A new mitochondrial pool of cyclin E, regulated by Drp1, is linked to cell-density-dependent cell proliferation

Spatiotemporal regulation of the cell cycle is crucial in various developmental processes, with cell cycle deregulation leading to cell proliferative disorders like tumorigenesis. As mitotic cells move through the growth (G1) phase to the DNA synthesis (S) phase followed by a short growth phase and mitosis (G2-M), they are governed by specific cyclins at each phase (Lodish et al., 2000). Cyclin E (CycE, SwissProt P24864 and O96020 in humans) governs the transition from G1 to S phase by initiating DNA replication (Hwang and Clurman, 2005). In a regulated mitotic cycle, CycE abundance and activity is at a maximum during the G1-S transition and progressively decreases through the S phase to a minimum at the G2 and M phases. CycE binds and activates its partner kinase cyclin-dependent kinase 2 (CDK2) to phosphorylate downstream targets in the cell nucleus to initiate DNA synthesis (Siu et al., 2011). In mouse models, CycE is dispensable for the development of the embryo proper but is indispensable for development of polyploidy, release from quiescence and oncogenic transformation, all of which require S phase entry (Geng et al., 2003). These functions of CycE have been found to be CDK2 independent (Geng et al., 2007), the details of which still remain to be investigated. Moreover, CDK2-independent CycE localization to the centrosome might be linked to S phase entry (Matsumoto and Maller, 2004). Apart from CDK2, CycE also binds to CDK1, although this interaction remains less characterized (Welcker and Clurman, 2005).

The CycE abundance is precisely regulated to restrict its activity to the G1-S phase transition. This is achieved as active CycE is phosphorylated and subsequently ubiquitylated to target it for degradation (Siu et al., 2011). Deregulation of CycE abundance and/or activity leads to deregulation of cell proliferation: constitutively elevated CycE throughout the cell cycle causes DNA damage and chromosome misalignment, ultimately leading to genomic instability and aneuploidy (Spruck et al., 1999; Hwang and Clurman, 2005). CycE overexpression has been noted in various tumors, which can arise from loss of function of components responsible for CycE degradation (Davis et al., 2014). Although elevated nuclear CycE levels are used as markers for tumorigenicity in various cancer types (Hwang and Clurman, 2005; Donnellan and Chetty, 1999), overabundance of an uncharacterized cytosolic pool of CycE has been noted in cancer tissues (Donnellan and Chetty, 1999). Specific low-molecular-weight forms of CycE have been proposed to be the direct cause behind CycE-driven tumorigenicity (Bedrosian et al., 2004; Bales et al., 2005).

The understanding of CycE regulation and function remains incomplete and shrouded with controversies. Interestingly, CycE stability requires ATP production from mitochondria in mammalian cells and in Drosophila melanogaster (Mandal et al., 2010, 2005). Enhancement of mitochondrial ATP production during the G1 to S transition in normal proliferating fibroblasts likely stabilizes CycE during this period of the cell cycle (Mitra et al., 2009, 2013). Mitochondrial ATP-synthesizing capability can be directly or indirectly influenced by a certain group of proteins that directly modulate the ability of mitochondria to undergo fusion or fission events with each other (Westermann, 2010; Galloway et al., 2012). Reduction of the levels or activity of the key mitochondrial fission protein, dynamin-related protein 1 (Drp1, SwissProt O00429 in humans) leads to elevation of CycE levels in various mammalian cells and in Drosophila (Mitra et al., 2012, 2009; Qian et al., 2012). The emerging model suggests that cell cycle regulators regulate Drp1 (Taguchi et al., 2007; Kashatus et al., 2011; Horn et al., 2011), which, in turn, further regulates CycE levels in proliferating cells (Mitra, 2013). We have previously shown that although loss of Drp1 deregulates CycE in various cell contexts, it promotes aberrant cell proliferation only in the presence of EGFR signaling (Mitra et al., 2012). Using mammalian cells and Drosophila model systems in parallel, we report that a new mitochondrial pool of CycE, which can be modulated by Drp1, likely through regulation of mitochondrial energetics, is linked to control of cell proliferation in a cell-density-dependent manner.

