Methylglyoxal couples metabolic and translational control of Notch signalling in mammalian neural stem cells
Nature Communications volume 11, Article number: 2018 (2020) Cite this article
Gene regulation and metabolism are two fundamental processes that coordinate the self-renewal and differentiation of neural precursor cells (NPCs) in the developing mammalian brain. However, little is known about how metabolic signals instruct gene expression to control NPC homeostasis. Here, we show that methylglyoxal, a glycolytic intermediate metabolite, modulates Notch signalling to regulate NPC fate decision. We find that increased methylglyoxal suppresses the translation of Notch1 receptor mRNA in mouse and human NPCs, which is mediated by binding of the glycolytic enzyme GAPDH to an AU-rich region within Notch1 3ʹUTR. Interestingly, methylglyoxal inhibits the enzymatic activity of GAPDH and engages it as an RNA-binding protein to suppress Notch1 translation. Reducing GAPDH levels or restoring Notch signalling rescues methylglyoxal-induced NPC depletion and premature differentiation in the developing mouse cortex. Taken together, our data indicates that methylglyoxal couples the metabolic and translational control of Notch signalling to control NPC homeostasis.
During the development of the mammalian brain, neural precursor cells (NPCs) self-renew and differentiate to give rise to appropriate numbers of neurons 1 . Key to this balance of self-renewal and differentiation is the crosstalk between gene expression and metabolism, two fundamental processes that co-ordinate NPC fate decision 2 , 3 , 4 . The expression of metabolic genes is known to be tightly controlled to support the metabolic shift during NPC differentiation. For example, activation of Notch signalling in NPCs induces the expression of pro-proliferative genes (e.g. basic helix-loop-helix transcription factor Hes1) to maintain the self-renewal of NPCs while at the same time upregulating the expression of glycolytic enzyme genes (e.g. hexokinase 2 and lactate dehydrogenase) that are required by NPCs for energy production and biosynthesis 5 , 6 . On the other hand, during neurogenesis NPCs express proneural genes to induce differentiation and downregulate the expression of glycolytic enzymes to support the metabolic transition from glycolysis to mitochondrial oxidative phosphorylation as the primary energy source of neurons 5 , 6 . Accumulating evidence suggests that NPC metabolism is not a simple adaptation to different cellular states but instead plays a more direct role in regulating self-renewal and differentiation. Although NPCs rely on glycolysis, their mitochondria exhibit an elongated morphology and are functional, with a forced metabolic switch to mitochondrial oxidative phosphorylation enhancing their differentiation 7 , 8 . These observations reveal the reciprocal nature of the relationship between metabolism and gene expression critical for NPC fate decision. However, while the pathways that regulate metabolic gene expression are well described, the signals and mechanisms that mediate the metabolic feedback control of gene expression for proper NPC fate decision remain poorly understood.
Glycolysis produces a wealth of metabolites. How do these metabolic cues instruct gene expression in NPCs? One mechanism used in the adult murine NPCs is mediated by a cyclic AMP-responsive element-binding protein (CREB)-dependent pathway that senses the level of glucose to direct Hes1 transcription and thereby the self-renewal of NPCs 9 . A second mechanism may involve glycolytic enzymes acting as RNA-binding proteins (RBPs) to regulate target mRNAs post-transcriptionally 10 , 11 , 12 . For example, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is a key glycolytic enzyme that catalyzes the conversion of glyceraldehyde-3-phosphate (G3P) into 1, 3-bisphosphoglycerate (1, 3-BPG) 13 . Interestingly, GAPDH can bind to the AU-rich element in the 3ʹ untranslated region (3ʹUTR) of mRNAs and subsequently alter their stability and translation 14 . This dual function of GAPDH is best described in immune cells. In T cells where oxidative phosphorylation serves as the primary energy source, GAPDH functions as an RBP to repress the translation of the interferon γ mRNA 10 . When T cells are activated and switch from oxidative phosphorylation to glycolysis, GAPDH is re-engaged in the glycolytic pathway and no longer represses the translation of interferon-γ mRNA 10 .
What controls the functional switch of metabolic enzymes is still largely unknown. One means of switching may involve feedback or feedforward control of their enzymatic activities by post-translational modifications with intermediate metabolites 15 , 16 . For example, methylglyoxal, an intermediate metabolite produced from G3P during glycolysis modifies GAPDH in a non-enzymatic manner, leading to inhibition of its enzymatic activities 17 . The competitive binding between the enzyme cofactor nicotinamide adenine dinucleotide (NAD) and RNA to the same domain on GAPDH suggests that its compromised activity for glycolysis may otherwise promote its engagement as an RBP to regulate target mRNAs 18 , 19 .
