Metabolic reprogramming during neuronal differentiation from aerobic glycolysis to neuronal oxidative phosphorylation

Abstract

How metabolism is reprogrammed during neuronal differentiation is unknown. We found that the loss of hexokinase (HK2) and lactate dehydrogenase (LDHA) expression, together with a switch in pyruvate kinase gene splicing from PKM2 to PKM1, marks the transition from aerobic glycolysis in neural progenitor cells (NPC) to neuronal oxidative phosphorylation. The protein levels of c-MYC and N-MYC, transcriptional activators of the HK2 and LDHA genes, decrease dramatically. Constitutive expression of HK2 and LDHA during differentiation leads to neuronal cell death, indicating that the shut-off aerobic glycolysis is essential for neuronal survival. The metabolic regulators PGC-1α and ERRγ increase significantly upon neuronal differentiation to sustain the transcription of metabolic and mitochondrial genes, whose levels are unchanged compared to NPCs, revealing distinct transcriptional regulation of metabolic genes in the proliferation and post-mitotic differentiation states. Mitochondrial mass increases proportionally with neuronal mass growth, indicating an unknown mechanism linking mitochondrial biogenesis to cell size.

Keywords: LDHA; developmental biology; glycolysis; human; metabolism; neuronal differentiation; stem cell; stem cells; tricarboxylic acid cycle.

Figure 1.. Gene expression of glycolysis and TCA during neuronal differentiation.

(A) Human BJ iPSC-derived NPCs showed homogeneous expression of the NPC markers Nestin and Sox2. (B) The left panel shows NPC-derived neurons after 3 weeks of differentiation; right panel shows staining of MAP2, a neuronal marker. (C) The top 200 up- and down-regulated genes during neuronal differentiation are depicted by a heatmap. Red and green intensities indicate fold increases and decreases, respectively, in gene expression (expressed as log2). (D) GO term analysis of genes up-regulated during neuronal differentiation. The top eight GO term biological process categories obtained are ranked by p-value. (E) The FPKM values of known neuron-specific genes are shown as log10-fold change. (F) The fold changes of FPKM values of proliferating NPC markers are shown. Bars represent mean ± SD of four RNA-seq replicates for NPCs and neurons differentiated at 1 and 3 weeks. (G, H, I and J) Relative expression levels of the key metabolic genes in glycolysis, tricarboxylic acid cycle (TCA), pyruvate dehydrogenase (PDH) complex and pentose phosphate pathway. Bars show the mean of FPKM values of differentiated neurons at 1 and 3 weeks relative to FPKM values of NPCs. Error bars represent SD of four RNA-seq replicates at each time point. Abbreviations, dihydroxyacetone phosphate (DHAP); 1,3-bisphosphoglyceric acid (1,3 BPG); 3-phosphoglyceric acid (3PG); phosphoenolpyruvic acid (PEP), α-ketoglutarate (α-KG); oxaloacetate (OAA). (Figure 1—sourcer data 1). DOI: http://dx.doi.org/10.7554/eLife.13374.003

Figure 1—figure supplement 1.. The outline of the protocol used to differentiate neurons from iPSCs (upper panel).

Representative pictures of fibroblasts, iPSCs, NPCs and neurons (lower panel). DOI: http://dx.doi.org/10.7554/eLife.13374.005

Figure 1—figure supplement 2.. Colonies containing neural rosettes, and type 4 colony produces NPCs of the best quality.

DOI: http://dx.doi.org/10.7554/eLife.13374.006

Figure 1—figure supplement 3.. Electrophysiological study of BJ 5-week neurons.

Representative results were shown. (a) Patched neuron filled with rhodamine; (b) Evoked voltage dependent sodium and potassium currents recorded in voltage-clamp (−70 mV). (c) Evoked action potential. (d) Spontaneous burst of action potentials. (e) Zoom of 'd'. DOI: http://dx.doi.org/10.7554/eLife.13374.007

Figure 1—figure supplement 4.. FPKM values of paralogous genes in glycolysis.

DOI: http://dx.doi.org/10.7554/eLife.13374.008

Figure 1—figure supplement 5.. Metabolites quantified by gas chromatography mass spectrometry (GC-MS).