The canonical function of CycE is in the nucleus, but CycE might re-locate to the centrosomes to promote centrosome duplication (Siu et al., 2011) and to the Golgi membrane to be ubiquitylated (Lu and Pfeffer, 2013). Here, we investigated whether the mitochondrial effect on CycE (Mandal et al., 2010, 2005; Mitra, 2013; Qian et al., 2012) can be mediated by recruitment of the molecule to mitochondria. Immortalized mouse embryonic fibroblasts (MEFs) had a distinct cytosolic pool of CycE (Fig. 1A, upper panel), a fraction of which colocalized with the mitochondrial marker Tom-20 (arrows in Fig. 1A, lower panel). This indicated the presence of a mitochondrial CycE pool, referred to as mtCycE hereafter. Multiple washes with PBS buffer before cell fixation reduced the level of the mtCycE pool, dramatically indicating its labile and transient nature (data not shown). We also detected the mtCycE pool in cancer cells (Fig. S1A,B). Next, we quantified the mtCycE pool in individual MEFs by colocalization analyses of CycE with Tom-20. Linear regression analyses of the mtCycE versus the non-mtCycE pool revealed a strong correlation between the two CycE pools, suggesting that the two pools influence each other and possibly interact (Fig. 1B). We further confirmed the presence of the new mtCycE pool in crude mitochondrial fractions from MEFs, which were devoid of the nuclear marker Sox-2 and the cytosolic marker actin, and strongly enriched for the mitochondrial marker Tom20 and modestly for mitochondria-associated endoplasmic reticulum (MAM) marker protein disulfide isomerase (PDI; also known as P4HB) (Fig. 1C). Although the non-mitochondrial fraction harbors the characteristic doublet bands of CycE, we consistently detected the lower band of CycE in the mitochondrial fraction. The specificity of the widely used anti-CycE antibody was confirmed in cells transfected with CycE small interfering RNA (siRNA) to reduce its protein level by only 20% to avoid affecting cell proliferation (Fig. S1C). CycE after binding to CDK2 gets auto-phosphorylated at degrons, which is required for the subsequent ubiquitylation and degradation of CycE (Siu et al., 2011). Therefore, CDK2-dependent phosphorylation of CycE reflects the active CycE pool that eventually gets degraded to keep CycE levels regulated. We found that the active pool of CycE phosphorylated at T62 [p-(T62)CycE; the N-terminal degron] mostly resided in the nucleus (Fig. 1D,E). In addition, colocalization of the transfected mitochondrial reporter mito-RFP with p-(T62)CycE was significantly lower than that with CycE in individual cells (Fig. 1E,F). These data suggest that the mtCycE pool is devoid of the CDK2-dependent T62 phosphorylation and thus deemed inactive. However, it remains to be examined whether there are other phosphorylated sites on mtCycE.

We next sought to detect the new mtCycE pool in vivo in the Drosophila follicle cell layer where we have previously demonstrated specific mitochondrial regulation of CycE (Mitra et al., 2012). The follicle cell layer is the epithelial cell layer encapsulating the Drosophila egg chambers. Using an antibody against the Drosophila CycE (DmCycE), we detected a distinct pool of DmCycE colocalizing strongly with the mitochondrial marker ATP-B (the ATP synthase β subunit) in the terminally differentiated follicle cell layer (Fig. 1G). The early follicle cells, after differentiating from the lineage-specific stem cells, undergo mitotic divisions during developmental stages 1 through 6. After stage 6, the follicle cells exit the mitotic cycle to terminally differentiate into the epithelial cell layer, which is further patterned into various cell types (Klusza and Deng, 2011). We have previously reported differential mitochondrial regulation in the mitotic follicle cells and the differentiated-patterned main body follicle cells (MBCs) and posterior follicle cells (PFCs) (Mitra et al., 2012). Here, we found that the mtDmCycE pool was significantly higher in the MBCs than the PFCs or the mitotic follicle cells (Fig. 1H; Fig. S2A), suggesting that the mtDmCycE pool is developmentally regulated in the Drosophila follicle cell layer.

Our novel observation of the existence of the mtCycE pool (revealed by two distinct antibodies against mammalian and DmCycE) likely underlies the mechanism behind a direct mitochondrial regulation of CycE. Based on the focal organization of mtCycE (Fig. 1A) that was identified in a cell fraction with modest enrichment of a MAM marker (Fig. 1C), we speculate that the mtCycE pool could reside at contact sites between mitochondria and endoplasmic reticulum.

The levels of mtCycE in the various cell types in the Drosophila follicle cell layer (Fig. 1H) negatively correlate with the previously reported status of Drp1-driven mitochondrial fission (Mitra et al., 2012), indicating that reduced Drp1 activity might elevate the mtCycE pool. We tested this possibility in MEFs obtained from the DRP1-knockout (DRP1-KO) embryos and thereafter immortalized with the SV-40T antigen (Ishihara et al., 2009). Comparison of the CycE and Tom-20 colocalization between the wild-type (WT) and the DRP1-KO MEFs revealed a significantly elevated mtCycE pool in the absence of Drp1 (Fig. 2A,B). Introduction of Drp1–GFP into the DRP1-KO MEFs reduced the mtCycE pool when compared to introduction of the EGFP vector (Fig. 2C), thus confirming that the levels of Drp1 regulate the levels of the mtCycE pool. We further validated the effect of Drp1 loss on the mtCycE pool in the Drosophila follicle cell layer by generating Drp1 functionally null clones to compare DmCycE localization between the clones and the background WT follicle cells. We have previously shown that Drp1-null follicle cell clones harbor hyperfused mitochondrial clusters (Mitra et al., 2012). Here, we found that the majority of the DmCycE pool localized to the mitochondrial clusters in the Drp1-null cells in the differentiated MBC region (arrows in Fig. 2D) or in early mitotic stages (Fig. S2B), confirming our observation in the MEFs.

