Highlights
Downregulation or inhibition of MCL-1 causes loss of OCT4 and NANOG expression
MCL-1 downregulation or inhibition causes elongation of the mitochondrial network
MCL-1 binds and modulates DRP-1 and OPA1 expression and activity
Interactions between MCL-1 and DRP-1/OPA1 are disrupted by MCL-1 inhibitor
Summary
Human pluripotent stem cells (hPSCs) maintain a highly fragmented mitochondrial network, but the mechanisms regulating this phenotype remain unknown. Here, we describe a non-cell death function of the anti-apoptotic protein, MCL-1, in regulating mitochondrial dynamics and promoting pluripotency of stem cells. MCL-1 is induced upon reprogramming, and its inhibition or knockdown induces dramatic changes to the mitochondrial network as well as loss of the key pluripotency transcription factors, NANOG and OCT4. Aside from localizing at the outer mitochondrial membrane like other BCL-2 family members, MCL-1 is unique in that it also resides at the mitochondrial matrix in pluripotent stem cells. Mechanistically, we find MCL-1 to interact with DRP-1 and OPA1, two GTPases responsible for remodeling the mitochondrial network. Depletion of MCL-1 compromised the levels and activity of these key regulators of mitochondrial dynamics. Our findings uncover an unexpected, non-apoptotic function for MCL-1 in the maintenance of mitochondrial structure and stemness.
Introduction
Considerable efforts have been made to identify the gene networks that regulate the pluripotent state and control the first steps of differentiation, when cells start to acquire a cell-lineage-specific identity. These studies have identified key transcriptional regulatory networks that determine the pluripotent state (Kumar et al., 2014). In contrast, we know relatively little about how different organelles, such as mitochondria, adapt to the changing environments they encounter during differentiation and reprogramming. While increased mitochondrial fragmentation is a fundamental feature of reprogramming (Prieto et al., 2016), the molecular mechanisms underlying this phenomenon are not fully understood.
The mitochondrial dynamics machinery is composed of dynamin superfamily guanosine triphosphatases (GTPases) that have roles in either fission (division) or fusion of mitochondria. Dynamin-related protein 1 (DRP-1) is required for mitochondrial fission. DRP-1 activation is mediated in part by phosphorylation, ubiquitination, and sumoylation, which allow for increased recruitment to various receptors (e.g., FIS1) (Wai and Langer, 2016). However, these mechanisms of enhanced DRP-1 recruitment to the outer mitochondrial membrane are not well characterized in pluripotent stem cells (PSCs), where they appear to be constitutively active. Once activated, DRP-1 forms oligomers and assembles into rings that constrict around the mitochondria, dividing the organelle into separate entities (Westermann, 2010). Fusion is mediated through the activity of the optic atrophy type 1 (OPA1) GTPase at the inner mitochondrial membrane and of mitofusins (MFN1 and MFN2) at the outer mitochondrial membrane (Chen and Chan, 2010). Mitochondrial dynamics is beginning to emerge as a crucial factor in regulating cell fate (Khacho et al., 2016, Longo and Archer, 2013, Martinou and Youle, 2011).
Many connections between the mitochondrial dynamics machinery and apoptosis have been made. Apoptosis is mediated by the BCL-2 family of proteins (Llambi et al., 2011). When cells are subjected to stress, the mitochondrial network has been shown to undergo DRP-1- and FIS1-mediated fragmentation in a process involving BAX translocation and co-localization with DRP-1 and endophilin B1 (Friedman and Nunnari, 2014, Karbowski et al., 2002, Sheridan et al., 2008). Previous reports suggested that BAX and BAK could also influence MFN2 function in healthy cells (Karbowski et al., 2006). The data shown in this report demonstrate that an anti-apoptotic protein belonging to the BCL-2 family, myeloid cell leukemia 1 (MCL-1), is a fundamental regulator of mitochondrial dynamics in human PSCs (hPSCs), independent of its role in apoptosis. MCL-1 is essential for embryonic development and for the survival of various cell types (Rinkenberger et al., 2000). Interestingly, recent studies have proposed that MCL-1 may also drive changes in cancer cell metabolic profiles to promote the biosynthesis of substrates needed for proliferation (Andersen and Kornbluth, 2012, Opferman, 2015).
To examine the function of MCL-1 in both undifferentiated and differentiated cells, we took advantage of various features of the hPSC model, including (1) high expression of MCL-1, (2) a predominantly fragmented mitochondrial network, and (3) the ability to induce differentiation into early progenitor stages. This model provided an ideal system to examine not only the potential role of MCL-1 in mitochondrial structure but also its effects on cell fate. Here, we report MCL-1 as a modulator of mitochondrial dynamics in PSCs and demonstrate the value of the stem cell model in studying the plasticity of the mitochondrial network during reprogramming and differentiation.