We have recently found that an increase in methylglyoxal levels depletes NPC numbers in the developing mouse cortex 20 , raising the possibility that methylglyoxal may serve as a metabolic signal to regulate specific genes for NPC homeostasis by modulating RNA-binding enzymes such as GAPDH. Here, we show that methylglyoxal induces feedback regulation of Notch signalling in NPCs by engaging GAPDH as an RBP. An increase in methylglyoxal levels reduces the enzymatic activity of GAPDH and promotes its binding to Notch1 mRNA in NPCs. This leads to the translational repression of Notch1 mRNA and a reduction in Notch signalling, ultimately causing premature neurogenesis. This study provides a mechanistic link for the metabolic regulation of gene expression in NPC homeostasis.
Excessive methylglyoxal depletes neural precursors
We have previously shown that methylglyoxal-metabolizing enzyme glyoxalase 1 (Glo1) maintains NPC homeostasis, thereby preventing premature neurogenesis in the developing murine cortex 20 . To determine whether Glo1 controls NPC differentiation by enzymatically modulating methylglyoxal, we initially assessed methylglyoxal-adduct levels in NPCs and neurons in the cortex 21 , 22 . Immunostaining of embryonic day 16.5 (E16.5) cortical sections for a major methylglyoxal-adduct MG-H1 showed only weak immunoreactivity in the cytoplasm of Pax6+ radial precursors in the ventricular and subventricular zones (VZ/SVZ) (Fig. 1a, b , Supplementary Fig. 1a ). MG-H1 production was gradually increased in newborn neurons migrating in the intermediate zone (IZ) and became highly enriched in the cortical plate (CP), where it accumulated in the nuclei of neurons expressing neuronal markers βIII-tubulin (cytoplasmic) and Brn1 (nuclear) (Fig. 1a, b , Supplementary Fig. 1a ). The gradual increase in methylglyoxal levels from NPCs to neurons was consistent with a previous study 23 and is in agreement with the higher expression level of Glo1 in NPCs than in neurons 20 . We next manipulated Glo1 enzymatic activity using S-p-bromobenzylglutathione diethyl ester (BBGD), a cell-permeable and reversible Glo1 inhibitor 24 . As expected, upon incubation with BBGD, methylglyoxal levels were significantly elevated in isolated E13.5 cortical tissues (Fig. 1c ). We then injected BBGD into the lateral ventricle at E13.5 followed by in utero electroporation of a plasmid encoding nuclear EGFP to label and track NPCs and the neurons they give rise to. The reversible effect of BBGD allows the manipulation of NPCs adjacent to the lateral ventricle, with a minimal impact on migrating newborn neurons in the IZ. Cortical sections were immunostained for EGFP and cell-type-specific markers three days after treatment. We found that BBGD exposure led to a reduction of EGFP+ cells in the VZ/SVZ (Fig. 1d, e ). In contrast, the proportion of EGFP+ cells in the CP was increased, with no change in proportions in the IZ (Fig. 1d, e ). In line with the altered cell distribution, we found fewer EGFP+ cells that also expressed the radial precursor marker Pax6, and more EGFP+ cells expressing the neuronal marker Satb2 in cortices exposed to BBGD (Fig. 1f, g ). To further confirm that the alterations of NPCs were due to an aberrant increase in methylglyoxal, we labeled proliferating NPCs with bromodeoxyuridine (BrdU) followed by injection of PBS or methylglyoxal into the lateral ventricle. Two days later, we found fewer BrdU+ cells in the VZ/SVZ as well as fewer BrdU+, Pax6+ radial precursors in the cortex received methylglyoxal (Supplementary Fig. 1b–g ). These results indicate that an increase in methylglyoxal shifts the balance of NPC homeostasis towards neurogenic differentiation.
