Characteristics of Mitochondria at G₁–S.

A single, tubular mitochondrial element has not been considered a normal mitochondrial phenotype because it is rarely seen within cells (11). Its specific occurrence at the narrow window of G₁–S transition, however, suggested it might serve particular cell cycle functions. So, we took a number of approaches to identify its specific characteristics.

We first investigated whether the tubular mitochondrial system at G₁–S has a continuous matrix. A rectangular box across the mitochondrial network of a mito-NRK cell in G₁–S or other stages/conditions was photobleached, and the kinetics of recovery of the expressed RFP-matrix molecule was measured (Fig. 2 A and B). Fluorescence recovers within 2 min to >60% of the prebleached level from adjacent mitochondrial areas in G₁–S cells and cells expressing DRP1m. Virtually no recovery occurs in proliferating or mitotic cells. Mitochondria at G₁–S thus have their outer and inner membranes fused sufficiently to give rise to a continuous matrix to permit the matrix markers to diffuse freely between mitochondrial elements.

Open in a separate window

Fig. 2.

Mitochondria in cells at G₁–S or in cells expressing DRP1m-expressing cells have a continuous matrix. (A) FRAP analysis for assaying mitochondrial matrix continuity in mito-NRK cells at different cell cycle stages or expressing DRP1m. Mitochondrial fluorescence was bleached in the rectangular box, and recovery of fluorescence into the bleached region from adjacent mitochondrial elements was monitored. Images at the prebleach, postbleach, and 120-s recovery point are shown. (B) Plot of recovery kinetics from the FRAP experiment shown in A. Comparable results were obtained from 3 or 4 separate experiments. (Scale bar: 5 μm.)

We next investigated whether the continuous mitochondrial system at G₁–S has electrically continuous inner membranes. Mitochondria were loaded with tetramethylrhodamine ethylamine (TMRE), which incorporates specifically into mitochondrial membranes because of their transmembrane potential (12). A small region of interest (1 × 1 μm) of the TMRE-loaded mitochondria was irradiated by using a 2-photon laser. The cell was then monitored for TMRE loss from mitochondrial branches directly connected to the point of irradiation, indicative of spread of depolarization (13). In cells arrested at G₁–S, depolarization at the irradiated point (Fig. 3A, arrow in preirradiation image) causes immediate TMRE loss in many mitochondrial elements throughout the cell (Fig. 3A, arrowheads in postirradiation image), indicating these elements are electrically continuous. As mitochondrial elements undergo fusion and fission, the depolarization spreads to most remaining mitochondrial elements (Movie S5). Measuring the amount of TMRE loss in the outlined box over time reveals the extent of TMRE loss after irradiation (Fig. 3B). On microirradiation in cells whose mitochondria are irreversibly fused together through expression of DRP1m, a similar, and even faster, spread of mitochondrial depolarization is observed (Fig. 3 A and B). By contrast, mitochondrial membrane depolarization is restricted to a small zone around the irradiated point in proliferating, mitotic, and G₀ cells (Fig. 3 A and B). Thus, only mitochondria in cells at G₁–S or in cells expressing DRP1m show rapid spread of membrane depolarization.

Open in a separate window

Fig. 3.

The giant tubular mitochondrial form at G₁–S is electrically continuous and associated with increased membrane potential and elevated ATP production. (A) Microirradiation of TMRE-loaded mitochondria to measure electrical continuity of mitochondria by rapid spread of depolarization from an irradiated point (see Results). (Scale bar: 5 μm.) (B) Plot for loss of TMRE fluorescence during the microirradiation experiment. Fluorescence associated with the rectangular boxes in A before and after irradiation is plotted in each cell condition. Comparable results were obtained from 3 to 4 separate experiments. (C) Scatter plot of mitochondrial TMRE/MitoTracker Green uptake in individual cells from populations enriched in different stages of the cell cycle or in cells treated with FCCP (10 μM, 16 h) after 6 h of release from G₀. Data points are color-coded according to their cell cycle stage. A total of 50 cells was counted for each condition in 3 independent experiments. (D) Contribution of mitochondrial ATP at different stages of cell cycle. At each stage, samples were treated with or without oligomycin (10 μg/mL for 4 h), and the absolute difference between the treated and untreated ATP values was then expressed as the percentage of the untreated value. Data derived from the average of 3 experiments are shown with the bars showing the standard deviation.

