The dynamin-related GTPase Dnm1 regulates mitochondrial fission in yeast

Mitochondrial membranes form nets in dnm1 mutant cells

We previously reported that mitochondrial membranes collapse to one side of the cell in a dnm1Δ mutant strain¹¹. Transmission electron microscopy indicated that these collapsed membranes might be organized in an unusual structure¹¹. To characterize this structure further, we studied mitochondrial morphology in dnm1Δ cells under several different conditions.

As reported previously, dnm1Δ mutants grown at 25 °C lack the highly branched mitochondrial network characteristic of wild-type cells (Figs 1a, b, 2; 98% of wild-type cells show such a network). Instead, mitochondrial membranes in the dnm1Δ strain collapse into one or more long tubular structures that contain fewer branches than in wild-type cells and remain near the cell cortex (Figs 1c, d, 2; 39%). These collapsed mitochondrial membranes often contain one or more expanded regions that appear as bright spots of variable size in the fluorescence microscope (Fig. 1d, 2; 49%). In a few cells, these bright regions expand further to reveal mitochondrial membranes organized as ‘nets’ (Figs 1e, f, 2; 11%). This net morphology is very unusual and has not been reported in other yeast mitochondrial-morphology mutants or cell types.

Genetic and biochemical studies indicate that the distribution of mitochondrial membranes in yeast is regulated, in part, by interactions with actin cables (bundles of actin filaments) which extend from the base of the mother cell into the growing bud (reviewed in ref. 15). We reasoned that such organelle–actin interactions might prevent mitochondrial nets from spreading out along the cortex, impeding their visualization in many cells. To test this idea, we combined the dnm1Δ mutation with a mutation in MDM20 (mitochondrial distribution and morphology, ref. 16) that disrupts actin cables. We reported previously that loss of actin cables in mdm20Δ cells causes moderate mitochondrial-inheritance defects at 25 °C, which become more severe at 37 °C (ref. ¹⁶). However, mitochondrial morphology in the mdm20Δ mutant remains wild-type at the permissive temperature (Figs 1g, h, ***********2; 72%). Mitochondrial nets occupy a greater volume of the cell and are more easily visualized in the dnm1Δ mdm20Δ double mutant (Fig. 1i, j). Moreover, in asynchronously growing cultures, the percentage of cells with visible mitochondrial nets increases from 11% in the dnm1Δ mutant to 81% in the dnm1Δ mdm20Δ double mutant (Fig. 2). We used the mdm20Δ mutation as a tool to enhance visualization and analysis of the mitochondrial nets formed in dnm1 mutant strains.

Transmission electron microscopy confirmed that mitochondrial membranes are organized as nets in dnm1Δ mdm20Δ cells (Fig. 3). Previous studies indicated that mitochondrial profiles in wild-type cells are usually oval or tubular in shape and are distributed evenly at the cell periphery^9,11. In contrast, mitochondrial membranes in dnm1Δ mdm20Δ cells are organized as flattened sheets perforated by a series of regularly spaced holes (Fig. 3a). Reconstruction of twelve serial sections containing these mitochondrial sheets reveals a curved mitochondrial net (Fig. 3b) resembling those visualized by fluorescence microscopy in the dnm1Δ and dnm1Δ mdm20Δ mutants (Fig. 1f, j).

The mitochondrial nets formed in dnm1 mutant strains could arise from a defect in mitochondrial fission. For example, the Dnm1 GTPase normally may function in vivo to ‘clip’ membranes that separate two adjacent holes in the mitochondrial net, releasing mitochondrial ends or tips and forming tubules.