We next isolated mitochondria from the WT and DRP1-KO MEFs to confirm the increase in the mtCycE pool upon Drp1 loss. The crude mitochondrial fractions from the DRP1-KO MEFs had significant levels of actin (Fig. 2E), which might be of functional relevance, as recently reported (Li et al., 2015). However, the MAM marker, PDI, was more enriched in the WT mitochondrial fraction than the DRP1-KO mitochondrial fraction. The mitochondrial fractions from the DRP1-KO MEFs indeed harbored higher levels of the low-molecular-weight band of CycE although the amount of the non-mitochondrial upper bands of CycE was comparable between the WT and DRP1-KO (Fig. 2F). Although, the mtCycE pool was devoid of the CDK2-dependent T62 phosphorylation in both WT and DRP1-KO cells (Fig. 2G), the non-mitochondrial active p-(T62)CycE pool was higher in the DRP1-KO cells, which likely resulted from higher CDK2 levels in those fractions (Fig. 2G). Furthermore, a distinct CDK2 pool was also detected in the mitochondrial fraction from both WT and DRP1-KO MEFs (Fig. 2G), with the latter showing an extra CDK2 band (Fig. S1D). Our data demonstrates that loss of Drp1 leads to enhancement of the mtCycE pool devoid of the p-T62 degron (inactive) with an associated increase in the amount of non-mitochondrial p-(T62)CycE (active).

Next, we investigated whether the enhanced mtCycE pool in the DRP1-KO MEFs is associated with aberrantly high CycE levels throughout the cell cycle. We detected only modestly higher CycE levels in the asynchronously proliferating DRP1-KO MEFs when compared to the WT. However, DRP1-KO MEFs synchronized in S phase or in G2-M phase had significantly higher levels of CycE than similarly synchronized WT MEFs (Fig. 2H). This finding is consistent with observations in mitotic breast cancer cells (Qian et al., 2012) and suggests that Drp1 loss reduces degradation of CycE throughout the cell cycle. Inhibition of CycE degradation in mitosis can sustain aberrantly higher levels of the mitotic cyclin cyclin B (Keck et al., 2007). Indeed, we observed that cyclin B levels were higher in the mitotic DRP1-KO MEFs in comparison to the WT cells, whereas cyclin A levels were not substantially different between them (Fig. 2H; Fig. S1E).

Taken together, our results suggest that Drp1 loss elevates the mtCycE pool and deregulates CycE by maintaining higher levels of CycE throughout the cell cycle. As the mtCycE pool lacks the p-T62 degron, our results imply that elevation of mtCycE pool in the absence of Drp1 impairs CycE degradation, which is the major cause of CycE deregulation (Siu et al., 2011).

CycE is actively degraded in Drosophila harboring a mutation in Cytochrome c oxidase subunit V that lowers mitochondrial ATP production (Mandal et al., 2010, 2005). Moreover, we have previously found that the level of canonical nuclear CycE pool is reduced upon mitochondrial depolarization (Mitra et al., 2009). Therefore, we investigated whether maintenance of the mtCycE pool is directly or indirectly mediated by mitochondrial energetics and how Drp1 could regulate it. Mitochondrial oxygen consumption is coupled to mitochondrial ATP synthesis by the mitochondrial transmembrane potential maintained by the electron transport chain (ETC), while ATP is synthesized utilizing the transmembrane potential (Nicholls and Ferguson, 2013). Inhibition of mitochondrial ATP synthase (by treatment with oligomycin) or disruption of the mitochondrial transmembrane potential (by treatment with the protonophore CCCP) reduced the mtCycE pool significantly within 3 h in both the WT and the DRP1-KO MEFs (Fig. 3A,B). The release of mtCycE happened without any gross change in mitochondrial shape, and mtCycE was removed from the mitochondrial tubules in the DRP1-KO MEFs, but it remained associated with swollen mitochondrial structures arising after bioenergetic perturbations (arrows in Fig. 3A, lower panel, compared with Fig. 2A, lower panel). Interestingly, DRP1-KO MEFs that had an elevated mtCycE pool (Fig. 2A,B) also had increased spare respiratory capacity as revealed by the enhanced stimulation of the oxygen consumption rate (OCR) upon CCCP treatment in the DRP1-KO MEFs, which have a basal OCR comparable to the WT MEFs (Fig. 3C). Such boosted mitochondrial energetics in the absence of DRP1 has also been found in other mouse and human cells (Serasinghe et al., 2015; Kashatus et al., 2015). Furthermore, we found that the enhanced spare respiratory capacity between the DRP1-KO and WT MEFs was abrogated after 2 h of serum depletion, whereas both WT and DRP1-KO MEFs had enhanced spare respiratory capacity in the presence of serum (Fig. 3D). We conclude that serum-dependent bolstered electron transport chain activity (indicated by enhanced spare respiratory capacity) in the DRP1-KO MEFs likely underlies the increase in the mtCycE pool observed in these cells.

Our results demonstrate that the mtCycE pool is directly or indirectly modulated by mitochondrial bioenergetics that is governed by Drp1. Perturbation of mitochondrial bioenergetics induced by physiological means, by drugs or by pathologic mutations might release mtCycE and make it susceptible to degradation, as observed in Drosophila harboring mitochondrial mutants (Mandal et al., 2010, 2005).

