Time-dynamics of mitochondrial membrane potential reveal an inhibition of ATP synthesis in mitosis

The energetic demands of a cell are believed to increase during mitosis 1–7. As cells transit from G2 into mitosis, mitochondrial electron transport chain (ETC) activity increases 4,8,9, and cellular ATP levels progressively decrease until the metaphase-anaphase transition 3,7,10, consistent with elevated consumption. The rates of ATP synthesis during mitosis, however, have not been quantified. Here, we monitor mitochondrial membrane potential of single lymphocytes and demonstrate that cyclin-dependent kinase 1 (CDK1) activity causes mitochondrial hyperpolarization from G2/M until the metaphase-anaphase transition. By using an electrical circuit model of mitochondria, we quantify the time-dynamics of mitochondrial membrane potential under normal and perturbed conditions to extract mitochondrial ATP synthesis rates in mitosis. We found that mitochondrial ATP synthesis decreases by approximately 50 % during early mitosis, when CDK1 is active, and increases back to G2 levels during cytokinesis. Consistently, acute inhibition of mitochondrial ATP synthesis failed to delay cell division. Our results provide a quantitative understanding of mitochondrial bioenergetics in mitosis and challenge the traditional dogma that cell division is a highly energy demanding process.


Abstract
The energetic demands of a cell are believed to increase during mitosis 1-7 . As cells transit from G2 into mitosis, mitochondrial electron transport chain (ETC) activity increases 4,8,9 , and cellular ATP levels progressively decrease until the metaphase-anaphase transition 3,7,10 , consistent with elevated consumption. The rates of ATP synthesis during mitosis, however, have not been quantified. Here, 5 we monitor mitochondrial membrane potential of single lymphocytes and demonstrate that cyclindependent kinase 1 (CDK1) activity causes mitochondrial hyperpolarization from G2/M until the metaphase-anaphase transition. By using an electrical circuit model of mitochondria, we quantify the time-dynamics of mitochondrial membrane potential under normal and perturbed conditions to extract mitochondrial ATP synthesis rates in mitosis. We found that mitochondrial ATP synthesis 10 decreases by approximately 50 % during early mitosis, when CDK1 is active, and increases back to G2 levels during cytokinesis. Consistently, acute inhibition of mitochondrial ATP synthesis failed to delay cell division. Our results provide a quantitative understanding of mitochondrial bioenergetics in mitosis and challenge the traditional dogma that cell division is a highly energy demanding process. 15

Main text
In animal cells, including most cancer cells, energy in the form of ATP is produced mainly through oxidative phosphorylation in mitochondria (reviewed in [11][12][13] ). However, largely due to a lack of quantitative single-cell approaches, little is known about ATP synthesis during mitosis. To study mitochondrial bioenergetics at the single-cell level, we combined suspended microchannel resonators 20 (SMR), a non-invasive single-cell buoyant mass sensor, with a fluorescence detection system. This allowed us to monitor cell mass-normalized fluorescence signals with a temporal resolution of 2 min and a fluorescence measurement error of 2 % without perturbing normal growth 14 Fig. 3). However, the mouse Fl5.12 pro-B lymphocytes did not display any change in TMRE at the end of cell cycle ( Supplementary Fig. 3A).
L1210 cells were utilized as model system in all further studies. 35 To validate that the spike-like TMRE increase reflects an increase in ∆Ψm, we used quenching concentrations (10 µM) of an alternative ∆Ψm probe, Rhod123 ( Supplementary Fig. 2B, Supplementary note 2). Rhod123 signal suddenly decreased approximately 1h prior to cell division ( Fig. 1B), consistent with a spike-like increase in ∆Ψm. In contrast, the ∆Ψm-insensitive MitoTracker Green probe did not display any changes at the end of cell cycle (Fig. 1C). We next examined changes 40 in plasma membrane potential (∆Ψp) using a 1 µM DiBAC4(3) probe. The DiBAC4(3) signal was reduced before and during the TMRE increase, indicative of increased ∆Ψp ( Supplementary Fig. 4A).