Localization of Dnm1

Time-lapse imaging studies of wild-type mitochondrial networks revealed that fission events usually occur within a tubule (creating two ends) or at branchpoints in the network (creating one end and a tubule)¹⁷. If Dnm1 is required for mitochondrial fission, it should be associated with sites on the mitochondrial compartment that are preparing to divide (on the sides of tubules), are undergoing division (at constriction sites in tubules), or have just completed division (at the tips/ends of tubules). Our previous indirect immunofluorescence studies localized a haemagglutinin-tagged Dnm1 protein (Dnm1–HA_cp) to punctate structures at the tips and sides of mitochondrial tubules and at branchpoints in the network¹¹. Although this localization pattern was consistent with a role for Dnm1 in fission, indirect immunofluorescence techniques lack the resolution necessary to detect constriction sites in mitochondrial tubules. To examine sites of Dnm1 localization in more detail, we performed immunogold labelling experiments.

Immunogold labelling was carried out on ultrathin cryosections of cells expressing the Dnm1–HA_cp protein (strain JSY1781; ref. 11). Consistent with previous immunofluorescence studies¹¹, 87% of the 5-nm gold particles were found at the tips/ends of mitochondrial tubules (Fig. 4b–f, h, Table 1) and at discrete sites on the sides of mitochondrial tubules (Fig. 4g, i–l, Table 1). Interestingly, 34.6% of the gold particles observed were clustered at constriction sites on mitochondrial tubules, indicating that these tubules may have been undergoing division (Fig. 4i–l, Table 1). In some sections, label was detected on the end of a mitochondrial tubule in close proximity to another unlabelled mitochondrial end (Fig. 4d, e; there were 13 pairs of closely spaced tips with gold particles on only one of the two tips; 157 gold particles were counted). In cases in which two tips were juxtaposed, gold particles were never found on both tips. This labelling pattern could represent a recently completed division event. Together, these results support the idea that Dnm1 is required for mitochondrial fission.

We also observed gold particles on the ends of mitochondrial tubules in close proximity to the plasma membrane (Fig. 4b, c; 9 tips with 76 gold particles). Thus, Dnm1 may also localize to sites at which the outer mitochondrial membrane is anchored at the cell cortex. This localization is consistent with our previous finding that disruption of DNM1 interferes with the cortical distribution of mitochondrial membranes¹¹. Although the localization of the yeast mitochondrial network at the cell periphery is well documented (reviewed in ref. 15), the role of Dnm1 in mitochondrial anchoring and cortical distribution is unclear.

dnm1 mutations block mitochondrial fragmentation in fzo1-1 cells

Maintenance of a tubular mitochondrial network is regulated by opposing fission and fusion events¹⁷. The fusion reaction requires the transmembrane GTPase Fzo1, which localizes to the outer mitochondrial membrane⁹. As reported previously, shifting a conditional fzo1-1 mutation to the non-permissive temperature (37 °C) for 10 minutes results in the rapid and complete fragmentation of the mitochondrial network (Figs 5c, d, 6; ref. 9). We postulated that this fragmentation was caused by ongoing mitochondrial fission events regulated by the Dnm1 GTPase. To test this possibility, we analysed mitochondrial morphology at 25 °C and 37 °C in fzo1-1 dnm1Δ double mutant strains. At 25 °C, fzo1-1 dnm1Δ cells contained collapsed mitochondrial membranes and nets (Figs 5e, f, 6). These mitochondrial morphologies were similar to those observed in the single dnm1Δ mutant at both 25 °C and 37 °C (Figs 1d, f, 2, 5f, 6). Shifting fzo1-1 dnm1Δ cells to 37 °C for up to 3 hours had little effect on organelle morphology; wild-type networks were rarely seen (Figs 5i, j, 6), and mitochondrial membranes remained tubular and exhibited dnm1-like morphologies despite the absence of FZO1 function at the higher temperature (Figs 5g, h, 6). Thus, maintenance of a pre-formed net does not appear to require ongoing mitochondrial fusion. We also observed mitochondria with dnm1-like morphologies (95%) in fzo1Δ dnm1Δ strains at 25 °C, although these cells did not contain detectable nets. These results indicate that the dnm1 mitochondrial-morphology defect is epistatic to the fzo1-1 mitochondrial-morphology defect, and that the loss of Dnm1 function in the fzo1-1 dnm1Δ double mutant blocks the mitochondrial fission and fragmentation normally observed in fzo1-1 cells at the non-permissive temperature.