Fig. 1: Inhibition of Glo1 perturbs the maintenance of NPCs in the developing cortex.
a, b E16.5 coronal cortical sections immunostained for MG-H1 (red). In a, sections were costained for Pax6 (green), and the ventricular surface and cortical boundaries between VZ/SVZ, IZ and CP are labelled with dotted white lines. b High-magnification images of cells in the VZ, IZ and CP from sections as in a. n = 2 experiments. c Relative methylglyoxal levels measured in cortical tissues isolated from the E13.5 cortex (fresh) or treated with BBGD or DMSO vehicle alone. n = 3 experiments. d–g BBGD or DMSO was injected into lateral ventricles of E13.5 cortices, followed by the electroporation with EGFP. Coronal cortical sections were analyzed 3 days later. Sections were immunostained for EGFP (green, d, f), and the relative location of EGFP+ cells was quantified (e). n = 4 embryos each. f Images of the VZ or CP of electroporated sections that were costained for Pax6 (red, left) or Satb2 (red, right). Arrows denote double-labelled cells. g Quantification of sections as in f for the proportion of EGFP+ cells that were also positive for Pax6 or Satb2. n = 4 embryos each. Sections were counterstained with Hoechst 33258 (blue, a, b, d and f). LV lateral ventricle, VZ ventricular zone, SVZ subventricular zone, IZ intermediate zone, CP cortical plate. Scale bars, 50 μm in a and d, 10 mm in f and 5 mm in b. Data are presented as mean values ± SEM and analyzed using two-tailed, unpaired students t-test with Bonferroni correction. **p < 0.01, ns = p > 0.05. Source data and p-values are provided as a “Source Data file”.
Full size image
Methylglyoxal regulates Notch signalling to affect NPCs
Notch signalling plays a crucial role in stem-cell maintenance 5 . Perturbing components of Notch signalling leads to NPC depletion and aberrant neurogenesis characterized by NPC apical detachment and mislocalization in the developing cortex 25 , which is phenocopied by Glo1 knockdown 20 . Therefore, we asked whether methylglyoxal alters Notch signalling in NPCs, using a Notch signalling reporter that contains the responsive element of the canonical Notch effector c-promoter binding factor 1 (CBF1) upstream of the EGFP gene (CBFRE-EGFP) 25 , 26 . We co-electroporated the Notch signalling reporter with control or Glo1 shRNA plasmids into E13.5 cortices and a plasmid encoding DsRed2 driven by the constitutive CMV promoter, used as an internal transfection control. After two days, almost 70% of DsRed2+ cells in the VZ/SVZ of the control cortex expressed EGFP, indicating active Notch signalling in the transfected cells (Fig. 2a, b ). However, following Glo1 knockdown, this number was reduced to less than 40%, suggesting that Notch signalling is suppressed by excessive methylglyoxal induced by Glo1 knockdown. To further assess the changes in Notch signalling, we examined the mRNA levels of downstream Notch targets in control and Glo1 knockdown cells. To this end, we co-electroporated control or Glo1 shRNAs with an EGFP plasmid into E13.5 cortices and collected EGFP+ cells 2 days later by fluorescence-activated cell sorting (FACS) for quantitative real-time polymerase chain reaction (qRT-PCR) analysis. The mRNA levels of the Notch1 targets, Hes1, Hey1, and Hey2, were significantly reduced following Glo1 knockdown (Fig. 2c ), suggesting a reduction of Notch signalling. To confirm these findings, we also examined the effect of Glo1 inhibition on Notch signalling in a relatively homogenous NPC population, in vitro cultured human embryonic stem-cell (H9)-derived NPCs (hNPCs). Treatment with BBGD caused a significant increase in methylglyoxal levels (Supplementary Fig. 1h ) and reduced the mRNA levels of HES1, HES2, HES5, and HEY2 genes (Fig. 2d ). The expression of MASH1, which is repressed by the HES protein family, was correspondingly upregulated (Fig. 2d ). Moreover, knockdown of GLO1 in hNPCs caused an increase in intracellular methylglyoxal levels (Supplementary Fig. 2a–d ), accompanied by a reduction in expression of Notch1-responsive genes (Supplementary Fig. 2e ). These results indicate that excessive methylglyoxal reduces Notch signalling in NPCs. Further support for this conclusion came from the direct application of exogenous methylglyoxal to hNPCs, which also increased intracellular methylglyoxal levels (normalized fold change, 10.16 ± 1.80; p < 0.01; n = 3) and suppressed Notch1-responsive genes (Supplementary Fig. 2f ).
