Drp1 regulates transcription of ribosomal protein genes in embryonic hearts

Originating from the engulfment of α-proteobacteria by precursors of modern eukaryotic cells, mitochondria have now evolved to become a vital organelle with multiple essential functions in eukaryotes, including supplying energy, regulating apoptosis, maintaining calcium homeostasis, sustaining redox homeostasis and generating crucial metabolites for cellular activities (Dorn, 2016; Dorn et al., 2015; Ernster and Schatz, 1981; Friedman and Nunnari, 2014; Sharma et al., 2019; Zhao et al., 2019). In addition to their roles in postnatal organs under normoxia, accumulated evidence has established that mitochondria are also required for fetal organogenesis, including embryonic heart development in hypoxia (Dorn et al., 2015; Zhao et al., 2019). Mitochondrial dysfunction in fetal hearts can lead to severe congenital cardiomyopathies and prenatal/neonatal lethality in human patients (Berardo et al., 2011; El-Hattab and Scaglia, 2016; Schiff et al., 2011). These heart defects cover a large spectrum, including hypertrophic, dilated and noncompaction cardiomyopathies, suggesting a complex regulatory role of mitochondria during heart development. To date, no specific treatment has been developed to target congenital cardiomyopathies caused by mitochondrial dysfunction. This clinical deficiency is at least partially due to the lack of sufficient knowledge of how mitochondrial defects affect normal heart development at the molecular, cellular and tissue levels.

The heart is the first functional organ to form in mammalian embryos (Buckingham et al., 2005; Christoffels et al., 2004; Sylva et al., 2014). In mouse embryos prior to embryonic day (E) 11.5, cardiac mitochondria are highly immature with few cristae protruding into the mitochondrial matrix, and cardiomyocytes rely almost exclusively on anaerobic glycolysis to acquire ATP (Dorn et al., 2015; Hom et al., 2011; Mackler et al., 1971). By E13.5, cardiac mitochondria are functionally mature with their electron transport and oxidative phosphorylation (OXPHOS) activities being indistinguishable from those in mitochondria of adult hearts (Beutner et al., 2014; Hom et al., 2011; Mackler et al., 1971). At this stage, the ultrastructure of cardiac mitochondria is very similar to fully matured mitochondria with tubular cristae connected to the periphery, and embryonic cardiomyocytes acquire energy through both anaerobic glycolysis and OXPHOS (Chung et al., 2007; Hom et al., 2011; Lopaschuk and Jaswal, 2010; Mackler et al., 1971; Zhao et al., 2019). By the end of gestation, >50% of total cardiac ATP is generated through OXPHOS with the remainder being generated through anaerobic glycolysis (Scarpulla, 2008; Zhao et al., 2019). Therefore, cardiac mitochondria already provide a large portion of the energy required to support pumping of the heart at mid- to late-gestation stages, and the cardiac metabolic shift from glycolysis to OXPHOS begins in embryos.

In response to extra-/intracellular stimuli, mitochondria undergo fission and fusion, leading to a highly dynamic tubular network in cells. Mitochondrial morphology is a major determining factor for the normal functions of mitochondria; disruption of mitochondrial fission or fusion through genetic and pharmacological approaches leads to defects in numerous in vitro and in vivo systems (Dorn, 2016; Dorn et al., 2015; Kageyama et al., 2011; Sharma et al., 2019; Tilokani et al., 2018; Zhao et al., 2019). A group of cellular dynamin-related GTPases are pivotal for mitochondrial dynamics, with MFN1, MFN2 and Drp1 (encoded by the Dnm1l gene) required for fusion and fission, respectively (Dorn, 2016; Dorn et al., 2015; Kageyama et al., 2011; Sharma et al., 2019; Tilokani et al., 2018; Zhao et al., 2019). The impact of mitochondrial dynamics on cardiogenesis was initially demonstrated by studies using the Nkx2.5-Cre driver to simultaneously inactivate Mfn1 and Mfn2 (Chen et al., 2011; Kasahara et al., 2013), which have partially overlapping functions in promoting mitochondrial fusion (Chen et al., 2003; Detmer and Chan, 2007). Double inactivation of Mfn1 and Mfn2 in embryonic cardiomyocytes aberrantly increases cytosolic calcium concentration and calcineurin activity, leading to enhanced Notch signaling, which impairs expression of multiple cardiomyocyte differentiation genes (Kasahara et al., 2013). Drp1 is essential for mitochondrial fission; it is recruited to the outer mitochondrial membrane by various adaptor proteins [including FIS1, MFF, MiD49 (MIEF2), and MiD51 (MIEF1)], oligomerizes, and subsequently leads to mitochondrial constriction and eventual fission (Sharma et al., 2019; Tilokani et al., 2018). In addition to promoting mitochondrial fission, Drp1 also is essential for peroxisomal fission (Kamerkar et al., 2018; Koch et al., 2003). Mice in which functional peroxisomes have been completely removed by deletion of Pex2 or Pex5, which are essential for peroxisome assembly, can survive to birth without an overt cardiac defect (Baes et al., 1997; Faust and Hatten, 1997), suggesting that peroxisomes are not essential for embryonic heart development. In previous studies, Dnm1l was inactivated in hearts using Myh6-Cre and MCK-Cre (Ishihara et al., 2015; Kageyama et al., 2014). Although these studies have demonstrated that Drp1 is essential for postnatal heart development, they did not address its role during embryonic heart development, as expression of Drp1 was not efficiently inactivated in fetal hearts in these studies.