We showed previously that mitochondrial fragmentation in fzo1 cells results in secondary defects in mitochondrial DNA (mtDNA) maintenance and the loss of mitochondrial genomes⁹. This mtDNA loss is prevented when a dnm1Δ mutation is introduced into fzo1-1 or fzo1Δ cells. The retention of mtDNA in these strains presumably results from the unusual mitochondrial morphology in dnm1Δ cells. Consistent with this idea, we find that dnm1Δ mutations also prevent or reduce mtDNA loss in several other yeast mitochondrial-morphology mutants (S.W. Gorsich and J.M.S., unpublished observations).

Mitochondrial fusion is not restored in dnm1 fzo1-1 cells

The finding that dnm1 mutations block mitochondrial fragmentation in fzo1-1 cells is consistent with a role for Dnm1 in mitochondrial fission. Alternatively, activation of a FZO1-independent fusion pathway could prevent mitochondrial fragmentation in the fzo1-1 dnm1Δ double mutant. To test this idea, we studied mitochondrial fusion in fzo1-1 dnm1Δ cells during mating. We assessed mitochondrial fusion by labelling mitochondria in cells of opposite mating type with either a mitochondrial-targeted form of the green fluorescent protein (mito-GFP) or MitoTracker CMXRos (Molecular Probes) and examining the distribution of the fluorescent markers in budded zygotes formed at the permissive (25 °C) and non-permissive (37 °C) temperatures¹⁷. In wild-type zygotes formed at 25 °C or 37 °C, mitochondrial fluorescent markers co-localized, indicating that the mitochondrial membranes had fused and the mitochondrial contents had mixed (Fig. 7a–d, Table 2). Mitochondrial fusion also occurred efficiently in fzo1-1 × fzo1-1 zygotes formed at 25 °C (Table 2). As shown previously, mitochondrial networks fragmented and failed to fuse in fzo1-1 × fzo1-1 zygotes formed at 37 °C, indicating that fusion requires FZO1 function (Fig. 7e–h, Table 2; ref. 9). Consistent with its proposed role in mitochondrial fission, DNM1 function is not required for mitochondrial fusion, as indicated by the mixing of mitochondrial content observed in dnm1Δ × dnm1Δ zygotes at both temperatures (Fig. 7i–l, Table 2). Mitochondrial fusion observed in dnm1Δ cells is, however, strictly dependent on FZO1 function. Mitochondrial fusion is blocked in zygotes formed from fzo1-1 dnm1Δ haploid cells at 37 °C, but not 25 °C (Fig. 7m–p, Table 2). Taken together, these results show that maintenance of mitochondrial tubules in fzo1-1 dnm1Δ mutants results from a block in mitochondrial fission rather than the activation of an alternate fusion pathway in these cells. Moreover, with respect to mitochondrial fusion during mating, the fzo1 mutation is epistatic to the dnm1 mutation.

Yeast methods

Standard methods were used for growth, transformation and genetic manipulation of Saccharomyces cerevisiae. All rich (YPDextrose, YPGlycerol) and synthetic (SDextrose, SGlycerol) media were prepared as described²².

Yeast strains

All yeast strains are isogenic to strain FY10 (ref. ²³). The following strains were used in this study: JSY1781, MATa/MATα leu2Δ1/leu2Δ1 HIS3/his3Δ200 ura3-52/ura3-52DNM1-HA_c/DNM1-HA_c; JSY3094, MATa leu2Δ1 his3Δ200 ura3-52 mdm20Δ::LEU2; JSY3095, MATa leu2Δ1 his3Δ200 ura3-52 mdm20Δ::LEU2 dnm1Δ::HIS3; JSY3096, MATα leu2Δ1 his3Δ200 ura3-52; JSY3097, MATα leu2Δ1 his3Δ200 ura3-52 dnm1Δ::HIS3; JSY3161, MATa his3Δ200 ura3-52 fzo1-1; JSY3162, MATα his3Δ200 ura3-52; JSY3163, MATa his3Δ200 ura3-52 fzo1-1 dnm1Δ::HIS3; and JSY3164 MATα his3Δ200 ura3-52 dnm1Δ::HIS3. JSY3094–JSY3097 and JSY3161–JSY3164 are sister spores from the same tetrad.

