Metabolic and cell cycle shift induced by the deletion of Dnm1l attenuates the dissolution of pluripotency in mouse embryonic stem cells

Mitochondria are versatile organelles that continuously change their morphology via fission and fusion. However, the detailed functions of mitochondrial dynamics-related genes in pluripotent stem cells remain largely unclear. Here, we aimed to determine the effects on energy metabolism and differentiation ability of mouse embryonic stem cells (ESCs) following deletion of the mitochondrial fission-related gene Dnml1. Resultant Dnm1l−/− ESCs maintained major pluripotency characteristics. However, Dnm1l−/− ESCs showed several phenotypic changes, including the inhibition of differentiation ability (dissolution of pluripotency). Notably, Dnm1l−/− ESCs maintained the expression of the pluripotency marker Oct4 and undifferentiated colony types upon differentiation induction. RNA sequencing analysis revealed that the most frequently differentially expressed genes were enriched in the glutathione metabolic pathway. Our data suggested that differentiation inhibition of Dnm1l−/− ESCs was primarily due to metabolic shift from glycolysis to OXPHOS, G2/M phase retardation, and high level of Nanog and 2-cell-specific gene expression.


Introduction
Mitochondria are versatile organelles with a variety of functions and can self-divide or change their morphology via fission and fusion.Mitochondrial fission and fusion occur dynamically in response to the cellular environment and, thus, are crucial for cellular function and survival [5,29].Accordingly, the number of mitochondria and their functions are influenced by mitochondrial biogenesis (division) and mitochondrial dynamics (fission and fusion) [41].Several proteins are involved in the mitochondrial fission and fusion processes in eukaryotes [61].In mammals, Mfn1/2 (mitofusin 1/2) and Opa1 (optic atrophy 1) are located in the outer and inner membranes, respectively, and regulate mitochondrial fusion.Dnm1l (dynamin 1-like), Fis1 (fission, mitochondrial 1), and Mff (mitochondrial fission factor) are located in the mitochondrial outer membrane and regulate mitochondrial fission [58,61].
Pluripotent stem cells (PSCs), such as embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs), represent the early developmental state, viz. the inner cell mass of the blastocyst [12].These PSCs have great potential for research in various fields and for clinical approaches owing to their self-renewal ability and potential to differentiate into all three germ layers [63].Mitochondrial function is closely associated with the self-renewal and differentiation abilities of PSCs [16,17,28,49,68,70].In general, adult somatic cells obtain energy, adenosine triphosphate (ATP), from aerobic metabolism based on oxidative phosphorylation (OXPHOS).Alternatively, PSCs are heavily dependent on anaerobic glycolysis; this may be correlated with mitochondrial shape and the number of mitochondria in the cytoplasm [17,26].Mitochondrial morphologies vary 302 Page 2 of 18 depending on cell types.Differentiated cells generally possess elongated and branched, rod-shaped mitochondria with well-developed cristae, and a denser matrix.PSCs possess globular-shaped mitochondria with poorly developed cristae and a loose matrix compared with differentiated cells [8,13,14].Furthermore, it has been reported that specific mitochondrial features, such as morphology, localization, and number, are crucial factors for maintaining pluripotency and cellular differentiation [30].
One of the main effector proteins associated with mitochondrial fission is Dnm1l (DRP1 in humans), which functions as a GTPase [4].Cytosolic Dnm1l proteins are recruited to the mitochondrial outer membrane and form multiple helical complexes to conduct mitochondrial fission by constricting the mitochondria [15].Loss of Dnm1l results in embryonic lethality as early as embryonic day (E) 11.5 [59], indicating the importance of Dnm1l in normal embryonic development and survival.Moreover, studies using knockout mouse models suggested that Dnm1l plays a critical role in the normal functioning of tissues such as the brain [21,59] and heart [50].These results were consistent with the observation in humans that the mitochondrial fission protein DRP1 is essential for neuronal function [40] and is involved in various types of cardiovascular diseases [35].However, the detailed mechanism and etiology of Dnm1l for diseases and embryonic development have not yet been elucidated.Ishihara et al. reported that Dnm1l knockout (Dnm1l −/− ) ESCs and fibroblasts grow slightly slower than their control counterparts [21].In addition, Dnm1l protein levels are correlated with cyclin-Cdk complexes and, thus, may be involved in cell cycle transition [47].The GTPase activity of Dnm1l is directly regulated by cyclin B-Cdk1 complexes and leads to mitochondrial fission during the G2/M cell cycle transition [51].Therefore, a deficiency of the Dnm1l protein causes G2/M cell cycle retardation [43].Although there have been several reports regarding the association between Dnm1l and the cell cycle, further studies on different cell types and cell conditions are necessary to understand the correlation between mitochondrial dynamics and metabolism and the cell cycle.Notably, knockout effects of the mitochondrial fission gene Dnm1l have been poorly reported in mouse ESCs.Recently, we reported that knockout of mitochondrial fission-related genes, including Dnm1l, Fis1, and Mff, could modulate energy metabolism in ESCs [48].Here, we aimed to investigate the effect of knockout of Dnm1l, which is the most effective gene for metabolic change in ESCs, on energy metabolism, the cell cycle, stress-related phenotypes, and differentiation ability in ESCs.

Chemicals
Chemicals were purchased from Sigma Chemical Co.(St.Louis, MO) unless stated otherwise.