In our current study, we aimed to test the effect of blocking mitochondrial fission in embryonic heart development during the crucial period when the cardiac metabolic shift starts to occur. To this end, we specifically inactivated Dnm1l using the Mef2c-AHF-Cre driver (Verzi et al., 2005). We show that deletion of Dnm1l in the second heart field (SHF) impairs mitochondrial morphology, ultrastructure and function, leading to severe myocardial wall defects in the right ventricle and embryonic lethality before E17.5. We thus provide the first genetic evidence in mouse to support the essential role of Drp1 during mammalian cardiogenesis. Furthermore, by characterizing conditional knockout (cKO) hearts, we reveal that Drp1 deficiency reduces transcription of a group of ribosomal protein (RP) genes through the AMPK-SIRT7-GABPB axis. Although mitochondrial dysfunction can repress mRNA translation through the integrated stress response (ISR) (Guo et al., 2020) or the mechanistic target of rapamycin (mTOR) signaling pathway (Dennis et al., 2001; Desai et al., 2002; Topf et al., 2019), our research has suggested a previously unknown mechanism, i.e. repression of RP gene transcription, whereby mitochondrial dysfunction leads to downregulation of protein synthesis. We thus provide further insight into how mitochondrial dysfunction results in pathological alterations during fetal organogenesis.

To understand the function of Drp1 during cardiogenesis, we used the Mef2c-AHF-Cre driver (Verzi et al., 2005) to inactivate Dnm1l in SHF-derived structures, including the outflow tract (OFT) and right ventricle (RV). We crossed Mef2c-AHF-Cre/Dnm1l^loxp/+ male mice with Dnm1l^loxp/loxp female mice (Wakabayashi et al., 2009) to obtain cKO (Mef2c-AHF-Cre/Dnm1l^loxp/loxp) and control (Dnm1l^loxp/+ or Dnm1l^loxp/loxp) embryos at various stages. We first determined at which stage expression of Drp1 was inactivated using immunostaining assays. At E9.5, Drp1 expression could be detected in ∼50% of cardiomyocytes in mutant RVs (Fig. S1), and at E10.5 its expression was observed in very few cardiomyocytes in the OFT and RV of mutant embryos (Fig. 1A), indicating that Dnm1l was efficiently inactivated at this stage in structures derived from the SHF. Expression of Drp1 in the common atria and atrioventricular canal region remained unchanged in mutant hearts, confirming the specificity of this Cre line. Mutant embryos were recovered at the expected Mendelian ratio (∼25%) until E15.5. At E16.5, only 9% of total living embryos were mutants and no mutant embryos were recovered at E17.5 (Fig. 1B), suggesting that embryonic lethality occurred between E15.5 and E17.5.

Whole-mount examination showed that the size of the mutant RV was reduced at E14.5 and this hypoplastic RV phenotype was confirmed in embryos from E12.5 to E15.5 by histological studies (Fig. 1C-E). No overt defect was observed in mutant RVs at E11.5. The size of the left ventricle (LV) was not altered in any of the stages examined. In addition to the reduced size of RVs, we also observed a noncompaction defect at E15.5 in both RVs and LVs of mutant hearts (Fig. 1D). As Dnm1l was specifically inactivated in SHF-derived structures, the defect in the LV of mutants must be secondary to the RV defect, supporting the idea that development of the two ventricular chambers is coordinated. The noncompaction defect observed in mutant hearts is consistent with the clinical observation that mitochondrial dysfunction is a primary cause of congenital noncompaction (Towbin and Jefferies, 2017; Weisz et al., 2010). No defect was observed in the OFT region, including aorta and pulmonary trunk. We thus show, for the first time, that Drp1 is required for normal cardiogenesis.

To understand how deletion of Dnm1l leads to hypoplastic RVs, we examined cell proliferation and apoptosis. No abnormality was observed in mutant RVs at E11.5. Beginning at E12.5, we observed significantly decreased cell proliferation and increased apoptosis in the RV of mutant hearts (Fig. 2). No cell proliferation or survival defect was observed in mutant LVs. During examination of cell proliferation and death, we noticed that cardiomyocytes in mutant RVs appeared to be less well organized than in control samples. We therefore examined expression of N-cadherin (cadherin 2), which mediates cell-cell interactions and is required for normal orientation of cardiomyocytes in mouse embryos (Li et al., 2016b; Miao et al., 2019). Our N-cadherin staining (Fig. 3A) confirmed that mutant cardiomyocytes were less well organized in mutant RVs. In addition, the N-cadherin immunostaining signal appeared be lower in cKO cardiomyocytes, and this result was confirmed through subsequent western blot analysis (Fig. 3B). Reduced expression of N-cadherin suggests abnormal formation of cell-cell contacts in mutant samples. We next directly examined the orientation of cardiomyocytes in the compact zone of LVs and RVs according to a published protocol (Miao et al., 2019). At E12.5, the majority of cardiomyocytes (38/50) oriented perpendicularly to the heart wall in control RVs, whereas most cardiomyocytes in mutant RVs (40/50) were parallel to the heart wall (Fig. 3C,D). The orientation of cardiomyocytes in mutant LVs was not altered. Our data collectively indicate that Drp1 is required for normal cell proliferation, survival and orientation in the RV.