Results and Discussion
Human Pluripotent Stem Cells Downregulate MCL-1 upon Differentiation
We previously reported that human embryonic stem cells (hESCs) become resistant to DNA damage as differentiation is engaged (Dumitru et al., 2012). This resistance is in part associated with changes in the activation status of BAX, as well as mitochondrial priming (Dumitru et al., 2012, Liu et al., 2013). We first sought to probe the status of the apoptotic machinery by measuring levels of various BCL-2 family members during differentiation, when this apoptotic resistance is acquired. hESC colonies were grown in suspension in the absence of a matrix layer as aggregates, also known as embryoid bodies (EBs) (Odorico et al., 2001). We compared protein expression between hESCs and 3-day EBs by western blot analysis. Since differentiation causes apoptotic resistance, we expected to detect an overall increase in the expression of anti-apoptotic proteins and a decrease in the expression of pro-apoptotic proteins. As expected, the pluripotency transcription factor NANOG was downregulated in EBs, and sirtuin-2 (SIRT2), a NAD+-dependent deacetylase, which is induced upon differentiation (Ramalho-Santos et al., 2002), was upregulated. BAX and BCL-XL showed no detectable changes in expression during differentiation, and BCL-2 was significantly upregulated. In stark contrast, MCL-1 was significantly downregulated after only 3 days of differentiation (Figure 1A). MCL-1 downregulation was also detected in mouse ESC-derived EBs (Figure S1A). As reported previously, these results indicate that BCL-2 may be critical mediator of cell death resistance in early differentiation (Liu et al., 2013). Counterintuitively, MCL-1 does not appear to be a major determinant of the apoptotic resistance characteristic of EBs (Dumitru et al., 2012).
To determine whether downregulation of MCL-1 was a particular trait of EB formation, we differentiated hESCs into trophoblast-like cells by addition of bone morphogenetic protein 4 (BMP4) and into neural progenitor cells using dual SMAD inhibition, as previously reported (Amita et al., 2013, Chambers et al., 2009) (Figures S1B–S1F). Differentiation to both lineages was accompanied by a significant downregulation of MCL-1 (Figures S1C and S1D), mimicking the EB formation results. This was unexpected, since EBs, trophoblast-like cells, and neural progenitor cells are more resistant to apoptosis than undifferentiated stem cells (Figures S1E and S1F).
To probe whether the downregulation of MCL-1 had a functional role in maintaining pluripotency, we examined the effect of inhibiting MCL-1 in hESCs. Cell-permeable, selective small-molecule antagonists of BCL-2, BCL-XL, and MCL-1, known as BH3 mimetics, have been identified (Ni Chonghaile and Letai, 2008). We first used the pan-BCL-2 inhibitor GX15-070 (Chiappori et al., 2012), which inhibits BCL-2, BCL-XL, and MCL-1, and compared it with the BH3-only mimetic ABT-737 (Oltersdorf et al., 2005), which only inhibits BCL-2 and BCL-XL. Since these small-molecule inhibitors target the main anti-apoptotic proteins, we expected treatment with etoposide in combination with either inhibitor to accelerate cell death onset compared to treatment with only etoposide. Not surprisingly, treatment with ABT-737 increased the sensitivity of hESCs to etoposide-induced DNA damage (Figure 1B). However, rather unexpectedly, treatment with GX15-070 had the opposite effect and resulted in reduced cell death sensitivity in response to etoposide (Figure 1B), pointing to a paradoxical effect of MCL-1 inhibition in mediating apoptotic resistance.
These results, together with the observed reduced MCL-1 expression, suggest that MCL-1 inhibition induces a partial protective effect that resembles the decrease in apoptotic sensitivity engaged during differentiation. Consistent with this hypothesis, treatment of hESCs with GX15-070 reduced the expression of pluripotency markers NANOG and OCT4, while treatment with ABT-737 did not affect the expression of these pluripotency factors (Figures 1C–1E). To confirm the effect of MCL-1 on pluripotency, we then turned to MIM-1, an MCL-1-specific small-molecule inhibitor (Cohen et al., 2012), which caused the significant decrease of both NANOG and OCT4 expression (Figures 1F–1H, S1G, and S1H). SOX2 expression was not affected by MCL-1 inhibition, suggesting that hESCs treated with MIM-1 may be differentiating to the default neuroectoderm lineage. These results indicate that MCL-1 may be necessary for hPSC maintenance of pluripotency.