CRISPR/cas9 construct design
Clustered regularly interspaced short palindromic repeats (CRISPR)/cas9 single guide RNA (sgRNA) design tool, developed by Dr. Zhang's laboratory (http:// zlab.bio/ guidedesign-resou rces), was used to design sgRNA for Cas9 nuclease targeting Dnm1l gene in the mouse genome.Target sequences of several genes in exon were selected following the analysis of each gene.The target sequence, 5′-AAG TGT CAG GTT GAC AAC GT-3′ (20 nucleotides), including the protospacer adjacent motif (PAM; 5′-NGG-3′), was selected for its predicted high score and reduced off-target effects.The sequence including antisense sequence and restriction enzyme site sequence was synthesized as oligomers for cas9/ sgRNA expression vector (px330 Cas9/sgRNA dual expression system from Hyongbum Kim's laboratory in Yonsei University (Republic of Korea)) cloning (Bionics, Seoul, Republic of Korea).The process of cloning was as per previous report [44].

Electroporation and establishment of knockout cell line
After E14tg2a mESCs were dissociated into single cells using trypsin-EDTA (Gibco), they (5 × 10 5 cells/50 µL) were resuspended in R buffer (Neon transfection system, Invitrogen) with 5 µg px330 plasmid vector, which contained dnm1l sgRNA and Cas9.Electroporation was performed with 1200 V, 10 ms width, and 1 pulse, according to the manufacturer's instruction.The electroporated cells were expanded in mESC medium for 3 days.Individual colonies from a single cell were picked and plated onto 96-well plates and expanded.Genomic DNA was extracted using Quick-Extract (Epicentre) according to manufacturer's instruction.The target region was amplified, cleaved by T7 endonuclease I (NEB), and subjected to paired-end read sequencing using Illumina MiSeq at LAS (Illumina, San Diego, CA, USA).

RNA isolation and real-time PCR analysis
Total RNA was isolated using a RNeasy Mini Kit (Qiagen, Venlo, Netherlands, http:// www.qiagen.com), and then treated with DNase to remove genomic DNA contaminants according to the manufacturer's instructions.cDNA was synthesized from 1 mg total RNA using SuperiorScript III cDNA Synthesis Kit (Enzynomics, Daejeon, Republic of Korea).For real-time PCR, standard curves were created for each target gene primer set using known quantities of total complementary DNA (cDNA), mitochondrial DNA (mtDNA), and nuclear DNA (nDNA) from other cells.PCR reactions were performed in triplicate using a TOPreal™ qPCR 2X PreMIX (Enzynomics, Daejeon, Republic of Korea) and Roche LightCycler 5480 following appropriate instructions.Target genes were amplified by 45 cycles of 95 °C, 60 °C, and 72 °C for 10 s each.We corrected the differences in PCR efficiency between the target and reference loci using efficiency correction in the Relative Quantification Software (Roche LC 480).The primers for real-time PCR are shown in Table 1.

Teratoma formation
In this study, the cells (5 × 10 5 ) were injected into BALB/c nude (5-week-old, male) mice that had been purchased from Orient Bio (Gyeonggi-do, Korea) for experimental purpose.Each cell line was injected into the testis capsule of the mouse.The teratomas formed were harvested from the mouse at 4 weeks post-injection and fixed in 4% paraformaldehyde (Sigma), embedded in paraffin, and sectioned.To analyze the three-germ layer differentiation potential in vivo, the sectioned slides were histologically stained with hematoxylin/eosin (Endoderm), Alcian blue (Mesoderm), and anti-TUJ1 antibody (Ectoderm).

Electron microscopy
For transmission electron microscopy (TEM), samples were fixed in 4% paraformaldehyde (Sigma) and 2.5% glutaraldehyde (Sigma) in a 0.1 M phosphate (Sigma) buffer for 24 h.After washing with 0.1 M phosphate buffer, the samples were post-fixed in 1% osmium tetroxide (Sigma) containing 0.1 M phosphate buffer for 1 h.Next, the samples were serially dehydrated with graded ethanol (Merck Millipore).The dehydrated samples were embedded in Epon 812 and incubated at 60 °C for 3 days for polymerization.Ultrathin Sections (60-70 nm) were obtained by ultramicrotome (Leica Ultracut UCT).Ultrathin sections collected on a grid (200 mesh) were examined by a TEM (JEM 1010) operating at 60 kV, and images therefrom were recorded by a chargecoupled device (CCD) camera (SC1000; Gatan).

Mitochondrial length analysis
Images from electron microscopy were analyzed and measured using the ImageJ 1.43 (NIH) software for calculating the maximum (Max)/minimum (Min) ratio of mitochondrial length.At least over fifty mitochondria were measured and analyzed per sample to obtain reliable data.

Flow cytometry analysis
Each cell line was enzymatically dissociated by 0.25% trypsin with mechanical pipetting and washed twice by phosphate-buffered saline (PBS, Welgene, Gyeongsan, Korea).
After centrifugation at 1500 rpm for 5 min and removal of supernatant, 70% ethanol was added slowly dropwise while vortexing moderately.The fixed cells were stored at 4 °C for 2 h.After centrifugation at 2000 rpm for 7 min and removal of ethanol, the fixed cells were washed twice with PBS.Thereafter, the fixed cells were resuspended in propidium iodide staining solution (50 µg/mL of propidium iodide (PI; Sigma) and 2 µg/mL of DNase-free RNase (Thermo Fisher Scientific)) and incubated for 10 min at 37 °C in the dark.Stained cells were filtered through a 70-µm nylon mesh (Falcon) to remove any cell clump.Flow cytometry was performed using a CytoFLEX Flow Cytometer (Beckman Coulter) following the manufacturer's instructions.

Oxygen consumption rate analysis
For measuring the oxygen consumption rate (OCR), we used the Seahorse extracellular flux (XFp) analyzer.A total 2 × 10 4 cells of the control or Dnm1l knockout ESC were attached in XFp Cell Culture Microplate pre-coated with FBS in mESC medium 24 h before the assay.After the medium change to XF base media supplemented with d-glucose (1 g/L), Sodium pyruvate (1 mM) and l-glutamine (4 mM), the assay was performed by using XFp Extracellular Flux Analyzer (Seahorse Bioscience, North Billerica, MA, USA).Five measurements were obtained under basal conditions and after the addition of several chemicals such as oligomycin (1 µM), FCCP (2 µM) and rotenone (1 µM)/ antimycin A (1 µM).The process after treatment was performed following the manufacturer's instructions.