To determine how deletion of Dnm1l affects mitochondrial morphology, we first examined cultured primary cardiomyocytes derived from cKO and control RVs at E10.5 (Fig. 4A). We co-stained cells with antibodies for a mitochondrial marker, TOM20 (Baker et al., 2007), and Drp1. In comparison with control cells, cKO cells exhibited efficient reduction in Drp1 expression and displayed characteristic tubular mitochondria (Fig. 4A), as expected (Kageyama et al., 2011). To confirm this observation in embryonic cardiac tissues, we stained E10.5 heart sections with TOM20 and Drp1 antibodies. We found that mitochondria in many cKO cells were clustered at one side of the nuclei in contrast to their even distribution in control cells (Fig. 4B,C). This mitochondrial clustering phenotype is similar to the morphology observed in Dnm1l null cells within tissue (Mitra et al., 2012; Udagawa et al., 2014). Therefore, Dnm1l deletion led to characteristic tubular mitochondrial morphology in developing cardiomyocytes.

We next quantified mitochondrial DNA levels using quantitative polymerase chain reaction (qPCR). Mitochondrial DNA in cKO RVs (E12.5) was not significantly different from control RVs (Fig. S2A). Consistent with this result, expression of TFAM, a key transcription factor for mitochondrial DNA transcription and replication (Dorn et al., 2015), was not altered by Dnm1l deficiency (Fig. S2B). This result is different from that obtained in adult cardiomyocytes, in which deletion of Dnm1l reduced TFAM expression and mitochondrial biogenesis (Song et al., 2015). Our results suggest that the interaction between mitochondrial dynamics and biogenesis has not been fully established in embryonic cardiomyocytes. We next examined mitochondrial ultrastructure using transmission electron microscopy (TEM). We found that the average length of mitochondria was significantly increased in cardiomyocytes of mutant RVs compared with controls (Fig. 4D,E), consistent with the idea that Drp1 deficiency leads to tubulation of mitochondria.

We speculated that abnormal morphology and ultrastructure would lead to reduction of mitochondrial activity in mutant cardiomyocytes. We thus examined electron transport chain (ETC) function by assessing the oxygen consumption of RV tissues (E14.5) (Fig. 4F,G), and calculated the mitochondrial respiration rate as described previously (Beutner et al., 2014). There was no significant difference in basal respiration rate (BL) between control and cKO samples, or when triggered by complex I substrates (malate+glutamate) (V₀) (Fig. 4F). However, the ADP-induced maximum respiration rate (V_max) was significantly lower in cKO samples than in control samples (Fig. 4F), suggesting that the OXPHOS capacity of complex I was impaired by Dnm1l deletion. The significant reduction in respiratory control ratio (RCR; V_max/V₀) with deletion of Dnm1l (Fig. 4G) suggests that the coupling of the ETC with OXPHOS was compromised in cKO samples. In both cKO and control samples, addition of an ADP/ATP translocase inhibitor, atractyloside (ATR), dramatically reduced the mitochondrial respiration rate (Fig. 4F), indicating no detectable defect in ADP or ATP translocation across the inner membrane. To test further the effect of impaired ETC activity on generation of ATP, we measured the total ATP level in LVs and RVs at E13.5. As shown in Fig. 4H, the ATP level in cKO RVs was significantly reduced compared with control levels, whereas no significant difference was observed between control and cKO LVs.

To understand better how deletion of Dnm1l affects cardiomyocyte development at the molecular level, we performed droplet-based single-cell RNA sequencing (scRNA-Seq) using cells isolated from control and cKO RVs at E13.5. A total of 7739 control cells and 6967 cKO cells passed quality control testing and were used for data analysis. Through the uniform manifold approximation and projection (UMAP) dimension reduction technique (Becht et al., 2018), these cells were grouped into 21 clusters (Fig. S3). Based on known molecular markers of cells in embryonic hearts (Hill et al., 2019; Li et al., 2016a), we identified cell populations corresponding to different cell types, including cardiomyocytes, endocardial cells, epicardial cells, endothelial cells, red blood cells and macrophages (Fig. 5A,B). To understand how deletion of Dnm1l specifically impacts gene expression in developing cardiomyocytes, we combined the two cardiomyocyte clusters into one group and compared gene expression within this combined cluster between control and mutant cells. We found 122 genes for which expression was significantly altered by at least 25% (adjusted P<0.05). We noticed that more than ten RP genes were among the most significantly downregulated genes (Fig. 5C, Table S1). We then performed gene ontology (GO) term enrichment analysis using Metascape (Zhou et al., 2019) (Fig. 5D, Table S2). The term ‘eukaryotic translation initiation’, which includes multiple RP genes, was most significantly enriched in the list, consistent with our observation in Fig. 5C. As expected, the pathway involved in OXPHOS was impaired by deletion of Dnm1l (Fig. 5D). In addition, genes involved in muscle structure development (including Tnnt1, Myl2, Actc1, etc.) were enriched in the list, suggesting that differentiation of cardiomyocytes was also compromised by Dnm1l deletion. We did not observe altered Notch signaling, in contrast to prior studies looking at Mfn1 and Mfn2 double-knockout cardiomyocytes (Kasahara et al., 2013), suggesting that the molecular defects caused by tubulation of mitochondria is different from that of mitochondrial fusion deficiency.

Few studies have linked mitochondrial dysfunction and altered transcription of RP genes; therefore, we decided to focus on this group of genes. We first examined the expression of ten RP genes by quantitative reverse transcription PCR (qRT-PCR) using RNA samples isolated from control and cKO RVs at E13.5. Our results confirmed reduced expression of these genes in cKO samples (Fig. 6A). Western blot analysis confirmed reduced expression of four RPs in mutant RVs at the protein level (Fig. 6B, Fig. S4). To test the impact of reduced expression of RP genes, we examined protein synthesis in cultured cardiomyocytes derived from control and mutant RVs at E13.5. Our results show that protein synthesis was significantly reduced in cKO cardiomyocytes (Fig. 6C,D), as expected from the reduced expression of multiple RP genes.

