.svs” file using the Import
   Optical Image tool. The overview images and the high resolution scan
   were co-registered using three anchor points. For untargeted analysis,
   a first feature selection was done by setting an intensity threshold of
   30.000 a.u. using the Sliding Window tool. A list of 2926 features was
   selected and used for bisecting k-means clustering in order to
   automatically separate the tissue area from the background area. A
   second feature list was generated using the Find Discriminating
   Features (ROC) tool with the threshold set to 0.65. A new list of 829
   features was used for further analysis. Bisecting k-means segmentation
   using the new feature list, weak de-noising and correlation distance
   metric generated four regions of interest: red = muscle, yellow = base
   of the villi, green = center of the villi and purple/blue = tip of the
   villi. The villi were manually annotated based on the H&E staining scan
   and the segmentation image. The base and tip of four villi from each
   condition (WT and KO) were selected for statistical analysis.
   Annotation of features was done using [251]HMDB database with 10 ppm
   mass error tolerance, adduct and isotope pattern matching through
   MetaboScape API inside SCiLS. All data was normalized by
   root-mean-square (RMS).

Cell culture

   E18.5 tail tip—derived fibroblasts were cultured on gelatin using the
   following culture media: Knockout DMEM, 10% bovine fetal serum, 1X
   penicillin/streptomycin, 1X non-essential amino acids, 2 mM Glutamine
   and 0,1 mM β-Mercaptoethanol. Atp5e rescue experiments were performed
   on fibroblasts derived from E13.5 embryos and cultured as described
   above. Tet3 knockout and wild-type fibroblasts were transduced with
   defective ecotropic retroviral particles coding for Atp5e or empty. The
   retroviral vector pMSCV-puro (Clontech) and the packaging pCL-Eco
   plasmid (Addgene #12371)^[252]67 were used to generate the retroviral
   particles. After transduction, fibroblasts were selected with 1 μg/mL
   puromycin during 5 days. All experiments were performed using
   fibroblasts at passage 3.

Metabolomic analysis

   Cells were fixed with 80% methanol, at 100 μL per 1 million cells.
   Cells were lysed with the aid of two metal balls at 30 Hz on a MM 400
   mill mixer for 3 min, followed by centrifugal clarification at 21,000 g
   for 10 min. The clear supernatants were collected for the following
   assays. The precipitated pellets were used for protein assay using a
   standardized BCA procedure. For NAD and nucleotide quantification, an
   internal standard (IS) solution containing ^13C or ^2H-labeled NAD,
   NADH, AMP and ATP was prepared in 50% methanol. Serially diluted
   standard solutions of the targeted metabolites were prepared in 80%
   methanol. 10 μL of the clear supernatant of each sample or each
   standard solution was mixed with 40 μL of IS solution. 10 μL aliquots
   of the resultant solutions were injected into a C18 column to run
   UPLC-MRM/MS with (−) ion detection on a Waters Acquity UPLC system
   coupled to a Sciex QTRAP 6500 Plus MS instrument, using a tributylamine
   buffer solution and acetonitrile as the mobile phase for binary-solvent
   gradient elution. For GSH and GSSG quantification, an internal standard
   (IS) solution containing isotope-labeled GSH and GSSG was prepared in
   water. Serially diluted standard solutions of GSH and GSSG were
   prepared in 80% methanol. 20 μL of the clear supernatant of each sample
   or each standard solution was mixed with 80 μL of the IS solution.
   10 μL aliquots of the resultant solutions were injected into a HILIC
   column to run UPLC-MRM/MS with (+) ion detection on an Agilent 1290
   UHPLC system coupled to an Agilent 6495B QQQ MS instrument with the use
   of 0.1% formic acid in water and in acetonitrile as the mobile phase
   for binary-solvent gradient elution. For analysis of glucose,
   glucose-6P and mannose-6P, 50 μL of the supernatant of each sample was
   mixed with 50 μL of a solution of ^13C6-glucose as internal standard,
   100 μL of 25 mM AEC solution and 20 μL of acetic acid. The mixture was
   allowed to react at 60 °C for 70 min. After reaction, 300 µL of water
   and 300 µL of dichloroform was added. The mixture was vortex mixed and
   then centrifuged. The supernant was injected in 10-μL aliquots to run
   UPLC-MRM/MS with positive-ion detection on an Agilent 1290 UHPLC system
   coupled to a 4000 QTRAP mass spectrometer, as previously
   described^[253]68. The analysis of the TCA cycle carboxylic acids was
   carried out as previously described^[254]69. Briefly, 20 μL of the
   supernatant was mixed with 20 μL of a D- or 13-labeled analogue for
   each analyte as the internal standard, 20 μL of 200 mM 3-NPH solution
   and 20 μL of 150 EDC-6% pyridine solution. The mixture was allowed to
   react at 30 °C for 40 min. After reaction, the solution was diluted 3
   fold with water. 20 μL was injected to run UPLC-MRM/MS with
   negative-ion detection on an Agilent 1290 UHPLC system coupled to a
   4000 QTRAP mass spectrometer. Finally, quantification of other
   metabolites were performed as follows. 50 µL of each supernatant was
   mixed with 50 µL of water containing 50 pmol/mL GTP-13C10 and 50 µL of
   chloroform. The mixture was vortex mixed for 1 min, followed by
   centrifugation. 50 µL of each clear supernatant was taken out and mixed
   with an equal volume of water. 10 µL was injected to run UPLC-MRM/MS
   with negative-ion detection using a custom-developed reversed-phase
   LC-MRM/MS method for binary-solvent gradient elution. Concentrations of
   the detected analytes in the above assays were calculated with internal
   standard calibration by interpolating the constructed linear regression
   curves of individual compounds. The metabolomic analysis was performed
   in UVic Genome BC Proteomics Centre (Canada).

Statistical analysis

   Data, unless otherwise stated, is represented as mean ± standard
   deviation. Two-tailed t-test was used to assess significant differences
   between groups.

Reporting summary

   Further information on research design is available in the [255]Nature
   Portfolio Reporting Summary linked to this article.

Supplementary information

   [256]Supplementary Information^ (5.6MB, pdf)
   [257]Peer Review File^ (5.3MB, pdf)
   [258]41467_2024_54044_MOESM3_ESM.pdf^ (36.8KB, pdf)

   Description of Additonal Supplementary Files
   [259]Supplementary Data 1^ (22.4KB, xlsx)
   [260]Supplementary Data 2^ (16KB, xlsx)
   [261]Reporting Summary^ (3.5MB, pdf)

Source data

   [262]Source Data^ (1.5MB, xlsx)

Acknowledgements