Extracellular acidification analysis
To measure the extracellular acidification rate (ECAR), we used a Seahorse extracellular flux (XF96) analyzer.A total of 1.5 × 10 4 cells were cultured in XF96 Cell Culture Microplate, pre-coated with Matrigel in mESC medium, 24 h before the assay.After a medium change to XF base media supplemented with L-glutamine (4 mM, Gibco, 25030-081), the assay was performed using XF96 Extracellular Flux Analyzer (Seahorse Bioscience, North Billerica, MA, USA).Four measurements were obtained under basal conditions and after the addition of several chemicals, such as D-glucose (10 mM), oligomycin (1 µM), and 2-deoxy-D-glucose (50 mM) (Agilent Technologies, 103020-100).The process after treatment was performed according to the manufacturer's instructions.

Direct measurement of adenosine triphosphate level
To measure the adenosine triphosphate (ATP) levels, we used a luminescent ATP detection assay kit (Abcam, ab113849) with GloMax 96 Microplate Luminometer (Promega, E6521, Madison, WI, USA).A total of 1 × 10 4 cells were cultured in 96-well cell culture plates (SPL life science, 30296, Gyeonggi-do, Korea) pre-coated with Matrigel in mESC medium, 24 h prior to the assay.The process after treatment was performed according to the manufacturer's instructions.

Direct measurement of the alpha-ketoglutarate amount
To measure the amount of alpha-ketoglutarate, we used a colorimetric alpha-ketoglutarate assay kit (Sigma, MAK054) with xMark Microplate Absorbance Spectrophotometer (Bio-RAD, 1681150, Hercules, CA, USA).A total of 1 × 10 4 cells were cultured in 96-well cell culture plates (SPL life science, 30196, Gyeonggi-do, Korea), pre-coated with Matrigel in mESC medium, 24 h prior to the assay.The process after treatment was performed according to the manufacturer's instructions.

Detection of reactive oxygen species
To detect reactive oxygen species, we used the cell-permeant reagent 2', 7' -dichlorofluorescein diacetate (DCF-DA) in the cellular ROS assay kit (Abcam, ab113851).A total of 1 × 10 4 cells were cultured in 96-well cell culture plates (SPL life science, 30196, Gyeonggi-do, Korea), pre-coated with Matrigel in mESC medium, 24 h prior to the assay.The process after treatment was performed according to the manufacturer's instructions.

RNA-seq analysis
Total RNA samples were converted into cDNA libraries using the TruSeq Stranded mRNA Sample Prep Kit (Illumina).Starting with 100 ng of total RNA, polyadenylated RNA (mainly mRNA) was selected and purified using oligo dT-conjugated magnetic beads.This mRNA was physically fragmented and converted into single-stranded cDNA using reverse transcriptase and random hexamer primers, with the addition of Actinomycin D to suppress DNA-dependent synthesis of the second strand.Double-stranded cDNA was created by removing the RNA template and synthesizing the second strand in the presence of dUTP (deoxy-ribo-uridine triphosphate) in place of dTTP (deoxythymidine triphosphate).A single A base was added to the 3 prime end to facilitate ligation of sequencing adapters, which contain a single T base overhang.Adapter-ligated cDNA was amplified by polymerase chain reaction to increase the amount of sequence-ready library.During this amplification, the polymerase stalls when it encounters a U base, rendering the second strand a poor template.Accordingly, the amplified material used the first strand as a template, thereby preserving the strand information.Final cDNA libraries were analyzed for size distribution and using an Agilent Bioanalyzer (DNA 1000 kit; Agilent), quantitated by qPCR (Kapa Library Quant Kit; Kapa Biosystems, Wilmington, MA, USA), then normalized to 2 nM in preparation for sequencing.Validity of sequenced reads were accessed with FastQC (v0.11.5) and the reads were aligned with STAR (v2.5.2b) to UCSC mouse mm10 genome assembly.StringTie (v1.3.3b) was used to process STAR results into transcript assemblies.Results from StringTie were used to calculate average expression of each gene in FPKM and the acquired values were used for drawing scatter plots in R (v3.6.1).Read counts were normalized for read depth and analyzed using the BioJupies package [54] with default parameters.A notebook containing the code, a list of differentially expressed genes and an overview of the analysis can be found at the following URL: https:// amp.pharm.mssm.edu/ bioju pies/ noteb ook/ CKbmU QhlJ.

Ethics statement for animal use
Experiments were performed following the approved guidelines and all experimental protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of Konkuk University.All mouse strains were bred and housed at the mouse facility of the Konkuk University or were purchased from Orient-Bio Inc. (Gyeonggi-do, Korea; http:// www.orient.co.kr).Animal welfare was under the control of local committees.Mice were housed in a temperaturecontrolled room with automated darkness-light cycle system and fed with a regular ad libitum feeding.Before testis 302 Page 6 of 18 harvesting, the mice were sacrificed by carbon dioxide inhalation.

Statistical analysis
All experiments were performed in triplicate and data represented as means ± standard deviation.The significance of differences was assessed by an unpaired two-tailed student's t test, one-way analysis of variance (ANOVA) with Tukey's Honestly Significant Difference post hoc for multiple comparisons.p values < 0.05 were considered statistically significant.