In the next set of experiments, we aimed to determine the mechanism by which Dnm1l deletion leads to reduced transcription of RP genes. SP1, YY1 and GABP are major transcription factors promoting RP gene transcription (Li et al., 2016c; Nosrati et al., 2014; Perry, 2005); however, our initial western blot analysis did not reveal reduced expression of these proteins in cKO samples (Fig. 7A, Fig. S5A). It is known that GABP is composed of two subunits, GABPA and GABPB, which contain the Ets DNA-binding domain and transcriptional activation domain, respectively (Rosmarin et al., 2004). We then performed a co-immunoprecipitation (co-IP) study to detect formation of the transcriptionally active GABP complex in mutant tissues and showed that the formation of the GABPA-GABPB complex was reduced by Dnm1l deletion (Fig. 7B). It has been well established that post-translational modification of GABPB, including phosphorylation and acetylation, can interfere with the interaction between GABPA and GABPB (Fromm and Burden, 2001; Hoffmeyer et al., 1998; Ryu et al., 2014). We therefore probed the immunoprecipitated samples with an antibody against phospho-serine or acetyl-lysine. Although the level of phospho-GABPB1 was not altered, the level of acetyl-GABPB1 was clearly increased (Fig. 7B), supporting the idea that increased acetylation of GABPB1 blocks its interaction with GABPA to form the active GABP complex.

To investigate further how acetylation of GABPB1 was increased in cKO samples, we examined expression of SIRT7, which is a deacetylase of GABPB1 (Ryu et al., 2014). As shown in Fig. 7C and Fig. S5B, expression of SIRT7 was downregulated by deletion of Dnm1l. Because the level of Sirt7 mRNA was not altered in mutant cells from our scRNA-Seq data (Table S1), we examined whether phosphorylation of SIRT7 was altered, as this post-translational modification is known to alter its protein level (Sun et al., 2016). Our data clearly showed that phosphorylation of SIRT7 was increased in cKO samples (Fig. 7D, Fig. S5C). As AMP-activated protein kinase (AMPK) can directly phosphorylate SIRT7 upon energy starvation (Sun et al., 2016), we examined the level of active AMPK in cKO RVs. We showed that active AMPK was increased by Dnm1l deletion and, furthermore, the level of phospho-SIRT7 in cKO RVs was restored to the control level by treatment with Compound C (an AMPK inhibitor) (Fig. 7D, Fig. S5B,C). These data support the role of AMPK on SIRT7 phosphorylation. Finally, we showed that blocking AMPK through treatment of cKO cardiomyocytes with Compound C could partially rescue the global protein synthesis defect caused by Dnm1l deletion (Fig. 7E,F), providing direct evidence to support the role of AMPK-SIRT7 in mediating the protein synthesis defect in cKO cells (see the model in Fig. S6).

To understand better the mechanism by which mitochondria regulate cardiogenesis, we have examined how deletion of Dnm1l affects RV development during the crucial embryonic metabolic shift period. In mouse embryos between E11.5 and E13.5, cardiac mitochondria undergo a maturation process enabling cardiomyocytes to obtain ATP from both anaerobic glycolysis and OXPHOS (Chung et al., 2007; Hom et al., 2011; Lopaschuk and Jaswal, 2010; Mackler et al., 1971; Zhao et al., 2019). We speculated that proper mitochondrial morphology would be important for mitochondrial maturation to support the embryonic metabolic shift, and therefore applied a conditional gene inactivation approach to delete Dnm1 in developing hearts.

Inactivation of Dnm1l in neonatal hearts results in animal lethality within the first 2 weeks after birth (Ishihara et al., 2015; Kageyama et al., 2014), indicating that Drp1 is essential for postnatal heart homeostasis. However, the effect of Dnm1l deletion on embryonic heart development has not been reported in the literature. We inactivated Dnm1l using the Mef2c-AHF-Cre driver, which starts to inactivate target genes in the precursor cells of the SHF (Verzi et al., 2005). Strong reduction in Drp1 expression was observed starting from E9.5 (Fig. S1), and by E10.5 Drp1 was only detected in a few cardiomyocytes in the OFT and RV regions (Fig. 1). No reduction in Drp1 expression was observed in LVs of cKO hearts, and therefore cardiomyocytes in LVs of mutant hearts can be used as an internal control. We thus provide an ideal model in which to test the role of Drp1 in cardiogenesis during cardiac mitochondrial maturation. Our immunofluorescence microscopy studies confirmed that mitochondria in mutant cardiomyocytes displayed signature tubulation defects (Fig. 4), including tubular mitochondria observed in cultured cells (Fig. 4A) and uneven clustering observed in tissue sections (Fig. 4B). Our subsequent examination of mitochondrial ultrastructure by TEM showed significantly increased mitochondrial length in cKO cardiomyocytes. Our functional tests have provided further support for the idea that abnormal mitochondrial morphology can lead to reduced mitochondrial activities (Fig. 4F,G). Results in this study are in contrast to what has been observed in proliferating mouse embryonic fibroblasts or T cells, in which Dnm1l ablation can sustain elevated respiration (Kashatus et al., 2015; Parker et al., 2015; Serasinghe et al., 2015; Simula et al., 2018). Therefore, the effect of Dnm1l deletion on mitochondrial activities is influenced by cell type.