NPCs at early passage (P3) and 3-week neurons were incubated in fresh medium for 12 hr, and metabolites in cells were extracted and analyzed by GC-MS. The relative amount of metabolites extracted from neurons versus NPCs was obtained after normalization to protein content. Bars represent mean ± SD of three biological replicates for NPCs and 3-week neurons. (Figure 1—figure supplement 5—sourcer data 1) DOI: http://dx.doi.org/10.7554/eLife.13374.009

Figure 1—figure supplement 6.. Oxygen consumption rate (OCR) analysis by Seahorse extracellular flux analysis.

FCCP (F) is a chemical uncoupler of electron transport and oxidative phosphorylation; Rotenone and Antimycin A (R&A) are complex I and III inhibitors. Error bars represent SD. None mitochondrial OCR has been subtracted. (Figure 1—figure supplement 6—source data 1). In this experiment, NPCs had to be grown at high density to avoid differentiation and ensure optimal proliferation. The large amount of lactate secreted by high-density NPCs under these conditions decrease the pH value of the medium, which lacks sodium bicarbonate, significantly, from 7.4 to 7.0, during incubation and measurement. The NPC ECAR readings were out of the linear measurement range and not shown here. DOI: http://dx.doi.org/10.7554/eLife.13374.011

Figure 1—figure supplement 7.. The glucose and lactate concentrations in the medium growing NPCs and 3-week neurons were quantified by YSI 2950 metabolite analyzer.

The lactate production/ glucose consumption ratio was calculated as the amount of lactate divided by used glucose. Bars represent mean ± SD. n= 3. In theory, if one glucose molecule is completely converted into lactate (no entry into mitochondrial TCA cycle), the ratio of lactate/glucose would be 2. (Figure 1—figure supplement 7—source data 1) DOI: http://dx.doi.org/10.7554/eLife.13374.013

Figure 2.. Neuron-specific splicing of PKM and OGDH.

(A) The upper panel shows the RNA-seq reads obtained from NPCs and neurons mapped to the PKM and OGDH chromosome locus using Integrative Genomics Viewer. The lower schematic diagram depicts the organization of exons near the cell type-specific splicing site. Red box and line represent exon splicing unique to NPCs; blue ones represent that unique to neurons. PKM2 and OGDH1 are unique to NPCs; PKM1 and OGDHneu are predominantly for neurons. (B) Validation of PKM, OGDH splicing by PCR. Primers were designed to amplify the unique splicing region and common region of PKM1/2 and OGDH1/neu mRNA. PCR was carried out with cDNA prepared from NPCs, neurons at 3 week and primary human astrocytes. (C) The fold changes of FPKM values of hnRNPI, hnRNPA1 and hnRNPA2 are shown. Bars represent mean ± SD of four RNA-seq replicates for NPCs and neurons differentiated at 1 and 3 week. (D) The RNA-seq reads obtained from purified mouse astrocytes and neurons mapped to the PKM and OGDH chromosome locus using Integrative Genomics Viewer. The original RNA-seq data were from Zhang et al. (2014). (E) Alignment of amino acid sequences encoded by the alternative exons of human and mouse OGDH1/neu. OGDH1 contains a calcium-binding motif underlined in red, absent in OGDHneu. (Figure 2—source data 1). DOI: http://dx.doi.org/10.7554/eLife.13374.015

Figure 3.. Characterization of HK2 and LDHA in NPCs, differentiated neurons and neurons directly converted from fibroblasts.