MCL-1 Is Induced during Reprogramming, and Its Inhibition Causes Changes in the Stem Cell Mitochondrial Network
We next examined whether MCL-1 expression upregulation was also engaged during the transition to pluripotency. We first reprogrammed human fibroblasts by delivering pluripotency factors using the non-integrating Sendai-based vector system (Figures S2A and S2B). Compared with fibroblasts, MCL-1 is expressed at high levels in human induced PSCs (hiPSCs) (Figure 2A) and hESCs (Figure 2B), both of which are highly sensitized to cell death (Dumitru et al., 2012) (Figure S2C). Along with high expression levels of MCL-1, we also observed that iPSCs have a fragmented mitochondrial network when compared with their parent human fibroblast line (Figure S2D), consistent with previous reports suggesting that mitochondria undergo vast changes during reprogramming in response to overexpression of pluripotency factors (Chen and Chan, 2017).
We next investigated whether MCL-1 has a role in the maintenance of mitochondrial dynamics in PSCs. We inhibited MCL-1 in hESCs using MIM-1 and examined its effects on mitochondrial structure. In response to MCL-1 inhibition, the mitochondria appear to fuse and become more elongated, as shown by cytochrome c staining (Figures 2C and 2D). We hypothesized that these changes in mitochondrial shape could be orchestrated through crosstalk between MCL-1 and the proteins involved in mitochondrial dynamics. We first interrogated the expression levels of active DRP-1 in response to MCL-1 inhibition. Phosphorylation of DRP-1 on Ser-616 enhances DRP-1 activity (Taguchi et al., 2007). Cells treated with MIM-1 displayed downregulated DRP-1 phosphorylation (p-DRP-1 S616) compared with vehicle control cells (Figures 2E and 2F), providing evidence for a role of MCL-1 in the regulation of DRP-1 activity.
To confirm that the effects of the small-molecule inhibitor MIM-1 were due specifically to MCL-1 inhibition, we performed loss-of-function experiments utilizing an RNAi approach. MCL-1 expression was knocked down in hESCs using small interfering RNA (siRNA). As seen with the small-molecule inhibitors of MCL-1, transmission electron microscopy images confirmed significant elongation of the mitochondria in MCL-1 knockdown hESCs in comparison with scramble siRNA controls (Figure 3A). Importantly, OCT4 and p-DRP-1 Ser-616 levels were also significantly decreased upon MCL-1 knockdown (Figures 3B and 3C), as seen in the presence of MIM-1. Therefore, MCL-1 appears to affect pluripotency, at least in part, through the regulation of DRP-1 activity.
Expression of MCL-1 at the Mitochondrial Matrix Delays Differentiation
It has been shown that MCL-1, besides localizing to the outer mitochondrial membrane, also localizes to the mitochondrial matrix in mouse embryonic fibroblasts and some human cancer lines. The mitochondrial targeting sequence of MCL-1 resides in an N-terminal region that is cleaved from the wild-type form, allowing MCL-1 to be trafficked into the matrix. We investigated the effect of these separate forms of MCL-1 on maintaining pluripotency encoded by three expression vectors (Figure 3D) (Perciavalle et al., 2012): wild-type Mcl-1 (Mcl-1WT), mitochondrial outer membrane-located Mcl-1 (Mcl-1OM), and mitochondrial matrix-located Mcl-1 (Mcl-1Matrix). MCL-1WT localizes to both the outer mitochondrial membrane and the matrix. MCL-1OM was obtained by mutating the arginine residues at positions 5 and 6 of MCL-1 to alanine. MCL-1OM localizes to the mitochondrial outer membrane, preserving MCL-1 anti-apoptotic function, but it cannot be imported to the mitochondrial matrix. MCL-1Matrix was obtained by fusing the mitochondrial targeting sequence of matrix-localized ATP synthase to N-truncated MCL-1 (lacking the first 66 amino acids). MCL-1Matrix localizes exclusively to the mitochondrial matrix.
Ectopic expression of MCL-1OM showed the expected decrease in OCT4 expression in response to BMP4 (Figure 3E). Interestingly both MCL-1WT and MCL-1Matrixcaused a delay in differentiation in response to BMP4, as shown by a decrease in OCT4-negative cells (Figures 3E and S3). These results indicate that MCL-1 may have an alternative function at the mitochondrial matrix in maintaining pluripotency.
MCL-1 Interacts with and Maintains the Stability of DRP-1 and OPA1
To confirm that MCL-1 is in fact localized to the mitochondrial matrix in hPSCs, we co-expressed a GFP-tagged MCL-1 construct (EGFP-MCL-1) and a DsRed-mito construct, which encodes a truncated form of cytochrome c oxidase subunit 2 (COX2) that localizes exclusively to the mitochondrial matrix (Figure 4A). Line-scan measurements of fluorescence show that MCL-1 co-localizes with the matrix marker, DsRed-mito (Figure 4B). The localization of MCL-1 at both the outer mitochondrial membrane and at the matrix in stem cells suggests that MCL-1 could be interacting with DRP-1 (at the outer membrane) to promote mitochondrial fragmentation and/or OPA1 (at the matrix) to repress fusion of the mitochondrial network in hESCs.