No defect in cKO cardiomyocytes was observed up to E11.5. Starting from E12.5, Dnm1l deletion in cardiomyocytes of RVs led to reduced cell proliferation, increased cell death and abnormal cell orientation (Figs 1–3). This timeline correlates well with the mitochondrial maturation process; the cellular defects start to present when mitochondria begin to generate ATP through OXPHOS. We therefore provide another piece of evidence to support the idea that the cardiac metabolic shift has already begun at mid-gestation and that this shift is important for normal fetal heart development. Abnormal orientation of cardiomyocytes is a new phenotype associated with Dnm1l deletion. Expression of N-cadherin in Dnm1l-deleted cardiomyocytes was clearly reduced, as shown from both immunostaining and western blot analysis (Fig. 3). Considering the result from a recent publication that N-cadherin is required for proper orientation of cardiomyocytes (Miao et al., 2019), we speculate that Drp1-mediated mitochondrial fission acts through N-cadherin to promote normal cell orientation. Our scRNA-Seq data failed to reveal reduced expression of N-cadherin at the RNA level in mutant samples, meaning that modulation of N-cadherin by Dnm1l deletion likely occurs through certain post-transcriptional regulatory mechanisms. This result is different from the observation made in P19 teratocarcinoma cells; in these cells, transcription of N-cadherin could be repressed by altering Drp1 expression (Vantaggiato et al., 2019). Thus, the mechanisms underlying the crosstalk between Drp1 and expression of N-cadherin varies depending on cell type. Interestingly, expression of N-cadherin could also be reduced by blocking mitochondrial fusion in mammary epithelial cells, the counter process of fission (Wu et al., 2019). Therefore, expression of N-cadherin appears to be commonly targeted in different cell types by altered mitochondrial morphology.

One of the most significant discoveries of our current study is to reveal the AMPK-SIRT7-GABP axis that relays the Dnm1l deficiency to reduced cytosolic protein translation in mutant cells. The model shown in Fig. S6 summarizes our major discoveries. Dnm1l deletion in embryonic cardiomyocytes of RVs leads to tubulation of mitochondria, which acts together with other cellular defects caused by Dnm1l deficiency to reduce ATP generation. The reduced level of ATP activates AMPK, a key sensor of the energy level in cells (Wu and Zou, 2020), which phosphorylates SIRT7, leading to SIRT7 degradation. The normal function of SIRT7 is to remove the acetyl group from GABPB1 to promote formation of the GABPA and GABPB complex, which activates transcription of multiple RP genes. The reduced level of SIRT7 due to increased phosphorylation by AMPK reduces RP gene transcription and represses cytosolic protein translation in cKO cardiomyocytes.

In addition to its essential role in mitochondrial fission, Drp1 also has other functions. Although the human DNM1L gene was originally cloned based on its homology to the yeast endocytosis gene Vps1p (Imoto et al., 1998), the function of Drp1 in regulating endocytosis is only related to a brain-specific isoform, Drp1_ABCD, which is reportedly only 5% of the total Dnm1l transcripts (Itoh et al., 2018, 2019). Only Drp1_ABCD, and not the standard Drp1 isoform expressed in other tissues, was found to associate with endosomes and to be involved in endocytosis. It has also been clearly shown that Drp1 is essential for peroxisomal fission (Kamerkar et al., 2018; Koch et al., 2003). Although we cannot exclude the possibility that impaired peroxisomal fission also contributed to the observed defects in mutant cardiomyocytes, we speculate that such a contribution would not be a major factor. Pex2 and Pex5 are essential genes for the assembly of peroxisomes, and complete inactivation of these two genes did not cause overt cardiac defects at the newborn stage (Baes et al., 1997; Faust and Hatten, 1997). Therefore, it is unlikely that a failure in peroxisomal fission is the primary cause of the severe cardiac defects and embryonic lethality observed in Dnm1l mutant mice. Our data in Fig. 4 clearly show both morphological and functional defects of mitochondria in Dnm1l mutant cardiomyocytes. We thus propose that mitochondrial anomalies are the initiating factor that activates the AMPK-SIRT7-GABP axis.

Two major pathways have been reported to mediate the crosstalk between mitochondrial stress and protein synthesis in mammalian cells. Mitochondrial dysfunction can act through the OMA1-DELE1-HER1 cascade to trigger the ISR (Guo et al., 2020), which represses global mRNA translation and at the same time enhances translation of specific mRNAs, including ATF4, ATF5 and DDIT3 (Topf et al., 2019). The resulting proteins activate transcription of cytoprotective genes to help restore the function of mitochondria (Lindqvist et al., 2018; Quiros et al., 2017b; Topf et al., 2019). In another pathway, reduced energy level in cells due to mitochondrial dysfunction represses the activity of the mTOR signaling pathway to inhibit protein synthesis (Dennis et al., 2001; Desai et al., 2002; Topf et al., 2019). To the best of our knowledge, our research provides the first example suggesting that mitochondrial dysfunction can act through downregulating transcription of RP genes to repress protein synthesis. In future studies, it will be of great interest to determine whether this signaling cascade can also function in other cell types. We noticed that blocking AMPK only partially rescued the protein synthesis defect in mutant cardiomyocytes (Fig. 7), suggesting the presence of another pathway, such as the ISR pathway, which acts independently of the AMPK-SIRT7-GABPB cascade to repress protein synthesis. Consistent with this idea, the pathway involved in transcription in response to stress was activated by Dnm1l deletion (Fig. 5D).