We speculated that this ∆Ψp change may be a feature of mitotic cell swelling 17,18 , which can be inhibited with 5-(N-ethyl-N-isopropyl)amiloride (EIPA) 17 . Indeed, treatment of L1210 cells with 5 µM EIPA partially inhibited the observed decrease in the DiBAC4(3) signal but did not affect the 45 TMRE signal increase ( Supplementary Fig. 4). This indicates that mitotic cell swelling associates with changes in plasma membrane potential, but the extent to which plasma membrane hyperpolarizes is not affecting the reliability of TMRE as a reporter for ∆Ψm (Supplementary note 2).
Next, we studied the exact timing of the ∆Ψm increase. Inhibition of mitotic entry using 2.5 µM CDK1 inhibitor RO-3306 completely eliminated the observed increase in TMRE signal, 50 despite the fact that cells continued to increase in size beyond the typical G2/M transition (Fig. 1D).
We then monitored single-cell density to compare TMRE signal increase to the timing of mitotic cell swelling, an event that is known to start in prophase 17, 18 . The increase in the TMRE signal was observed immediately following the onset of density reduction, indicating that mitochondrial hyperpolarization begins shortly after mitotic entry (Fig. 1E). This timing of TMRE increase was 55 further validated using biophysical markers of G2/M transition ( Supplementary Fig. 5) 14,15 . We next compared the timing of TMRE signal increase to the degradation of the protein Geminin using L1210 FUCCI cells, which express fluorescently labelled Geminin (Geminin-mAG) 15,19 . The Geminin-mAG signal was fully degraded in approximately 8.6 minutes at the metaphase-anaphase transition ( Fig. 1F, Supplementary Fig. 6), and the loss of Geminin aligned exactly with the maximal TMRE 60 signal (Fig. 1G). In most cells, the TMRE signal declined back to typically G2 levels before the final abscission of the daughter cells (Figs. 1E, 1G). Together, these results indicate that the mitochondrial hyperpolarization begins shortly after the G2/M transition, reaches a maximum at the metaphaseanaphase transition and returns to G2 levels during cytokinesis.
Previous work has shown that the CDK1/Cyclin B complex localizes to mitochondria 65 during mitosis and directly phosphorylates components of the mitochondrial ETC 4 . CDK1 also associates with other metabolic proteins, including the α subunit of ATP synthase 20 . CDK1 activity is minimal before mitotic entry, after which the switch-like activation of CDK1/cyclin B complex results in high CDK1 activity until the onset of anaphase 21-23 . Since the timing of mitochondrial hyperpolarization coincided exactly with the reported CDK1 activity, we hypothesized that the 70 switch-like CDK1 activity was causally responsible for mitochondrial hyperpolarization. To test this, we first arrested cells in a CDK1 active-state (prometaphase and metaphase) using three different chemicals: the kinesin motor inhibitor S-trityl-l-cysteine (STLC), the microtubule polymerization inhibitor nocodazole, and the anaphase-promoting complex inhibitor proTAME. In response to any of these three chemicals we observed that TMRE signal increased following mitotic entry and 75 plateaued to a high level during the mitotic arrest, indicative of mitochondria reaching a steady, hyperpolarized state. Next, we partially inhibited CDK1 with RO-3306 (1 µM) or with an alternative CDK1 inhibitor BMS-265246 (400 nM) and examined the level of TMRE during STLC mediated prometaphase arrest. Note that complete inhibition of CDK1 blocks mitotic entry, but it is possible to partially inhibit CDK1 while allowing mitotic entry and progression 15,17 . We observed that CDK1 Unexpectedly, when we treated L1210 cells in the G2 cell cycle phase with 1 µM oligomycin and monitored their growth using the SMR, the cells still proceeded through mitosis and divided symmetrically (Fig. 3A), although the magnitude of the TMRE signal increase in mitosis was reduced 95 (Fig. 3B). To further quantitatively analyze the role of mitochondrial ATP synthesis in mitotic entry and progression, we synchronized cells to G2 using RO-3306, treated the cells with 1 µM oligomycin for 15 min, released the cells to enter mitosis in the presence of oligomycin, and collected samples for cell cycle analysis at different timepoints. Surprisingly, mitochondrial ATP synthesis inhibition had little effect on mitotic entry and the subsequent appearance of G1 cells (Figs. 3C, 3D). Similar 100 results were observed in BaF3 and DT40 lymphocytes ( Supplementary Fig. 8). To further examine the extent to which ATP synthesis inhibition influences L1210 cell behaviour, we monitored singlecell mass accumulation (growth) rates using a serial SMR 15,24 . We observed that oligomycin treatment caused a major reduction in cell growth rates that persisted for several hours (Fig. 3E).
