Metabolic Control over mTOR-Dependent Diapause-like State

Abstract

Regulation of embryonic diapause, dormancy that interrupts the tight connection between developmental stage and time, is still poorly understood. Here, we characterize the transcriptional and metabolite profiles of mouse diapause embryos and identify unique gene expression and metabolic signatures with activated lipolysis, glycolysis, and metabolic pathways regulated by AMPK. Lipolysis is increased due to mTORC2 repression, increasing fatty acids to support cell survival. We further show that starvation in pre-implantation ICM-derived mouse ESCs induces a reversible dormant state, transcriptionally mimicking the in vivo diapause stage. During starvation, Lkb1, an upstream kinase of AMPK, represses mTOR, which induces a reversible glycolytic and epigenetically H4K16Ac-negative, diapause-like state. Diapause furthermore activates expression of glutamine transporters SLC38A1/2. We show by genetic and small molecule inhibitors that glutamine transporters are essential for the H4K16Ac-negative, diapause state. These data suggest that mTORC1/2 inhibition, regulated by amino acid levels, is causal for diapause metabolism and epigenetic state.

Keywords: H4K16Ac; LKB1; amino acids; diapause; epigenetics; glutamine transporter; lipolysis; mTOR; metabolism; pluripotent stem cells.

Figure 1.. Gene Expression and Splice Variants Separate Diapause Stage from ICM and Post-implantation Stages

(A) Schematic diagram depicting outline of the experiment. (B) PCA of our RNA-seq data showed that diapause is in a distinct transcriptional state compared to pre- and post-implantation epiblasts. (* = Boroviak et al., 2015) (C) Scatter plot of gene contributions to PC1 and PC2 in the PCA plot in (B) showing genes that are specifically up- or down-regulated in diapause compared topre-/post-implantation. (D) Metabolic pathway enrichment of genes differentially expressed between diapause and ICM. (E) PCA plot using transcript splicing rate of our RNA-sequence samples clearly separates diapause and pre-/post-implantation samples. (F) Sashimi plot of Lkb1 exons 8–10. Bar height represents expression level (FPKM). Arcs connect two exons that are spliced together. (G) qPCR analysis of Lkb1 splice variant expression in ICM, dICM, and epiblast. n = 2 for each group, p = 0.001965 for ICM versus dICM, and p = 0.000102 for dICM versus Epi. (H) Immunostaining of pre-implantation and diapause embryos with antibodies against both Lkb1 isoforms (red), OCT4 (magenta), or stained with DAPI (blue). Scale: 22 μm.