(A) Immunoblotting analysis of the representative metabolic genes in glycolysis, tricarboxylic acid cycle (TCA) pathways. 20 µg of protein lysate from NPCs and from neurons differentiated for 3, 7 and 21 days (D3, D7, D21) were loaded. (B) Immunostaining analysis of HK2 and LDHA in NPCs and 3-week neurons. (C) Effects of HK2 or LDHA knockdown on NPC proliferation. NPCs at early passage (P2) were seeded in 24-well plates one day before infection. The NPC number was determined at 5 days after infection with lenti-shRNA virus against HK2 or LDHA. Two effective shRNA lentiviral vectors targeting different regions of HK2 or LDHA were used. Scramble shRNA vector was used as control. Error bars represent ± SD, n= 3. The knockdown efficiency was confirmed by immunoblotting. (D, E) Immunostaining analysis of LDHA in neurons directly converted from fibroblasts. Two protocols were applied: one was to knockdown PTB1, a single RNA-binding protein, and the other was to overexpress proneuronal transcription factors Ngn2 and Ascl1. Tuj1 (ß-III tubulin) and Tau were stained as early and mature neuronal markers, respectively. The above experiments were repeated at least three times. (F) ChIP analysis of HK2 and LDHA promoters using anti-cMYC or N-MYC antibodies and rabbit IgG as control. Chromatin were prepared from NPCs. The enrichment values are shown as percentage normalized to input. N.C. stands for non-specific control. Bars are mean ± SD, n= 3. Immunoblotting and real time PCR analysis of HK2 and LDHA expression in NPCs with inducible c-MYC. mRNA expression levels of HK2 and LDHA relative to those from non-induction control were calculated after normalization to β-actin. Bars are mean ± SD, n= 3. (G) Immunoblotting analysis of c-MYC, N-MYC and Max in NPCs and 3-week neurons. (Figure 3—source data 1). DOI: http://dx.doi.org/10.7554/eLife.13374.017

Figure 3—figure supplement 1.. Immunoblotting analysis of mouse HK1, HK2 and LDHA in mouse NPCs derived from embryonic stem cell (ES-E14TG2a), 2-week differentiated neurons and embryonic neurons at E18.

The relative mRNA expression levels of HK1, HK2 and LDHA in neurons were calculated against the levels in NPCs. Bars are mean ± SD, n= 3. (Figure 3—figure supplement 1—source data 1). DOI: http://dx.doi.org/10.7554/eLife.13374.019

Figure 3—figure supplement 2.. Immunoblotting and real time PCR analysis of HK2 expression in neurons with inducible c-MYC.

mRNA expression levels of HK2 relative to those from non-induction control were calculated after normalization to β-actin. Bars are mean ± SD, n= 3. (Figure 3—figure supplement 2—source data 1). DOI: http://dx.doi.org/10.7554/eLife.13374.021

Figure 4.. Shutoff of aerobic glycolysis is critical for neuronal differentiation.

(A) Immunoblotting analysis of HK2 and LDHA in NPCs and 3-week neurons constitutively expressing HK2 and LDHA (Flag-tagged). (B) Immunostaining of NPCs with anti-FLAG antibody (green), and nuclear staining was done with Hoechst (red). The percentage of Flag-positive cells were quantified, and 100 cells were counted for each group. (C) Immunostaining analysis of MAP2 and GFAP in 3-week neurons. The percentage of GFAP and MAP2 cells were quantified, and 100 cells were counted for each group, and three times of neuronal differentiation were included. Bars are mean ± SD, n= 3. The GFAP mRNA abundance in the RNA extracted from neuronal culture was quantified by real-time PCR and normalized to β-actin, and presented as a fold increase compared to neurons differentiated from control NPCs. Bars are mean ± SD, n = 3. (D) Nuclear staining with Hoechst in NPCs and 3-week neurons. The percentages of condensed nuclear were quantified, and 50 cells were counted for each group. Bars are mean ± SD, n= 3. (E) Immunostaining analysis of LDHA, MAP2 and GFAP in 3-week neurons differentiated from NPC constitutively expressing HK2 and LDHA. (F, G) Colocalization of irregular puntated staining of LDHA (green) with MAP2 (red, in F) or condensed nuclear stained with Hoechst (red, in G) in 3-week neurons differentiated from NPC constitutively expressing HK2 and LDHA. (H) Immunostaining analysis of anti-FLAG and GFAP in 3-week neurons differentiated from NPC constitutively expressing HK2 and LDHA. (I) Immunoblotting analysis of AMPK T172 phosphorylation in the cell lysate extracted from day 4 and day 21 neuronal culture differentiated from NPCs expressing HK2 and LDHA. The AMPK T172 phosphorylation was quantified after normalized to non-phosphoAMPK, and presented as fold increase. (J) Lactate in medium from day-4 neuronal culture differentiated from NPC expressing HK2 and LDHA were quantified and normalized by protein content, and presented as percentage compared to those from control neurons, bars are mean ± SD, n= 3. (K) Immunostaining analysis of LDHA, MAP2 and GFAP in 3-week neurons differentiated from NPC constitutively expressing HK2 and LDHA. The neuronal differentiation medium used contained 5 mM sodium pyruvate, ten fold of the standard concentration (0.5 mM). The above experiments were repeated at least three times. (Figure 4—source data 1). DOI: http://dx.doi.org/10.7554/eLife.13374.023

Figure 5.. Increased neuronal PGC-1α /ERRγ maintain the transcription of metabolic and mitochondrial genes during neuronal differentiation.