Given that Drp1 loss elevates the potentially active CycE pool (Fig. 2G), maintains higher levels of CycE throughout the cell cycle (Fig. 2H) and recapitulates key CycE overexpression phenotypes (Mitra, 2013; Qian et al., 2012), we wanted to investigate whether Drp1 loss causes any of the cell cycle aberrations seen upon CycE overexpression. CycE overexpression advances S phase entry, consequently reducing G1 length, but slows S phase progression after entry, consequently increasing S phase length (Hwang and Clurman, 2005). We used four different cell cycle analysis approaches to compare various aspects of cell cycle status between the WT and the DRP1-KO MEFs (Fig. 4A). Flow cytometry analyses of DNA content indicated that the DRP1-KO MEFs have ∼15% more cells in S phase in comparison to the WT MEFs (Fig. S3B). This observation was confirmed by quantifying the proportion of cells with PCNA–GFP in DNA replication foci, which is seen in the S phase, and the proportion of cells with diffused nuclear PCNA–GFP, which is seen in cells in any other cell cycle phase (Leonhardt et al., 2000) (Fig. 4B). Monitoring the alteration in PCNA–GFP foci distribution as the cells progress through S phase (Leonhardt et al., 2000) (Fig. S3C, right panel) revealed higher numbers of DRP1-KO MEFs only in the early S phase (Fig. 4B). We did not observe any difference in the G1 phase using a flow cytometry assay (Fig. S3B). However, with a FUCCI cell cycle reporter, which is based on Cdt-1- and Geminin-degron-tagged fluorescent proteins marking G1 as red, G1-S as yellow and S, G2 and M as green (Sakaue-Sawano et al., 2008), we found that the number of cells in G1 or G1-S was 50% or lower in the DRP1-KO MEFs compared to the WT MEFs, suggesting progression through G1 is different in DRP1-KO MEFs (Fig. 4C). We ignored the green population to avoid the ambiguity between the S, G2 and M status of these cells. We also used Aurora-B–GFP to identify the mid-body proper or remnants to be able to quantify cells having undergone cytokinesis to commence G1 (Fig. S3C, left panel) and found that DRP1-KO MEFs had 22% fewer cells in early G1 when compared to WT. Therefore, the combination of the various cell cycle analyses suggest that loss of Drp1 causes cells to spend longer time in early S phase, likely due to early entry into S phase.

Stem cells characteristically have constitutively enhanced cyclin E activity (Orford and Scadden, 2008), a long S phase and short G1 phase, which is reversed during the process of differentiation (Singh and Dalton, 2009). Given that DRP1-KO MEFs have the above properties, we compared the expression of the stem cell marker Sox-2 between the WT and the DRP1-KO MEFs. Intriguingly, we found that the DRP1-KO MEFs had higher levels of Sox-2 (Fig. 4D,E), which showed a trend towards a significant reduction within 24 h of re-introduction of Drp1 (Fig. S4A). We reasoned that if the higher expression of Sox-2 in the DRP1-KO MEFs were dictated by the altered cell cycle status, which has been proposed to be a factor for maintaining stemness (Singh and Dalton, 2009), loss of Drp1 in other cells would cause an increase in other relevant stem or progenitor cell markers. We chose to downregulate Drp1 in an ovarian cancer cell line (A2370) and a glioblastoma cell line (U87MG) using Drp1-specific short hairpin RNAs (shRNAs). Using two distinct shRNAs against Drp1, we indeed saw an increase in the level of the well-studied stem and progenitor cell marker Aldh1 in the ovarian cancer cell line (Flesken-Nikitin et al., 2013; Landen et al., 2010) (Fig. 4F) and of Sox-2 in the glioblastoma cell line (Ying et al., 2011; Eyler et al., 2011) (Fig. S4B). This data is consistent with the report that reduction of Drp1 activity prevents cell differentiation in embryonic stem cells (Todd et al., 2010). However, in Drosophila, the Drp1-null adult follicle cell stem cells can successfully differentiate into follicle cells (expressing the lineage specific marker, FasIII) that fail to undergo terminal differentiation in certain signaling contexts (Mitra et al., 2012).

Our data demonstrate that loss of Drp1, which leads to CycE deregulation due to aberrant elevation of the mtCycE pool, leads to alterations in the cell cycle that might promote the expression of stem and progenitor cell markers.

Given that Drp1 loss leads to CycE deregulation and stem cell marker enrichment, both of which can modulate the rate of cell proliferation, we performed detailed investigations of the cell proliferation status of the DRP1-KO MEFs. A growth curve analysis revealed that, although the WT MEFs slowed down proliferation at a high cell density, the DRP1-KO MEFs maintained cell proliferation in high-cell-density cultures (Fig. 5A). Furthermore, comparison of colony forming assays revealed that the DRP1-KO MEFs proliferated slower than the WT when there were fewer cells in the colonies (low density), whereas they formed bigger colonies when the colonies grew larger with higher cell numbers (high density) (Figs 5B and 6C,D). Interestingly, the DRP1-KO MEFs had an inner core with a surrounding lightly stained region of cells in comparison to a uniform colony in the WT (Fig. 5B), indicating a possible alteration of cell properties during the growth of the DRP1-KO MEF colonies. The DRP1-KO MEFs grew by more than tenfold, from low to high cell densities, in comparison to the twofold change for the WT MEFs (Fig. 5C). In addition, the DRP1-KO MEFs proliferated further with addition of fresh medium whereas the WT MEFs remained contact inhibited in high-cell-density cultures (Fig. 5C), confirming the loss of contact inhibition in the DRP1-KO MEFs. Flow-cytometry-based cell cycle analyses indicated that distribution of DRP1-KO population in G1 and S phases (as compared to the WT) showed the opposite trend when seeded in low versus high cell densities (Fig. S3E) (cell numbers are given in the Materials and Methods).