Characterization of the pluripotency of Dnm1l −/− embryonic stem cells (ESCs)
We previously reported that knockout of mitochondrial fission genes, Dnm1l, could maintain pluripotency; however, loss of this gene induced metabolic changes in ESCs [48].Dnm1l biallelic knockout (at Exon 5) mouse ESCs were generated by the CRISPR/cas9 gene-editing system (Fig. S1A).Since monoallelic Dnm1l ± ESCs showed a negligible difference in several aspects compared with wild-type ESCs (data not shown), biallelic Dnm1l −/− ESCs were used for further analyses.Thereafter, we assessed whether any off-target mutations occurred in biallelic Dnm1l −/− ESCs.Whole genome sequencing analysis revealed that there was an indel on-target site (Fig. S1B).However, there were no possible indels in various predicted off-target sites (Fig. S1C).
To test the differentiation potential of all three germ layers [6], the teratoma formation assay and in vitro differentiation analysis were conducted using Dnm1l −/− and control ESCs.The Dnm1l −/− ESCs differentiated into all three germ layers, including smooth muscle actin (SMA, mesoderm), endothelium (Sox17, endoderm), and neuron (Tuj1, ectoderm), in vitro (Fig. S2B).The teratoma formation assay revealed that Dnm1l −/− ESCs could form teratomas (Fig. 1E) containing endodermal (gut epithelium), mesodermal (cartilage), and ectodermal (Tuj1 + neuron) tissues (Fig. 1F).The size of teratoma formed from Dnm1l −/− ESCs was significantly smaller than that of teratoma formed from control ESCs (Fig. 1E).Collectively, these findings indicated that knockout of Dnm1l in ESCs could affect the expression pattern of pluripotency genes, but the ability of ESCs to selfrenew and differentiate potential to the three germ layers was not severely affected.

Intracellular characteristics of Dnm1l −/− ESCs
As Dnm1l is one of the mitochondrial fission-related genes, changes in the mitochondrial morphology of Dnm1l −/− ESCs were examined using electron microscopy (Fig. 1G).As expected, the morphology of the mitochondria of Dnm1l −/− ESCs was distinct from that of control ESCs (Fig. 1G).Generally, the morphology of mitochondria in ESCs and other pluripotent stem cell types is globularshaped with immature cristae [8,13,14].The mitochondria of Dnm1l −/− ESCs were more elongated and rod-shaped and showed more mature cristae than those of the control ESCs (Fig. 1G, left).Next, we measured the calculated-maximal (c-max) and calculated-minimal (c-min) lengths of mitochondria in the control and Dnm1l −/− ESCs (Fig. 1G, right).The ratio values of c-max/c-min from control ESCs (1.41) and Dnm1l −/− ESCs (2.06) indicated that the mitochondria of Dnm1l −/− ESCs were more elongated than those of control ESCs (Fig. 1G).

Dnm1l deletion in ESCs results in metabolic changes and excessive reactive oxygen species (ROS) production
We performed a series of analyses to clarify how the deletion of Dnm1l alters the cellular metabolic processes in ESCs.Considering the coupling of mitochondrial fission and mitochondrial metabolic pathways such as glycolysis, tricarboxylic acid cycle, and oxidative phosphorylation [18], we hypothesized that Dnm1l deficiency might influence overall cellular metabolism.As previously reported [48], Dnm1l −/− ESCs are more dependent on the OXPHOS than on glycolysis (Fig. 2A, B).Likewise, XF cell energy phenotype analysis showed that Dnm1l −/− ESCs moved to the energetic phase, indicating that Dnm1l −/− ESCs utilized OXPHOS upon oxidative stress or excessive energy demand, whereas control ESCs moved to the glycolytic phase (metabolic state focused on producing cellular compounds).These results indicate that knockout of Dnm1l in ESCs triggers a metabolic shift from glycolysis to OXPHOS (Fig. S3A).In particular, Dnm1l −/− ESCs displayed higher spare respiratory capacity (SRC) and lower glycolytic capacity than control ESCs (Fig. 2B and D).Also, Dnm1l −/− ESCs displayed lower proton leak than control ESCs (Fig. S3B).Thereafter, total cellular ATP concentration was determined using the luminescent ATP detection assay (Fig. S3C).As expected, (from the higher SRC levels), Dnm1l −/− ESCs produced more intracellular ATP than the control ESCs (Fig. 2B, Fig. S3C).Besides, Dnm1l −/− ESCs displayed higher non-mitochondrial oxygen consumption (Fig. S3D).
As Dnm1l −/− ESCs used more aerobic metabolism than control ESCs, we hypothesized that the mitochondria of Dnm1l −/− ESCs may produce more ROS.To this end, superoxide levels were measured using MitoSOX™ Red, which exhibits red fluorescence when oxidized by superoxide radicals [24]; the more superoxide levels, the more red fluorescence is emitted (Fig. 2E).As expected, the mitochondria of Dnm1l −/− ESCs produced more superoxide than the control ESCs (Fig. 2E).Thereafter, fluorescence-activated cell sorting (FACS) analysis also confirmed increased MitoSOXreactive fluorescence in Dnm1l −/− ESCs than in control ESCs (Fig. 2F).
Next, we analyzed the expression patterns of ROS-related genes.Nrf2 regulates the gene expression of endogenous antioxidant synthesis and ROS-eliminating proteins [23].To this end, genes related to the glutathione metabolic process, such as Keap1 and Nfe2l2 (Nrf2), were overexpressed in Dnm1l −/− ESCs (Fig. 2G).To gain a better understanding of the Nrf2 signaling cascade, expression patterns of Nrf2 target genes that were reported previously [31,39,53] were investigated (Fig. 2H).Forty Nrf2 target genes were found to be differentially expressed between the Dnm1l −/− and control ESCs.Notably, mitochondrial superoxide dismutase 2 (Sod2) was overexpressed in Dnm1l −/− ESCs, while the expression levels of Sod1 and Sod3 were similar (Fig. 2I).Collectively, these results demonstrated that Dnm1l deficiency in ESCs caused an increase in OXPHOS, which is closely related to excessive ROS production and subsequently triggers the Nrf2 signaling cascade.