Figure 2.. Lkb1 Activation by Starvation Induces Diapause-like State

(A) Starvation moved mESC samples (R1/Lkb1) toward a diapause state. Cell line samples are projected in the PCA of in vivo samples using a previously described method (Kho et al., 2004).* = Boroviak et al., 2015. (B) Model of Lkb1 and starvation. AMPK is activated by starvation and is a target of Lkb1. Activated AMPK inhibits mTOR and induces diapause. (C) Immunostaining of day 3.5 and diapause embryos with antibody against pAMPKa2(Thr172) (green) and OCT4 (blue). Scale: 20 μm. (D) 1-h exposure to AICAR, starvation, or 2-DG (10 min) increased the activity of AMPK compared to untreated cells. beta-actin 1: pmTOR, pAMPKand pS6; betaactin 2: mTOR, AMPK and S6. (E) Phosphorylation of mTOR, Akt(S473), S6, ULK1(S757), and acetylation of H4K16 in R1(LL) cells decreased after 24 h in starvation, whereas p4EBP1 did notchange. Beta-actin 1, pmTOR and pS6; beta-actin 2, mTOR and S6; beta-actin 3, H4K16ac; beta-actin 4, pAkt(S473) and p4EBP1; beta-actin 5, Akt. (F) Immunostaining of mouse ESCs with or without starvation. Staining was performed to detect pluripotent marker, OCT4 (green). The nuclei of all cells were stained blue with DAPI. Scale: 21 μm. (G) Lkb1 siRNAs effectively knock down expression of Lkb1 at the protein level compared to the control luciferase siRNA. The efficiency of the transient siRNA approach was assessed at the protein level by western blotting coupled with densitometry for Lkb1, mTOR, and S6 in transiently transfected R1(LL) cells. (H) mTOR, and S6 phosphorylation were significantly reduced in luciferase siRNA control under starvation compared to Lkb1 siRNA under starvation. Beta-actin 1: Lkb1 and pmTOR; beta-actin 2: pS6; beta-actin 3: mTOR and S6. n = 3 for pmTOR, p = 0.010 for control versus KD; n = 3 for pS6, p = 0.0039 for control versus KD; n = 3 for Lkb1, p = 0.035 for control versus KD, p = 0.0068 for control+starv versus KD+starv, p = 0.0081 for control versus KD+starv, p = 0.014 for control+starv versus KD. (I) A simplified diagram of the mTOR signaling pathway and mTOR Complexes 1 and 2 inhibitor, INK-128. See text for details. (J) Phosphorylation of mTOR, Akt(S473), S6, 4EBP1, ULK1(S757), and acetylation of H4K16 in R1(LL) cells decreased after INK-128 treatment for 24h. Beta-actin1: pmTOR, pAkt(S473), pS6, and p4EBP1, beta-actin 2: mTOR, S6, and 4EBP1, beta-actin 3: Akt, beta-actin 4: pULK1 and H4K16ac.

Figure 3.. Reversible Diapause-like State Can Be Induced In Vitro by Starvation or Inhibition of mTOR

(A) Generation of Lkb1 short only isoform. A schematic of Talen-mediated gene editing of the endogenous Lkb1 locus using donor construct that expresses Lkb1-short form-specific exon 9A (Figure 1F) fused directly in frame with exon 8 followed by 2A peptide. This modified Lkb1 locus expresses only short form of Lkb1. (B and C) Pharmacological inhibition of mTOR and starvation can induce a reversible diapause-like state. (B) mTOR inhibition by INK-128, is reversible as indicated by the rephosphorylation of mTOR, S6 and 4EBP1 24 h after removal of INK-128 in R1(LL) and R1(SS) cells. Beta-actin 1: pmTOR, pS6 and p4EBP1, beta-actin 2: mTOR, S6 and 4EBP1. (C). Starvation abolishes phosphorylation of mTOR and its substrates, S6 kinase, S6, ULK1, and 4EBP1 in 24 h. This inhibition is reversible as indicated by the rephosphorylation of those proteins 24 h after removal of starvation media from culture both in R1(LL) and R1(SS) cells. Intriguingly, the mTOR signal in the untreated and the reversed R1(SS) samples is reduced compared to their respective R1(LL) samples. Beta-actin 1: pmTOR, pS6 and p4EBP1, beta-actin 2: mTOR, S6 kinase, S6 and 4EBP1, beta-actin 3: pS6 kinase, beta-actin 4: pULK1, beta-actin 5: H4K16ac. (D) Quantification of pmTOR protein bands from C using ImageJ software. n = 3 for each sample, p = 0.0027 for R1(LL) versus R1(SS) 24h after starvation removal. (E) Western blot of pAMPK in mESCs expressing different forms of Lkb1 splice variants. The phosphorylation of AMPK is increased in response to 2DG both in R1(LL) and R1(SS). However, higher levels of AMPK phosphorylation are observed in R1(SS) even in the absence of 2-DG. (F) A schematic representation of the two splice forms of Lkb1.