We reasoned that the mitotic stage of the Drosophila follicle cell layer development, where the WT follicle cells are mitotically active, might be similar to the low-cell-density environment of the in vitro MEF culture. Supporting this notion, we found that Drp1-null mitotic follicle cells in Drp1-null clones underwent fewer divisions than the WT follicle cells (Fig. 5E). In addition, 27% of the mitotic DRP1-null cells had detectable Geminin expression (which marks S, G2 and M cells) in comparison to only 17% of the WT background cells. This indicates that cell cycle aberration might underlie the slower proliferation of the Drp1-null mitotic follicle cells, which was possibly reflected in our previous observation of Drp1-null mitotic clones having higher phospho-histone 3 levels and elevated BrdU incorporation (Mitra et al., 2012). The high-cell-density environment of the contact-inhibited WT MEFs and aberrantly proliferating DRP1-KO MEFs might be reflected in the post-mitotic differentiated Drosophila follicle cell monolayer layer where the WT follicle cells have exited mitosis and Drp1-null clones aberrantly proliferate by employing local EGFR signaling in the PFC region; Drp1-null MBCs lacking EGFR activation can exit the mitotic cycle and enter differentiation successfully (Mitra et al., 2012; Fig. 5D). As expected, the Drp1-null PFCs continue to express Geminin, signifying mitotically cycling cells (Fig. 5F).

Taken together, our data suggest that Drp1 loss attenuates the cell proliferation rate in low-cell-density environments and certain signals in the high-cell-density environments might switch the cell properties to allow them to escape contact inhibition of cell proliferation.

Drp1-null PFCs in the Drosophila follicle cell layer have elevated levels of phosphorylated ERK [known as Rolled in Drosophila, and ERK1/2 (MAPK3 and MAPK1) in mammals] (Fig. S4D), possibly due to activation of the downstream components of the EGFR signaling pathway (Yarden and Shilo, 2007). The DRP1-KO MEFs also showed higher levels of phosphorylated ERK (p-ERK) in comparison to the WT MEFs, but only when seeded at high cell densities (Fig. 6A), further highlighting the similarity between the effects of Drp1 loss in MEFs at high cell density and in Drosophila PFCs. In the high-cell-density DRP1-KO MEF population, p-ERK was markedly enhanced in a subset of cells and was mostly intracellular, as opposed to being at the plasma membrane as in the WT-MEFs (Fig. 6B).

ERK is phosphorylated by MEK proteins (also known as MAPK2K proteins) and the phosphorylation is abrogated with MEK inhibitors (Akinleye et al., 2013). We chose the widely used MEK inhibitor PD98059 to investigate the effect of abrogation of ERK phosphorylation on cell proliferation in the presence or absence of Drp1. We first confirmed that PD98059 lowers the p-ERK signal both in WT and DRP1-KO MEFs (Fig. S4E). The DRP1-KO MEFs were increasingly sensitive to MEK inhibition within a dose range of 0 to 10 μM PD98059 (Fig. 6C) compared to the WT MEFs in a colony-forming assay. Quantification of the Crystal Violet staining (Fig. 6D) and colony diameter (Fig. 6E) after treatment with 5 μM PD98059 shows that PD98059 reduced cell proliferation of the DRP1-KO cells in high cell densities, whereas the WT cells were affected mostly at low cell densities. Consistent with this, PD98059 treatment also reduced the Sox-2 signal in the DRP1-KO cells at high cell densities (Fig. S4C). Next, we tested whether MEK inhibition with PD98059 would attenuate the aberrant cell proliferation of the p-ERK-positive Drp1-null PFCs in the differentiated follicle cell layer in Drosophila. We found that the presence of 100 μM PD98059 in the food rescued the aberrant cell proliferation in 37% (n=41) of Drp1-null PFC clones as they formed a single layer of differentiated cells in comparison to 13% (n=41) with the DMSO control (Fig. 6G). Therefore, the data suggest that enhanced MEK–ERK signaling in the absence of Drp1 allows the cells to escape contact inhibition to maintain aberrant cell proliferation in high-cell-density environments both in vitro and in vivo.

Next, we investigated how the mtCycE pool is linked to the cell proliferation status in the presence and absence of Drp1. Our data thus far show that DRP1-KO MEFs with higher mtCycE proliferate slower than the WT MEFs at low cell densities. Interestingly, we noted that DmCycE did not colocalize with the mitochondrial marker in the aberrantly proliferating Drp1-null PFCs in Drosophila (Fig. S2C). Therefore, we hypothesized that the increase in the mtCycE pool in the absence of Drp1 is lowered or released by the elevated ERK signaling at high cell densities to support aberrant cell proliferation. To test this hypothesis, we investigated the effects of cell density and MEK inhibition on mtCycE in growing colonies. The cells in the periphery of the growing colonies do not contact each other and are in a low-cell-density environment, whereas the cells within the colony are in a high-cell-density contact-inhibited environment (Gumbiner and Kim, 2014). In the WT MEFs the mtCycE pool was reduced in the contact-inhibited (high cell density) compared to proliferating state (low cell density) (Fig. 6F). Moreover, PD98059 reduced the mtCycE pool of the WT cells in low cell densities as it reduced cell proliferation. As expected, PD98059 did not affect the mtCycE pool of the contact-inhibited WT MEFs as it did not affect their proliferation. Therefore, we conclude that mtCycE in the WT cells is positively linked to their cell proliferation status. However, the elevated mtCycE pool in the DRP1-KO MEFs was associated with lower cell proliferation at low cell densities and the mtCycE pool was lowered to the levels of the WT cells at the high cell densities where the DRP1-KO cells proliferated aberrantly (Fig. 6F). Furthermore, PD98059 prevented the loss of mtCycE in the DRP1-KO MEFs at high cell densities as it attenuated their proliferation (Fig. 6F). PD98059, as expected, did not affect the mtCycE levels of the DRP1-KO cells in low cell densities as it did not affect their proliferation. This suggests that aberrantly elevated mtCycE in the DRP1-KO MEFs is inhibitory towards cell proliferation in low-cell-density environments and is released in high-cell-density environments in a MEK–ERK-dependent manner to allow aberrant cell proliferation.