Dnm1l deletion in ESCs induces telomere shortening and the activation of 2-cell-stage-specific genes
Several studies have reported that loss of Dnm1l results in cell cycle perturbation in Drosophila melanogaster [32], breast cancer cells [43], and normal rat kidney cells [34].
To broaden the knowledge of the effect of Dnm1l deficiency in the cell cycle of pluripotent stem cells, the cell cycle was investigated in Dnm1l −/− ESCs.Dnm1l −/− ESCs formed a smaller colony than the control ESCs after 3 days of passaging (Fig. 3A).In addition, the proliferation assay revealed that the proliferation rate of Dnm1l −/− ESCs was much lower than that of control ESCs (Fig. 3B).As cell cycles were extended in response to telomere shortening to repair telomeres [38], telomere lengths in the control and Dnm1l −/− ESCs were compared.Telomere length in Dnm1l −/− ESCs was found to be shorter than that in control ESCs (Fig. 3C).Thereafter, we further analyzed detailed cell cycle phases of Dnm1l −/− ESCs using FACS analysis after propidium iodide staining.Normal ESCs are known to display extremely rapid and distinct cell division cycles compared with other somatic cell types [62].The cell cycle of Dnm1l −/− ESCs was compared with that of control ESCs and fibroblast cells as experimental controls (Fig. 3D).Fibroblasts primarily remained in the G0/G1 phase (80.08%), while control ESCs showed a much lower ratio of the G0/G1 phase (43.39%) and much higher ratios of the S (34.51%) and G2/M phases (22.34%) (Fig. 3D, left and middle), confirming the rapid and distinct cell division cycle of pluripotent ESCs.Although the S phase ratio of Dnm1l −/− ESCs (33.26%) was indistinguishable from that of control ESCs (34.51%), the G2/M phase ratio of Dnm1l −/− ESCs (32.88%) was much higher than that of control ESCs (22.34%) (Fig. 3D, middle and right).The ratios of the G0/G1 phase of Dnm1l −/− ESCs (33.65%) were much lower than those of control ESCs (43.39%) (Fig. 3D, middle and right).Considering the slower proliferation of Dnm1l −/− ESCs than the controls, Dnm1l deficiency in ESCs may be affected by cell cycle length of the G0/G1 and G2/M phase.
Telomere repair and elongation systems are activated upon telomere shortening, and telomeres could be effectors in cell cycle change [19,37,38,64,67,69].Thus, we subsequently analyzed the expression levels of telomere elongation-related genes, such as Zscan4, Tcstv1/3, and Tert, in Dnm1l −/− ESCs.As expected, the Zscan4 family, Tcstv1/3, and Tert genes were overexpressed in Dnm1l −/− ESCs (Fig. 3E).Zscan4 becomes active after sensing shortened telomeres, which extends the cell cycle length owing to the repair of the telomere presumably at the G2/M phase [38].The Zscan4 protein is known to directly elongate the telomere via homologous recombination following the recruitment of SPO11 [67].In addition, ectopic expression of Tcstv1/3 extends telomere length in ESCs with Zscan4 activation [69].As Zscan4 is a marker for the 2-cell-stage embryo [46], we further analyzed the expression levels of other 2-cell-stage-specific genes.Diego et al. reported that a small number of cells expressing 2-cell-stage-specific genes are present among cultured ESCs [46].Notably, most 2-cell-stage-specific genes were found to be differentially expressed in Dnm1l −/− ESCs (Fig. 3F).Collectively, these data suggest that deletion of Dnm1l in ESCs results in a change in the expression levels of 2-cell-stage-specific genes and reduced proliferation rates, which may be induced by G2/M retardation and telomere shortening.

Dnm1l deletion in ESCs attenuates the dissolution of pluripotency under differentiation induction
As mentioned above, Dnm1l −/− ESCs have distinct cell cycle and differential expression of 2-cell-stage-specific genes.Previous studies reported that ESCs have an increased susceptibility to differentiation during the G1 phase [9,42].Thus, we hypothesized that the lower ratio of the G0/ G1 phase in Dnm1l −/− ESCs might attenuate the dissolution of pluripotency or hamper the differentiation from the pluripotent state (Fig. 3D, right).To confirm this, the Dnm1l −/− and control ESCs were differentiated for 14 days via embryoid body (EB) formation in vitro.Notably, most Dnm1l −/− ESCs maintained Oct4 expression, while Oct4expressing cells were not examined in differentiationinduced cells from control ESCs (Fig. 4A), indicating the suppressive effect of Dnm1l deficiency on differentiation.Moreover, Dnm1l −/− ESCs maintained high levels of Nanog; however, control ESCs did not express Nanog after 14 days of differentiation (Fig. 4B).When these differentiated cells from Dnm1l −/− ESCs were replated on ESC culture medium (+ LIF), AP-positive ESC colonies were formed (Fig. 4C), confirming that several Dnm1l −/− ESCs maintained a pluripotent state even after 14 days of differentiation induction.
Recently, Santosha et al. demonstrated that α-ketoglutarate is related to Nanog expression and enhances the self-renewal ability of ESCs [55].Thus, the amount of α-ketoglutarate was measured in Dnm1l −/− ESCs.As expected, Dnm1l −/− ESCs contained a higher amount of α-ketoglutarate than control ESCs (Fig. 4D), indicating the correlation between Nanog upregulation and the level of α-ketoglutarate in Dnm1l −/− ESCs.Thereafter, we investigated whether the gene expression levels of Nanog could affect the dissolution of pluripotency in Dnm1l −/− ESCs.To induce the upregulation of Nanog, we treated Dnm1l −/− and control ESCs with two small molecule inhibitors (2i), PD032591 (MEK inhibitor) and CHIR99021 (GSK3 inhibitor), which are known to positively regulate Nanog [33].PSCs cultured in a medium containing 2i without a feeder layer were maintained in a naïve pluripotent state [2].Thus, we tested whether Dnm1l −/− ESCs, which expressed high Nanog levels, could self-renew without 2i.Control ESCs self-renewed and formed colonies in naïve pluripotency medium in the presence of 2i; however, they could not maintain ESC colonies in the absence of 2i (Fig. 4E, F).Surprisingly, Dnm1l −/− ESCs retained colonies over five passages in the naïve pluripotency medium even without 2i (Fig. 4E, F), indicating that higher levels of Nanog in Dnm1l −/− ESCs could compensate the effects of MEK and GSK3 inhibition induced by 2i treatment (Fig. 4E, F).These results suggest that upregulated Nanog expression may be the cause of the attenuation of pluripotency dissolution in Dnm1l −/− ESCs.