Fig. 2: Methylglyoxal perturbs NPC by suppressing Notch signalling.
a, b E13.5 cortices were co-electroporated with CBFRE-EGFP and DsRed2, and control or Glo1 (shGlo1) shRNAs and analyzed two days later. a Images of sections immunostained for EGFP (green) and DsRed2 (red). Arrows denote double-labelled cells, and arrowheads denote cells with reduced Notch signalling. b Quantification of sections as in a for EGFP+, DsRed2+ cells. n = 4 and 5 embryos each. c qRT-PCR analysis of FACS sorted EGFP+ cells from cortices co-electroporated with EGFP and control or Glo1 shRNAs. n = 3 experiments. d qRT-PCR analysis of low passage hNPCs treated with BBGD or DMSO for 48 h. n = 3 experiments. e, f E13.5 cortices were co-electroporated with CBFRE-EGFP and DsRed2 plus empty vector control or a plasmid expressing Glo1 (Glo1) and analyzed two days later. Images (e) and quantification (f) of EGFP (green) and DsRed2 (red) double-positive cells. n = 3 and 5 embryos each. g–j A low dose of plasmids (0.5 µg µl−1) expressing NICD was co-electroporated with Glo1 shRNA and EGFP into E13.5 cortices. Cortical sections were analyzed three days later. g Images of sections immunostained for EGFP (green). Dotted white lines denote the ventricular surface and boundaries between VZ/SVZ, IZ, and CP. h Quantification of the relative localization of EGFP + cells. n = 4 embryos each. i Images of the VZ or CP of electroporated sections that were immunostained for EGFP (green) and Pax6 (red, left panels) or Satb2 (red, right panels). Arrows denote double-labelled cells. j Quantification of the proportion of EGFP+ cells that were also positive for Pax6 or Satb2. n = 4 embryos each. LV lateral ventricle, VZ ventricular zone, SVZ subventricular zone, IZ intermediate zone, CP cortical plate. Scale bars, 25 µm in a, e and g; 10 µm in i. Data are presented as mean values ± SEM and analyzed using two-tailed, unpaired students t-test with Bonferroni correction. **p < 0.01, *p < 0.05, ns = p > 0.05. Source data and p-values are provided as a “Source Data file”.
Full size image
To determine if physiological levels of methylglyoxal regulate Notch signalling, we overexpressed Glo1 in the cortex by in utero electroporation to deplete intracellular methylglyoxal and at the same time, co-electroporated the Notch signalling reporter. The analysis of the cortex after two days showed that Glo1 overexpression significantly increased the number of VZ/SVZ cells that had active Notch signalling (Fig. 2e, f ), in agreement with a concomitant increase in NPC numbers (Supplementary Fig. 2g–j ). Consistently with these results, ectopic expression of human GLO1 in cultured hNPCs reduced methylglyoxal levels (Supplementary Fig. 1h ) and caused an increase in the expression of Notch1-responsive genes (Supplementary Fig. 2k ).
Taken together, our results indicate that methylglyoxal regulates Notch signalling in NPCs and raise the possibility that impaired Notch signalling might account for the depletion of cortical NPCs induced by increased methylglyoxal. If so, then restoring Notch signalling in NPCs might antagonize the effect of methylglyoxal and rescue Glo1 knockdown-induced NPC depletion. Notch signalling is initiated when the Notch1 receptor binds to its ligands, followed by protease cleavage and release of the Notch intracellular domain (NICD) that translocates into the nucleus to activate downstream targets 5 , 26 . To test our hypothesis, we constitutively activated Notch signalling by ectopically expressing NICD in the E13.5 cortex in which Glo1 was also knocked down. After 3 days, our analysis of cortices showed that NICD expression completely reversed the phenotype induced by Glo1 knockdown, resulting in significantly more EGFP+ cells in the VZ/SVZ as well as EGFP + Pax6+ radial precursors and a reduction in Satb2+ neurons (Supplementary Fig. 3a–c ). Given the important role of Notch signalling in NPC maintenance, we wondered whether the phenotypic rescue by NICD was due to an unspecific overactivation of Notch signalling. We therefore titrated the dose of NICD used in electroporation to a level at which NICD itself did not change cortical development (Supplementary Fig. 3d–f ). This low dose of NICD was sufficient to normalize the aberrant distribution of EGFP+ cells in the VZ/SVZ induced by Glo1 knockdown (Fig. 2g, h ) and rescued the proportions of EGFP+, Pax6+ radial precursors and Satb2+ neurons to control levels (Fig. 2i, j , Supplementary Fig. 3g ). These results demonstrate that the downregulation of Notch signalling mediates methylglyoxal-induced NPC depletion and premature neurogenesis. Interestingly, NICD expression did not rescue the reduction of EGFP+ cells in the CP (Fig. 2g, h , Supplementary Fig. 3a, b ), suggesting that methylglyoxal regulates neuronal migration independent of Notch signalling.