Figure 4.. Lkb1 Short Does Not Respond Dynamically to Starvation, and Diapause-like State Cells Have Higher Glycolytic Activity Compared to Control Metabolic Flux in Mouse ESCs with Different Lkb1 Splice Variants

(A–F) Short Lkb1 does not respond dynamically to starvation. Metabolic flux of mouse ESCs with different Lkb1 splice variants using Seahorse analyzer. (A) Representative trace of OCR changes is shown under a MitoStress protocol. (B) Quantification of A (4 independent experiments). (C) Representative trace of OCR changes of mouse ESCs with R1(SS) or R1(SL) lines have reduced OCR changes in response to FCCP compared to R1(LL) splicevariant of Lkb1. (D) Quantification of C (4 independent experiments). n = 28 per group, p = 0.0076 R1(LL) versus R1(SL), p < 0.0001 R1(LL) versus R1(SS). Mouse ESC lines with short Lkb1 splice variant cannot respond dynamically to stress. (E) mESCs with Lkb1 short splice variant show high mitochondrial beta-oxidation when substrate, fatty acid palmitate (Palm) is offered for oxidation in normalconditions. n = 7 per group, p = 0.0088 R1(LL) versus R1(SL), p = 0.062 R1(LL) versus R1(SS). (F) The mESC lines with Lkb1 SL and SS splice variants show lower fatty acid beta-oxidation levels than R1(LL, blue). n = 7 per group, p = 0.073 R1(LL) versus R1(SL). (G) MitoTracker Green staining of R1(LL) and R1(SS). No change in mitochondrial mass was detected between R1(LL) and R1(SS) in normal or starvationconditions. (H) Representative trace of ECAR changes in response to glucose, oligomycin, and 2-DG is shown under a glucose stress protocol. (I and J) Using starvation and INK-128, naive mESCs, R1(LL), and R1(SS), under starvation or INK-128, have higher glycolytic capacity (I) and glycolytic reserve (J) compared to naive mESCs under normal conditions. (I) n = 12 for R1(LL) and R1(SS), n = 9 for R1(LL)+INK-128, n = 15 for R1(LL)+starv, n = 22 for R1(SS)+starv, n = 8 for R1(SS)+INK-128. p < 0.0001 for R1(LL) versusR1(LL)+INK-128, p < 0.0001 for R(LL) versus R1(LL)+starv, p < 0.0007 for R1(LL)+starv versus R1(LL)+INK-128, p < 0.0026 for R1(SS) versus R1(SS)+INK-128, p < 0.0001 for R1(SS) versus R1(SS)+starv, p = 0.0088 for R1(LL) versus R1(SS), p = 0.0109 for R1(LL)+starv versus R1(SS)+starv. (J) n values are the same as Figure 4I. p = 0.0063 for R1(LL) versus R1(LL)+INK-128, p < 0.0001 for R(LL) versus R1(LL)+starv, p < 0.0001 for R1(LL)+starv versus R1(LL)+INK-128, p = 0.0022 for R1(SS) versus R1(SS)+starv, p = 0.0030 for R1(SS)+starv versus R1(SS)+INK-128, p = 0.0176 for R1(LL) versus R1(SS). (K) No glycolytic capacity differences were observed between R1(LL) and R1(SS) under starvation when AICAR is added to R1(LL). n = 11 for R1(LL)+starv, n = 12 for the rest, p = 0.537 for R1(LL)+starv+AICAR versus R1(SS)+starv, p = 0.0171 for R1(LL)+starv versus R1(LL)+starv+AICAR, p = 0.5267 for R1(SS)+starv versus R1(SS)+AICAR, p = 0.0350 for R1(LL)+starv versus R1(SS)+starv. (L) Compound C, an inhibitor of AMPK, reduces the glycolytic capacity observed under starvation for both Lkb1 splice variants. n = 4 for R1(SS), n = 5 for R1(LL) and R1(LL)+10 mM compound C, rest, n = 6. p = 0.0027 for R1(LL)+starv versus R1(LL)+starv+10 mM compound C, p < 0.0001 for R1(SS)+starv versus R1(SS)+starv+10 mM compound C, p = 0.5978 for R1(LL) versus R1(LL)+10 mM compound C, p = 0.0056 for R1(SS) versus R1(SS)+10 mM compound C. (M) Model of AMPK pathway. AMPK, once phosphorylated by Lkb1, can stimulate glucose uptake and inhibit mTOR while INK-128 inhibits mTOR.