G2/M stage retardation in ESCs causes attenuation of the dissolution of pluripotency
Thereafter, we determined whether G2/M stage retardation could be an effector for the inhibition of differentiation of ESCs.Accordingly, we tested whether cell cycle retardation in the G2/M phase, which is a phenotype of Dnm1l deletion, leads to the inhibition of differentiation in wild-type ESCs.Charles et al. reported that RO-3306, an ATP-competitive and selective CDK1 inhibitor, and ABT-75, a microtubule polymerization inhibitor, are reversible cell cycle arrest inducers that can be used for cell cycle synchronization via the G2/M block [65].Excessive concentration of these inhibitors caused cell cycle arrest and ESCs remained in a single cell state without colony formation (Fig. S4A).Thus, to induce cell cycle retardation, which may induce the formation of smaller colonies, wildtype ESCs were treated with permissive concentrations of RO-3306 and ABT-75 in ESC medium (Fig. 5A).ABT-751 treatment (100 and 300 nM) for 3 days could induce cell cycle retardation (forming smaller colonies); however, it did not affect the expression levels of Nanog (Fig. 5B).RO-3306 treatment at higher concentrations of 3 and 5 µM induced cell cycle retardation with upregulation of Nanog, approximately 2-and 3.5-fold, respectively (Fig. 5A and  B).Collectively, treatment with 3 µM RO-3306 simulated the effect of Dnm1l deletion, i.e., G2/M phase retardation and twofold expression of Nanog.Thereafter, we investigated whether RO-3306 treatment (3 µM) could affect the differentiation potential of ESCs.Notably, 3 µM RO-3306-treated ESCs showed a tendency to not differentiate; undifferentiated colonies were maintained even after 7 days of differentiation induction (Fig. 5C).RO-3306 treatment groups (3 µM) maintained much higher levels of Nanog than those in the control, even after day 7 of differentiation (Fig. 5D).These results indicate that G2/M retardation without genetic modification can inhibit the differentiation of ESCs, which is a similar phenomenon as observed in Dnm1l −/− ESCs.

Chemically induced metabolic shift from glycolysis to OXPHOS attenuates the dissolution of pluripotency
One of the distinct phenotypes of Dnm1l deletion in ESCs was metabolic shift from glycolysis to OXPHOS.Hence, we tested whether chemical-induced metabolic shift without genetic modification could affect the differentiation ability of wild-type ESCs.Dichloroacetate (DCA) is a drug that diverts metabolism from glycolysis to OXPHOS [52].First, we determined the effect of DCA on the metabolic shift in wild-type ESCs by measuring ROS production.Wild-type ESCs were treated with DCA (1, 3, 5 and 10 mM) and ROS production was analyzed using dichlorofluorescein diacetate (DCF-DA) staining (Fig. S4B).The higher the DCA concentration, the more ROS was detected.Thereafter, we assessed the differentiation ability of DCA-treated ESCs to investigate whether the metabolism shift from glycolysis to OXPHOS could inhibit the differentiation of ESCs.Notably, DCA-treated ESCs were not differentiated; they maintained undifferentiated colonies even after 7 days of differentiation induction, indicating that metabolic shift from glycolysis to OXPHOS could attenuate the dissolution of pluripotency of ESCs (Fig. 5E).Nanog expression increased with an increase in DCA concentration (Fig. 5F).Therefore, we assumed that maintaining glycolytic metabolism in a pluripotent state or at the onset of differentiation may be required for the dissolution of pluripotency of ESCs.Given that differentiated somatic cells primarily use OXPHOS rather than glycolysis, the metabolic shift from glycolysis to OXPHOS might be a subsequent event after differentiation has progressed.Taken together, Dnm1l deficiency in embryonic stem cells (ESCs) led to massive changes in metabolic shift from glycolysis to OXPHOS, ROS increase, 2-cell-specific gene expression, G2/M arrest, which were the main causes of attenuation of the dissolution of pluripotency (Fig. 5G).