Methylglyoxal regulates translation of Notch1 mRNA
Our data suggest that methylglyoxal may target-specific component(s) of the Notch pathway to regulate NPC homeostasis. Therefore, we asked whether one target could be the Notch1 receptor itself since it initiates the signalling cascade and is under extensive regulation for precise cell fate decision 27 , 28 , 29 , 30 , 31 , 32 . To test whether methylglyoxal regulates Notch1 expression in NPCs, we electroporated control or Glo1 shRNAs and a nuclear EGFP plasmid into the E13.5 cortex and examined Notch1 protein expression by immunostaining EGFP+ cells 2 days later. The quantification of Notch1+ cells in VZ/SVZ showed a 35% reduction following Glo1 knockdown (Fig. 3a, b ), in line with our observation of attenuated Notch signalling (Fig. 2a–c ). To determine if the reduction in Notch1 protein abundance was due to a decrease in its mRNA levels, we sorted EGFP+ cells from electroporated cortices by FACS and examined Notch1 mRNA by qRT-PCR. Glo1 knockdown did not affect Notch1 mRNA levels (Fig. 3c ). Similarly, we found that GLO1 knockdown caused a reduction in NOTCH1 protein abundance in hNPCs without changing its mRNA levels (Supplementary Fig. 4a–c ).
Fig. 3: Methylglyoxal suppresses Notch1 mRNA translation.
a–c E13.5 cortices were co-electroporated with nuclear EGFP and control or Glo1 (shGlo1) shRNAs and analyzed two days later. Images (a) and quantification (b) of EGFP + cells (green) also positive for Notch1 (red). White boxes in a show higher magnification at the bottom. Arrows denote double-labelled cells. n = 3 embryos each. c qRT-PCR analysis for Notch1 mRNA in FACS sorted EGFP+ cells from cortices as in a. n = 3 embryos each. d–i hNPCs were treated with DMSO or BBGD for 48 h. Images (d) and quantification (e) from western blots of hNPCs probed for NOTCH1 and ACTb. n = 3 experiments. f qRT-PCR analysis for NOTCH1 and NESTIN (NES) mRNAs in hNPCs. n = 3 experiments. g Polysome profile of hNPCs. Fractions corresponding to 40 S and 60 S ribosome subunits, the 80 S monosomes and polysomes are labelled. h Relative distribution of the NOTCH1 mRNA (left) and GAPDH mRNA (right) across all fractions measured by qRT-PCR. Each point corresponds to the value of each fraction normalized to the total NOTCH1 or GAPDH mRNA. i Quantification of the relative enrichment of NOTCH1 and GAPDH mRNAs in the heavy polysome fractions (#8–12). n = 3 experiments. j–l Dual-luciferase reporters containing the full-length (FL) or truncated Notch1 3ʹUTR (dARE) were co-electroporated into E13.5 cortices with control or Glo1 shRNAs. j Luciferase activity of the reporter. n = 6 embryos each. k Luciferase mRNA levels quantified by qRT-PCR. n = 3 embryos each. l Luciferase activity from the 3ʹUTR-dARE reporter. n = 8 embryos each. m, n hNPCs were transfected with dual-luciferase reporters and treated with DMSO or BBGD for 48 h. m Luciferase activity of the reporters.; n = 3 experiments. n Luciferase mRNA levels quantified by qRT-PCR. n = 3 experiments. Scale bars, 25 µm. Data are presented as mean values ± SEM and analyzed using two-tailed, unpaired students t-test. **p < 0.01, *p < 0.05, ns = p > 0.05. Source data and p-values are provided as a “Source Data file”.