Quantification of mitochondrial morphology phenotypes

Mitochondrial morphology was scored in wild-type and mutant cells expressing mitochondrial-targeted forms of GFP from plasmids (pDO12, ref. 11, and pVT100UGFP, provided by B. Westermann and W. Neupert). pVT100UGFP efficiently targets GFP to both rho⁺ (containing mtDNA) and rho° (lacking mtDNA) mitochondria. Alternatively, 100 μl culture was stained with 100 nM 3,3′-dihexyloxacarbocyanine (DiOC₆) (Molecular Probes) as described¹⁶. Cells were grown overnight in YPD or synthetic medium at 25 °C, diluted to 0.5 optical-density units per ml in fresh medium and incubated at 25 °C or 37 °C for an extra 3 h before viewing. Numbers in Figs 2, 6 represent the percentage of an asynchronously growing culture with a particular mitochondrial morphology (n=100–300). Comparable results were obtained using both staining methods, indicating that the mitochondrial-morphology phenotypes observed in mutant cells were not due to the presence of the Cox4–GFP fusion protein. Digital microscopic images of cells were acquired as described^9,11.

Mitochondrial-fusion assay

To examine mitochondrial fusion during mating, we transformed haploid strains JSY3161–JSY3164 of opposite mating type with the mito-GFP plasmid (pVT100UGFP) or labelled them with MitoTracker CMXRos (Molecular Probes); the strains were mated at 25 °C or 37 °C, and analysed as described¹⁷.

Electron microscopy

dnm1 mdm20Δ cells were fixed and embedded as described⁹. Serial sections (75 nm) were viewed with a Hitachi H-7100 transmission electron microscope. Digital images obtained from 12 consecutive sections were used to produce three-dimensional reconstructions of mitochondrial nets using Voxblast software (VayTek, Fairfield, IA).

Immunoelectron microscopy was performed essentially as described²⁴. Exponentially growing cells (30 °C) were fixed in suspension for 15 min by adding an equal volume of freshly prepared 8% formaldehyde in 1 × PBS, pH 7.4. The cells were pelleted, resuspended in fresh fixative (4% formaldehyde, 1 × PBS, pH 7.4) and incubated for an extra 18–24 h at 4 °C. The cells were washed briefly in PBS and resuspended in 1% low-gelling-temperature agarose. The agarose blocks were trimmed into pieces of 1 mm³, cryoprotected by infiltration with 2.3 M sucrose/20% polyvinyl pyrrolidone (10 K)/PBS (pH 7.4) for 2 h, mounted on cryopins and rapidly frozen in liquid nitrogen. Ultrathin cryosections were cut on a Leica UCT ultramicrotome equipped with an FC-S cryo-attachment and collected onto formvar/carbon-coated nickel grids. The grids were washed through several drops of 5% fetal calf serum (FCS) plus PBS containing 10 mM glycine (pH 7.4), blocked in 10% FCS for 30 min and incubated overnight in 20 μg ml^–1 monoclonal anti-haemagglutinin antibody (Covance Research Products, Richmond, CA). After washing, the grids were incubated for 2 h in 5-nm gold donkey anti-mouse conjugate (Jackson ImmunoResearch). The grids were washed through several drops of PBS followed by several drops of ddH₂O. Grids were then embedded in an aqueous solution containing 3.2% polyvinyl alcohol (10 K)/0.2% methyl cellulose (400 centiposes)/0.1% uranyl acetate. The sections were examined and photographed on a Philips 420 transmission electron microscope at 80 kV.