A connection between Drp1 and RP genes was also reported previously, although contrary to what we report here (Liang et al., 2020; Spurlock et al., 2021; Tanwar et al., 2016). In particular, higher Drp1 expression was found to correlate with lower expression of various ribosomal genes in tumor tissues of various cancer types (Tanwar et al., 2016) and almost complete depletion of Drp1 in keratinocytes led to an increase in ribosomal gene expression (Spurlock et al., 2021). Similarly, a population of quiescent hematopoietic stem cells with higher Drp activity was found to be enriched in negative regulators of translation and RNA processing (Liang et al., 2020). Therefore, tubulation of mitochondria may increase or decrease RP gene transcription in a context-dependent fashion.

In summary, we provide convincing mouse genetic evidence to show that Drp1 is essential for normal development of embryonic cardiomyocytes of RVs. We further revealed a AMPK-SIRT7-GABPB signaling pathway that relays the reduced cellular energy level to reduced RP gene transcription and protein synthesis. Our study enhances our understanding of the role of mitochondria in the regulation of cardiogenesis and could contribute to the development of novel clinical applications aimed at inborn cardiomyopathies caused by mitochondrial dysfunction.

This study conforms to the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication no. 85–23, revised 2011). All protocols were approved by the Institutional Animal Care and Use Committee at the University of Alabama at Birmingham. Euthanasia of mice was achieved through inhalation of CO₂ followed by cervical dislocation. Establishment of Mef2c-AHF-Cre mice (Verzi et al., 2005) and Dnm1l^loxp/loxp mice (Wakabayashi et al., 2009) and their genotyping strategy were described previously. Mef2c-AHF-Cre mice were maintained on the C57BL/6 and 129S6 mixed background, whereas Dnm1l^loxp/loxp mice were maintained on the C57BL/6 background. Mef2c-AHF-Cre male mice were used to cross Dnm1l^loxp/loxp mice to obtain Mef2c-AHF-Cre/Dnm1l^loxp/loxp male mice, which were then crossed with female Dnm1l^loxp/loxp mice to acquire control and cKO embryos at various stages. During mating, one male mouse and one or two female mice were put in the same cage, and the noon of detection of a plug in the vagina of a female mouse was designated as E0.5. Mice aged 2-8 months were used for mating to obtain embryos.

Standard procedures were performed as described in previous studies (Liu et al., 2014; Peng et al., 2016; Yan et al., 2020). Briefly, embryos or embryonic hearts were dissected in PBS and fixed in 4% paraformaldehyde overnight at 4°C. The next day, samples were dehydrated with ethanol, cleared with Histoclear, and embedded in paraffin. Wax blocks were sectioned at 10 μm to acquire paraffin sections, which were kept at 4°C until use. Sections were stained with Hematoxylin and Eosin (HE) for histological examination. For immunofluorescence staining, paraffin sections were dewaxed, rehydrated, treated with blocking buffer [10% serum in TBST (TBS with 0.1% Tween 20)] for 60 min, incubated with primary antibody overnight at 4°C, and incubated with an Alexa-conjugated secondary antibody (Thermo Fisher) at room temperature for 1 h. Slides were then thoroughly washed with TBST and briefly stained with DAPI (Thermo Fisher) to visualize the nuclei. If necessary, the TSA Plus Cyanine 3.5 System (Perkin Elmer) was used to amplify the signal. Immunostained samples were observed using a Zeiss Axio fluorescence microscope or a Zeiss laser-scanning confocal microscope (LSM700). The primary antibodies used for immunostaining were purchased from Abcam (Drp1, ab56788, 1:100), BD Biosciences (N-cadherin, 610920, 1:100), Cell Signaling Technology (cleaved caspase 3, 9661, 1:500), Iowa Hybridoma Bank (MHC, MF20, 1:500), Millipore (phospho-Histone H3, 06-570, 1:500) and Santa Cruz Biotechnology (TOM20, sc-11415, 1:100). The anti-MLC2A antibody was kindly provided by Dr S. Kubalak (Medical University of South Carolina, SC, USA; 1:1000). Measuring the orientation of cardiomyocytes in the compact zone was performed following the procedures described in a published study (Miao et al., 2019). E12.5 heart sections were stained with wheat germ agglutinin (WGA, Thermo Fisher) to label the cell membrane, as described previously (Harmelink et al., 2013). Sections were co-stained with the anti-MHC antibody to label cardiomyocytes. Cardiomyocyte orientation was measured following the procedures described previously (Miao et al., 2019). If the ratio of the length of a cell to its width was greater than 120%, this cell was considered oriented, and its orientation was determined by the angle of the longitudinal axis to the myocardial wall surface reference line (see Fig. S7 for illustration). Cells with angles between 60° and 90° were classified as perpendicular and those with angles between 0° and 30° were regarded as parallel. Cells oriented between 30° and 60° were considered non-classified.