Full size image
We next examined NOTCH1 protein and mRNA levels in cultured hNPCs treated with BBGD or directly with methylglyoxal. Western blot and qRT-PCR analyses showed that while BBGD and methylglyoxal treatment caused a reduction in NOTCH1 protein abundance (Fig. 3d, e , Supplementary Fig. 4d, e ), NOTCH1 mRNA levels again were unaltered (NESTIN mRNA used as a control) (Fig. 3f , Supplementary Fig. 4f ). Given that glycolysis is the primary source of intracellular methylglyoxal, we asked whether an increased glycolytic flux induced by high glucose condition could lead to similar changes in hNPCs. However, high glucose medium did not affect methylglyoxal levels in hNPCs (Supplementary Fig. 4g ). In contrast, ectopic expression of GLO1 that reduced methylglyoxal in hNPCs (Supplementary Fig. 1h ) caused an increase in NOTCH1 protein abundance (Supplementary Fig. 4h, i ) but did not affect NOTCH1 mRNA levels (Supplementary Fig. 4j ).
These results suggest that the regulation of NOTCH1 expression may occur at the translational level. To test this, we performed polysome profiling to measure mRNA translational status by separating non-translated mRNAs from those being associated with multiple ribosomes (heavy polysomes) for active translation (Fig. 3g ). We observed that BBGD treatment did not alter the polysome profile in hNPCs (Fig. 3g ), suggesting that global translation was not perturbed by excessive methylglyoxal. This was consistent with the lack of changes seen in total protein synthesis measured by puromycin metabolic incorporation (Supplementary Fig. 4h, i ). We then examined the polysomal distribution of NOTCH1 mRNA by qRT-PCR. BBGD treatment induced a robust shift in NOTCH1 mRNA towards lighter polysome fractions, with a concomitant increase in fractions containing non-translated mRNA (Fig. 3h, i ), indicating reduced translation. On the contrary, the highly expressed GAPDH mRNA was similarly engaged in translation in both conditions (Fig. 3h, i ). Together, these results indicate that methylglyoxal negatively regulates NOTCH1 expression in NPCs at the level of translation.
Suppression of Notch1 translation requires an AU-rich motif
Translational regulation of Notch1 mRNA has been described in C. elegans during germline development 33 , 34 , and recently in mouse T cells during thymocyte development 35 . In these cases, the translational regulation of Notch receptors is mediated by the 3ʹUTR of Notch1 mRNA. Therefore, we asked whether the Notch1 3ʹUTR mediated methylglyoxal-induced translational repression. Using luciferase reporter assay, we included the full-length Notch1 3ʹUTR downstream of the firefly luciferase gene in a dual-luciferase vector and electroporated it with control or Glo1 shRNAs into the E13.5 cortex. Analysis of the cortex 2 days later showed that Glo1 knockdown markedly suppressed firefly luciferase activity (Fig. 3j ), while luciferase mRNA levels remained unchanged (Fig. 3k ), suggesting the presence of a translational regulatory component within the Notch1 3ʹUTR. The AU-rich element within the Notch1 3’ UTR mediates the translational regulation of Notch1 by interacting with RBPs in T cells 35 . We found that the deletion of this AU-rich region in the Notch1 3ʹUTR completely abolished Glo1 knockdown-induced reduction in luciferase activity in the cortex (Fig. 3l ). Similarly, in cultured hNPCs, BBGD treatment significantly suppressed the luciferase activity from the reporter with the full-length Notch1 3ʹUTR but not the construct lacking the AU-rich region (Fig. 3m, n ), suggesting that the AU-rich region present in the Notch1 3ʹUTR mediates the translational regulation of Notch1 mRNA.
GAPDH binds to the Notch1 3ʹUTR to represses its translation
Given the involvement of the AU-rich region, RBP(s) may interact with this region to mediate methylglyoxal-induced translational repression of Notch1 mRNA. We reasoned that methylglyoxal might modulate this interaction by changing the availability or enzymatic activity of an RBP since the post-translational addition of methylglyoxal moieties can change protein stability and functions 21 , 36 , 37 , 38 . We focused our search of RBPs to those known to bind AU-rich elements and to be modified by methylglyoxal. The glycolytic enzyme GAPDH is a known target of methylglyoxal modification, and this modification inhibits its enzymatic activity for glycolysis 17 . Interestingly, when its enzymatic function in glycolysis is disengaged in T cells, GAPDH acts as an RBP to suppress the translation of interferon γ mRNA after binding an AU-rich sequence within interferon γ 3ʹUTR 10 . This raises the possibility that GAPDH mediates the effect of methylglyoxal on Notch1 translation in NPCs.