We further examined mitochondrial membrane potential (assessed by TMRE uptake) per unit of mitochondrial mass (determined by using MitoTracker Green) at different cell cycle stages (Fig. 3C). Normalization of TMRE signal with MitoTracker Green in this assay is essential because mitochondrial mass increases during the cell cycle. A predictable decrease in mitochondrial potential after membrane depolarization by the protonophore carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP) (14) is observed. Notably, when mitochondrial potential is monitored in cells at different cell cycle stages, the membrane potential is greatest at G₁–S.

Oxygen consumption by mitochondria is reported to increase from early to late G₁ (3). To test whether there is also a change in mitochondrial ATP production at G₁–S, we measured levels of ATP at different cell cycle stages. Total ATP increases at G₁–S (Fig. S4A), with mitochondrial ATP output (quantified as the fraction of total cellular ATP sensitive to oligomycin) significantly higher at G₁–S compared with other stages of the cell cycle (Fig. 3D).

Mitochondrial Depolarization Specifically Blocks G₁-to-S Cell Cycle Progression in a p53-Dependent Manner.

Previous work in flies has shown that reduction of cellular ATP caused by a mutation in a mitochondrial electron transport chain component triggers the p53-dependent G₁–S checkpoint (4). Therefore, we investigated whether depolarizing mitochondria by FCCP treatment, which prevents the G₁–S-associated increased mitochondrial ATP output, will trigger the p53-dependent G₁–S checkpoint. Cells released from G₀ without treatment show BrdU incorporation and Aurora B expression (Fig. 4A and Fig. S5), indicative of S-phase entry. FCCP-treated cells, however, neither incorporate BrdU nor express Aurora B in the nucleus, similar to cells in G₀ (Fig. 4A). When cells in late G₁ are treated with FCCP, proliferating cell nuclear antigen (PCNA) foci (indicative of S-phase entry) do not form, even after 10 h, unlike that in untreated cells, which form PCNA foci within 4 h (Fig. 4B). FACS analysis of DNA content in the FCCP-treated cells reveal cell enrichment of the G₁ population, indicative of a G₁–S block (Fig. S6A). Cells treated with FCCP on exit from mitosis, by contrast, continue further into G₁ (Fig. 4C), and cells in S phase treated with FCCP still progress through this phase (albeit more slowly than control cells) (Fig. S6 B and C). This suggests that reducing mitochondrial potential by FCCP treatment specifically blocks progression from late G₁ to S.

Open in a separate window

Fig. 4.

Depolarization of mitochondria causes a specific block in G₁-to-S transition in a p53–p21-dependent manner. (A) Effect of FCCP on BrdU incorporation and Aurora B expression in a cell population at G₀ (G₀), or released from G₀ for 6 h and maintained with or without in FCCP (10 μM) for 18 h (G₀→control; G₀→FCCP). The effect was measured in approximately ≈400 cells per condition. (B) Time-lapse images of mito-NRK cells expressing PCNA-GFP to detect G₁-to-S progression in cells treated with or without FCCP. (Scale bar: 5 μm.) (C) FACS analysis of DNA content of NRK cells arrested in/before mitosis, and subsequently shifted into control medium with or without FCCP. Markers are M₁ for G₁; M₂ for S phase; and M₃ for G₂; numbers represent the proportion of cells in these stages. (D) FACS analysis of DNA content in HCT116 cells (p53+/+ and p53−/−) on treatment with FCCP. Proliferating cells were arrested in G₀ and then released in the presence of FCCP (10 μM, 18 h) (G₀→FCCP). Numbers denote percentage of cells in G₁, G₂, and S. (E) Immunoblot analysis showing FCCP's effect on cyclin E and p21 levels, and its relationship to p53. Proliferating HCT116 cells (p53+/+ or p53−/−) were treated with FCCP for 18 h, and cell lysates were immunoblotted with relevant antibodies. Tom20 served as the loading control. Comparable results were obtained from 3 to 4 separate experiments.