To test this possibility, we performed co-immunoprecipitation experiments using both total cell fractions and mitochondrial purifications to pull down MCL-1 and check for binding to DRP-1 and OPA1 (Figure S4A). Western blot analysis identified a shifted band roughly 20 kDa larger than the expected DRP-1 band (Figure S4B). We speculate that this could be a modified, active form of DRP-1, presumably due to its phosphorylation and sumoylation when in its active state (Anderson and Blackstone, 2013, Chang and Blackstone, 2010). In the total cell fraction, we did not detect binding between MCL-1 and OPA1. However, we detected strong binding in mitochondrial preparations of hESCs (Figure S4B). These in vitro biochemical assays suggest that MCL-1 is binding to both DRP-1 and OPA1 in human embryonic stem cells.
We then used a proximity ligation assay (PLA) to confirm binding of these proteins in situ (Figure S4C). We first confirmed MCL-1 interaction to the BH3-only protein, BIM. BIM is known to bind MCL-1 by inserting its BH3 domain into MCL-1’s surface groove (Martinou and Youle, 2011). We detected the expected interaction of MCL-1 and BIM, as indicated by red fluorescent puncta (Figure S4D). We then probed for binding of MCL-1 with both OPA1 and DRP-1. Indeed, as seen in the immunoprecipitation experiments, we detected MCL-1 interaction with OPA1 (Figure 4C) and DRP-1 (Figure 4D). We then used a derivative of a recently reported MCL-1 inhibitor, S63845 (Kotschy et al., 2016), which works by competing for binding to the BH3 domain of MCL-1. We confirmed that S63845 also caused changes to the mitochondrial network (Figure S4E). As expected, S63845 effectively disrupted the interaction between MCL-1 and BIM (Figure S4D). Interestingly, as indicated by the decrease in the number of puncta, S63845 (MCL-1i) disrupts interactions with mitochondrial dynamics machinery at both the outer mitochondrial membrane (MCL-1:DRP-1) and the matrix (MCL-1:OPA1) (Figures 4C and 4D). These results suggest that MCL-1 associates with both GTPases through its characteristic hydrophobic binding groove (Billard, 2013).
Previously, we determined that MCL-1 downregulation affected the activity of DRP-1, but not its stability (Figure 3B). Interestingly, MCL-1 downregulation did cause a significant decrease in OPA1 expression (Figure 4E), indicating that MCL-1 influences OPA1 stability, but not necessarily its function, since the mitochondria in hESCs are still capable of fusing when MCL-1 is inhibited (Figure 3A). Mitochondrial fusion is key for efficient assembly of the electron transport chain (ETC). Interestingly, OPA1 has also been implicated in maintaining oxidative phosphorylation in cells. Thus, our results suggest that preventing OPA1 activity (either in fusion or in OXPHOS) could be crucial for maintaining stem cell pluripotency.
Taken together, our findings imply that MCL-1 modulates mitochondrial dynamics and pluripotency through interactions with mitochondrial fission and fusion regulators. Based on our observations, we propose a model in which MCL-1 at the outer mitochondrial membrane regulates cell death and promotes DRP-1 activity and mitochondrial fragmentation (Figure 4F). Our results with S63845 (Letai, 2016) suggest that these interactions are modulated through the MCL-1 BH3 binding groove. It would be interesting to test whether there are two pools of MCL-1, one that is binding to the pro-apoptotic proteins and inhibiting apoptosis and another capable of binding and regulating DRP-1, promoting mitochondrial fragmentation. It will also be necessary to follow up the effects of MCL-1 inhibition-induced mitochondrial changes on the metabolic state of PSCs. In addition, examining the metabolic requirement for MCL-1 during reprogramming could be intriguing in light of recent reports suggesting a multi-stage reprogramming process (Lee et al., 2016).
The process of differentiation into progenitors and committed cells is likely regulated, at least in part, by a shift in dependence to MCL-1 anti-apoptotic activity. It will be crucial to examine MCL-1’s function in adult stem cells from different tissues. Likewise, identifying the signaling pathways by which mitochondria-nucleus crosstalk occurs upstream and downstream of MCL-1 in stem cells and their progenitors, as has been shown with murine neural progenitor cells (Khacho et al., 2016), is an important future direction. By studying the role of MCL-1 in mitochondrial dynamics, we can increase our understanding of the fundamental mechanisms governing pluripotency and self-renewal.