(A) mtDNA copy number was measured during neuronal differentiation. Real-time PCR was done with NPCs and neurons differentiated at 1, 3, and 7 weeks. Bars represent mean ± SD n= 3. (B) Measurement of protein mass content of NPCs and neurons at 3 weeks normalized as in per million cells. Bars represent mean ± SD, n =3. (C) Immunoblotting analysis of the representative component of each mitochondrial respiratory complex. 10 µg protein lysate from NPCs and neurons at 1 and 3 weeks were loaded in SDS-PAGE gel. (D) Expression changes of main transcription factors involved in the transcription of metabolic and mitochondrial genes. Bars show the mean of FPKM values of differentiated neurons at 1 and 3 weeks relative to those of NPCs. Error bars represent SD of four RNA-seq replicates at each time point. Immunoblotting analysis confirmed the upregulation of PGC-1α and ERRγ. (E) Relative expression levels of genes encoding the mitochondrial respiratory complexes. Bars show the mean of FPKM values of differentiated neuron at 1 and 3 weeks relative to those of NPCs. Error bars represent SD of four RNA-seq replicates at each time point. (F) Effects of PGC-1α or ERRγ knockdown on the gene expression of glycolysis, tricarboxylic acid cycle (TCA) and mitochondrial respiratory complexes. Neurons differentiated at 3 week were infected with lenti-shRNA virus against PGC-1α or ERRγ, and the total RNA was extracted 5 days after infection. Two effective shRNA lentiviral vectors targeting different regions of PGC-1α or ERRγ were used. Scramble shRNA vector was used as control. In real-time PCR experiments, the relative mRNA expression levels in neurons depleted of PGC-1α or ERRγ to scramble control were calculated after normalization to β-actin. Bars are mean ± SD, n= 3. Similar results were obtained for both shRNA knockdown constructs. (G) A hypothetical model depicting distinct transcriptional regulation of metabolic genes in the proliferation and post-mitotic differentiation states. (Figure 5—source data 1). DOI: http://dx.doi.org/10.7554/eLife.13374.025

Figure 5—figure supplement 1.. The fold changes of FPKM values of UCP2 are shown.

Bars represent mean ± SD of four RNA-seq replicates for NPCs and neurons differentiated at 1 and 3 weeks. (Figure 5—figure supplement 1—source data 1). DOI: http://dx.doi.org/10.7554/eLife.13374.027

Figure 6.. A model depicting transcriptional changes of metabolic genes underlying the switch from aerobic glycolysis in NPCs to oxidative phosphorylation in neurons.

The genes with decreased expression are dimmed. The width of the arrows indicates increased and decreased pyruvate and lactate utilization at different steps in NPCs and neurons. Abbreviations: glucose 6-phosphate (G6P); fructose 1,6-bisphosphate (FBP); glycerol 3-phosphate (G3P); dihydroxyacetone phosphate (DHAP); 1,3-bisphosphoglyceric acid (BPG); 3-phosphoglyceric acid (3PG); phosphoenolpyruvic acid (PEP). DOI: http://dx.doi.org/10.7554/eLife.13374.029

Author response image 1.. Neuronal differentiation in the presence of 5'-aza-2'-deoxycytidine (2 µM) or DMSO as control.

NPC-derived neurons after 2 weeks of differentiation were stained of MAP2, a neuronal marker. DOI: http://dx.doi.org/10.7554/eLife.13374.031

Author response image 2.. Extracellular acidification rate (ECAR) by Seahorse extracellular flux analysis on NPC and 3-week neurons.

Error bars represent SD. DOI: http://dx.doi.org/10.7554/eLife.13374.032