FCCP treatment of NRK cells reduces total ATP levels by ≈30% (Fig. S4B), which could trigger the p53-dependent metabolic checkpoint at G₁–S (15). To test whether FCCP-induced G₁–S arrest was p53-dependent, we used isogenic HCT116 lines expressing or lacking p53 (16). Whereas both cell lines progress through S phase in a similar manner in the presence of FCCP (Fig. S6D), when FCCP is added after G₀ release, G₁–S arrest is observed in p53+/+ cells but not in p53−/− cells (Fig. 4D). The p53/p21 stress-sensing module is known to induce a G₁–S block in stressed cells by preventing cyclin E accumulation/activity (17), so we monitored cyclin E and p21 expression levels in FCCP-treated, p53+/+ HCT116 cells. Both reduced cyclin E levels and increased p21 expression are observed (Fig. 4E and Fig. S7A for NRK cells). Hence, reducing mitochondrial transmembrane potential through FCCP treatment results in a p53-dependent G₁–S arrest involving p21.

The Presence of Hyperfused Mitochondria Induces Cyclin E Buildup.

Levels of cyclin E rise in G₁–S and then fall again in S phase in coordination with other cyclins (18). This allows cyclin E to play a specific role in S phase, including initiation of DNA replication. To investigate whether the transient formation of hyperfused mitochondria at G₁–S and its subsequent breakdown into isolated tubular elements in S is linked to cyclin E regulation, we induced mitochondrial hyperfusion and then examined cyclin E levels within cells. Mitochondria were induced to become hyperfused through treatment with mdivi-1, a drug that tubulates mitochondria by specifically inhibiting the mitochondrial fission protein, DRP1 (19). We found that ≈50% of cells in an asynchronous population exhibit highly tubular mitochondria within 3 h of treatment (Fig. S8A). Longer mdivi-1 treatment (>5 h) results in more cells with highly fused mitochondria, but there is also increased cell death, so we restricted our experiments to 4 h of treatment. Notably, immunoblot analysis of mdivi-1-treated cells reveals a time-dependent increase in cyclin E levels over 4 h, with no change in cyclin A levels (Fig. 5A). Cyclin E levels also increase when hyperfused mitochondria are induced by DRP1m-GFP overexpression for 24 h in either NRK (Fig. 5B) or HCT116 cells (Fig. S7B). Thus, cellular cyclin E levels rise whenever mitochondria are hyperfused.

Open in a separate window

Fig. 5.

A hyperfused mitochondrial form leads to G₀-to-S transition by increasing cyclin E levels independent of growth factors. (A) Immunoblot analysis for cyclin E and cyclin A during treatment with mdivi-1. Proliferating HCT116 cells were treated with mdivi-1 (50 μM) for the indicated times, and cell extracts were immunoblotted with the indicated antibodies. Tom20 and Tubulin served as controls. (B) Immunoblot analysis of cyclin E in NRK cells overexpressing DRP1m or EGFP. Tom20 served as loading control. (C) Immunofluorescence showing BrdU incorporation after treatment of cells in G₀ with mdivi-1. NRK cells at G₀ (serum starved) were treated with mdivi-1 (50 μM, 4 h) and assessed for BrdU incorporation (see Methods). (Scale bar: 20 μm.) (D) Quantification of BrdU incorporation induced by mdivi-1 treatment compared with that by serum treatment in HCT116 cells maintained at G₀ (serum starved). BrdU incorporation was quantified in G₀ (G₀) cells treated with 10% FBS (G₀+FBS) and in G₀ cells treated with mdivi-1 (G₀+mdivi-1). (E) Immunoblot analysis showing cyclin E induction by mdivi-1 treatment of G₀ cells is independent of growth factor signaling. Analysis was performed by using HCT116 cells that were serum starved (G₀); serum starved followed by addition of 10% FBS (G₀+FBS); or serum starved followed by 50 μM mdivi-1 treatment for 4 h (G₀+mdivi-1). Parallel G₀ populations were first treated with PD098059 (50 μM, 45 min) before addition of FBS (G₀+FBS+PD098059) or mdivi-1 (G₀+mdivi-1+PD098059). Cell extracts were immunoblotted for cyclin E and cyclin D with tubulin serving as loading control.

Hyperfused Mitochondria Induce G₀ Cells to Enter S Phase.

We next investigated whether cyclin E buildup in response to hyperfused mitochondria affects cell cycle progression. Normally, serum-starved cells remain in G₀ indefinitely unless growth factors are added to induce signaling pathways for cyclin E buildup (20). Because inducing hyperfused mitochondria leads to increased cyclin E levels, we tested whether this was sufficient to drive serum-starved G₀ cells into S phase in the absence of growth factors.