Global gene expression patterns of Dnm1l −/− ESCs
Lastly, RNA-seq analysis was performed to compare global gene expression patterns between control and Dnm1l −/− ESCs.At first, we performed principal component analysis (PCA) to validate sample variation and clustering as a part of quality control.PCA confirmed that the control and Dnm1l −/− ESCs were evidently separated (Fig. 6A).The volcano plot showed the differentially expressed genes (log 2 |fold change|> 1.5, p > 0.05) between the control and Dnm1l −/− ESCs (Fig. 6B).Forty-two genes were downregulated, and sixty-nine genes were upregulated in Dnm1l −/− ESCs compared with those in control ESCs.A total of 5940 genes were further quantified using differential gene expression methods by comparing Dnm1l −/− ESCs with control ESCs (log 2 |fold change|> 0) (Fig. 6C).Thereafter, we investigated the biological function of the genes differentially expressed in Dnm1l −/− ESCs compared with those in control ESCs using five bioinformatic approaches, including Gene Ontology of biological process and molecular function, Kyoto Encyclopedia of Genes and Genomes, REACTOME, and Wikipathway using Biojupies [54], a bioinformatical platform (Fig. 6D).Notably, glutathionerelated terms were enriched in all categories of upregulation (Fig. 6D).We also confirmed that amino acid-related terms were enriched in four categories of downregulation, except wikipathways (Fig. S5).Therefore, the most prominent characteristic feature of Dnm1l deficiency in ESCs may be the functional change to reduce ROS production induced by metabolic shift.

Discussion
In this study, we demonstrated that deletion of Dnm1l, a mitochondrial fission-inducing gene, in ESCs caused changes in mitochondrial morphology, cellular metabolism, telomere length, ROS production, expression levels of ROS-related genes, cell cycle, and differentiation ability (dissolution of pluripotency).We focused on the mechanism of attenuation of dissolution of pluripotency via Dnm1l deletion.Dnm1l −/− ESCs could differentiate into three germ layers, both in vitro and in vivo; however, the size of the teratoma was smaller than that of control ESCs.Moreover, Dnm1l −/− ESCs maintained Oct4 expression and dome-like colonies even after differentiation induction for 14 days through EB formation in vitro.Here, we suggested that G2/M phase retardation, high levels of Nanog and differentially expressed 2-cell-specific genes, and metabolic shift from glycolysis to OXPHOS were the main causes for the inhibition of the differentiation of Dnm1l −/− ESCs.
Several studies have reported an association between Dnm1l and the G2/M phase of cell cycle.Zunino et al. showed that SENP5, a SUMO proteinase, modulates DNM1L-dependent fission during the G2/M phase [71].We also showed that the cell cycles of Dnm1l −/− ESCs were retarded at the G2/M phase and the length of the G1 phase in Dnm1l −/− ESCs was shorter than that in control ESCs.As differentiation from ESCs primarily occur during the G1 phase [9,42], an extended G2/M phase and a reduced G1 phase could be the cause for inhibition of the differentiation of Dnm1l −/− ESCs by reducing its time window.
It is well known that upregulation of Nanog and 2-cellstage-specific genes affects the differentiation of ESCs [38,46].Practically, no changes were observed in the expression patterns of pluripotency-related genes after Dnm1l deletion; however, the upregulation of Nanog and differentially expressed 2-cell-stage-specific genes was observed, indicating that these genes could be associated with the inhibition of differentiation of Dnm1l −/− ESCs.Nanog-overexpressing ESCs could maintain stemness without LIF and feeder cells [20,38], which may be a type of differentiation inhibitory effect of high levels of Nanog.Diego et al. suggested that less than 1% of ESC culture expressed 2-cell-stage-specific genes and these rare 2-cell-like cells show higher developmental plasticity and exit pluripotent state through a different way than that of common ESC differentiation [46].In addition, ESCs expressing 2-cell-stage-specific Zscan4 express higher levels of Nanog [46], indicating that Nanog expression is correlated with Zscan4.Therefore, it will be interesting to compare the characteristics of Dnm1l −/− ESCs with 2-celllike cells, as they share specific gene expression patterns, such as high levels of Nanog and 2-cell-stage-specific genes.Metabolic shift from glycolysis to OXPHOS could be a reason for the inhibition of the differentiation of Dnm1l −/− ESCs.We showed that Dnm1l deletion results in an increase in ROS production via mitochondrial elongation and increased OXPHOS metabolism.Chemically induced metabolic shift from glycolysis to OXPHOS by DCA treatment could also inhibit the differentiation of ESCs.Collectively, the metabolic shift from glycolysis to OXPHOS, which is induced by Dnm1l deletion or DCA treatment, could attenuate the dissolution of pluripotency.PSCs, particularly ESCs, have globular-shaped mitochondria with immature cristae [7].However, our results revealed that Dnm1l −/− ESCs had more elongated mitochondria and a relatively mature form of cristae compared with those of control ESCs.Notably, Dnm1l −/− ESCs showed increased levels of ROS compared to control ESCs.ROS are primarily produced in the electron transport chain [3,27].ROS and glycolysis inhibit each other; ROS inhibits several glycolysisrelated proteins and glycolysis reduces ROS via NADPH and glutathione production [36].Therefore, the increase in ROS production in Dnm1l −/− ESCs must be due to the changes in energy metabolism induced by elongated mitochondria.Several genes related to the Keap1-Nrf2 pathway and the SOD family, which are the major regulators of cytoprotective responses to cellular stresses [22], were also activated in Dnm1l −/− ESCs.This phenomenon might also be a counteraction against excessive ROS produced in Dnm1l −/− ESCs.In addition, considering the glycolytic function data, overall, Dnm1l −/− ESCs appear to be metabolically less active, but carry a higher spare respiratory capacity.This is probably a result of larger mitochondria or increased fusion.
One of the phenotypes of Dnm1l −/− ESCs was telomere shortening.Zscan4 is associated with telomere length [67].Several researchers have demonstrated that Zscan4 is activated by sensing telomere shortening and subsequently facilitating telomere elongation via inhibition of DNA methylation [10,38].In the present study, we showed that telomere shortening and activation of Zscan4 were induced in response to Dnm1l deletion in ESCs.Telomere shortening can also be induced by excessive ROS levels [25,56,57].Therefore, increased ROS production via mitochondrial elongation and increased OXPHOS metabolism in response to Dnm1l deletion may lead to telomere shortening through undefined mechanisms.
This study was performed using pluripotent mouse ESCs, which have rarely been reported for studying the relationship between mitochondrial dynamics-related genes and cellular differentiation.Among the mitochondrial dynamics-related genes, Dnm1l was the most effective in mitochondrial morphology and energy metabolism, and the deletion of Dnm1l eventually attenuated the dissolution of pluripotency.Although the mechanism is unclear, Dnm1l, as a mitochondrial fragmentation and apoptosis inducer, is associated with several diseases in humans.The unbalanced post-translational modification of Drp1 (common name of Dnm1l in humans) leads to various neurological diseases such as Alzheimer's, Huntington's, and Parkinson's diseases [1,11,45].Thus, treatment with Mdivi-1, which is a well-known inhibitor of mitochondrial fission (or Drp1), is beneficial for treatment of such diseases [60].Moreover, cardiovascular diseases are also affected by Drp1-mediated mitochondrial dynamics [35].According to our current data, it is possible that these Drp1-related human diseases may be due to any defect in the differentiation process in the early stage of embryonic development.Further studies on Drp1 in humans need to be conducted to elucidate the etiology of neurological and cardiovascular diseases.