To test this idea, we first asked whether GAPDH interacts with Notch1 3’UTR. We performed an RNA-electrophoretic mobility shift assay (REMSA) on in vitro transcribed and radioactively-labelled Notch1 3ʹUTR with different amounts of recombinant GAPDH (rGAPDH) protein, followed by the separation of RNA-protein complexes on native polyacrylamide gels. Multiple shifted bands of labelled RNA were detected in the presence of rGAPDH in a dose-dependent manner, indicating a direct interaction between GAPDH and the Notch1 3ʹUTR (Fig. 4a ). This interaction was specific, as an excess of unlabeled Notch1 3ʹUTR RNA out-competed the binding of rGAPDH and abolished shifted bands of labelled Notch1 3ʹUTR RNA, while yeast tRNA was unable to out-compete the binding of rGAPDH to Notch1 3ʹUTR RNA (Fig. 4b ).
Fig. 4: GAPDH binds Notch1 mRNA 3ʹUTR in response to excessive methylglyoxal and regulates Notch1 translation.
a–c REMSA was performed using the labelled (hot) full-length (FL) or AU-rich element-deleted (dARE) Notch1 3ʹUTR RNA probe in the presence of incremental amounts of recombinant GAPDH protein and the presence of unlabeled (cold) specific or unspecific RNA probes. Arrows and arrowheads indicate free RNA probes and RNA-GAPDH complexes, respectively. n = 3 experiments. d Luciferase activity from a reporter containing the full-length (FL) or AU-rich element-deleted (dARE) Notch1 3ʹUTR co-transfected with a plasmid expressing EGFP control or GAPDH in HEK293 cells. Firefly luciferase activity values were normalized to Renilla luciferase activity in the same samples. n = 3 experiments. e–i hNPCs were treated with DMSO or BBGD for 48 h. e GAPDH activity in the cell lysates. The same number of cells were seeded prior to treatment and the GAPDH activity was further normalized by total protein mass. n = 3 experiments. f Western blots probed for GAPDH and reprobed for ACTb as a loading control. n = 3 experiments. g qRT-PCR analysis of mRNA enrichment by RNA immunoprecipitation (RIP) with control IgG or an anti-GAPDH antibody (normalized to the total RNA input). n = 3 experiments. h qRT-PCR analysis of the relative distribution of the HIF1a and c-MYC mRNAs across all fractions from polysome profiling. i Quantification of the relative enrichment of HIF1a and c-MYC mRNAs in the heavy polysome fractions. Each point corresponds to the value of each fraction normalized to the total HIF1a or c-MYC mRNA. n = 3 experiments. Data are presented as mean values ± SEM and analyzed using two-tailed, unpaired students t-test. **p < 0.01, ns = p > 0.05. Source data and p-values are provided as a “Source Data file”.
Full size image
We next asked whether the AU-rich region within the Notch1 3ʹUTR mediates this interaction. We repeated the REMSA using an in vitro synthesized Notch1 3ʹUTR containing a deletion of the AU-rich region and found no shifted bands in the presence of rGAPDH (Fig. 4c ). Moreover, this mutant form of Notch1 3ʹUTR was unable to compete for the binding of rGAPDH to the full-length Notch1 3ʹUTR (Fig. 4c ). Given that the AU-rich region in the Notch1 3ʹUTR can be bound by rGAPDH and was critical for the translational suppression of Notch1 mRNA in cortical precursors (Fig. 3j, l ), we speculated that GAPDH could suppress Notch1 mRNA translation. To test this, we co-transfected the Notch1 3ʹUTR luciferase reporters into the human embryonic kidney (HEK) 293 cells with plasmids overexpressing GAPDH or EGFP as a control. Indeed, the luciferase assay showed that ectopic expression of GAPDH caused a 30% decrease in the luciferase activity from full-length Notch1 3ʹUTR, but not the mutant 3’UTR lacking the AU-rich region (Fig. 4d ), suggesting that GAPDH acts as a translational repressor on the Notch1 3’UTR via the AU-rich region.
Methylglyoxal engages GAPDH to suppress Notch1 mRNA
Our data suggest that the translation of Notch1 mRNA is controlled by its interaction with GAPDH. It is known that methylglyoxal can post-translationally modify GAPDH, and this modification inhibits its enzymatic activity for glycolysis 17 . Therefore, we explored the model that methylglyoxal modulates the dual function of GAPDH in NPCs by altering its enzymatic activity and subsequently engaging it as an RBP to bind Notch1 mRNA, leading to translational repression. To test this, we first assessed GAPDH activity in hNPCs. Following BBGD treatment, we found that while GAPDH protein levels remained unchanged, GAPDH enzymatic activity was suppressed by approximately 80% (Fig. 4e, f ).