Thus, mitochondrial ATP synthesis is required to support cell growth, but not cell division. This 105 finding is consistent with prior observations that mitochondrially localized dominant negative form of CDK1 did not affect G2/M progression despite reducing mitochondrial respiration 4 , and that cells devoid of mitochondrial DNA and mitochondrial ATP production can proliferate despite significantly reduced growth rates 25 .
Next, to more directly examine bioenergetics and oxidative stress in mitosis, various 110 fluorescence based metabolic reporters were expressed in both L1210 and BaF3 cells. However, the expression of these exogenous proteins resulted in the loss, or even the reversal of the normal mitotic mitochondrial hyperpolarization ( Supplementary Fig. 9, Supplementary note 3), indicating that these genetic tools can bias quantitative analyzes of mitotic mitochondrial bioenergetics in our model system.

115
As an alternative approach to understand mitotic bioenergetics, we developed a model to derive ATP synthesis rates from the TMRE signal dynamics (Supplementary Fig. 10,   Supplementary note 4). First, we converted the TMRE signal to approximate ∆Ψm using existing measurements of tetramethylrhodamidine ester dye accumulation and membrane potentials 26 .
Second, we assumed that the voltage across the inner mitochondrial membrane (∆Ψm) is determined 120 by the currents through ATP synthesis (IATP), voltage-dependent leakage 27 (ILeak) and the ETC (IETC).
Third, we modelled the inner mitochondrial membrane as an electrical circuit with voltage (∆Ψm), capacitance (C) and resistances (R) for each one of the currents (Fig. 4A, Supplementary note 4), constituting a simple model that is consistent with the biochemical view of mitochondria 16,28 . We assumed that the circuit behaves in a switch-like manner between two distinct states whether CDK1 125 is activated (CDK1 on) or inactivated (CDK1 off). By fitting our model's analytical solution to the ∆Ψm data we derive RC values, which reflect the time constant of the ∆Ψm change (Figs. 4B, 4C, Supplementary notes 5 and 6). We derived these RC values separately for each control and oligomycin-treated cell during CDK1 on and CDK1 off states (Fig. 4D, Supplementary Fig. 11).
Comparing the RC values between control and oligomycin-treated cells, we extracted the resistance 130 of ATP synthase (RATP) during the CDK1 on and CDK1 off states (Fig. 4B). We found that RATP is higher during the CDK1 on state than during CDK1 off state ( Supplementary Fig. 11C, Supplementary note 4). In addition, comparing RC values during the CDK1 on and off states in each control cell revealed that in order for RATP to increase, RLeak and RETC are required to cumulatively decrease during the CDK1 on state (Supplementary note 7). We then derived the current through ATP synthase (IATP), i.e. the ATP synthesis rate, throughout mitosis using Ohm's law (IATP=V/RATP) (Supplementary notes [7][8][9]. Surprisingly, our modelling revealed that the mitochondrial ATP synthesis rate is inhibited by 54 % ± 11 % (mean±s.e.m.) during prometaphase and metaphase, when compared to G2 ATP synthesis rates (Figs. 4E, 4F). During anaphase, only a minor increase (<10%) in comparison to G2 levels was observed (Fig. 4E). Overall, this temporal control of ATP synthesis results in 40 % 140 decrease in total mitochondrial ATP production during early mitosis (between G2/M transition and metaphase-anaphase transition) when compared to a situation where mitochondrial ATP synthesis would remain at G2 levels (Fig. 4G).