Here, we provide the first evidence of a direct mitochondrial control of CycE mediated by its recruitment to the organelle. Our findings are based on the comparison of CycE properties in the presence or absence of the mitochondrial fission protein Drp1 in MEFs (in vitro) and the Drosophila follicle cell layer (in vivo). Our data support a model whereby a certain level of mtCycE, lacking the N-terminal phospho-degron (T62) and maintained by Drp1, likely through modulation of mitochondrial bioenergetics, is positively linked to cell proliferation, with deregulation of such an mtCycE pool deregulating cell proliferation (Fig. 7). We speculate that the mtCycE pool might be under local bioenergetic control at the contact sites between mitochondria and endoplasmic reticulum as crosstalk between these organelles modulates mitochondrial bioenergetics (Cárdenas et al., 2010). Correlation of the mtCycE pool with the non-mtCycE pool suggests that part of the total CycE pool might get recruited to the mitochondria in a controllable fashion. We propose that the mitochondrial recruitment of CycE upon enhancement of the bioenergetic capability and its controlled release upon lowering of mitochondrial bioenergetics might be part of the normal life cycle of CycE. Our data suggest that bioenergetics-driven mitochondrial docking of CycE might prevent phosphorylation of the N-terminal degron and thus prevent degradation of the molecule. Indeed, loss of mitochondrial bioenergetics, which would release mtCycE, leads to active degradation of the molecule (Mandal et al., 2005). Therefore, we hypothesized that increasing the steady-state mtCycE-pool would prevent degradation and deregulate CycE. Indeed, loss of Drp1 boosts mitochondrial energetics and enhances the potentially degradation-resistant mtCycE pool to maintain elevated levels of the molecule throughout the cell cycle. Such deregulation of CycE attenuates cell proliferation in low-cell-density environments but supports aberrant cell proliferation in high-cell-density environments because the mtCycE pool undergoes MEK–ERK-dependent release (Fig. 7).

It remains to be examined whether reduction of the excess mtCycE pool is sufficient to allow aberrant cell proliferation in the absence of Drp1 in high-cell-density environments or whether release of the excess mtCycE in the absence of Drp1 actively promotes S phase entry at high cell densities. Nonetheless, our current model implies that certain differentiated states maintained with reduced Drp1 would allow mitochondria to act as a reservoir of inactive CycE that could be released by undue MEK–ERK activation allowing aberrant cell proliferation and possibly initiating tumorigenic events. This concept explains why the Drosophila MBCs with elevated mtCycE (Fig. 1F) and reduced Drp1 activity only differentiate in the absence of EGFR–MEK–ERK signaling (Klusza and Deng, 2011; Mitra et al., 2012), which would otherwise release the mtCycE pool to aberrantly maintain cell proliferation and prevent differentiation. By contrast, the Drosophila PFCs having lower mtCycE (Fig. 1F) and higher Drp1 activity differentiate under the influence of EGFR–MEK–ERK signaling (Klusza and Deng, 2011; Mitra et al., 2012).

MEK–ERK signaling might release mtCycE by altering mitochondrial bioenergetics or by directly impacting upon the mitochondrial recruitment of CycE as p-ERK can re-localize to mitochondria (Rasola et al., 2010; Horbinski and Chu, 2005). Recently, Ras-driven MEK–ERK signaling has been shown to activate Drp1-dependent mitochondrial fission (Serasinghe et al., 2015; Kashatus et al., 2015) in line with our previous finding that EGFR signaling (mediated by downstream activation of MEK–ERK signaling) promotes mitochondrial fission (Mitra et al., 2012). We have further shown that loss of Drp1-driven mitochondrial fission aberrantly enhances EGFR signaling to support aberrant cell proliferation (Mitra et al., 2012). Consistent with this, our current findings show that Drp1 loss maintains ERK-dependent cell proliferation only at high cell densities. Activation of the Ras–Raf–MEK–ERK pathway brought about by oncogenic Ras causes CycE-dependent cell transformation (Geng et al., 2003, 2007) only when CycE is susceptible to degradation (Minella et al., 2005). Our study, based on DRP1-KO MEFs that are immortalized (using SV-40T antigens) (Ishihara et al., 2009) and a Drosophila model system, predicts that lowering of Drp1 activity would prevent oncogenic Ras-driven transformation at low cell densities because Ras-driven MEK–ERK signaling would fail to release mtCycE for degradation. Indeed, recent reports attempting to transform cells in the absence of Drp1 (using combinations of transforming agents; Kashatus et al., 2015; Serasinghe et al., 2015) have demonstrated that Ras-driven transformation is attenuated when Drp1 is lowered. However, Ras-driven MEK–ERK signaling in cells with reduced Drp1 at high cell densities would release mtCycE to support aberrant cell proliferation and possibly lead to cell transformation.