We next asked if the suppression of GAPDH enzymatic activity induced by BBGD could lead to enhanced interaction between GAPDH and NOTCH1 mRNA. We performed RNA immunoprecipitation (RIP) using antibodies against GAPDH or control isotype IgG from hNPCs extracts treated with or without BBGD. qRT-PCR analysis of immunoprecipitated RNA showed that the GAPDH antibody was enriched for NOTCH1 mRNA as well as other known GAPDH target mRNAs encoding regulators of NPC fate decision, including HIF1a 39 , 40 and c-MYC 41 (Fig. 4g ). The interaction was specific, as these target mRNAs were not immunoprecipitated by control IgG, and the GAPDH antibody was not able to immunoprecipitate NESTIN mRNA or 18 S RNA used as negative controls (Fig. 4g ). Interestingly, following BBGD treatment, the amount of GAPDH-enriched NOTCH1 mRNAs, but not HIF1a or c-MYC mRNAs, were markedly increased, suggesting the existence of functional selectivity (Fig. 4g ). In agreement with this observation, we found that HIF1a and c-MYC mRNAs were actively engaged in translation regardless of BBGD treatment (Fig. 4h, i ).
GAPDH knockdown rescues methylglyoxal-induced NPC depletion
The above results support a model whereby the modification of GAPDH by methylglyoxal engages it as an RBP to selectively suppress Notch1 mRNA translation, leading to NPC depletion and premature neurogenesis. If the model is correct, a reduction in GAPDH abundance should ameliorate the impact of excessive methylglyoxal on NPCs in the developing cortex by releasing the repression on Notch1 translation. To test this, we used a GAPDH shRNA that efficiently knocked down GAPDH expression in transfected cultured NPCs (Fig. 5a, b ). We co-electroporated this shRNA with EGFP plus control or Glo1 shRNAs into E13.5 cortices. Analysis 3 days later showed that while Glo1 knockdown caused a robust reduction in EGFP+ cells in the VZ/SVZ and CP, concurrent knockdown of GAPDH and Glo1 partially rescued the distribution of EGFP+ cells in the VZ/SVZ but not the CP (Fig. 5c, d ). Moreover, GAPDH knockdown normalized the proportions of EGFP+ cells that were also positive for Pax6 or Satb2 (Fig. 5e, f ). Together, these data suggest that methylglyoxal controls NPC homeostasis, at least in part, by engaging GAPDH as an RBP to regulate Notch signalling.
Fig. 5: Reducing GAPDH abundance partially rescues the NPC depletion induced by excessive methylglyoxal.
a, b Cultured E12.5 cortical precursors were co-transfected with EGFP and control or GAPDH shRNAs (shGAPDH) for three days, immunostained for EGFP (green) and GAPDH (red) (a) and quantified for EGFP-positive cells with detectable GAPDH (b). Arrows and arrowheads denote EGFP-positive and -negative cells, respectively. unpaired t-test; n = 3 experiments. c–f E13.5 cortices were co-electroporated with control or Glo1 shRNAs together with or without a shRNA against GAPDH, and coronal cortical sections were analyzed three days later. c Images of electroporated sections immunostained for EGFP (green). Dotted white lines denote boundaries between VZ/SVZ, IZ and CP. d Quantification of sections as in c for the relative location of EGFP+ cells. n = 3 (Control), 8 (shGlo1), 6 (shGAPDH) and 6 (shGAPDH + shGlo1) embryos. e Images of the VZ or IZ of electroporated sections that were immunostained for EGFP (green) and Pax6 (red, top panels) or Satb2 (red, bottom panels). f Quantification of sections as in e for the proportion of EGFP+ cells that were also positive for Pax6 or Satb2. n = 3 (Control), 8 (shGlo1), 6 (shGAPDH) and 6 (shGAPDH + shGlo1) embryos. Sections in a and c were counterstained with Hoechst 33258 (blue). Scale bars, 100 µm in c, 10 µm in a and e. Data are presented as mean values ± SEM and analyzed using two-tailed, unpaired students t-test with Bonferroni correction. **p < 0.01, *p < 0.05, ns = p > 0.05. Source data and p-values are provided as a “Source Data file”.