It is important to recognize the limitations of our approach. Especially, i) the TMRE signal is a proxy for ∆Ψm, subject to systematic errors, ii) we assumed that oligomycin only perturbs Supplementary note 10). Since our modeling relies on comparisons between control and oligomycintreated cells, any systematic bias that affects TMRE in both samples will not affect our ATP synthesis results.
Our findings are compatible with existing literature. Our observation that ∆Ψm can increase in mitosis in the presence of oligomycin (Fig. 3B) is supported by the reported mitotic 160 activation of the ETC 4,8,9 . Both ETC activation and ATP synthesis inhibition cause ∆Ψm to radically increase, and high ∆Ψm is known to promote mitochondrial protein import 9 , reactive oxygen species (ROS) generation 29 , proton leakage 27 and heat production 30 . Consistently, mitochondrial protein import, ROS levels and cellular heat output have been reported to increase during mitosis in a CDK1dependent manner 9,31-33 . Furthermore, the ATP synthesis dynamics we discover can explain the 165 reported ATP level dynamics in mitosis 3,7,10 .
Overall, our work reveals the previously unknown dynamics of ∆Ψm and mitochondrial ATP synthesis during mitosis. Considering that mitochondria are responsible for the majority of cellular ATP synthesis [11][12][13] and that ATP consumption in interphase consumes cellular ATP pools within minutes 32 , our discovery that mitochondrial ATP synthesis is inhibited during mitosis suggests 170 a much lower rate of ATP consumption during mitosis than previously assumed. Notably, cells maintain ATP at concentrations near 4 millimolar 7,34 , but most enzymes have Michaelis constants (Km) for ATP in the micromolar range. Thus, even a major decrease in cellular ATP levels will not affect enzymatic reaction rates, and the intracellular ATP pools may be able to fulfill the energetic needs of mitosis even in the absence of additional ATP synthesis, as suggested by early work on 175 antephase 35,36 . The decreased ATP levels may even promote cell division by facilitating cellular reorganization and chromatin condensation 7,37 .

Competing interests
Scott R Manalis is a co-founder of Travera and Affinity Biosensors, which develops techniques relevant to the research presented.    Boxplots on the right indicate the level to which TMRE increases during mitotic arrest. p-values were obtained using ANOVA followed by Sidakholm test.     The colored areas reflect total amounts of ATP synthesized during early mitosis (red) and anaphase 365 (light yellow).
(F) Quantifications of relative ATP synthesis rates (IATP) in G2 (dark blue) and mitosis (light blue) as modelled on a single-cell level in non-arrested cells (left), and as estimated based on oxygen consumption rates (OCRATP) in cell populations arrested to G2 and mitosis (right). Note that both approaches rely on comparing control and oligomycin treatment conditions to derive ATP synthesis 370 rates, but the model accounts for changes in ∆Ψm and does not require cell cycle synchronizations.
(G) Quantifications of the total amount of mitochondrial ATP synthesis during early mitosis (from G2/M to M/A transition, red) and during anaphase (from M/A transition to end of anaphase where cell elongation is complete, light yellow), when compared to a null-hypothesis where ATP synthesis 375 rate remains at G2 levels throughout mitosis (dashed horizontal line). Data depicts mean ± s.e.m. of the RATP values (N=40-32, n=85-32).