Our findings clearly point towards involvement of contact inhibition pathways in modulating the interplay of Drp1-driven mitochondrial fission and the EGFR–Ras–MEK–ERK pathway. Cell–cell contact signals through various pathways impinge on the Hippo pathway (McClatchey and Yap, 2012). Interestingly, Drp1 loss in the Drosophila follicle cell layer closely phenocopies loss of Hippo signaling in the increase in CycE levels and maintenance of aberrant cell proliferation in defined cell contexts (Meignin et al., 2007; Mitra et al., 2012). Moreover, recent studies indicate that various aspects of the Hippo pathway involve mitochondria (Sing et al., 2014; Ohsawa et al., 2012; Del Re et al., 2014; Nagaraj et al., 2012). Therefore, directed studies to investigate the interplay between Drp1 and components of EGFR–Ras–MEK–ERK and Hippo pathways will facilitate understanding of the exact role of mitochondrial fission in cell proliferation and tumorigenesis.

Biochemicals were obtained from Fisher Biochemicals or Sigma. Other materials used were as follows: oligomycin, CCCP, aphidicolin, nocodazole and PD98059 (Sigma); Drp1 shRNA (Dharmacon); Premo FUCCI Cell Cycle Sensor (Life Technologies); Flouromount G (SouthernBiotech); FuGENE^® HD Transfection Reagent (Promega); Vectashield (Vectorlabs); Micro BCA Protein Assay Kit, RIPA buffer (Pierce); Luminata Forte Western HRP substrate (Millipore); 4% paraformaldehyde aqueous solution and Triton X-100 (FisherScientific); Labtek chambers (Nalgene Nunc International); Millicell EZ slides (Millipore); DMEM (GIBCO); Graces Medium w/L-Glut (Bioworld); Nutri-fly BF (Genesee Scientific); Oligofectamine (Invitrogen); and CycE and Allstar siRNA (Qiagen).

Primary antibodies were against: 44/42 MAPK (ERK1/2) (western blotting, 1:1000; immunofluorescence, 1:100) phosphorylated p44/42 MAPK (western blotting, 1:1000; immunofluorescence, 1:100) CycE1 (HE12) (western blotting, 1:1000; immunofluorescence 1:100) Cyclin A (BF683) (1:300) Cyclin B1 (1:1000) p-CycE1(T62) (western blotting, 1:500; immunofluorescence 1:100) and CDK2 (1:10000) (Cell Signaling); actin AB-5 (1:10000; BD Biociences), Tom20 (western blotting, 1:10000; immunofluorescence, 1:200) and DmCycE (d-300; 1:100) (Santa Cruz Biotechnology), Sox2 (western blotting, 1:1000; immunofluorescence, 1:300; BD Bioscience), HSP-60 (1:200; BD Bioscience), ATP-B (1:100; Abcam) and Geminin (1:100; a gift from Mary Lilly, Section of Gamete Development, NICHD, Bethesda, USA). Secondary antibodies were from Jackson ImmunoResearch Laboratories.

Cells used were immortalized WT and DRP1-KO MEFs (a gift from Katsuyoshi Mihara, Dept. of Molecular Biology, Kyushu University, Fukuoka, Japan) the U87MG glioblastoma cell line, cells isolated from the GBM xenograft line D456MG, the A2780 ovarian cancer cell line and the DU145 prostate cancer cell line. All cells were maintained in Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum (FBS) using standard cell culture techniques. 1×10⁶ and 1.8×10⁵ MEFs were seeded per well of a 12-well plate to obtain high and low cell densities, respectively. Cells were harvested for immunoblot analyses 16 to 24 h post seeding.

Cells were transfected using FuGENE^® with 500 ng of DNA per well of an eight-well Labtek chambers and analyzed after 24–48 h, or using Oligofectamine with 120 pmols of CycE siRNA and analyzed after 72 h. For transduction of the FUCCI cell cycle probe, the manufacturer's protocol was followed after optimization. The transduction of the DRP1 shRNAs (5′-AATAAGTTGGAGTAAAGTAGC-3′ and 5′-TTCAATAACCTCACAATCTCG-3′) or scrambled-shRNA-expressing lentiviral particles were performed following standard methods and cells were harvested after 72 h.

Cells were synchronized in S phase using aphidicolin (2.5 µg/ml, 16 h) and in the G2-M phase using nocodazole (1 µg/ml, 16 h). Propidium iodide staining (50 µg/ml, 1 h) was performed following our previous protocol (Mitra et al., 2009) and stained cells were analyzed by flow cytometry at the Center for Aids Research Flow Core Facility, UAB using a Becton Dickinson LSRII flowcytometer. Data were acquired using software Becton Dickinson FACSDiva ver. 8.0 and analyzed using Becton Dickinson Verity House ModFit software.

Mitochondria were isolated from MEFs by modifying a published protocol (Frezza et al., 2007). Briefly, cells were lifted from the plate, centrifuged and then homogenized. The homogenate was centrifuged at 600 g to obtain the non-mitochondrial fraction 1 (F1). Supernatant, containing mitochondria, was centrifuged at 10,000 g to obtain the mitochondrial pellet and the non-mitochondrial fraction 2 (F2) as the supernatant. F1 and F2 were pooled to obtain the total non-mitochondrial fraction. Protein estimation was performed using a standard micro BCA kit.