Figure 5.. Overall Changes of Lipids and Polar Metabolites between Diapause and Pre-implantation Blastocyst

(A) Schematic diagram depicting outline of the metabolomics experiment for pre-implantation or diapause embryos. (B) Chemical enrichment analysis of diapause blastocyst compared to pre-implantation blastocyst show dramatic differences in lipid composition. Node size reflects the number of metabolites and color indicates direction of change, blue is decreased and red increased. The y axis indicates enrichment p values calculated by Kolmogorov-Smirnov-test and x axis show the polarity of chemical clusters. (C) Relative lipid classes between diapause and pre-implantation blastocyst. (D) Log2 fold change of a total of 243 polar metabolites that were identified between diapause and pre-implantation blastocyst. (E) Schematic diagram of sphingolipid metabolism showing metabolites and genes that are upregulated in diapause compared to pre-implantation. D, diapause;PI, pre-implantation; TAG, Triacylglycerol; DAG, Diacylglycerol; ASAH1, N-acylsphingosine amidohydrolase 1; ACSS1, acyl-CoA synthetase short-chain family member 1; FASN, fatty acid synthase. (F) One-carbon metabolism pathway indicating metabolites that are increased or decreased and their connections with other pathways.

Figure 6.. Lipolysis Is Upregulated in Embryonic Diapause, in Diapause-like State, and in Rictor Knockout mESCs

(A) Abundance of TAG, DAG, FA, and AC in pre-implantation and diapause blastocyst. (B) Schematic diagram depicting outline of the Lkb1 splice variant metabolomics experiment and abundance of TAG, DAG, FA, and AC in Lkb1 splice variants,R1(LL) and R1(SS), in control and starvation conditions. (C) A simplified diagram of the mTOR signaling pathway, showing components, and downstream targets of mTORC1/2. (D) Schematic diagram of CRISPR-Cas9 system–mediated homozygous Rictor (mTORC2 specific component) KO in mESCs. Sequence and chromatogram(homozygote) of Rictor KO clone 14 mESCs. (E) Screening of Rictor KO clones (top panel) by western blotting and analysis of Rictor knockout effect on mTOR and its downstream targets (bottom panel). (Beta-actin-1: pAkt and pmTOR, Beta-actin-2: pS6 Kinase and pS6, Beta-actin-3: Akt, S6 and 4EBP1, Beta-actin-4: mTOR, Beta-actin-5: pULK1 and p4EBP1, Beta-actin-6: S6 Kinase). (F) Schematic diagram depicting outline of the Rictor KO metabolomics experiment and Rictor KO effect on lipolysis. TAGs are downregulated in Rictor knockout clones compared to wildtype.

Figure 7.. Inhibition of the Glutamine Transporters Leads to Exit from Diapause

(A) Glutamine transporters, SLC38A1, SLC38A2, and SLC1A5 are upregulated in diapause and starvation-induced diapause-like state. (B) Schematic diagram depicting outline of the glutamine transporter inhibition in mouse diapause embryos. (C) Immunostaining of diapause embryos with or without GPNA treatment (1h) for epigenetic, H4K16ac mark (magenta), and DAPI (blue). n = 4 diapaused embryos per treatment, p < 0.0001 for diapaused embryos versus diapaused embryo+10 mM GPNA. Scale: 20 μm. (D) Hypothetical model of mouse embryos in diapause state. Pre-implantation embryos enter the diapause state through Lkb1-AMPK dependent down regulation of mTOR. Diapause is associated with upregulation of SLC38A1/2 and glycolysis, low mitochondrial activity, low fatty acid beta-oxidation (FAO), increased NF-κB activity, and downregulation of mTOR. Downregulation of mTOR leads to upregulation of lipolysis, resulting in breakdown of TAGs. Glutamine transporters (SLC38A1/2) are required for epigenetic diapause state, plausibly through prolonged mTOR inhibition. PM, plasma membrane.