Analyses were performed as we described previously (Yan et al., 2020). For western blotting, cells or tissues were lysed using 1.5× Laemmli buffer. Protein concentration was determined using the DC protein assay kit (Bio-Rad). Equal amounts of protein (10-30 μg) were separated by SDS-PAGE. Proteins were transferred to a PVDF membrane (Bio-Rad). The membrane was blocked with 10% bovine serum albumin (BSA) in TBST for 1 h and incubated with primary antibody at 4°C overnight. The following day, the membrane was incubated with a horseradish peroxidase-conjugated secondary antibody (Thermo Fisher). The signals were detected using the Immobilon ECL Ultra Western HRP Substrate (Millipore) following instructions provided by the manufacturer. The images were acquired and the signal intensity was quantified using the LiCOR Odyssey imaging system 2800. For co-IP analyses, tissues were lysed with lysis buffer [50 mM Tris-HCl (pH 7.4), 0.1% Triton X-100, 0.1% IGEPAL, 100 mM NaCl, 10% glycine]. Lysates were first incubated with an anti-GABPB antibody or an anti-SIRT7 antibody for 6 h at 4°C and were then added to 30 μl Protein G Magnet beads (Pierce) for incubation at 4°C overnight. The next day, beads were thoroughly washed with lysis buffer and proteins were eluted by incubation with 1.5× Laemmli buffer at 70°C for 5 min. Supernatants were subjected to western blot analysis using the appropriate antibodies. To test the impact of blocking AMPK on the expression of phosphorylated SIRT7, E13.5 embryonic hearts were isolated and cultured in DMEM (with 10% BSA) containing DMSO or 10 μM Compound C (Millipore, 171620) for 6 h in a cell culture incubator (37°C, 5% CO₂). The RVs were then removed and subjected to IP using an anti-SIRT7 antibody. The primary antibodies used for western and IP analyses were purchased from Abcam (GABPA, ab224325; Phosphoserine, ab9332, 1:100), Abclonal (RPS28, A17937, 1:100), BD Biosciences (N-cadherin, #610920, 1:100), Cell Signaling Technology (acetylated-lysine, 9441, 1:100; AMPKα, 2532, 1:100; phospho-AMPKα, 40H9, 1:100; SIRT7, D3K5A, 1:100; SP1, D4C3, 1:100; YY1, D5D9Z, 1:100), Iowa Hybridoma Bank (beta-tubulin, E7), Proteintech (RPL35, 14826-1-AP, 1:100; RPL37A, 14660-1-AP, 1:100; RPS21, 16946-1-AP, 1:100) and Santa Cruz Biotechnology (GABPB, E-7, 1:100).

Cardiomyocytes from RVs of E13.5 embryos were isolated using the neonatal heart dissociation kit purchased from Miltenyi Biotec. To help remove fibroblast contamination, isolated cells were pre-plated on a 100 mm Petri dish four times in a cell culture incubator for 30 min. Remaining cells were then suspended in culture medium [DMEM with 10% FBS and 1×Pen/Step (Thermo Fisher)], plated in a 48-well culture plate, and cultured for 48 h in a cell culture incubator followed by subsequent procedures. De novo protein synthesis was measured using the Global Protein Synthesis Assay Kit (Abcam, ab235634) according to instructions provided by the manufacturer. Briefly, cells were incubated with 1× Protein Label (containing O-propargyl-puromycin) added to the culture medium for 4 h followed by signal detection. Total nuclei were visualized by DAPI staining. In a negative control experiment, cardiomyocytes isolated from control embryos were treated with cycloheximide (20 μg/ml) for 30 min before adding Protein Label. Images were taken using AxioCam HRc high-resolution digital camera connected to a Zeiss Axio fluorescence microscope under identical settings between control and mutant samples. Signal intensity was quantified using ImageJ. To test the effect of blocking the AMPK activity, cultured cardiomyocytes were treated with 10 μM Compound C for 16 h followed by measuring protein synthesis.

E13.5 embryonic hearts were isolated from control and mutant embryos, treated with Ca²⁺-free HEPES buffer (130 mM NaCl, 5.4 mM KCl, 0.33 mM NaH₂PO₄, 0.5 mM MgCl₂, 22 mM glucose and 20 mM HEPES, pH 7.4) for 3 min, and then fixed with 2% paraformaldehyde/2.5% glutaraldehyde (EM grade, Sigma-Aldrich) for 24 h at room temperature. Samples were then submitted to the UAB High Resolution Imaging Facility for osmification, dehydration and embedding according to the procedures described previously (Larson-Casey et al., 2016). Embedded samples were cut at 100 nm and observed using the Tecnai Spirit T12 Transmission Electron Microscope (Thermo Fisher, formerly FEI). Images were taken using an AMT (Advanced Microscopy Techniques) Bio Sprint 29 megapixel digital camera. Mitochondria length was measured as described previously (Wang et al., 2013). The straight line tool in ImageJ was used to mark and measure the longest axis of each mitochondria.

The RVs of control and cKO animals at E14.5 were dissected and weighed using an AB54-S/FACT Analytical scale (Toledo). We usually combined two or three cKO RVs into one tube as the weight of a cKO RV was about half of a control RV (∼1.2 mg). Tissues were permeabilized, as described previously (Hunter et al., 2017). Briefly, samples were transferred to ice-cold transport buffer (B1-200 mM sucrose, 0.5 mM Na₂EDTA, 10 mM Tris, 1 g/l BSA) and were immediately submitted to the UAB Bioanalytical Redox Biology (BARB) Core. Tissues were permeabilized by gentle dissection on ice with needle-tip forceps, washed in MiR03 buffer (0.5 mM EGTA, 3 mM MgCl₂.6H₂O, 20 mM taurine, 10 mM KH₂PO₄, 20 mM HEPES, 200 mM sucrose, 1 g/l BSA) by gentle rotation for 15 min at 4°C, and then transferred to fresh MiR03 buffer for high-resolution respirometry (HRR). HRR was performed by measuring oxygen consumption in 2 ml of MiR03 buffer, in a two-chamber respirometer (Oroboros Oxygraph-2k with DatLab software; Oroboros Instruments) with constant stirring at 750 rpm. Respiration rates were measured using the protocols as described previously (Beutner et al., 2014). Substrate-mediated respiration (State 2 respiration or V₀) was measured in the presence of complex I substrates (3 mM malate and 5 mM glutamate), followed by 1 mM ADP (State 3 or V_max). The RCR V_max/V₀ (State 3/State 2) was calculated using these measures. To evaluate inner mitochondrial membrane coupling, 10-100 µM ATR (added in 10 µM increments) was added to block ATP translocase and the oxygen consumption rate was then measured to obtain V_max+ATR.