NRK cells in G₀ were treated with mdivi-1 to prevent mitochondrial fission and thereby maintain mitochondria in a highly fused form. The cells were then monitored for BrdU incorporation to determine initiation of DNA replication. A considerable fraction of the serum-starved cells incorporate BrdU within 4 h of mdivi-1 treatment, in contrast to untreated, serum-starved cells (Fig. 5C). The increase in cells incorporating BrdU under mdvi-1 treatment is similar to that observed after growth factor (FBS) replenishment for 4 h, as measured in HCT116 cells (Fig. 5D).

The increased BrdU incorporation in mdivi-1-treated cells is correlated with increased cyclin E levels in both HCT116 (Fig. 5E) and NRK cells (Fig. S8B). The MEK kinase inhibitor PD098059 (21) prevents cyclin E buildup in growth factor-stimulated cells but not in mdivi-1-treated cells (Fig. 5E), suggesting cyclin E accumulation in mdivi-1-treated G₀ cells occurs downstream of growth factor signaling. Consistent with this, there is no increase in the growth factor cyclin, cyclin D, in mdivi-1-treated cells, in contrast to growth factor-stimulated cells (Fig. 5E).

Cell Culture and Imaging.

Cells were maintained by standard tissue culture techniques, and G418 (1 mg/mL) was used to generate stable lines. Immunofluoresent staining was performed following standard techniques. For BrdU incorporation experiments, cells were incubated with 100 μM BrdU for 10 min, which was detected by immunostaining with anti-BrdU antibody. Imaging was performed by using a laser-scanning confocal microscope (LSM510; Carl Zeiss MicroImaging). In live-cell experiments, cells were imaged in buffered medium on a preheated microscope stage (37 °C). Appropriate laser lines for each fluorophore were used.

High-resolution images were acquired by using the 63×, 1.4 N.A. Plan Neofluar oil-immersion objective. Optical slices were taken along the z axis covering the whole depth of the cell.

For fluorescence recovery after photobleaching (FRAP) analyses, the same 63× objective was used. The fluorescence was bleached and recovery was monitored every second for 2 min. Quantification of recovery kinetics was performed according to ref. 30.

The microirradiation experiment was performed using the 63× Plan Fluor oil objective of the Zeiss LSM 510 META system. Two-photon laser light at 800 nm and 25% power was used to irradiate a region of interest of 15/15 pixels (1 × 1 μm) on TMRE-loaded mitochondria. Images were acquired with the 543-nm line every 3 sec after irradiation until 18 sec or 2 min, as required.

Image processing was performed using the Zeiss LSM software (version 3.4). See SI Methods for details.

Analyses of Mitochondrial Properties.

Cells were first incubated with MitoTracker Green (100 nM) for 15 min followed by TMRE (50 nM) for another 15 min. A 40× Plan Neofluar oil-immersion objective was used to acquire images. Mitochondrial potential per unit mass was assessed as a ratio of the TMRE/MitoTracker Green signal. Total cellular ATP was assayed using the ATP determination kit (Molecular Probes), and the mitochondrial contribution was assayed using oligomycin.

For morphometric analyses, a 3D stack of high-resolution confocal images was used for calculating the volume of individual mitochondrial elements within a single cell from different cell cycle stages. Within the merged 3D stack, mitochondrial elements were segmented (31). After segmentation, surface points were extracted for each identified mitochondrial object in the 3D image. The minimum distance between the surface points of the objects and a selective set of quantitative features was computed for each object and used for subsequent statistical analysis. Using all the extracted features, an unbiased volume quantification was used to quantify mitochondrial tubulation. See SI Methods for details.

Cell Cycle Synchronization and Identification of Cell Cycle Stage.

A combination of synchronization methods including serum starvation, aphidicolin, and nocodazole treatment was used. See SI Methods and Fig. S3 for details.

In fixed cells, cell cycle stages were identified by BrdU incorporation, Aurora B staining, and propidium iodide staining. In live cells, cell cycle stages were identified through expression of cell cycle markers, as described in detail in SI Methods.

We thank R. J. Youle, M. McNiven, C. Cardoso, Y. Wang, J. Chen, B. Vogelstein, and J. Nunnari for reagents and S. Dwarkapuram for help in the morphometric analysis and R. Hegde, R. J. Youle, C. Smith, M. Lilly, A. Arnaoutov, and members of the Lippincott-Schwartz Laboratory for valuable suggestions. B.R. acknowledges National Institutes of Health Grant R01 EB005157 and National Science Foundation Grant EEC-9986821.