For immunoblot analyses, asynchronous or synchronized cells (extracted in Laemmli buffer), or mitochondrial and non-mitochondrial lysates (extracted in RIPA buffer) were run on 12% polyacrylamide gels and transferred onto PVDF membranes following standard techniques. After blocking with fat-free milk, membranes were probed with relevant primary antibodies followed by horseradish peroxidase (HRP)-tagged secondary antibodies. The signal was detected using a chemiluminescence system.

Previously published Drosophila stocks (Mitra et al., 2012) were maintained and crosses performed at room temperature in vials containing NutriFly food. Drp1-null or control FRT40A clones were generated by crossing males of the genotype hsflp; ubiquitin nls GFP (UbiGFP) FRT40A/CyO with virgin females of the genotype drp1^KG03815 FRT40A/CyO or FRT40A/CyO. The reciprocal cross with the drp1^KG03815 FRT40A/CyO was avoided as it yielded fewer progeny. The progenies selected against CyO with genotype hsFLP; drp1^KG03815FRT40A/ubiGFP FRT40A were collected within 5 days of eclosion. The adult flies were heat pulsed in a 38°C water bath for 1 h to generate follicle cell clones and were maintained in food with granulated yeast for the following 8 days and the ovaries were dissected thereafter. For the experiment using PD98059, the heat-shocked flies were maintained in food containing 100 µM PD98059 or DMSO control for 8 days and the ovaries were dissected thereafter. To quantify the clonal divisions, the total number of nuclei in each clone was counted from confocal micrographs and expressed as Log (base 2) values.

Cells were seeded in eight-well Millicell chambers or glass-bottom Matek dishes (for colony assay), fixed using fresh 4% paraformaldehyde and permeabilized with freshly prepared 0.1% Triton X-100. Importantly, we avoided any PBS wash before fixation as that decreased the mtCycE pool. Fixed and permeabilized cells were blocked in 1% BSA in PBS (5% BSA for CycE staining) followed by incubation with primary and subsequently Cy3- or Cy5-tagged secondary antibodies with PBS washes after each incubation. Finally, slide chambers were mounted with Fluoromount G with Hoechst 33342 (10 µg/ml) to stain DNA and analyzed using confocal microscopy.

Drp1-null mosaic Drosophila ovaries were isolated and immediately fixed in fresh 4% paraformaldehyde and then dissected into ovarioles, which were then permeabilized in PBS with Triton X-100 (0.5%). Only longer permeabilization (30 min) allowed us to identify the mtCycE pool. Permeabilizied ovarioles were blocked in 2% BSA in PBS with Triton X-100 followed by incubation with primary antibodies and then conjugated secondary antibodies with PBS Triton X-100 washes after each incubation. Ovarioles were stained with Hoechst 33342 and mounted with Vectashield and analyzed using confocal microscopy.

300 cells were seeded in each 35-mm tissue culture dish and allowed to grow for 7 (low cell densities) to 10 days (high cell densities) without replenishing the medium. PD98059 or DMSO was added in fresh medium after 24 h of seeding as required. Colonies were stained using Crystal Violet solution (0.25%) and quantified by measuring Crystal Violet absorbance on a Bio-Tek Powerwave HT multiplate reader with Gen 2.05 software or by measuring colony diameters using Carl Zeiss's Zen software. For the purpose of immunostaining, the growing colonies were replenished with fresh medium 24 h before fixation to avoid the confounding influence of nutrient or growth factor deprivation.

High-resolution fluorescence microscopy was performed using the LSM 700 laser scanning confocal microscope (Carl Zeiss). Optical slices were acquired with a 1.2 N.A. 40× Plan Neofluar oil objective using the appropriate microscope settings. Image processing and colocalization analyses were performed on 30 or more cells using the Zen Black software (Carl Zeiss). Colocalization was expressed as a Pearson's coefficient (R) or as a fraction of the colocalized pool, as appropriate. For regression analysis, the total colocalized and non-colocalized pools were obtained by multiplying the mean colocalized or non-colocalized signal by the number of pixels in each category. For quantification of signals of p-ERK or Sox-2, more than 300 cells were included.

Oxygen consumption rates (OCRs) were obtained using the Seahorse Extracellular Flux Analyzer (Seahorse Biosceince). 60,000 cells were seeded in each 24-well chamber and measurements obtained after 16–20 h. Three readings were obtained from each well for basal or CCCP-stimulated oxygen consumption. The optimal dose of CCCP (5 µM or 1 µM in the presence or absence of serum, respectively) was used after proper titration (Fig. S3A).

Student's t-test was used for assessing statistical significance of the difference in mean values. Linear regression analyses and its statistical significance was performed using R (version 3.1.2). The R² goodness of fit was calculated by fitting the linear model of the data and the P-value was obtained using Pearson's correlation two-sided test.

We acknowledge K. Mihara for sharing the immortalized DRP1-KO and WT MEFs; M. Lilly for sharing Geminin antibody; Kristin Spraggins and Onna Marie Baldwin for help with data analyses; Marion Spell for the help in the Flowcytometry analyses; Michael Sack, NHLBI, for allowing access to Seahorse extracellular flux analyzer and Richa Rikhy for valuable discussions.