The relative mitochondrial DNA to nuclear DNA ratio was measured as described previously (Quiros et al., 2017a; Singh et al., 2015). Total DNA was isolated from RVs of control and cKO hearts (E12.5) using the QIAamp DNA mini kit (QIAGEN). qPCR analysis was performed in a Roche LightCycler 480 using primers specific for mitochondrial DNA (mMitoF1: 5′-CTAGAAACCCCGAAACCAAA-3′; mMitoR1: 5′-CCAGCTATCACCAAGCTCGT-3′) and for nuclear DNA (mB2MF1: 5′-ATGGGAAGCCGAACATACTG-3′; mB2MR1: 5′-CAGTCTCAGTGGGGGTGAAT-3′). To measure ATP content, RVs and LVs were isolated from control and cKO embryos at E13.5 and weighed, followed by measuring ATP content using the Luminescent ATP Detection Assay kit (Abcam, Ab113849). Luciferase activity was measured using the Synergy H1 microplate reader (BioTek). The raw numbers were normalized against tissue weight.

For scRNA-Seq, RVs of control and mutant embryos at E13.5 were isolated and treated with trypsin to acquire single cells. Samples were pooled from three or four hearts of the same genotype. Cells were stained with 7-amino-actinomycin (7-AAD, Thermo Fisher) to label nonviable cells and then sorted using a FACS Aria II sorter (BD BioSciences) to acquire viable cells. Single-cell transcriptome libraries were prepared using 10X Genomics Single Cell 3′ v3 Reagent Kits according to the standard protocol outlined in the manual. The single-cell libraries were sequenced using Illumina Next Seq500 with the goal of a minimum of 2000 total reads per cell. 10X Genomics Cellranger software (version 3.0.0), ‘mkfastq’, was used to create the fastq files from the sequencer. Following fastq file generation, Cellranger ‘count’ was used to align the raw sequence reads to the reference genome using STAR. The ‘count’ software created three data files (barcodes.tsv.gz, genes.tsv.gz, matrix.mtx.gz) from the ‘filtered_feature_bc_matrix’ folder that were loaded into the R (version 3.6.0) package Seurat version 4.0.1, which allows for selection and filtration of cells based on QC metrics, data normalization and scaling, and detection of highly variable genes. We followed the two Seurat vignettes (https://satijalab.org/seurat/v3.0/pbmc3k_tutorial.html and https://satijalab.org/seurat/v3.0/immune_alignment.html) to create the Seurat data matrix object. In brief, we kept all genes expressed in at least >3 cells and cells with at least 200 detected genes. Cells with mitochondrial gene percentages over 7.5% and unique gene counts >5000 or <200 were discarded as well. The data were normalized using Seurat's ‘NormalizeData’ function, which uses a global-scaling normalization method, LogNormalize, to normalize the gene expression measurements for each cell to the total gene expression. The result is multiplied by a scale factor of 1e4 and the result is log-transformed. Highly variable genes were then identified using the function ‘FindVariableGenes’ in Seurat, which returned 2000 features per dataset. We then merged the two datasets by identifying anchors using ‘FindIntegrationAnchors’ in Seurat using 30 dimensions followed by integrating the two datasets with ‘IntegrateData’. Following integration, we regressed out the variation arising from library size and percentage of mitochondrial genes using the function ‘ScaleData’ in Seurat. A UMAP was created to visualize the clusters of the combined datasets. We next identified conserved cell type markers with ‘FindConservedMarkers’ for each cluster. To identify differentially expressed genes in cell clusters across the two datasets, we used the function ‘FindMarkers’ in Seurat.

For qRT-PCR analysis, total RNA was isolated from RVs of control and mutant embryos at E13.5. Total RNA was reverse transcribed to acquire cDNA followed by real-time PCR as we described previously (Yan et al., 2020). Hprt was used as the loading control. The primers for qRT-PCR are provided in Table S3.

We thank Dr G. A. Porter (University of Rochester Medical Ctr.) for suggestions on isolating and culturing embryonic cardiomyocytes; Dr M. Wu (University of Houston) for suggestions on measuring cardiomyocyte orientation; Dr S. Kubalak (Medical University of South Carolina) for providing the MLC2A antibody; and Dr S. H. Litovsky (UAB) for helping with analysis of TEM images. We thank the UAB Genomics Core Facility for performing deep-Seq and Sanger sequencing; E. Phillips and M. Foley at the UAB High Resolution Imaging Facility for performing TEM studies; and K. Smith-Johnston and M. J. Sammy at the UAB Bioanalytical Redox Biology (BARB) Core for performing high-resolution respirometry. We thank the members of Dr Jiao's lab for their suggestions that helped support this project. The UAB BARB Core is supported by the National Institute of Diabetes and Digestive and Kidney Diseases [DK P30DK079626, DK056336], UAB Center for Exercise Medicine (UCEM), Comprehensive Diabetes Center (UCDC), Center for Free Radical Biology (CFRB) and Comprehensive Neuroscience Center (UCNC).