Fig. 1
Fig. 1 Generation and characterization of Dnm1l −/− embryonic stem cells (ESCs).A Bright field images of control and Dnm1l −/− ESCs.B Immunocytochemical analysis of the pluripotency markers, NANOG (top) and OCT4 (bottom) in control and Dnm1l −/− embryonic stem cells (ESCs).Scale bars: 200 µm.C Real time-polymerase chain reaction (PCR) analysis of core pluripotent genes, including Pou5f1 (Oct4), Sox2, and Nanog, in control and Dnm1l −/− ESCs.Data are presented as the mean ± SEM of the values from three independent experiments.D Western blot analysis of core pluripotent and Dnm1l proteins in control and Dnm1l −/− ESCs.Data are presented as the mean ± SEM of values from three independent experiments.E, F Teratoma formation and analysis of in vivo differentiation potential.E Teratoma formation in control and Dnm1l −/− ESCs.Control

Fig. 2 302 Page 10 of 18 Fig. 3
Fig. 2 Metabolic analysis of control and Dnm1l −/− embryonic stem cells (ESCs).A Measurement of oxygen consumption rate (OCR) in control and Dnm1l −/− ESCs.B Measurement of basal respiration, maximal respiration, spare respiratory capacity in control and Dnm1l −/− ESCs.Oligomycin; an ATP synthase inhibitor, p-trifluoromethoxyphenylhydrazone (FCCP); an uncoupling agent that collapses the proton gradient, R/A (rotenone/antimycin); a complex I inhibitor/a complex III inhibitor.C Measurement of extracellular acidification rate (ECAR) in control and Dnm1l −/− ESCs.D Measurement of glycolysis, glycolytic capacity, glycolytic reserve in control and Dnm1l −/− ESCs.2-DG (2-deoxy-glucose); a glucose analog, inhibiting glycolysis through competitive binding to glucose hexokinase.E Representative image of control and Dnm1l −/− ESC lines observed using mitoSOX.F Flow cytometry analysis of the proportion of control and Dnm1l −/− ESCs stained with mitoSOX.G Real time-polymerase chain reaction (PCR) analysis of ROS-related genes, including Keap1 and Nfe2l2 (Nrf2), in control and Dnm1l −/− ESCs.An unpaired t test: ***p < 0.001.Data are presented as the mean ± SEM of values from three independent experiments.H Heatmap of Nrf2 target gene expression patterns in control and Dnm1l −/− ESCs.I Real time-PCR analysis of superoxide dismutase 1, 2 (mitochondria-specific), and 3 in control and Dnm1l −/− ESCs.An unpaired t test: ***p < 0.001, n.s not significant.Data are presented as the mean ± SEM of values from three independent experiments ◂

302 Page 12 of 18 Fig. 4
Fig. 4 Dnm1l deficiency in mouse embryonic stem cells (ESCs) decreases the population of differentiated cells under stimulation of exit from pluripotency.A Immunocytochemical analysis of the Oct4 protein in control and Dnm1l −/− ESCs after 14 days of spontaneous differentiation.Scale bars: 200 µm.B Real time-polymerase chain reaction (PCR) analysis of Nanog in control and Dnm1l −/− ESCs at day 14 of spontaneous differentiation.An unpaired t test: **p < 0.01.Data are presented as the mean ± SEM of values from three independent experiments.C Representative image of control and Dnm1l −/− ESC lines observed after 14 days of spontaneous differentiation (left)

Fig. 5
Fig. 5 Chemical-induced G2/M retardation attenuates the dissolution of pluripotency in mouse embryonic stem cells (ESCs).A Representative image of control ESCs observed after 3 days of cell expansion in mESC medium with RO-3306 (1, 3, and 5 µM) or ABT-751 (100 and 300 mM).Scale bars: 200 µm.B Real time-polymerase chain reaction (PCR) analysis for Nanog expression in control ESCs, Dnm1l −/− ESCs, and control ESCs treated with RO-3306 or ABT-751 for 3 days.One-way analysis of variance (ANOVA) t test: ***p < 0.001.Data are presented as the mean ± SEM of values from three independent experiments.C Representative image of control ESCs observed after 7 days of spontaneous differentiation in differentiation condition w/ or w/o RO-3306.Scale bars: 200 µm.D Real time-PCR analysis of Nanog in control and ESCs at day 7 of spontaneous differentiation in differentiation condition w/ or w/o RO-3306.One-way ANOVA t test: *p < 0.05.Data are presented as

Table 1
Primer sets used for real-time PCR Forward Reverse